See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245161156

Damage and failure pattern of prefabricated structures after major

earthquakes in Turkey and shortfalls of the Turkish Earthquake code

Article  in  Engineering Failure Analysis · June 2006

DOI: 10.1016/j.engfailanal.2005.02.006

CITATIONS READS
43 622

3 authors:

Musa Hakan Arslan Hasan Korkmaz

Selcuk University Selcuk University
65 PUBLICATIONS   458 CITATIONS    16 PUBLICATIONS   430 CITATIONS   

SEE PROFILE SEE PROFILE

Fatma Gulten Gulay

Istanbul Technical University
26 PUBLICATIONS   243 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Steel fibre View project

Investigation of pre-damaged shear and bending beams View project

All content following this page was uploaded by Musa Hakan Arslan on 12 September 2019.

The user has requested enhancement of the downloaded file.

 Engineering Failure Analysis 13 (2006) 537–557
www.elsevier.com/locate/engfailanal

Damage and failure pattern of prefabricated structures after

major earthquakes in Turkey and shortfalls of the Turkish
Earthquake code
a,* a,1 b,2
M.H. Arslan , H.H. Korkmaz , F.G. Gulay
a
Selcuk University, Engineering and Architecture Faculty, Department of Civil Engineering, Konya, Turkey
b
Istanbul Technical University, Civil Engineering Faculty, Department of Civil Engineering, Istanbul, Turkey

Received 10 December 2004; accepted 11 February 2005

Available online 13 April 2005

Abstract

The cost of the damage of the earthquakes in the last 10 years has been serious especially on prefabricated industrial
structures. Therefore in this study, extensive, serious damage and failures in prefabricated reinforced concrete structures
in the August-17, 1999 Adapazari (Turkey) and November-12, 1999 Duzce (Turkey) earthquakes have been examined.
By performing a cause and effect assessment of the damage types experienced, some strengthening recommendations
have been made. To investigate the causes of the damage beside the assessment of practical imperfections, a comparison
between the sets of criteria encompassed in the Turkish Code, UBC-97 and Eurocode-8-98 Codes especially on the cri-
teria of base shear force, displacement and loads acting on the nodes/connections has been made. For the sake of a
more comprehensible illustration of the distinctions, the design of a pre-constructed industrial structure to the three
mentioned codes has been conducted. In order to compare the above-mentioned codes a sample prefabricated structure,
which totally failed after 1999 Marmara earthquake is analysed.

2005 Elsevier Ltd. All rights reserved.

Keywords: Failure characteristics; Industrial structures; Prefabricated; Earthquake codes

*
Corresponding author. Tel.: +90 332 223 1971; fax: +90 332 241 0635.
E-mail addresses: mharslan@selcuk.edu.tr (M.H. Arslan), hhk73@selcuk.edu.tr (H.H. Korkmaz), gulayg@itu.edu.tr (F.G. Gulay).
1
Tel.: +90 332 223 1965.
2
Tel.: +90 212 285 3337.

1350-6307/$ - see front matter
2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfailanal.2005.02.006
538 M.H. Arslan et al. / Engineering Failure Analysis 13 (2006) 537–557

Nomenclature

Vt base shear force

W weight of the structure
C lateral loading coefficient
A(T1) spectral acceleration coefficient
Ra(T1) earthquake load reduction coefficient
T1 first natural period of the ground
A 0, Z earthquake zone coefficient
I structural importance factor
S(T) spectrum coefficient
TB corner period of the ground
T natural period of the structure
Di inter-storey drift
H story height
l effective column height
m total mass
E·I total elastic rigidity
Fp force of the architecture member
Wp weight of the architecture member
Cv velocity spectrum coefficient
Ca acceleration spectrum coefficient
Nv earthquake proximity coefficient (fault proximity coefficient)
s elastic displacement ratio (elastic spectral drift)
de elastic displacement
R structural behaviour coefficient
Cp lateral load coefficient
Vb earthquake load acting on structure
Cd the design coefficient
S ground type coefficient
TC right corner period of the ground
R0 structural behaviour coefficient
kd quality level of structure
kp a coefficient related with the locus of joints
c a parameter in relations with the importance factor
r security parameter
F lateral force
hc column height
EcIc column elastic rigidity
d1, d 2 story displacement

1. Introduction

Industrialisation in Turkey affected the construction sector, and in the early 1960’s ‘‘prefabricated con-
struction’’ became a prominent sub-sector of the construction industry. The fundamental reasons lying be-
neath the use of prefabricated construction can be enumerated as: (a) reduction of construction period; (b)
 M.H. Arslan et al. / Engineering Failure Analysis 13 (2006) 537–557 539

improved quality-control of member fabrication in the factory environment; (c) increased rate of produc-
tion and economy compared to conventional construction techniques.
Contrary to the advantages that prefabricated technology provides, failures and damages observed in the
structures erected using this technique in destructive earthquakes that hit the country especially in the last
10 years (such as 1992 Erzincan (Mw 6.8), 1996 Adana-Ceyhan (Mw 6.3), 1999 Adapazari-Izmit _ (Mw 7.4),
1999 Düzce (Mw 7.2)), showed the need for the re-examination of the criteria in the Turkish Earthquake
Code (TEC) and revision of the code deficiencies relative to the criteria in UBC and Eurocode-8 (the codes
in the United States and the European Union, respectively).
This study, focusses on the analyses of single-storey prefabricated industrial structures, according to the
criteria deficiencies in the Turkish Code which are apt to cause damage. In order to compare the above-
mentioned codes a sample prefabricated structure, which totally failed after 1999 Marmara earthquake
is analysed. Moreover, damage and failure types observed in the prefabricated reinforced concrete buildings
after the earthquakes that hit Turkey in recent years will be examined and recommendations for strength-
ening will be put forward.

