Engineering Failure Analysis 125 (2021) 105378

Contents lists available at ScienceDirect

Engineering Failure Analysis

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

Progressive collapse of a steel structure under expected snow loads

Zeynep Fırat Alemdar *, Fatih Alemdar
Civil Engineering Department, Yıldız Technical University, 34220 Istanbul, Turkey

A R T I C L E I N F O A B S T R A C T

Keywords: The structural design process has many vital roles in preventing the collapse of structures and
Progressive collapse should be conducted according to national obligatory standards. However; structural failures or
Steel structure even catastrophic collapses based on improper design, underestimated design loads, unexpected
Structural deficiency
extreme loads, and improper erection are reported worldwide on large-span structures. A steel
Buckling analysis
Snow
structure of the factory partially collapsed under expected snow loads in the western part of
Turkey was investigated in this study. The possible reasons for the collapse of the system were
examined by visual inspections and analytical results. The load-carrying capacities of the main
girders and purlins were determined using the buckling analyses of the structural member. The
nonlinear geometric analysis of the entire steel structural system was defined to evaluate the
vertical load-carrying capacity of the main girder. The progressive collapse mechanism of the
system was determined. The obtained results indicate that the snow load is not the main reason
for the collapse and mistakes in the project design, constructional defects and inadequate capacity
of structural elements are the primary reasons behind the collapse.

1. Introduction

Structural design requires “an ability to create a cost-efficient load-bearing scheme in accordance with a set of ‘rules’ prescribed by
building codes, for minimal design cost” [1]. Design is, a process of synthesis [1], which utilizes assumptions relating to probable loads,
structural behavior, and the capacity of material properties. The design process has many crucial roles in the overall response to
structural failure.
Some structural failures are caused due to poor communication between the various design professionals involved in conceptual
design and the supervision of the execution of works, poor communication between the fabricators and erectors, and bad work­
manship, which is often the result of failure to communicate the design decisions to the persons. Other common causes of structural
failure can be defined as lack of appropriate professional design and construction experience, the complexity of codes and specifi­
cations leading to misinterpretation and misapplication, unwarranted belief in calculations, and specified extreme loads and
properties.
Structural failures or even catastrophic collapses based on improper design, underestimated design loads, unexpected extreme
loads, improper manufacturing and poor workmanship are reported worldwide on structures having steel roof construction [2–7].
Industrial halls, exhibition centers, and airport terminals are typically large-span structures having a flat or low sloped steel roof
construction. Similar to conventional steel structures, these lightweight roofs can be subjected to extreme weather conditions such as
incidental, short-lasting heavy rain showers as well as strong heavy snowfalls. Additionally, these roofs subjected to unbalanced snow

* Corresponding author.
E-mail addresses: [email protected] (Z. Fırat Alemdar), [email protected] (F. Alemdar).

https://doi.org/10.1016/j.engfailanal.2021.105378
Received 2 November 2020; Received in revised form 20 February 2021; Accepted 14 March 2021
Available online 23 March 2021
1350-6307/© 2021 Elsevier Ltd. All rights reserved.
Z. Fırat Alemdar and F. Alemdar Engineering Failure Analysis 125 (2021) 105378

accumulation and ponding may experience failure or collapse.

