Engineering Failure Analysis 36 (2014) 301–313

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier.com/locate/engfailanal

Site investigation of damages occurred in a steel space truss roof

structure due to ponding
Filiz Piroglu a, Kadir Ozakgul a,⇑, Hikmet Iskender b, Levent Trabzon c, Ceyhan Kahya d
a
Department of Civil Engineering, Istanbul Technical University, Maslak 34469, Istanbul, Turkey
b
Department of Chemical Engineering, Istanbul Technical University, Maslak 34469, Istanbul, Turkey
c
Department of Mechanical Engineering, Istanbul Technical University, Gumussuyu 34439, Istanbul, Turkey
d
Department of Meteorological Engineering, Istanbul Technical University, Maslak 34469, Istanbul, Turkey

a r t i c l e i n f o a b s t r a c t

Article history: In this study, a steel space truss roof structure which was partially collapsed after a local
Received 8 September 2013 meteorological event that produced strong winds and heavy rains in Marmara region
Received in revised form 17 October 2013 was investigated. For this purpose, damage reconnaissance studies were conducted and
Accepted 18 October 2013
findings observed in the site were interpreted according to the current Turkish steel build-
Available online 28 October 2013
ing design codes based on allowable stress design. For the structural members taken from
debris, material tests were conducted. This steel space truss roof structure covering an
Keywords:
industrial facility was supported on reinforced concrete columns of 11 m height, sheathed
Steel space truss roof
Siphonic roof drainage
by using fibro panel with membrane, having a slope of 1% to monitor the waterflow to the
Roof edge parapet siphonic roof drainage system and surrounded by the parapets of 15 and 25 cm heights at
Ponding the roof-edges. As a result of the investigation, it was estimated that during this extreme
Partial roof collapse rainfall event, due to any deficiency or blockage of the roof drainage system in the edge
zones with parapets of the roof, rainwater accumulation occurred at these regions of the
roof fully and eventually due to the ponding which is a particular loading case for the
design of the structures that is however not defined by the Turkish design standards, roof
collapse of this industrial hall building occurred partially.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction

The roof structures can be generally damaged under the unforeseen extreme meteorological events or in case of poor de-
sign or construction under normal load conditions. The roof collapse is a failure case especially for steel structures which can
occur worldwide. The causes of such collapse events have been investigated by many researchers [1–8]. In these studies, the
structural deficiencies and defects in the damaged or collapsed roof systems have been researched in detail by carrying out
experimental tests and computer analysis.
In addition, each of the roof components such as the roof sheathing panel, drainage system and parapet has been indi-
vidually scrutinized by several researchers. It may be said that the performance and strength of these components may
be so important in terms of load-carrying capacity of the roof structures under extreme loads. The wind effects on the various
roof covering systems have been estimated by experimental and analytical studies [9–13]. Henderson et al. [14] conducted
an experimental study on the performance and failure mechanisms of roof sheathing under fluctuating wind loads. Massar-
elli et al. [15] performed an experimental study based on dynamic tests in order to determine the seismic characteristics of
the steel deck roof diaphragms. Blaauwendraad [16,17] presented numerical methods and models for the analytical solution

⇑ Corresponding author. Tel.: +90 212 2856572; fax: +90 212 2856587.
E-mail address: ozakgulk@itu.edu.tr (K. Ozakgul).

1350-6307/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engfailanal.2013.10.018
302 F. Piroglu et al. / Engineering Failure Analysis 36 (2014) 301–313

of rainwater ponding problem on flat or nearly flat one-way and two-way roofs. Mans et al. [18] carried out a study related
with wind-induced loads acting on parapets situated on low-rise buildings. In this paper, the wind-induced pressures on par-
apet surfaces were measured and then, the wind loads on the parapets were obtained by using these measurements. In an-
other study, Suaris and Irwin [19] investigated the effectiveness of different parapet geometries by reducing the extreme
suctions on the low-rise building roofs experimentally and analytically. Also the performance of the roof drainage systems
has been investigated using numerical models and laboratory tests by several researchers [20–24].