2. TEC’98, UBC’97 and Eurocode 8-98

2.1. Turkish Earthquake Code, TEC’98 (TDY’98, Turkish Acronym)

‘‘The Turkish Code for the Structures in Disaster Regions’’ [1] which came into operation in 1975, has
become inadequate due to advancements in structural technology and earthquake engineering over the
years. Aiming at improving the code, a code more consistent with the current requirements took effect
in 1997. The base shear force is computed by using Eqs. (1) and (2) as
W
AðT 1 Þ W
A0
I
SðT Þ
Vt ¼W
C ¼ ¼ ; ð1Þ
Ra ðT 1 Þ Ra ðT 1 Þ

AðT Þ A0
I
SðT Þ
C¼ ¼ ; ð2Þ
Ra ðT Þ Ra ðT Þ
with respect to the TEC’98.
Here, A0, denoting the effective ground acceleration coefficient, takes the values 0.10 g for the 4th and
0.40 g for the 1st degree seismic risk zones, respectively. I, the structural importance factor, takes values
varying in between 1 and 1.5 and takes 1 for a conventional reinforced concrete structure. S(T), the spec-
trum coefficient, is represented by a curve which gives the values of design acceleration spectrum varying
with the natural period, T, of the structure. The type of the curve differs depending on the ground charac-
teristics and each curve type gives a value of 2.5 at maximum
S ¼ 2:5 T < T B;
S ¼ 2:5ðT B =T Þ0:8 T > T B; ð3Þ
S P 0:10R T > T B .
TB, in Eq. (3), is the corner period appointed regarding the ground type Ra, defined as the earthquake
reduction or behavioural coefficient, is the indicator of the structural ductility adopting values varying be-
tween 3 and 8. For prefabricated structures with fixed connections at the ground level and hinges at upper
ends of the members, the coefficient becomes 5.
One of the fundamental changes in the 1998 Code is the dual classification of the structural systems
namely those with high ductility and normal ductility. With regulations on detailing given in the Code,
540 M.H. Arslan et al. / Engineering Failure Analysis 13 (2006) 537–557

structural systems possessing high ductility values can be designed. With higher ductility values, conditions
given in the Code for detailing, which are directly in correlation with the structural ductility, such as the
stirrup spacing, calculations of column-beam intersection zones, arrangement of the compression reinforce-
ment in the cross-section, and concrete quality have been becoming rigorous to satisfy the reduction of the
C coefficient in the earthquake load computation. Thus the structure has been rewarded for ductile behav-
iour. The converse holds true for systems, in which ductility levels are normal.
Under intensive seismic motions, due to the elasto-plastic deformation owing to ductility, large lateral
displacements form, resulting in the formation of secondary moments. To keep secondary moments at a
minimum, inter-storey drifts, in other words, the relative storey displacements, have been limited in the
Code. Eq. (4) shows TEC’98 displacement criteria.
ðDi Þmax 6 0:0035h;
0:02 ð4Þ
ðDi Þmax 6 h:
R
Since the R-value equals 5 in the buildings where all the earthquake loads are met by single-storey frames
(which have fixed supports at basement level, are hinged at the top, and can transmit moments) this limit
can be expressed as follows:
ðDi Þmax 6 ð0:0035h; 0:004hÞmin : ð5Þ
Given that the columns of the single pendulum type structures operate as corbel beams, the elastic nat-
ural period in single-storey industrial structures can be calculated as follows:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
l3
m
T ¼ 2p ð6Þ
3
E
I
In Eq. (6), m stands for the total mass borne by a single-axis system; l is the effective column height (the
distance from below the gutter to the socket entrance), and E · I is the total elastic rigidity of all the mem-
bers on the axis in question.
For lateral and axial seismic loads acting on the corbel-like members operating in a single floor such as
balconies, parapets, and chimneys and in non-structural members such as siding and division panels, the
following formula is used:
F p ¼ A0
I
W p : ð7Þ
Here, Wp refers to the weight of the architectural member.