Local or global failures of various steel structures were investigated by several researchers. The collapse of Mero space truss roof
structure was studied both experimentally and numerically and it was concluded that the snow load intensity and some mistakes made
in the elastic design of the truss were the main reasons for the collapse [2]. The reasons for the total collapse of the Katowice Fair
Building structure were presented based on the results of engineering investigations [3]. It was determined that the collapse of the
Katowice Fair Building was due to the lack of strength and rigidity of the main structural elements, the incorrect arrangement of the
structure and the overloading action of a thick ice-snow layer. The collapse of an industrial steel shed under a 10 cm layer of fresh snow
was investigated and it was seen that this collapse was a result of a chain of errors; in the design phase, during its assemblage and in the
final inspection and control phase [4]. After the catastrophic August 17, 1999 Kocaeli Earthquake, the buckled brace system of an
existing industrial building in Bursa was investigated by generating three-dimensional finite element analysis. The static nonlinear
analysis was carried out to evaluate the structure under earthquake loads and buckling analyses were performed to determine buckling
forces. It was concluded that the buckling was not occurred due to the earthquake motion [5].
Nonlinear structural analysis was carried out to study the structural collapse of a self-weighted metallic roof with and with-out
skylights located in Cabañaquinta’s schoolyard (Principality of Asturias, Northern Spain) at 550 m above sea level [6]. The reason
for the collapse of the cable-arch structure was found to be huge snow loads in winters of 2004 and 2006. A damage assessment based
on the nonlinear post-buckling analysis of the entire self-weighted metallic roof by FEM was proposed and structural design recom­
mendations were given to avoid the type of failures in the future. In January 2015, the western regions of Turkey were struck by heavy
snowfall and vehicles, roofs of markets and buildings especially those with a long-span roof such as factory buildings were either
partially or completely collapsed under snow loads. Lateral buckling failures of laterally unrestrained steel cantilever roof of a tribune
collapsed due to snowfall on October 25, 2015 were considered as a case study. The results of experimental, numerical, and analytical
works indicated that snow load, inadequate section properties, errors in design and bad workmanship during construction of the roof
were the reasons for failures [8].
Hail protection canopies of a car dealership collapsed due to snowfall on the 1st of January 2016, in Istanbul, Turkey were analyzed
to investigate the collapse. The site inspections and results of the numerical analysis showed that the combination of improper design
assumptions relating to the extreme snow load, poor design and inefficient construction of structural members as well as structural
connections were the reasons for the failure [9]. Another case study of a steel roof collapse of a closed market in Turkey was considered
by evaluating the numerical results of the linear and nonlinear static push-down analysis of one truss steel girder in the collapsed
portion of the market. The urgent need to upgrade the national standards regarding snow loads to comply with the changes in snowfall
density and frequency due to accelerated climate change was emphasized [10].
Although the loads to be considered in the design of steel structures are calculated correctly, progressive collapses may occur in the
structures due to possible errors in the design or construction stages. Progressive collapse is a dynamic failure process in which loss of
local load-carrying capacity propagates through a system and precipitates widespread collapse. In this study, a factory building where
the purlins first exceeded the capacity and then the main girders collapsed and finally almost the entire structure demolished under the
expected snow load was investigated as an example of progressive collapse. The load-carrying capacities of the main girders and
purlins were determined by the buckling analyses of the structural elements. Furthermore, using SAP2000, a finite element numerical
model of the entire steel structural system was generated and analyzed under the snowfall observed at the studied site. The problems in
the connection regions, such as inadequate supports in the column-footing regions, improper connections between the main structural

Fig. 1. 3-D view of the structure.

2
Z. Fırat Alemdar and F. Alemdar Engineering Failure Analysis 125 (2021) 105378

elements, and missing details were investigated. The reasons for the collapse were identified.

2. Description of the industrial structure

The steel industrial factory structure under investigation is located in Isparta, Turkey and established for filling natural spring water
in 2018. The factory is 110 km away from the city center and the altitude of the city is 1150 m. The structure was erected only a year
ago before collapse and the details of the designing solutions could not obtain by the manufacturer. However; the structure was
assumed to be designed according to national standard. The plan dimensions of the structure in the shop drawing are 20 m × 84 m, and
starting from half of the long direction of the factory, the short span direction decreases linearly from 20 m to 12.05 m towards the
North as shown in Fig. 1. The load-bearing system of this structure consists of trapezoidal portal frames formed by steel columns, steel
girders and steel purlins repeating through 14 spans with 6 m intervals. The roof structure is supported on steel columns of 8 m height
on one side of the long direction and starting from 5 m to variable height on the other side of the direction to provide the 15% slope of
the roof (Fig. 1). The columns and girders are hot rolled sections of HEA 300 and IPE 400, respectively. The portal frames are connected
using corbels at the top of the columns along the frame direction. The steel material of the columns and the main girders is determined
as S275-JR on the site investigations. Trapezoidal portal frames in the system are connected at the top end of the column with beam-
column members along the out-of-plane direction (Fig. 2).
The purlins are made of cold-formed steel having a cross-section of C175 having a flange width of 45 mm, a height of 175 mm, shear
lip length of 20 mm, and a thickness of 2.5 mm. The purlins are formed with a distance of 1.0 m in place and connected to the girders by
using L profile with 2 M12 bolts. Tie rods of 8 mm diameter are welded through the center of the purlins to accommodate out-of-plane
forces arising from the inclination of the roof plane.
On the façade, HSS square 100*4 profiles for cladding are used as seen in the drawing. Sandwich panel elements having 38 mm
thickness are used as roof and side covering. The sandwich panels were connected with the purlins by using screws at 0.5 m intervals.
The column-base plates of 500 * 500 * 20 mm are shown on the project drawing and connected to reinforced concrete footing using
6 M22 bolts having 300 mm anchorage length with anchor plate as seen in Fig. 3.
The steel main girders having 20 m length along the portal frames are connected using 12 M20 10.9 bolts on the flanges and 20 M16
10.9 bolts on the web regions as shown in Fig. 4.