2. Description of the steel space roof structure

The industrial hall building investigated due to the partially collapses of the roof structure in this study consisted of an
administration building and six halls with different geometrical sizes and column configurations where an industrial produc-
tion was made. Plan view of this structure is given in Fig. 1. The load-carrying system for the roof structure of each hall was
constructed by using a 2 m high double layer grid steel space truss system which is one of the most commonly used form of
space truss systems for more than 25 years in Turkey.
Mero system consisting of a steel sphere with screw thread holes is used as joint assembly at the truss connection joints.
Additionally chord and diagonal members having a steel tube sections with different diameters and thicknesses belong to
the load bearing members of this Mero system, as shown in Fig. 2. The screwed cone ends of these pipes were connected
by the steel sphere joints. These spheres with flat facets and tapped holes for bolts are hot-pressed forging nodes which
are constituted as connectors, i.e. assemblies by tightening the bolts by means of hexagonal sleeve and dowel pin arrange-
ment for the straight bearing circular hollow sections of the space truss system without causing any joint-eccentricity. At
every node-point the axes of all the joining truss members can pass through the center of these connectors so that only axial
forces can develop. Thus tensile forces will be resisted by the space truss members and transmitted along the longitudinal
axis of the bolts on their end cone. However there will be no stress on the bolts if they are subjected to compression forces
which will be then distributed through the hexagonal sleeves to these connectors. In the design C45 steel material having
38–49 kN/cm2 yield stresses was chosen and galvanized for these Mero connectors according to the Turkish Standard TS
2525-2 [25] by considering the sphere diameters between 75 mm and 154 mm.
In this investigated industrial hall building, six steel space truss roof structures with different spacings on the individual
halls covered totally 30,000 m2 production area and were separated by the structural expansion joints from each other. This
light weight roof system on every hall was rigidly supported in vertical direction on reinforced concrete columns of 11 m
height. At the support points spheres were welded on square or circular bearing plates which were mounted on the top
of the reinforced concrete columns, as seen in Fig. 3.
As can be seen in Fig. 1, Hall A covered 5244 m2 area in the south-west region of the industrial building between the A–G
and 1–4 axes in plan view. 4567 steel pipe members were used in the space truss roof system which was supported on 28
reinforced concrete columns with different spacings. Hall B in the south-east region covered 1440 m2 area between the G–I

Fig. 1. Scheme of the industrial structure subjected to ponding due to heavy rainfall.
 F. Piroglu et al. / Engineering Failure Analysis 36 (2014) 301–313 303

Fig. 2. Schematic view of members of a Mero type steel space roof structure.

Fig. 3. Scheme of a steel support sphere welded on a baseplate.

and 2–4 axes in plan view. 1152 tube members were used for the space truss which was supported on 17 columns. Hall C
was in the middle of the west part of the building and covered 4896 m2 area between the B–E and 4–8 axes in plan view.
4129 pipe members were used for the space truss which was supported on 20 columns. Hall D was in the middle of the east
part and covered 5760 m2 area between the E–I and 4–8 axes in plan view. 4707 members were used for the space truss
which was supported on 41 columns. Hall E in the north-west part covered 5256 m2 area between the B–E and 8–13 axes
in plan view. 4536 members were used for the space truss which was supported on 24 columns. Finally, Hall F in the north-
east part covered 7200 m2 area between the E–I and 8–13 axes in plan view. 5976 members were used for the space truss
which was supported on 50 columns.

3. Evaluation of the steel space truss roof structure according to current Turkish design standards

For the light weight space roof trusses steel pipes with 13 different diameters varying between 42.4 mm and 168 mm
with different thicknesses were selected as bearing members. High strength low carbon steels corresponding to the Turkish
Standard TS 301-2-3 [26], Turkish European Standards TS EN 10219-2 [27] and TS EN 10305-3 [28] as well as German
Standard DIN 2458 [29] were considered in the design for material quality and required weldability feature. Based on the
galvanized steel S235JR that has an ultimate strength, ru, 340–470 N/mm2 and yield strength, ra, 235 N/mm2 the allowable
304 F. Piroglu et al. / Engineering Failure Analysis 36 (2014) 301–313