2.2. United states building code, UBC’97

For the United States Building Code [2], C, the lateral earthquake load coefficient, is computed using the
formula:
C ¼ C v
I=R
T : ð8Þ
The values computed using the formula, however, should not lie outside the maxima and minima given
below:

C max ¼ 2:5C a
I=R
T ;

C min ¼ 0:11
C a
I; ð9Þ
C min ¼ 0:8
Z
N v
I=R:
 M.H. Arslan et al. / Engineering Failure Analysis 13 (2006) 537–557 541

In Eq. (9), Z refers to the earthquake zone coefficient and corresponds to A0 in TEC’98. The structural
behaviour coefficient, R takes the value of 2.2 for prefabricated structures. Nv is the earthquake proximity
coefficient which has never been mentioned in the Turkish Earthquake Codes. Cv and Ca are the velocity
and acceleration spectrum coefficients, respectively.
There are formulations in UBC’97, too, for limiting the inter-storey drifts. The respective elastic dis-
placement ratio among storeys is defined as
de
s¼ ð10Þ
h
and it is kept in between the following limit,
0:025
s6 ;
0:7
R ð11Þ
0:020
s6 :
0:7
R
For the earthquake loads acting on the joints, the formulation
F p ¼ Cp
W p ð12Þ
is used. For the lateral load coefficient, Cp;
C p ¼ 4C a I 6 0:70C a I ð13Þ
is proposed.

2.3. Eurocode-8-98

In the code [3], for the computation of the base shear force
V b ¼ Cd
W : ð14Þ
Here, W is the total weight of the structure, while Cd stands for the design spectrum coefficient. The design
spectrum coefficient Cd is
C d ¼ 2:5
A0
S=R T B 6 T 6 T c;
2=3
C d ¼ 2:5
A0
ðT c =T Þ =R T c 6 T 6 3; ð15Þ
2=3 5=3
C d ¼ 2:5
A0
ðT c =T D Þ ðT D =T Þ =R T P 3:
Here, A0 is the maximum ground acceleration coefficient for design purposes, while S is the ground type
coefficient (equal to 1 for medium-class earth, and 0.9 for loose sands and medium clays). The coefficient
R is obtained by the multiplication of the base value R0 by the factors presented below:
R ¼ k d k p R0 ð16Þ
kd, given in the equation, refers to the quality level of the structure’s ductility. In the design computa-
tions of structures possessing high ductility, kd is taken as 1 while for medium-ductility it equals 0.5. kp,
however, is a coefficient related with the locus of joints (connections) and varies between 1.0 and 0.75.
The Code assumes the value of the structural behaviour coefficient, R0, as 2.0 for prefabricated
structures.
Eurocode-8 limits the respective displacements of floors using the relation below:
0:004
c
s6 ð17Þ
k d
R0
I
542 M.H. Arslan et al. / Engineering Failure Analysis 13 (2006) 537–557

c, given in Eq. (17), is a parameter in relation to the importance factor and varies between 2.0 and 2.5. For
lateral earthquake loads acting on the connections at the roof level in the reverse-pendulum type of prefab-
ricated industrial building:
F p ¼ ð3A0 rI=R0 ÞW p ð18Þ
is proposed. Being a security parameter, r is taken as 2 in medium and high ductility regions bearing critical
loads, and 1.5 in low-ductility locations.

3. Comparison of proposed codes for a model building

As a model building, an industrial structure, which completely failed in the 1999-Kocaeli Earthquake,
will be analysed. The reasons for the collapse of the building will be given in detail later. The erection year
of the building is 1999, the same year as the Earthquake. In other words, the building had been designed
according to the criteria embodied in TEC-1998. In Fig. 1, views of the building two months before failure
and in its failed state are given. The type of building is known as a lambda type, i.e., a reinforced concrete
prefabricated production industrial structure possessing a two 20-m transverse spacing and 6-metre height
with connections transmitting moments. The enclosed area is 40 · 42 m (1680 m2) (Figs. 2 and 3). One axis
composed of two spacings weighs 200 kN. An 85 kN snow load is added to the overall weight. Having a
30 MPa ultimate strength (C30), the concrete has a modulus of elasticity of 32,000 MPa. The dimensions
of the cross-sections of the columns and beams are given in Table 1 and Fig. 4.

Fig. 1. The state of the building before and after the Kocaeli Earthquake.

Fig. 2. Dimensions of the model building.

 M.H. Arslan et al. / Engineering Failure Analysis 13 (2006) 537–557 543

Fig. 3. The column application scheme of the model building.

Table 1
Sections and corresponding area of reinforcement
Section b (cm) d (cm) As (cm2)
a,g 35 55 29.14
b 35 109 48.80
c,e 35 109 48.80
d 35 50 13.19
f 35 55 26.75

Fig. 4. Sections and mass accumulation in the model building.

The model factory is located in the Adapazari Organized Industrial Zone, thus the maximum value
(A0 = 0.4 g) for the Seismic Region Coefficient given in the Code was chosen. The structural importance
factor, I, is taken as 1 since the building is of the industrial type.
544 M.H. Arslan et al. / Engineering Failure Analysis 13 (2006) 537–557

The equivalent static lateral earthquake loads, ratios, and upper limits of the inter-storey drifts com-
puted in an elastic analysis and the loads estimated to act on the connection regions have been, respectively,
calculated for the earthquake codes TEC-98, UBC-97 and Eurocode 8-98 and the results and assumptions
are given in Tables 2–4.