3. Conformity checks of the roof structure according to current Turkish design standards

The factory structure was exposed to 0.917 kN/m2 snowfall on January 6, 2020 and the average external temperature was 0.2 ◦ C as
the information received from the Isparta Meteorology Directorate [11]. Belonging to the characteristic region II as defined in national
standards TS EN 1991-1-3 [12], the design ground snow load (GSL) is properly selected as given in Table 1. Roof snow load (RSL) value
is calculated in regards to the national standard (Eq. (5.2)) by considering the roof slope, the exposure coefficient due to the topo­
graphic condition, and the temperature effect. Accordingly, the snow load destroying the steel structure is less than the snow load
value used in the design project as given in Table 1.

4. Deficiencies in the design of the structure

Damages in the structural elements of steel building systems may cause some or all of the structure to collapse. These damages are
either formed by errors in the shop drawings or deficiencies occurred during the on-site implementation of the project. In the examined
building, first of all, the damages on the structural members are taken into consideration.
The main girder of the trapezoidal portal frame lost the stability of the web and flange regions. The reason was transverse stiffener
members not observed in the connections of column, girder and corbel in accordance with the requirements in Turkish Design
Specification for Structural Steel Buildings (TDSSSB-2018) [13] and American Institute of Steel Constructions (AISC) 360–16 Section
J10.7 and 10.8 [14]. The web and flanges of the IPE 400 profile are compact per TDSSSB-2018 and AISC 360–16 standards, thus there
are two limit states, yielding and lateral torsional buckling. The moment capacity varies with proportional to the lateral torsional

Fig. 2. Girder-column connection detail.

3
Z. Fırat Alemdar and F. Alemdar Engineering Failure Analysis 125 (2021) 105378

Fig. 3. Column-footing connection design in the project.

Fig. 4. Girder-to-girder connection details.

Table 1
Snow load values for the factory structure.
Design GSL Altitude coefficient Roof slope Topographic condition Temperature effect Design RSL Measured
(kN/m2) for GSL coefficient coefficient coefficient (kN/m2) RSL

1.05 1.10 0.80 1.00 1.00 0.92 0.73

buckling effect. In this case, it is understood that the loss of stability of the web and flanges of the profile occurred when the mid-span of
the girder rotated about out-of-axis direction. This behavior is then considered to lead to problems also in the web and flanges stability
on the column area supporting the main girder, such as failure to maintain the angle between web and flanges and local buckling of the
web, as shown in Fig. 5.
Furthermore, to reduce the stress concentrations on the tops of the columns (Fig. 5), especially at the connection level of the
corbels, the column-corbel regions must be designed using transverse stiffeners. However, there was no such elements exist in the
regions.
The moment capacity of the main girders having 20.0 m span length is lower than that of the girder having a span length of 12.05 m.
Hence, the longer main girder rotated more than the shorter ones and almost all of the longer girders were collapsed due to lateral

4
Z. Fırat Alemdar and F. Alemdar Engineering Failure Analysis 125 (2021) 105378

Fig. 5. Damages on the girder support region.