stress limit was defined as 144 N/mm2 for the design calculations. Due to the Turkish earthquake resistant building code [30]
requirements every load bearing members’ sections should be checked for local buckling which is defined as a slenderness
ratio of the pipe diameter to its thickness by also considering the ductility level chosen in the design. It was observed that
the requirement for a compact pipe section was fulfilled by all truss members, because the upper limit value given as
0.05  E/ra, in this case 0.05  2100/2.35 = 43.75 was satisfied, where E is elasticity modulus of the steel material.
The pipe ends were tapered for the bolts of 10.9 material quality according to the Turkish Standard TS 2162 [31] as well as
German Standard DIN 17100 [32]. The allowable stress limit for the welds on the pipe member ends was considered as
11 kN/cm2 based on the Turkish Welding Standard TS 3357 [33] without regarding the strength of the welding consume
used. However, this Turkish standard has not been updated for more than 30 years although the new Turkish Earthquake
Code released in 2007 [30] strictly recommends welding consumes with strengths which should not be lower than the main
steel member. Therefore instead of considering 11 kN/cm2 as an allowable stress limit in the design calculations this should
be obtained depending on the welding consume’s strength. However in the design documentation and drawings it was ob-
served that there was no note on the welding consumes used.
According to the new Turkish Earthquake Code published in 2007 [30] only high strength bolts should be used by apply-
ing the relevant full torque values. Thus 10.9 quality bolts having 100 kN/cm2 ultimate tension strength and 90 kN/cm2 yield
stress were used in the design by also considering the Turkish Code TS 3576 [34]. According to these characteristic values
allowable stress limit was assumed as 54 kN/cm2 by dividing the yield stress to the safety factor 1.67. However this safety
factor is widely used for the main hot-rolled steel products. For the bolts it is recommended to consider higher safety factors
due to the unknowns. However in the Turkish design code for steel structures TS 648 [35] there are no requirements for the
high strength bolts. Especially in Turkey where the workmanship quality is inappropriate and sometimes poor in practice,
safety factors of 2–2.5 are recommended to assume, so that allowable stress limit should be then decreased from 54 kN/cm2
to 36 kN/cm2. The difference is not small therefore it is obvious that this decrease may dramatically control the design checks
as well as the collapse mode of the building.
As mentioned before in the design calculations the support spheres were vertically fixed (see Fig. 3). Although all the
needed calculations were not present in the design documentation it was believed that the lateral flexibilities were obtained
for every support point depending on the reinforced concrete columns’ strengths and configuration, because spring con-
stants for two lateral directions were defined in the design calculations of the steel space truss roof. However for every hall
a difference was detected between the number of types for the supports in the design calculations and constructional draw-
ings. It is sure that such a difference between the design assumptions and drawings for the fabrication is very serious/fatal
especially for such a light-weight space truss roof structure. For a structural engineer it should be a necessity to define the
boundary conditions without developing any discrepancies between the design and fabrication. Thus, relevant and appropri-
ate design drawings and calculations are significant necessities for every support assembly.
According to the Turkish TS 5317 [36] steel boxes having a 60  120  3 mm cross section were used as purlins by con-
sidering S 235 JR material quality.
In the design documentations there were no information, no design calculations and no drawings for the reinforced con-
crete bearing columns as well as for the steel columns which were constructed to transmit the wind loads acting on the
windward gable wall or sidewalls of the building. So it was not possible to check the lateral spring constants assumed in
the design calculations of the roof structure in the support points depending on the reinforced concrete columns’ strengths
and configuration.
In the design, 13 different types were undertaken for dead, live, snow, wind, temperature and earthquake loads. Depend-
ing on these loads static analysis were carried out by considering 40 different load combinations for the superimposed loads
in accordance with the Turkish codes. For the dead load of roofing + purlin + installations 500 N/m2 was totally considered in
the design where the dead load was calculated as 131 N/m2 for the Hall A and 122 N/m2 for Hall F which both were partially
collapsed. According to the Turkish Code for loads TS 498 [37], the snow load was defined as 0.75 kN/m2 regarding the build-
ing location measured 300 m from the sea level. Despite this an increase on this amount was undertaken for the snow load
and thus considered as 1 kN/m2 in the design. This load was uniformly distributed on the whole roof structure symmetrically
as well as antimetrically, i.e. on one side of the roof only.
Depending on the design considerations it was seen that no rain load or no load caused by an excessive accumulation of
water had been foreseen and assumed in the design. On the other hand, Turkish codes for loads TS 498 [37] as well as for
steel structure design TS 648 [35] do not address this unusual loading condition. The requirements for the roof serviceability
and its stability due to the excessive water accumulation are not given. As is well known ponding relates the strength as well
as stiffness of the roof structure. The instability due to ponding is caused by the affected water weight and deflection of the
roof. Therefore, more attention should be payed to superimposed load due to the rainwater especially on low-slope or flat
roofs which allow water to impound seriously due to its hydraulic gradient and higher parapets mounted around the roof.
Therefore AISC Specification [38] defines limits for the stiffness in order to ensure a stable roof structure, so that a structural
engineer can design stiffer bearing elements than would be required by this loading. However the accumulation of water is a
matter of discharge capacity of the main drainage system and maintenance issues. IBC [39] and IPC [40] recommend a sec-
ondary discharge system that should be built for the emergency situations to prevent the destructive roof damages due to
any blockage of drains of the primary drainage system. It is also suggested to design the flow capacity of the secondary drains
or scuppers used in emergency not less than that of the primary siphonic’s system.
 F. Piroglu et al. / Engineering Failure Analysis 36 (2014) 301–313 305