Table 2
Comparison of base shear forces calculated concerning different codes
TEC-98
C = A0SI/R Ground type – Z3 C = 0.1112
TB = 0.60 s
T = 1.25 s
A0 = 0.4
R=5
S = 2.5(TB/T)0.8
Cmin > 0.10A0I W = G + 0.3K Cmin P 0.04
UBC-97
C = CvI/RT Ground type – SD C = 0.232
Cv = 0.64
Ca = 0.44
T = 1.25 s
Z = 0.4
R = 2.2
Nv = 1
Cmin P 0.11CaI W = DL Cmin P 0.0484
Cmin P 0.8ZNvI/R Cmin P 0.145
Cmax 6 2.5CaI/R Cmin 6 0.500
Eurocode-8-98
C = 2.5A0S(Tc/T)2/3/R Ground type – type-B C = 0.306
A0 = 0.30
S=1
Tc = 0.60 s
T = 1.25 s
R0 = 2.2
kp = 0.75
kd = 1
R = kp Æ kd Æ R0 = 1.5
CminP0.20A0 W = G + 0.2K Cmin P 0.06

Table 3
Comparison of displacement amounts
Code Elastic displacement, de s = de/h (ratio) Upper bound formula Parameter Upper bound
TEC-98 2.030 0.00338 0.0035 R=5 0.0035
0.02/R 0.004
UBC-97 4.204 0.00700 0.02/(0.7R) R = 2.2 0.013
Eurocode-8-98 5.586 0.00931 0.004c/(kd Æ R0 Æ I) R0 = 2 0.0053
kd = 2
c = 0.75
 M.H. Arslan et al. / Engineering Failure Analysis 13 (2006) 537–557 545

Table 4
Comparison of the estimated loads acting on connections
TEC-98
Fp = A0IWp Ground type: Z3 Fp = 0.40 A0
A0 = 0.40
I=1
UBC-97
Fp = CaWp Ca = 0.44 Fp = 0.44 A0
Eurocode-8-98
Fp = (3A0rI/R0)Wp A0 = 0.4 Fp = 1.2 A0
r=2
I=1
R0 = 2

3.1. Comparison of base shear forces

The values of base shear force calculated using the code criteria summarised above are presented below
and in Table 2. The C-coefficient, which is a factor of the earthquake loads acting on the building, varies
significantly among the Codes. These variations are strongly interrelated with the R coefficient specified in
the Codes. Furthermore, in the fault-proximity norms given in UBC’97, coefficient values demonstrate vari-
ations. Keeping in mind that the lateral earthquake load is directly proportional to the C values computed
above, one can see from Table 2 that TEC-98 possess a much lower lateral load in design calculations with
respect to UBC-97 and Eurocode-8-98.

3.2. Comparison of inter-storey drifts and loads estimated at the connection regions

With the multiplication of the C-coefficient by the mass, the lateral earthquake load on the structure is
calculated. In Table 3, the lateral drift generated by this load at the storey level and the ratio obtained from
the division of this drift by the storey height for each Code are illustrated. The values given under the ‘s’
column stand for the numerical values of a ratio obtained by proportioning the elastic displacement to sto-
rey height for each case.
In Table 4, loads acting on connections (especially the nosing members like balconies, parapets, and cor-
bels) are listed for the Codes assessed. Both in Tables 3 and 4, TEC-98 underestimates elastic displacement
(de) and load acting on nosing members.

4. Behaviour of the structure and loads presented in the codes

The capacity curves obtained in the non-linear analyses of structures offer critical information about the
seismic behaviour of structures. The most popular one among these analysis methods is the Pushover Anal-
ysis, which is also named the Non-linear Static Analysis. In recent studies that have based on structural
behaviour, the pushover analysis method has become a more powerful method for seismic design [4–6].
In the previous section, a study comparing the Codes for a prefabricated structure which collapsed in
the 1999-Kocaeli Earthquake was conducted. The connection of the collapsed structure is given in Fig.
5. The lateral load/weight-crown displacement curves of the frame system using the Pushover Analysis tech-
nique conducted via SAP2000 [7] are presented in Fig. 7. In the analysis stage, the modelling of the con-
nection region is done according to the previous studies [8] and the rigidity of the section is decreased
(Fig. 6).
546 M.H. Arslan et al. / Engineering Failure Analysis 13 (2006) 537–557

Fig. 5. The upper beam-column connection in lambda system.

Fig. 6. The modelling of prefabricated structures for analyses.

The load versus displacement relationship of the structure is given in Fig. 7. As can also be seen in
the diagram, major differences in axial frame strength are observed. From Fig. 7 one can see that frame
lateral load capacity is higher than all codes (UBC-97, Eurocode-8-98, TEC-98) in the direction of the
main axes. On the other hand, in a direction orthogonal to the main axes, the capacity falls consider-
ably below the capacity required for UBC-97 and Eurocode-8-98, although it exceeds the required
capacity of TEC-98.

Fig. 7. Structural behaviour and loading values for respective codes.