torsional buckling (Fig. 6). As a result of excessive deformation the profile cross-section in the mid-span rotated around 90 degrees due
to the loss of stability in the support region of the girder (see Fig. 6). Since the vertical and horizontal cladding elements supporting the
side with a span of 20 m are supported on the columns and girders on the axis adjacent to this façade, they partially prevented the
destruction of the main girder on this axis as seen in Fig. 7. Another reason for this girder to rotate less is that the girder has half the roof
load.
The purlin-girder connection is built on-site using two bolts as seen in Fig. 8 and this region is insufficient. The web region of purlin
profile is not fully supported with the L profile and the bottom flange of the purlins does not contact with the flange of girders at the
connection level. Thus, the stability of the purlin profile is only provided by the bolts. However, the stability of the purlin profile should
be maintained by connections with as many members as possible. In purlin connections, it is important to arrange the bolt connections
in one or two rows instead of side by side in terms of ensuring the stability of the purlins.
Starting from the side with 12.05 m span girder, trapezoidal portal frames on the 4th axis and afterward (Fig. 9) have completely
collapsed by losing their vertical positions. The main girder span on the 4th axis is 15.38 m.

5. Improper erections

In the on-site examinations of the partially collapsed building, the incompatibilities between the project and the built structure are
determined and the reasons for the failure based on these findings are explained. The column-footing connection, the purlin support
regions, applications of tie rods and welding works can be shown as the main problems for improper constructions.
The column-base plates of 600 * 600 * 24 mm are measured on-site and connected to reinforced concrete footing using 8 M24 bolts
having 115–140 mm anchorage length as seen in Fig. 10. The length of the anchorage rods is approximately one-third to half the length
of the design value. The connection is formed without the anchor plates inside the footing even they are considered in the shop

Fig. 6. Damage due to lateral torsional buckling.

5
Z. Fırat Alemdar and F. Alemdar Engineering Failure Analysis 125 (2021) 105378

Fig. 7. The partial collapse of the vertical cladding members.

Fig. 8. Purlin connection detail at the roof plane.

drawings. Therefore, the anchorage rods were pulled out from the footing as inspected in Fig. 10.
The profile in the purlin connections has been deformed excessively (Fig. 11). It is understood that the reason for the excessive
deformation is a pointwise connection formed in the support areas of the purlins. The nominal tearout strength of the purlin members
is calculated according to Eq. (6c) in AISC 360-16 Section J3 and found as 42.12 kN. In the FE analysis results to be defined later, the
maximum normal force carried by the purlin is obtained as 90.46 kN. Therefore, bolt tear-out failures occurred in the some connections
of the purlins as shown in Fig. 11.
Looking at the photographs in Fig. 12, it is seen that the tie rods are incorrectly or incompletely supported on the profile section.
The tie rods should not be constructed as continuous, or if they are to be continuous, they should limit the out-of-plane movement of
two adjacent profiles towards each other. The tie rods observed on site are not even included in the project drawing.
On the roof plane and as well as along the short and long directions, brace elements for wind stability are observed on-site but not
considered in the shop drawing. The cross-sections are L50*5 on the roof level and along the façades, although HSS 40*30*2.5 profiles
are formed along the façades. The connections of the profiles are made using weld on the site. In most of the weld connections of the
structure, welding works are poorly done as seen in Fig. 13. Profiles on the roof diagonals were welded only on one-side, hence in case
of compression loading in one direction, local buckling and rotations were observed in the profiles (Fig. 13b). In addition, it is seen that
the groove weld applied to the HSS square profile is very poor quality and not sufficient since the connection plate is welded only one

6
Z. Fırat Alemdar and F. Alemdar Engineering Failure Analysis 125 (2021) 105378

Fig. 9. Loss of verticality of the columns.

Fig. 10. Column-footing connection detail.

Fig. 11. Damages on the purlins.

side of the HSS profile (Fig. 13c). If HSS type of profiles are to be combined with a single plate, knife plate should be used in the
connection. Otherwise, double plate connections should be considered to provide an appropriate detail.