According to the same code TS 498 [37] it is allowed to use 0.80 kN/m2 for the wind loads if the building is 20–90 m high
from the ground. An increment should be also anticipated due to the weather conditions which appropriately suit the build-
ing site. Thus in the design calculations this value was increased by an amount of 25% to 1.0 kN/m2. Due to the small value of
the roof slope (1%) the aerodynamic factor on the roof was defined as 0.4 according to the TS 498 [37], so that a wind suc-
tion of 0.4 kN/m2 was expected to effect on it. However, in the design documentation package the design calculations were
not present for the wind effects on the windward gable wall and sidewalls where the aerodynamic factors should be under-
taken as 0.8 pressure on the windward, simultaneously 0.4 suction on the leeward side.
For the design against seismic loads the industrial building was considered in the 2nd seismic critical region i.e. earth-
quake acceleration should be 0.3g, where g is the ground acceleration, although in the updated Turkish Earthquake Code re-
leased in 2007 [30] the building was located on the 3rd seismic region, i.e. earthquake acceleration should be taken 0.2g.
Thus it was concluded that the earthquake loads were overestimated for this building. However the building response or
behavior factor was considered as R = 4 which are given for reinforced concrete buildings in the code where R = 3 is recom-
mended for the prefabricated buildings. It was obvious that the earthquake loads affecting on the construction were under-
estimated in this case. Thus, whether this over and underestimated design values may compensate each other it is a matter
of analysis.
The meteorological data was obtained from a station in the region for the day of the roof collapse and before. It was eval-
uated that the wind was weaving with 30 km/h from the north and north–northeast direction dominantly. Until the collapse
hour of the steel roof trusses of A and F-halls seen in Fig. 1 the total rainfall was measured 22.6 mm, where max. 68.6 mm
rainfall was experienced in the history of this region. However the night guards expressed very stormy weather conditions
with heavy rainfalls and lightening during the night. In recent years heavy local rainfalls occur very often, thus it is not al-
ways possible to explain such an event depending on the data obtained from a station only.
Totally 42 siphonics were designed in order to function as a water drainage mechanical system on 4 edges of this indus-
trial building roof where the water followed a slope of 1%, as seen Fig. 1. The parapet heights all around the roof structure
were different so that 15 cm was measured along the longer side of the building, however 25 cm on the smaller side. In the
history of this only 2 years old building rainwater ponding problem had been seriously experienced several times. Thus later
2 scuppers which were 15 cm high above the lowest roof level were constructed along the smaller wall and 4 along the long-
er wall as an emergency discharge system. It was believed that for the parapets heights of 15 cm there was no need to build
scuppers. As can be seen from Fig. 4 it did not help too much.
In the design calculations of the siphonics it was observed that only 5 cm water height and corresponding load were taken
into account. Depending on this design value in the hall F the total water accumulation area for one siphonic was obtained
18  20 = 360 m2 by considering the 1% roof slope. In this calculation any permanent deflections or ponding effects were
avoided. So, if it is assumed that there is no blockage in the drainage system and it is functioning perfectly then the discharge
rate was calculated as 0.024 m3/s and 0.040 m3/s regarding the abovementioned 5 cm design water height as well as 15 cm
accumulated water till to the scupper’s holes additionally. Thus corresponding water weights were calculated as 179.1 kN

Fig. 4. Rainwater ponding.

306 F. Piroglu et al. / Engineering Failure Analysis 36 (2014) 301–313

and 537.5 kN respectively. So for the discharge time of the 5 cm water height 748 s were needed, where 1343 s for the addi-
tional 15 cm water accumulation.
Although a catastrophic collapse of a light weight roof structure is expected only in certain particular situations, water
accumulation and corresponding water weight on a low-slope roof can dramatically cause a destructive collapse partially
or even totally if they are not considered in the design.
During the site investigations it was seen that the discharge of the rainwater was inappropriate. The owner of the building
explained that routine maintenance of the roof was carried out weekly, sometimes twice a week due to the intensive birds’
feathers and droppings in this region. Thus, in order to solve the blockage of the drains pyramidal filters were attached
around the siphonics, but it did not help too much (see Fig. 4). It was believed that this catastrophic event in this building
could be addressed by the discharge system capacity and its effectiveness as well as the maintenance issues. Additionally it
was clearly seen that the roofs should be also designed for the maximum depth of water to prevent them from collapse. The
maximum allowable value should be easily obtained from the structural engineer. Thus a healthy communication is needed
among the professionals to solve this significant problem to overcome such a catastrophic result [40].

4. Site investigations

It was observed that in the Hall A the steel space truss roof on the 1st and 2nd axes was collapsed totally where on the 3rd
axes partially, as seen in Fig. 5. Similarly in the Hall F the roof on the 12th and 13th axes was collapsed totally however on the
11th axes partially, as seen in Fig. 6. No failure was investigated for the space truss roofs of the other remaining four halls
called as Hall B, C, D and E (see Fig. 1).
The steel space truss roof structure was mounted on the reinforced concrete columns having 11 m height and
80 cm  80 cm cross sections. The number of columns and their configuration can be seen in the plan view of the building
(see Fig. 1). No failure, no deformation and no cracks were observed on these columns during the investigations. Only some
broken wings were observed on their top section which was estimated to be happened during the collapse of the roof con-
struction. Thus it was concluded that the columns did not play a significant role on the collapse mechanism of this building.
In the halls A and F where local collapses of the steel space truss roof structures were observed a large amount of buckled
pipe members of different diameters were lying on the floor. Additionally the bearing members of the space truss subjected
to compression forces were also buckled in the regions close to the partially collapsed parts of the roof. The buckling behav-
iour was especially obvious for the pipes having high slenderness ratios so that it was concluded that they were subjected to
higher compression forces than considered in the design stage.
Although there was no information of the welding consumes used for the fabrication in the design documentation, no
failures and no propagated cracks were observed on the weld seams which connect the tapered cones to the pipe ends. Thus
it was believed that the welding quality did not play a significant role to trigger or even cause the roof collapse. Similarly no
failures and no propagated cracks on the node spheres were detected. Only bolts and pins were slipped and broken due to
the shear as can be seen in Fig. 7. Thus node spheres were also avoided by evaluating the collapse mechanism. The broken
high strength bolts and pins lying on the floor where roof collapses had occurred were investigated and then tested.
Based on the site investigation the workmanship of the fabrication of the support spheres were unfortunately very poor
(see Figs. 8 and 9). Different types of applications and fabrications were observed for the support joints. Some construction
details were very poor according to the basic design principles, therefore they are strictly forbidden in every design codes.
As can be seen in Figs. 8 and 9 space roof trusses’ supports were embedded on the top of the reinforced concrete column
by using 1–3 lapping circular or rectangular base plates with different sizes. Indiscriminate fabrication and/or outpouring
base plates which caused eccentricities in the support joints and non-uniform stress distribution under the base plates were
observed. Poor and interrupted welding seams were investigated in some support details so that very low welding quality