 M.H. Arslan et al. / Engineering Failure Analysis 13 (2006) 537–557 547

5. Earthquake damage in prefabricated structures

Depending on the supervision of the concrete fabricated and reinforcement qualities of the prefabricated
members, no material-based deficiency has been observed in the prefabricated structures [9,10]. Satisfaction
of sufficient strength, lateral rigidity and ductility, is, however, a crucial concern for such structures.
Previous research [11–13] demonstrated that the most vital complication in prefabricated structures is
sustaining the lateral loads exerted during earthquakes. The industrial-type structures especially with their
fixed-base and crown-hinged (reverse-pendulum type) design tend to convert the entire lateral load to dis-
placement by Eq. (19).
F
h3c
d¼ : ð19Þ
3Ec I c
In Fig. 8a, a system possessing monolithic joints and in Fig. 8b a crown-hinged system are presented.
From Eq. (19), the displacement in the latter system far exceeds that of the former. Another significant dis-
tinction among the systems is the variation in the location of the potential plastic hinge formation. Whereas
in Fig. 8a hinging occurs at both tips of the columns with a moment having a value of ‘u’, in Fig. 8b, it
forms at the bottom ends of the columns with a moment reaching a value of ‘U’, where ‘U’ is twice ‘u’.
The weakness of the crown-hinged prefabricated system regarding both lateral rigidity and ductility can
clearly be seen (U > u, M > m, d2 > d1). In the light of the above explanation, the pictures of failed prefab-
ricated structures are worth examining (Fig. 9).

Fig. 8. Monolithic and hinged systems (prefabricated) [14].

Fig. 9. Overall collapse of the prefabricated systems (a,b).

548 M.H. Arslan et al. / Engineering Failure Analysis 13 (2006) 537–557

TEC-98 states that the structural behaviour coefficient R may be assumed as 5 for hinged-connection
frames. When this value is compared with that of in situ cast frames, i.e., 8, it corresponds to a 60% increase
in lateral loads for the former. The inter-storey drift is limited to a maximum value of 0.0035. Furthermore,
while the Code allows only 1 storey construction for crown-hinged structures, it also insists on the require-
ment of in situ cast shear wall application for higher storey construction.
Following the earthquakes, numerous totally failed and many highly damaged prefabricated buildings
were reported [15,16]. An inventory illustrating the damage status of the buildings constructed in the region
by the Turkish Prefabricated Construction Association member firms is given in Table 5. Even though the
number of damaged structures in the table seems low, tens of failed buildings were excluded from these re-
cords. These structures are those constructed by firms not supervised by the Association and constructed
with serious mistakes at both design and application stages. In Fig. 10, the post-earthquake view of a struc-
ture constructed by a non-member firm is portrayed. The erection of the building had been completed only
three month before the Earthquake hit (this building had the same properties as the model building men-
tioned in Section 3).
Common damages and their types encountered in single-storey hinged-connection prefabricated frames
are itemized below.

(I) Insufficiencies in detailing in the regions where inclined roof-beams meet with columns: In the system
shown in Fig. 9b, the connection between columns and beams is a moment-free hinged-connection.
The friction force holding the beam on the support, in other words, preventing the beam from lon-
gitudinal slippage, is dependent on the weight of the beam, friction coefficient between the supporting
surfaces, and shear resistance provided by the transverse section.
Inclined roof-beams rest on the corbels attached to the columns. A pair of 12–20 mm diameter rein-
forcements extended outside each corbel crossing the holes at the head of the beam and finally the
remaining ring-like gaps in the holes are filled with mortar. Roof-beams are connected perpendicularly

Table 5
The distribution of damaged structures after the 1999 earthquake (among TPCA members) [9]
Total number of precast buildings 481
Heavily damaged 17
Partially damaged 14
Non-damaged 450

Fig. 10. Overall failure in prefabricated systems.

 M.H. Arslan et al. / Engineering Failure Analysis 13 (2006) 537–557 549

to each other only via secondary concrete beams extending in the longitudinal direction of the struc-
ture. These beams rest on 10–12 mm reinforcement extensions on the inclined roof-beams with a sim-
ple detailing. These reinforcement bars are covered and the remaining voids are filled with mortar
upon finishing. The seismic lateral inertia forces acting perpendicular to the inclined roof-beams
(out-of-plane loading) could not be transferred to the column tops. The reason for that is inadequacy
of the supports. As a result these beams overturned and leaned upon each other (in the out-of-plane
direction) like dominoes.
(II) In the design calculation of these connection members, the overturning moments and shear forces
generated due to inertia forces acting orthogonal to the frame must also be accounted for and the
diameters and anchorage lengths of the reinforcement members designated accordingly. The anchor-
age fasteners connecting the roof-beam to the column, in Fig. 11, are displaced and the beam is
detached from the column.
(III) Another type of failure usually occurs due to miscomputations in the roller-support detailing. In the
dilatation region the beams must be located so that a definite amount of displacement is released. In
doing that, during earthquakes these beams must not slip from the supporting points when the sep-
arate parts of the structure are displaced relative to each other. But these details were not applied and
were failures observed as a result.
(IV) Formation of bending cracks at lower sections of the columns is a widespread failure type. This type
of failure is the indicator of columns exceeding the ultimate elastic moment bearing strength at lower
sections. The inadequacy of the column cross-sections especially in the out-of-plane direction (the
asymmetrical approach in column-design) and exceedingly dense arrangement of the reinforcement
are the principal reasons for column lower section failures. The bending cracks investigated at the
socket level of the columns of a coffee-processing factory erected in 1996 in Izmit can be seen in
Fig. 12.
(V) Vertical cracks have mostly been observed on the lateral facades of the beams resting on the corbels, in
one of the planes cutting the holes, located at the beam-tip and through which the reinforcement
extended from the corbels. In some structures, it was seen that one of the tips of the beam extending
in between the adjacent columns had fallen down slipping out of the reinforcement bars. Meanwhile,
the tip of the beam also bruised the concrete tip of the corbel. Moreover, reinforcement bars extending