6. Finite element analyses of the structural elements

The characteristic load-carrying capacities of the purlin and the main girder in the studied structure are obtained in regards to the

7
Z. Fırat Alemdar and F. Alemdar Engineering Failure Analysis 125 (2021) 105378

Fig. 12. Welding of a tie rod through the purlin section.

Fig. 13. Brace elements for wind stability (a) L profiles along the façades (b) L profiles on the roof level (c) HSS square profiles along the façades.

standards and compared with the self-weight of the system members and the snow load in the design.
The steel structure, which was built on-site, transfers the load to the girders of the portal frame with purlins and roof covering
panels. The failure mechanism defined over the observations, examinations and evaluations made above will be verified by the
modeling of the whole system and each structural member separately. The reasons generating the different analyses are the simply
supported connections between the girder and purlin and the moment transfer connections between the columns and the girders. The
finite element (FE) model of the purlin section is established and the Euler buckling analysis is performed. The buckling load values are

8
Z. Fırat Alemdar and F. Alemdar Engineering Failure Analysis 125 (2021) 105378

obtained by considering the local and general buckling conditions. Material strength values taken into account in the calculations are
the values specified in the project and determined in the field. No laboratory tests have been performed. The analyses are repeated with
different boundary conditions. Before modeling, following TDSSSB-2018 [13] and AISC 360-16 [14] standards, the characteristic
moment capacities of the purlin and main girder are found and how much distributed load carried by the members is calculated by a
reverse method. This distributed load is compared with the sum of the girder, purlin, roof covering weight and the snow load.

6.1. Calculation of purlin capacity

According to the standards, whether the purlin can safely carry a load or not are studied depending on the ratio of the amount of
load affecting the purlin profile at the time of collapse to the amount of load to be carried safely based on the capacity by two different
analyzes. The first analysis is conducted using CUFSM [15] buckling analysis program. In this analysis, the cross-section of the purlin
member is defined as finite strip elements. Simple support boundary conditions are assigned for the section. The buckling load factor is
obtained by an application of unit loading. The buckling deformed shape of the purlin is illustrated in Fig. 14a. The load factors for the
section are calculated depending on the length of the purlin and shown in Fig. 15. The load factor of 0.178 is determined for the span
length of the purlin in the structure.
The second buckling analysis is carried out in ABAQUS FE software [16] as given in Fig. 14b. In the analysis, an 8-node linear brick
element, C3D8R type, is used. The size of the mesh is determined as approximately 15 mm after the mesh optimization generated by
comparing the buckling load factors. Boundary conditions are defined as pin along the contact surfaces between the bolts and the
profile. The material strength is used as the characteristic yield stress of S235 steel in both analyses. The buckling load factor is
determined under the unit loading. The characteristic moment capacity of the section is calculated by reducing the plastic moment
capacity of the section.
According to the first buckling analysis, the characteristic moment capacity calculation is given in Table 2 and the calculation of the
distributed load safely carried based on the characteristic moment capacity is given in Table 3. In the second buckling analysis, the
distributed load is calculated using the FE program results and indicated in Table 4. The control of whether the load at the time of
collapse can be carried safely is presented in Table 5. Therefore, it is found that the load at the time of collapse cannot be carried by the
purlins.

6.2. Calculation of main girder capacity

The amount of load safely carried by the girder section based on the capacity is determined according to the standards as defined in
Section 6.1. The calculations are presented in tables, in which the characteristic moment capacity calculation of the main girder for
lateral torsional buckling according to TDSSSB-2018, and AISC 360-16 is in Table 6 and the load calculation at the time of collapse is in
Table 7. It is seen in Table 6 that the main girder moment capacity changes depending on the girder length. Therefore, it is required to
calculate the capacity of the girders with respect to the variable unbraced length (lateral torsional buckling length).
According to Figs. 6 and 9, even when the purlins lose their capacity due to buckling, the roof covering elements are still loaded
with the roof snow load, thus the girders still carry the load at the time of collapse. However, the roof covering becomes unable to bear
the snow load as a result of the connection failure of between purlins and girders. Since the out-of-distortion in the purlins affects the

Fig. 14. Buckling deformed shape of the purlin profile a) CUFSM program b) FE program.