Fig. 5. Views of the partially collapsed roof and totally collapsed south sidewall of Hall A.
 F. Piroglu et al. / Engineering Failure Analysis 36 (2014) 301–313 307

Fig. 6. A view of the collapsed part in the Hall F.

Fig. 7. A view of the broken bolts at a node sphere.

Fig. 8. A view of the different types of support spheres’ assemblies.

was expected than assumed in the design. In some support assemblies the sphere’s base plates were anchored to the rein-
forced concrete columns by using 2 or 3 small plates welded on base plate sides and bolted to the column. Insufficient
anchorages of support spheres, sometimes lack of anchorage, i.e., reinforcement bars of the column were folded down
and used as anchorage bolts. Sometimes base plates of the support spheres were welded on the folded reinforcement bars.
It was observed that on the same base plate different hole sizes were drilled, but some of them were not used.
As mentioned before it was investigated that these fabricated support details did not represent the general and basic
design principles of a steel construction. Additionally there was inappropriate design calculations and drawings for these
308 F. Piroglu et al. / Engineering Failure Analysis 36 (2014) 301–313

Fig. 9. A view of the damaged support spheres’ assemblies.

different types of assemblies in the design documentation. Therefore it was believed that the eccentricities caused by not
precisely mounting the support assemblies, were not considered in the design calculations. Additionally it was concluded
that such kind of inappropriate fabrication could lead to very fatal results because these were the most critical joints of a
roof structure where the vertical reaction forces have to be transmitted safely. Besides that transferring the lateral forces
induced by earthquake motions as assumed in the design considerations was very significant, too. Due to the roof slope it
was estimated that the wind load affected on this roof was suction therefore it was concluded that it did not play a signif-
icant role as a superimposed load for the analysis. Additionally the maximum wind velocity was measured as 30 km/h, i.e.
lower than the design value.
It is obvious that the calculations should be executed for every steel connection detail considered in the design and they
should be also presented in the design documentation package. Furthermore these assemblies should be fabricated by
strictly following the design drawings and additionally by using a precise and rigorous workmanship to obtain the quality
foreseen in the design.
It was investigated that the design sizes for purlins of box type cross sections were matching the drawings given. It was
estimated that the bolts connecting the purlins to the Mero nodes of the space truss system were subjected to shear forces
and bending during the roof collapse.
Based on the site investigation trapezoidal steel sheeting with glass wool for thermal isolation on it was spanning be-
tween the primary structural members of this light-weight low-slope space truss roof. Additionally for roofing a PVC based
waterproof membrane was applied on top of the insulation. The panel width, plate thickness and height etc. were measured
on the site in order to calculate the load carrying capacity of the roofing which could not find in the design documentation
and additionally to evaluate the ponding effects on the roof collapse realistically. The name of the supplier for the cladding
material was observed from the debris. Based on the characteristic values given by this supplier it was calculated that the
load bearing capacity was inappropriately chosen by the architect. So it was clear that ponding could happen on this roof
structure and trigger or even accelerate the collapse mechanism, as shown in Fig. 10.
It was observed that there was no failure on the north-sidewall of the hall F (see Fig. 6), while the south-sidewall of the
hall A was totally collapsed (see Fig. 5b), although the wind was weaving from the north direction in the day of collapse. In
order to understand this phenomenon the structure was investigated in detail and seen that the sidewalls consisting of lat-
eral steel girders and vertical steel columns were designed as an individual bearing system. The wind affecting on this side-
wall was then resisted by the steel columns which were rigidly supported on the bottom, but could freely move on the top
laterally, so that no wind loads affecting on the sidewalls could be transmitted to the space truss roof structure.
It was explained by the owner of the building that HEA 140 columns of 13.55 m height on the north sidewall had been
later strengthened by welding a box column of 11 m height due to problems faced under service conditions, because the
dominant wind direction experienced from north in this region (see Fig. 6). However HEA 140 columns of hall A which were
anchored on a reinforced wall of 1.85 m height with 5 m spacing and 12 m height were not strengthened. It was observed
that the anchorages of these steel columns were also not rigorous. It was investigated that the steel columns found in the
junk yard were undergone permanent plastic deformations (see Fig. 11). Although there was no design calculations for these
 F. Piroglu et al. / Engineering Failure Analysis 36 (2014) 301–313 309

Fig. 10. A view of the damaged roof covers induced by rainwater ponding.