Fig. 11. The shear of the inclined roof-beam from the crown cap.
550 M.H. Arslan et al. / Engineering Failure Analysis 13 (2006) 537–557

_
Fig. 12. Bending cracks witnessed in a coffee-processing factory in Izmit.

from the corbels slipped out of the beams upon bending. The inadequacies of the reinforcement cross-
section and anchorage length in the roof-beam and column connection beside the usage of substan-
dard filling materials are the main reasons lying beneath this type of failure.
(VI) The excessive disproportionate displacements, exceeding the values in the Code, of the column tips
due to slenderness and consequently the damage caused by the secondary members failing to generate
the displacements in comparable amounts are the sources of a different failure type.
(VII) The presence of exterior walls besides non-existence of interiors particularly in buildings having
two or more spans resulted in differential displacements out-of-plane. In such structures, columns
located in the span experience more out-of-plane displacement compared to the column-couples
at the sides. In some prefabricated systems, damages concerning this type of failure were
observed.

Figs. 13 and 14 have been shot from a double-span prefabricated industrial structure constructed in Ada-
pazari Organised Industrial Zone. Columns in the middle were damaged (and failed) significantly compared
to the columns in the edges.
Fig. 15, shows the post-erection and post-earthquake state of the interior-column and interior-beam con-
_
nection section of a triple-span one-storey industrial site constructed in Izmit.
In the mentioned structures, no rupture, bruising, or cracking has been observed in the filling concrete
poured between the column and the socket foundation. Only following the 12 November 1999 Duzce
Earthquake, in the foundation sockets of a damaged structure, located in Bolu-Kaynasli province, were
some cracks detected. In almost none of the damaged structures in Adapazari Organised Industrial Zone
has a rupture in sockets observed [9,10]. What is more is that observing bending failures in the lower sec-
tions of the columns embedded in the sockets is another crucial piece of evidence of the fixed-behaviour of
the column-foundation connections (Figs. 12 and 16).
 M.H. Arslan et al. / Engineering Failure Analysis 13 (2006) 537–557 551

Fig. 13. An axis surviving thanks to the exterior walls.

Fig. 14. Damage in the middle axis (central column failure in double-span structures – Adapazarı).

In the direction of the material tests conducted in the existing prefabricated structures, some Schmidt
Hammer readings had been taken and these tests illustrated that strengths of concrete members tested
are in accordance with the limitations given in the Code. The strength values for concrete qualities sug-
gested in the design, namely C25 and C30, satisfy the empirically obtained strength values. In mechanical
and chemical analyses of the construction steel, it was seen that the carbon content considerably exceeded
the upper limit thus giving the reinforced concrete a brittle character [9,10].
Meanwhile, in many prefabricated structures, in beams of which no falling down or toppling occurred,
cracks in the beams and also, bruising in the edges of the upper facade of the corbels has been observed.
552 M.H. Arslan et al. / Engineering Failure Analysis 13 (2006) 537–557

Fig. 15. Central column failure in double-span structures (undamaged state, damaged state) – Adapazarı.

Fig. 16. No damage was observed in the socket.

Concentration of the damages in some particular regions illustrates the significance of both the better des-
ignation of the local ground subgroups and intensifying effect of the alluvial soil.
Beside many damaged structures, there are many prefabricated industrial buildings which resisted the
earthquake without even showing a trace of damage. This state of events brought out the fact that seismic
hazards may bring about significant damages not only in in situ cast reinforced concrete structures but also
in the prefabricated structures the design and erection of which are not vigilantly conducted and do not
conform with the computations and connection detailing criteria given in the TEC-98.