9
Z. Fırat Alemdar and F. Alemdar Engineering Failure Analysis 125 (2021) 105378

Fig. 15. Load factor variation in CUFSM analysis.

Table 2
Characteristic moment capacity according to first buckling analysis.
Characteristic yield stress (MPa) Elastic section of modulus (mm3) Load factor Strength coefficient Characteristic moment capacity (N.m)

235 38,790 0.178 0.9 1367.640

Table 3
Distributed load calculation in the first buckling analysis.
Characteristic moment Moment under uniform load for a Purlin length Uniform distributed Purlin interval Distributed load based on
capacity (N.m) simply supported beam (mm) load (N/mm) (mm) capacity (kN/m2)

1367.640 Q*L2/8 6000 0.300 1000 0.300

Table 4
Distributed load calculation in the second buckling analysis.
Load Vertical uniform area load in the The vertical uniform load caused Purlin flange width Distributed area load based on
factor analysis (kN/m2) buckling (kN/m2) (mm) capacity (kN/m2)

7.13 * 1000 7.130 45 0.321

3
10−

Table 5
Control of safety for purlins.
Purlin weight Roof cover Roof snow Total load at the time of Distributed load (CUFSM Distributed load (FE Safety
(kN/m2) weight (kN/m2) load (kN/m2) collapse (kN/m2) analysis) (kN/m2) analysis) (kN/m2) condition

0.058 0.088 0.734 0.880 0.300 0.321 Unsafe

Table 6
Characteristic moment calculation of the girder.
Characteristic yield stress Span length Resistance factor for Plastic moment (N. The characteristic moment for lateral torsional
(MPa) (mm) flexure m) buckling (N.m)

235 20,000 0.9 276,430 53,700

235 12,050 0.9 276,430 276,430

10
Z. Fırat Alemdar and F. Alemdar Engineering Failure Analysis 125 (2021) 105378

Table 7
Load calculation at the time of the collapse.
Main girder weight (kN/m2) Purlin weight (kN/m2) Roof cover weight (kN/m2) Roof snow load (kN/m2) Total load at the time of collapse (kN/m2)

0.111 0.058 0.088 0.734 0.991

moment capacities of the girders, this situation should be taken into account in the calculations in Table 6. Since there is no other
structural element used to prevent lateral buckling of the girders, the lateral capacity completely depends on the condition of the
purlins.
Considering the capacity calculations above, the contribution of the purlins on the main girder to the unbraced length of the girder
are investigated based on the F1. Commentary section of AISC360-16 [14] and the research conducted by Yura and Helwig [17]. Since
the number of the purlins collapsed on the girders affects the unbraced length of the main girder, the moment carrying capacity and
thus the load-bearing capacity of the girder will change. Depending on the number of purlins collapsed, lateral torsional buckling
conditions are examined according to the unbraced lengths of the main girder.
The three-dimensional finite element model of the entire structure is defined by using the SAP2000 program [18] as shown in Fig. 1
to consider the column-girder connection and determine the system response. In the modeling, column supports are considered as
simple supports due to the insufficient in-situ anchorage details. The column-girder and girder-purlin regions are modeled as moment
connections. The characteristic strength values are defined for the steel material properties. Brace members in the model are not
defined since the analysis is conducted under only the vertical loading. Due to the positions of the main beams on the columns, the axial
forces of the beams are allowed to generate moment in the modeling. The nonlinear geometric analysis was conducted to obtain the
moment values in the mid-span of the main girders by applying the vertical unit load over the roof area in Fig. 16.
The number of the purlins collapsed on the main girders was taken into account in the analysis and the unbraced lengths were
considered in the calculation of capacities of the girder. The maximum uniformly distributed load values safely carried by the girder is
then obtained by taking into account the moment capacities and the moment results in FE analysis. The distributed load values that can
be carried depending on the number of purlins failed are shown in Fig. 17 for the main girder having different lengths. In the load
calculation of the main girders on the sides, the effective roof area is taken as half and thus these girders are not shown in the Fig. 17.
The calculations show that the capacities of the girders and purlins are adequate without roof snow load. However, during the
snowfall the bearing strength of the girders is lost due to the lateral torsional buckling and the load carrying capacity of the system is
decreased. Consequently, the progressive collapse of the structure occurred.