Fig. 11. A view of the deformed steel column of the south-sidewall of the Hall A.

columns in the documentation package a quick analysis was made and it was concluded that the south sidewall should be
collapsed later due to the superimposed load caused by the wind pressures on the windward side as well as suction
inside the same wall. The people also approved this conclusion and claimed that the south wall of the hall A was collapsed
after 2–3 h later than the collapse of the roof truss.
Although the parapets heights were measured 15 cm and 25 cm on the building edges it was not possible to conclude
why the locally collapses had been occurred in two halls only. Thus the current wind direction from the north was evaluated
but it did not helped too much, because taking this wind direction on the day of collapse into account and simultaneously
wind driven rainflow and accumulation it was not possible to explain the collapses in opposite north-east and south-west
corners. It was found that in these opposite corners the roof structures of hall A and F were supported on a lower level so that
the parapets’ heights were between 35 cm and 40 cm. Thus in these regions the rainwater accumulation was 2.5–3.9 kN/m2
higher than expected values as 1.5–2.4 kN/m2 due to the parapet heights measured before. As a result of this it was con-
cluded that the water increased in these regions dramatically, covered more area and weighted more due to this superim-
posed load under intensive rainfall event experienced. On the other hand as mentioned before the load carrying capacity of
the roofing was also inappropriately chosen, thus it was estimated that ponding occurred, too. Depending on these it was
concluded that the steel space truss roofs of the halls A and F were undergone a destructive rainwater accumulation which
did not foreseen in the design stage. As a result of this catastrophic partial collapses occurred in these halls’ roof structures
dramatically.
310 F. Piroglu et al. / Engineering Failure Analysis 36 (2014) 301–313

Fig. 12. Results of the mechanical tests of the pipes, stress–strain diagrams.

Table 1
Results of the chemical decomposition tests of the pipes.

Elements C (%) Si (%) Mn (%) P (%) S (%) Cr (%) Mo (%) Ni (%) Cu (%) V (%) Results
B1 D = 114.3 mm t = 4.5 mm 0.048 0.0025 0.251 0.007 0.021 0.026 0.0022 0.0358 0.034 0.0002 St-37
B2 D = 88.9 mm t = 4.05 mm 0.054 0.0186 0.252 0.0092 0.0087 0.069 0.0129 0.127 0.263 0.0003 St-33
B3 D = 76.1 mm t = 3.0 mm 0.048 0.0173 0.176 0.0135 0.009 0.059 0.0126 0.128 0.315 0.0006 St-33
B4 D = 48.3 mm t = 3.25 mm 0.082 0.0082 0.480 0.0098 0.0104 0.019 0.0013 0.032 0.016 0.0005 St-33

Fig. 13. Specimens for high strength bolts.

 F. Piroglu et al. / Engineering Failure Analysis 36 (2014) 301–313 311

Fig. 14. Results of the mechanical tests of the bolts, stress–strain diagrams.

Table 2
Results of the chemical decomposition tests of the bolts.

Elements C (%) Si (%) Mn (%) P (%) S (%) Cr (%) Mo (%) Ni (%) Cu (%) V (%) Results
C1 D = 30 mm 0.456 0.138 0.827 0.005 0.005 1.04 0.0067 0.534 0.049 0.0027 C-45
C2 D = 27 mm 0.425 0.076 0.762 0.0128 0.0128 1.02 0.010 0.037 0.060 0.003 C-45
C3 D = 20 mm 0.356 0.0668 0.736 0.005 0.005 0.162 0.0228 0.125 0.132 0.0026 C-35
C4 D = 16 mm 0.356 0.0639 0.818 0.0068 0.0043 1.12 0.0062 0.051 0.048 0.0021 C-35
C5 D = 12 mm 0.205 0.113 0.938 0.0051 0.0064 0.055 0.005 0.049 0.073 0.003 –

5. Material tests

In order to find out the quality of the materials used for fabrication materials’ test were undertaken and the results were
evaluated to see whether there was any decrease in the expected characteristic values as well as any mismatch of the values
considered in the design.
Depending on the material test carried out for pipes it was seen that while mechanical properties of B1 and B4 specimens
matched S235 (St-37) steel quality considered in the design, B2 and B3 specimens matched S275 (St-44) steel quality, as gi-
ven in Fig. 12. However, the chemical decomposition showed only for the pipes having the diameter 114.3 mm a good agree-
ment, for the others the quality was obtained as S185 (St-33), as given in Table 1.
From the tests for the bolts M30 the quality was obtained as 10.9, however M27, M20, M16 bolts (see Fig. 13) showed
good agreement with the quality 8.8, as given in Fig. 14. Besides the chemical decomposition showed the quality C45 for
the bolts M30 and M27, and C35 for the bolts M20 and M16, as given in Table 2.
312 F. Piroglu et al. / Engineering Failure Analysis 36 (2014) 301–313