6. Retrofitting proposals for existing prefabricated reinforced concrete industrial structures

After the 1999-Marmara Earthquakes (17 August Kocaeli and 12 November Düzce), it was seen [17]
that some of the single-storey prefabricated industrial structures possessing pin-connected supports had
been razed and many more damaged. Furthermore, many of those having no failures are prone to connec-
 M.H. Arslan et al. / Engineering Failure Analysis 13 (2006) 537–557 553

tion deficiencies. Since many of these buildings had been erected before the enforcement date of the TEC-
98, they are far from satisfying the criteria asserted in the Code.
As mentioned above, the initial term that ought to be satisfied in a retrofitted structure is the procure-
ment of the required lateral rigidity. Satisfaction of this term will be possible with bracings located in ver-
tical frame planes or via shear walls. However, when each frame unit of the industrial framework is
presupposed to be pin-connected to each other, this method would be insufficient to solve the problem
of its own accord. Expecting rigid diaphragm behaviour from the roofing material, which is of aluminium
panels or eternit in general, seems quite unreasonable. Thus a shear wall or vertical bracings would only
stiffen the section or frame where it is located but cause no effect on the frame planes parallel to it.
Since displacements in central columns, particularly in those located in the middle of double-span
frames, are comparably higher than those observed in peripheral columns, connecting the central-columns
from the lowermost level of the gutter to each other in order to form a monolithic framework cell is highly
recommended. Connection of the trapezoidal beams to each other in a laterally supporting manner can be
applied just as performed in the steel roof erection. The moment bearing capacity of column-crown beam
connections, principally in lambda systems, can be increased by wrapping the connection with strengthen-
ing members (Fig. 17). One of the post-earthquake structural retrofitting techniques is the one performed
by jacketing each column-member. The crucial aspects of the technique are the increased risk at the column

Fig. 17. Peripheral column-crown beam joint strengthening.

Fig. 18. Interior column failure.

554 M.H. Arslan et al. / Engineering Failure Analysis 13 (2006) 537–557

sections where the jacketing ends, with generation of extra moments at the foundation level (Figs. 18
and 19).
In all prefabricated and conventional reinforced concrete buildings, walls laid with filling material,
whether they are bearing walls or not, support loads. Particularly when damaged multi-span structures
are investigated, it is seen that whereas columns located in the peripheral axes were held by the walls laid
in between, the framework cells on the interior axes, the inside of which were left unfilled for the sake of
easier in-factory production, deformed easily and displaced more. Accordingly, to solve this interior-span
handicap, retrofitting is applied on the columns located in such axes by connecting them to each other via
steel beams just below the column-crowns (Figs. 20 and 21).

Fig. 19. Interior column retrofitting – Adapazarı.

Fig. 20. Interior column retrofitting.

 M.H. Arslan et al. / Engineering Failure Analysis 13 (2006) 537–557 555

Fig. 21. Interior column retrofitting.

7. Results and assessment

Prefabricated structure systems and all members constituting these systems, as structures constructed
using conventional techniques, should satisfy the terms of earthquake-resistant construction such as ade-
quate strength, ductility and rigidity. Nevertheless, the recent earthquakes, i.e., 97-Ceyhan, 99-Marmara
and 99-Düzce, illustrated that even in the cases where one complies with all the criteria given in the Code
problems might occur.
Due to the fact that there is intensive demand for the optimum use of space in industrial structures espe-
cially in between the galleries, no spatial partitioning is applied and the use of partition walls or curtain
walls is comparatively limited owing to architectural concerns. Hence, it is the task of prefabricated col-
umns to stabilise the earthquake loads. The base shear force computed from the Turkish Earthquake Code
is less than half of the values computed from the procedures given in the other Codes. As a result, estimated
column cross-sections are lower compared to those calculated with the guidance of other codes.
As illustrated in the analysis of a typical prefabricated industrial structure, for 1 unit lateral load com-
puted in accordance with the regulations given in TEC-98, UBC-97 and Eurocode-8-98 one would estimate
2.29 and 2.62 units, respectively. The parameter having the ultimate effect on such variations is the struc-
tural behaviour coefficient, R. Reaching a value of 5 in TEC-98, this parameter varies from 1.5–2.5 in UBC-
97 and Euorocode-8-98.
The formula proposed in TEC-98 for the design of connection locations for non-structural members
such as ledges, corbels and architectural members is inadequate compared with the other Codes, particu-
larly the Eurocode-8-98 [18–22].
When the lateral loading–displacement correlation of the model structure is analysed, it is observed that
whereas the capacities especially of the framework cells orthogonal to the main axis satisfy the required
values computed according to TEC-98, they failed to satisfy the limits designated according the Codes
UBC-97 and Eurocode-8-98.
Most intensive damages observed in the recent earthquakes on prefabricated industrial structures are
those of the single-storey industrial structures with fixed connections at the ground level and hinges at
the upper ends. The incompatibility of the structural systems relative to the selected system’s inappropri-
ateness with the limits put forward in the Codes regarding lateral rigidity, strength and ductility, raises
doubts about the earthquake resistance of these structures.
Even though prefabricated systems possess the advantages of high material strengths (exceeding
30 MPa) and production under intensive control, deficiencies in in situ mounting and connection of mem-
bers resulted in increased failures.
556 M.H. Arslan et al. / Engineering Failure Analysis 13 (2006) 537–557

Enquiries conducted on the damaged and collapsed buildings in the Earthquake Region illustrated that
buildings fail especially in the direction orthogonal to the main axis. The fundamental reason lying beneath
this outcome is the insufficiency of the column lengths of corbel (reverse-pendulum) type buildings in this
direction.
The plastic-hinge formation loci would be at the ground level of the column, just above the socket entry
for the crown-hinged system. When the potential for column-lower-sections to form plastic hinges is con-
sidered, the necessity for increasing the frequency of the stirrups in the lower regions of the columns and the
column-length of the stirrup densification emerges.