Fig. 16. Moment results of the girders in FE analysis.

11
Z. Fırat Alemdar and F. Alemdar Engineering Failure Analysis 125 (2021) 105378

Fig. 17. The distributed load values of the main girder for variable lengths.

7. Discussions

The roof snow load value calculated according to the meteorological data is less than the snow load value that should be taken into
account in the design project. Thus, it is expected that the snowfall on January 6, 2020 may not destroy the structure. The partial
collapse of the structure reveals that the failure occurred due to deficiencies in the design or improper construction of the steel
structure. These problems will be discussed below.
The load safely carried by the girders is studied depending on the length of the girder as seen in Fig. 17. Since the distributed load
capacity of the girders is lower than the total load at the time of the collapse, failure is observed on the girders as a result of
vulnerability to lateral torsional buckling. This type of failure was inspected on-site as seen in Figs. 6 and 9. The reason why the girders
remaining in their positions despite excessive deformation in some spans, and observing total destruction of the girder with the column
in other spans is thought to be the column supports where the anchorages are insufficient. The main girders, which maintain their
positions earlier, have then become load-free due to local and general buckling.
Although the connections of purlins to the main girders and their arrangement positions are not constructively sufficient to provide
required lateral rigidity (see Fig. 8); the change in the unbraced length of the girder is taken into account in Fig. 17 according to the
assumption that the purlins prevent lateral torsional buckling of the main girders. Although it may be thought in the project design that
the main girders provide sufficient lateral rigidity with purlins, perhaps the use of connection stiffeners between the girders would be
of vital importance for this considered structure.
Depending on the inability to support the main girder laterally by losing the bearing strength of the purlins step by step as seen in
Fig. 17, the load-bearing capacity of the main girder decreases with the increase in the unbraced length of the main girder. According
to the number of purlins collapsed, the load capacity of the girders in other span decreases approximately from 0.30 to 0.60 kN/m2 to
0.10 kN/m2. Thus, the main girders only carry the self-weight. The total load at the time of collapse cannot be safely carried by the
main girders.
The columns also facilitated the deformation of the girders due to the inadequacy of connections on the column-base plate regions.
It can be said that it is easier for the system to lose its total rigidity. Besides, the failure of the column-girder connections accelerated the
loss of stability of the girders (Fig. 5).

8. Conclusions

In this study, the reasons for the collapse of a steel structure of the factory are investigated under expected snow loads. The collapse
mechanism obtained as a result of on-site examinations and observations is defined and the structural capacities of the main girder and
the purlin are obtained to reveal their effects on the collapse behavior. The following conclusions can be drawn:

1. In the lights of the analytical results and observations, it is thought that the progressive collapse mechanism of the considered
structure occurred with purlins first losing the stability due to torsion and then the main girders being unable to withstand loads
due to lateral torsional buckling.
2. It is detected that the snowfall could be trigger of the collapse but if the built structure was constructed according to the standards,
the structure would not have collapsed.
3. Structural details are missing in the shop drawing, especially details to ensure local and general stability, for example, not designed
transverse stiffener plates at the end of the column, girder and corbel connection, the absence of tie rods to increase the out-of-plane
stability for the purlins, the lack of brace elements to ensure the out-of-plane stability of the trapezoidal portal system.
4. There are workmanship errors or deficiencies in the structural system built on-site, especially, the anchor rod lengths in the support
of the columns are not suitable with the project drawing, the anchor plates inside the footing are missed, and the connections of the
tie rods to prevent the out-of-plane movement of the purlins are not made or made incompletely along the web region of purlin
elements.
5. It is determined from the analyses that the profiles selected for purlin and main girders are not sufficient to carry the structural
system weight and the design snow load at their elevation.

12
Z. Fırat Alemdar and F. Alemdar Engineering Failure Analysis 125 (2021) 105378

6. The national code provides accurate calculation of the snow load which was observed at the studied site.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to
influence the work reported in this paper.