6. Summary and conclusions

Design of the roof structure investigated in this study was generally appropriate. Only in seismic design an overestimated
value was taken into account, simultaneously an underestimated value was considered. Whether this contradiction results
positive in the design it needs an additional analysis. The test results showed also good agreement with the design consid-
erations. In this industrial building consisting of six halls it is believed that the fabrication was not precisely done and did not
follow the basic design principles especially for the support assemblies.
As is well known collapse of the roof structures may happen due to the ponding especially if the roof has a low-slope or is
flat, because the roof surface has then a large area for ponding which causes higher effects of rain or melting snow water
weights and results in failure. Additionally if the cladding is not designed properly the failure mechanism will be accelerated.
To obtain the ponding effects an iterative method should be used to study the collapse mechanism by eliminating buckled
members and broken bolts at every step of the progressive analysis. Considering the new equilibrium each time this analysis
should be carried out until achieving the instability mode due to the excessive deflection of the roof. In the next study it is
aimed to carry out this analysis by considering the test results as well as the site investigations.
In practice the roof structure is designed by structural engineers. The roof configuration, shape and size, roof geometry
and slope, roofing materials and parapets of the building are selected by the architects. Additionally plumping consultant
choose the mechanical system for the discharge of water, drain sizes and locations for providing a primary drainage system
by mounting siphonics and a secondary one consisting of scuppers as used in this building. Thus any lack of information,
communication or coordination among these design professionals mentioned above can be very fatal and lead to failures
or deficiencies as well as catastrophic collapses of the roof structures when establishing rain waters like experienced in this
study. Besides the maintenance of the roof play also a critical role on this catastrophic event. This responsibility however
belongs to the building owner. The roof should be free of loose debris which can easily block the roof drains. Although ser-
viceability requirements in the structural codes do not address maintenance issues and drain system it is necessary for the
team involved in the success of the roof drainage system to agree on the overflow conditions due to the heavy rainfalls or
rapid melting snow.
However the rain loads as well as related design requirements needed are not addressed in the current Turkish codes that
are not updated for 25–30 years. On the other hand Turkish engineers carry out very sophisticated design calculations and
analysis for the earthquake resistant design for 6 years. Snow accumulations due to any barrier are also considered in the
design. Unfortunately less attention is payed to problems which can be caused by the water accumulation faced commonly
and seriously in serviceability conditions of the buildings. The most of the structural engineers believe that uniformly dis-
tributed 750 N/m2 design snow load corresponds a water height of 7.5 cm. Thus they conclude automatically that this value
should be considered as a very safe design value. However a water accumulation which occurs within a short period of time
causes a triangular load distribution where the maximum value occurs always at the roof edges where the parapet heights
control this peak value. Additionally as mentioned before low-slope or flat roofs have more area for excessive accumulation
of water, thus more water raising capacity and water weights than assumed in the design should be expected. So the devel-
oped ponding effects will result in deflection of the roof structure and can trigger or accelerate its collapse.
Similar to the Turkish seismic design code the Turkish code for applied loads TS 498 [37] as well as design code for steel
structures TS 648 [35] should be updated immediately by providing contour maps for 1 h rainfall with a 100 year return per-
iod storm [39,41] as well as by defining rain loads and additionally by giving the requirements against ponding and stability
of the roof for the design needs.

References

[1] Caglayan O, Yuksel E. Experimental and finite element investigations on the collapse of a Mero space truss roof structure – A case study. Eng Fail Anal
2008;15:458–70.
[2] Biegus A, Rykaluk K. Collapse of katowice fair building. Eng Fail Anal 2009;16:1643–54.
[3] Biegus A, Madry D. Failure hazard and strengthening of roof in steel industrial building after explosion of electric furnace. Eng Fail Anal
2009;16:2427–32.
[4] Brencich A. Collapse of an industrial steel shed: a case study for basic errors in computational structural engineering and control procedures. Eng Fail
Anal 2010;17:213–25.
[5] Pinto J, Varum H, Ramos L. Two roofs of recent public buildings, the same technological failure. Eng Fail Anal 2011;18:811–7.
[6] Coz Diaz JJ del, Alvarez Rabanal FP, Garcia Nieto PJ, Roces-Garcia J, Alonso-Estebanez A. Nonlinear buckling and failure analysis of a self-weighted
metallic roof with and without skylights by FEM. Eng Fail Anal 2012;26:65–80.
[7] Biegus A, Kowal A. Collapse of halls made from cold-formed steel sheets. Eng Fail Anal 2013;31:189–94.
[8] Piskoty G, Wullschleger L, Loser R, Herwig A, Tuchschmid M, Terrasi G. Failure analysis of a collapsed flat gymnasium roof. Eng Fail Anal 2013;35:104–13.
[9] Chen Y, Baskaran A, Lei W. Wind load resistance of modified bituminous roofing systems. Constr Build Mater 1998;12:471–80.
[10] Granne F, Bjork F. Joints between roofing felt and sheet metal flashings-short- and long-term tests. Constr Build Mater 2000;14:375–83.
[11] Katnam KB, Van Impe R, Lagae G, De Strycker M. Modelling of cold-formed steel sandwich purlin-sheeting systems to estimate the rotational restraint.
Thin-Wall Struct 2007;45:584–90.
[12] Baskaran A, Ko SKP, Molleti S. A novel approach to estimate the wind uplift resistance of roofing systems. Build Environ 2009;44:723–35.
[13] Gavanski E, Kordi B, Kopp GA, Vickery PJ. Wind loads on roof sheathing of houses. J Wind Eng Ind Aerod 2013;114:106–21.
[14] Henderson D, Williams C, Gavanski E, Kopp GA. Failure mechanisms of roof sheathing under fluctuating wind loads. J Wind Eng Ind Aerod
2013;114:27–37.
[15] Massarelli R, Franquet JE, Shrestha K, Tremblay R, Rogers CA. Seismic testing and retrofit of steel deck roof diaphragms for building structures. Thin-
Walled Struct 2012;61:239–47.
[16] Blaauwendraad J. Ponding on light-weight flat roofs: strength and stability. Eng Struct 2007;29:832–49.
 F. Piroglu et al. / Engineering Failure Analysis 36 (2014) 301–313 313