8. Conclusions and recommendations

From this study it is observed that the capacity dictated from TEC-98 is lower than the capacity of the
other two codes. The reason for this is related to the economic situation of the relevant countries. Turkey is
a less-developed country so economical considerations tended to weigh strength considerations.
The studied structures which failed in the Marmara Earthquake were constructed according to designs
imported from non-earthquake countries. Future designs must be performed considering the EQ reality of
the region. Also new building and connection types should be developed considering the previous failure
and damage pattern of older buildings.

Acknowledgements

The authors thank Prof. Adnan Cakiroglu and Metin Aydogan (Istanbul Technical University), asst.
Prof. Sevket Ozden (Kocaeli University), asst Prof. Ahmet Yakut (Middle East Technical University)
_
and the engineers at YESA company (Istanbul) for their help. This study is supported by Selcuk Univer-
sity-Scientific Research Projects Office (S. U. BAP).

References

[1] Turkish Earthquake Code (TEC), Regulations on Structures constructed in Disaster Regions, Ministry Of Public Works And
Settlement, Ankara; 1998.
[2] UBC-97 Uniform Building Code. In: International conference of buildings official, USA; 1997.
[3] Eurocode-8-98. Design Provisions for Earthquake Resistance of Structures, European Union, European Prestandarts, Brussel;
1998.
[4] Mwafy AM, Elnashai AS. Calibration of force reduction factors of RC buildings. J Earthquake Eng 2002;6(2):239–73.
[5] Mwafy AM, Elnashai AS. Static puhover versus dynamic collapse analysis of RC buildings. Eng Struct 2001:407–24.
[6] Paulay T, Priestly MJN. Seismic design of reinforced concrete and masonry buildings. New York: John Wiley & Sons
Interscience Publication; 1992.
[7] SAP2000. Structural Analysis Program, Nonlinear Version 7.12, Computer and structures, Inc. Berkeley, CA,USA; 2000.
[8] Aydogan M, Akoz Y. A numerical approach to define the rotational stiffness of a prefabricated connection and experimental
study. Eng Struct 1995;57:999–1007.
[9] Atakoy H. The August 17th Earthquake and the prefabricated structures built by the members of the Prefabric Union. Concrete
Prefabrication, 2000, No. 52–53, October 1999–January [in Turkish].
[10] Arslan MH. The behavior of prefabricated industrial plants under earthquake loading. Graduate Thesis for the fulfillment of
Master’s Degree. Istanbul Technical University, Istanbul; June 2000 [in Turkish].
[11] Tezcan SS, ve Bodurolu MH. A Reconnaissance Report June 27, 1998 Adana-Ceyhan Earthquake, Turkish Earthquake
Foundation Publications. TDV/DR 98-026 Istanbul, August, 10; 1998.
[12] Tankut T, Ersoy U, Ozcebe U. Seismic performance of pre-cast concrete connections. In: 11th European conference on
earthquake engineering. Rotterdam: Balkema; 1998.
 M.H. Arslan et al. / Engineering Failure Analysis 13 (2006) 537–557 557

[13] Tankut T, Ersoy U, Ozcebe G. Pre-production structural failures in 1999 Marmara and Düzce Earthquakes. In: 10th
Prefabrication symposium, Istanbul; 2000.
[14] Toniolo G. The seismic design of pre-cast concrete structures in the new Eurocode 8. In: 17th international congress of the pre-
cast concrete industry, 1–4 May, Istanbul, Turkey; 2002.
[15] Zorbozan M, Barka G, Sarifakioglu F. Failures detected in prefabricated structures, their causes and solution proposals in the
Ceyhan Earthquake. Concrete Prefabrication 1998;48:5–8.
[16] Ersoy U, Ozcebe G, Tankut T. Pre-production structural failures in 1999 Marmara and Düzce Earthquakes, Istanbul, Turkey;
May 2000.
[17] Karadogan F, Yuksel E, Yuce S, Vatansever C.Several pilot tests for retrofitting of damaged prefabricated columns. In: 17th
International congress of the pre-cast concrete industry, Istanbul; 2002.
[18] Posada M, Wood SL. Seismic performance of pre-cast industrial buildings in Turkey. In: 17th International congress of the pre-
cast concrete industry, 1–4 May, Istanbul, Turkey; 2002.
[19] Iverson JK, Hawkins NM. Performance of pre-cast/pre-stressed concrete building structures during the Northridge Earthquake.
PCI J 1994.
[20] Yuzugullu O, Erdik M. Recent developments in the seismic performance of connections in precast concrete buildings. Rotter-
dam: Balkema; 1994.
[21] Ozturk I, Tonguc Y, Basaran H. Restoration and retrofitting applications on a prefabricated structure. In: ECAS2002
international structural and earthquake engineering symposium, METU, Ankara; 2002.
[22] Tezcan SS, Colakoglu HK. Weak points of the TEC’98 precast buildings code provisions. In: Fifth national conference on
earthquake engineering, Istanbul; 2003.

View publication stats