Acknowledgments

The authors remain grateful for the assistance of the Allianz Insurance Incorporated Company, Istanbul, Turkey.

Funding sources

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

References

[1] G.R. Bell, Engineering investigation of structural failures, in: R.T. Ratay (Ed.), Forensic Structural Engineering Handbook, McGraw-Hill, New York, 2000, 6.6,
6.11 (Chapter 6).
[2] O. Caglayan, E. Yuksel, Experimental and finite element investigations on the collapse of a Mero space truss roof structure-a case study, Eng. Fail. Anal. 15 (5)
(2008) 458–470, https://doi.org/10.1016/j.engfailanal.2007.05.005.
[3] A. Biegus, K. Rykaluk, Collapse of katowice fair building, Eng. Fail. Anal. 16 (5) (2009) 1643–1654, https://doi.org/10.1016/j.engfailanal.2008.11.008.
[4] A. Brencich, Collapse of an industrial steel shed: a case study for basic errors in computational structural engineering and control procedures, Eng. Fail. Anal. 17
(1) (2010) 213–225, https://doi.org/10.1016/j.engfailanal.2009.06.015.
[5] K. Ozakgul, O. Caglayan, O. Tezer, Investigation of buckled brace system of an existing industrial building, Eng. Fail. Anal. 18 (1) (2011) 455–463, https://doi.
org/10.1016/j.engfailanal.2010.09.037.
[6] J.J. del Coz Díaz, P.J. García Nieto, J.A. Vilán Vilán, J.L. Suárez Sierra, Non-linear buckling analysis of a self-weighted metallic roof by FEM, Math. Comp.
Model. 51 (3-4) (2010) 216–228, https://doi.org/10.1016/j.mcm.2009.08.032.
[7] G. Piskoty, L. Wullschleger, R. Loser, A. Herwig, M. Tuchschmid, G. Terrasi, Failure analysis of a collapsed flat gymnasium roof, Eng. Fail. Anal. 35 (2013)
104–113, https://doi.org/10.1016/j.engfailanal.2012.12.006.
[8] A.C. Altunişik, Ş. Ateş, M. Hüsem, Lateral buckling failure of steel cantilever roof of a tribune due to snow loads, Eng. Fail. Anal. 72 (2017) 67–78, https://doi.
org/10.1016/j.engfailanal.2016.12.010.
[9] C. Vatansever, E. Seçkin, G. Yazıcı, Snow induced collapse of hail protection canopies, Eng. Fail. Anal. 79 (2017) 186–197, https://doi.org/10.1016/j.
engfailanal.2017.04.032.
[10] F. ALMahdi, A. Doğangün, F. Genç, W. Rasekh, M.Ö. Timurağaoğlu, Investigation of snow-induced collapse in Bozüyük market, Engin. Fail. Anal. 118 (2020)
104822, https://doi.org/10.1016/j.engfailanal.2020.104822.
[11] Isparta Meteorology Directorate, Meteorology Reports. https://mgm.gov.tr/?il=Isparta, 2020 (accessed 10 April 2020).
[12] TS EN 1991-1-3, Eurocode 1. Actions on Structures, Part 1–3: General Actions - Snow Loads, TSE Turkish Standards Institution, Ankara, Turkey, 2007.
[13] Turkish Design Specification for Structural Steel Buildings, Ankara, Turkey, 2018.
[14] ANSI/AISC 360-16, American Institute of Steel Construction, Specification for Structural Steel Buildings, AISC Committee on Specifications, Illinois, USA, 2016.
[15] CUFSM Cross-section elastic buckling analysis program, Thin-walled Structures Group, Civil and Systems Engineering, Johns Hopkins University, 2020.
[16] ABAQUS/Standard User’s Manual, Version 6.14. Providence, RI: Dassault Systèmes Simulia Corp, 2014.
[17] J. Yura, T.A. Helwig, Buckling of beams with inflection points, in: Proceedings Structural Stability Research Council, Annual Stability Conference, 2010, pp.
761–780.
[18] SAP2000 Advanced 21.0.0, Structural Analysis Program, Computers and Structures Inc., Berkeley, California, 2019.

13