[17] Blaauwendraad J. Modified method for rainwater ponding on flat roofs. J Constr Steel Res 2009;65:559–68.
[18] Mans C, Kopp GA, Surry D. Wind effects of parapets on low buildings: Part 3. Parapet loads. J Wind Eng Ind Aerod 2005;93:857–72.
[19] Suaris W, Irwin P. Effect of roof-edge parapets on mitigating extreme roof suctions. J Wind Eng Ind Aerod 2010;98:483–91.
[20] Arthur S, Swaffield JA. Siphonic roof drainage system analysis utilizing unsteady flow theory. Build Environ 2001;36:939–48.
[21] Arthur S, Wright G, Swaffield J. Operational performance of siphonic roof drainage systems. Build Environ 2005;40:788–96.
[22] Wright GB, Jack LB, Swaffield JA. Investigation and numerical modelling of roof drainage systems under extreme events. Build Environ
2006;41:126–35.
[23] Wright GB, Arthur S, Swaffield JA. Numerical simulation of the dynamic operation of multi-outlet siphonic roof drainage systems. Build Environ
2006;41:1279–90.
[24] Arthur S, Wright GB. Siphonic roof drainage systems-priming focused design. Build Environ 2007;42:2421–31.
[25] TS 2525-2. Specification for quenched and tempered steels Part 2 – technical delivery conditions for unalloyed quality steels. Ankara (Turkey): TSE
Turkish Standards Institution; 1977.
[26] TS 301. Specification for seamless and welded steel pipes with thread. Ankara (Turkey): TSE Turkish Standards Institution; 1983.
[27] TS EN 10219-2. Specification for cold-formed welded structural hollow sections of non-alloy and fine grain steels – Part 2: tolerances, dimensions and
sectional properties. Ankara (Turkey): TSE Turkish Standards Institution; 1999.
[28] TS EN 10305-3. Specification for steel tubes for precision applications – technical delivery conditions Part 3: welded cold sized tubes. Ankara (Turkey):
TSE Turkish Standards Institution; 2004.
[29] DIN 2458. Standard for welded steel pipes and tubes. Berlin (Germany): German Institute for Standardization; 1981.
[30] Turkish Earthquake Code (TEC). Specification for buildings to be constructed in earthquake areas. Ankara (Turkey): Goverment of Republic of Turkey,
Ministry of Public Works and Settlement; 2007.
[31] TS 2162. Specification for steels for general structural purposes. Ankara (Turkey): TSE Turkish Standards Institution; 1986.
[32] DIN 17100. Steels for general structural purposes, quality standard. Berlin (Germany): German Institute for Standardization; 1987.
[33] TS 3357. Building code for the design and execution of welded connections in steel structures. Ankara (Turkey): TSE Turkish Standards Institution;
1979.
[34] TS 3576. Specification for fasteners, bolts and studs – steel – properties and test methods. Ankara (Turkey): TSE Turkish Standards Institution; 1989.
[35] TS 648. Specification for building code for steel structures. Ankara (Turkey): TSE Turkish Standards Institution; 1980.
[36] TS 5317. Specification for welded square and rectangular steel tubes. Ankara (Turkey): TSE Turkish Standards Institution; 1987.
[37] TS 498. Specification for design loads for buildings. Ankara (Turkey): TSE Turkish Standards Institution; 1987.
[38] ANSI/AISC 360-05. Specification for structural steel buildings. Chicago (USA): American Institute of Steel Construction, Inc.; 2005.
[39] IBC 2009. International building code. Chicago (USA): International Code Council, Inc.; 2009.
[40] IPC 2009. International plumbing code. Chicago (USA): International Code Council, Inc.; 2009.
[41] AISC-ASD89. Manual of steel construction: allowable stress design 9th ed. Chicago (USA): American Institute of Steel Construction, Inc.; 1989.