Thin-Walled Structures 143 (2019) 106259

Contents lists available at ScienceDirect

Thin-Walled Structures
journal homepage: www.elsevier.com/locate/tws

Full length article

Experimental analysis of the effect of dent variation on the buckling capacity T

of thin-walled cylindrical shells
Fatma Merve Aydın Korucuk, Mahyar Maali, Mahmut Kılıç, Abdulkadir Cüneyt Aydın*
Ataturk University, Engineering Faculty, Department of Civil Engineering, 25030, Erzurum, Turkey

ARTICLE INFO ABSTRACT

Keywords: Tanks, silos, and most large steel-shell structures consist of smaller pieces connected together during the man-
Theoretical and experimental ufacturing process. This causes several types of malformations on the shell walls. Furthermore, thin-walled
Cylindrical shell members can be easily deformed in wall surfaces owing to the thickness of the structure. Fourteen thin-walled
CFRP strips cylindrical shell specimens in two groups with different dent depths and various longitudinal dent numbers
Longitudinal dent number
subject to hydrostatic pressure were tested in the present work. The models were designed to demonstrate how
Hydrostatic pressure
repairing dents by using carbon-fibre-reinforced polymer can recover lost capacity. The results of testing under
different theories and codes were compared. This study shows the decreasing effects of the longitudinal dent
number on the buckling capacity of the shells. Using carbon-fiber-reinforced polymer strips resulted in softening
or stiffening behaviour in the models. Furthermore, to obtain the initial and overall buckling according to
theoretical formulas, coefficients were predicted to obtain the initial, overall, and collapse buckling without an
experiment for the models that were beyond the scope of the theories.

1. Introduction cylindrical shells against elephant-foot buckling, in which a small

For a decade, cylindrical geometry has been considered as an im-
portant thin-walled structure type for shells. The member structure
plays the main role in the deformations and disturbances of the wall
surfaces. The manufacturing process for thin-walled cylindrical shells is
critical for the designed member. During the manufacturing process,
the panels, which are used for tanks and silos, exhibit geometrical de-
fects. The thin-walled shells, a classical structural analysis problem, are
a continually increasing concern for modern industries. Thus, it is im-
portant to examine the problem of instability in this type of structure.
Many steel structures such as high-water tanks, water and oil re-
servoirs, marine structures, and pressure vessels, including shell ele-
ments, are widely under stress. Furthermore, shell elements are subject
to instability owing to the loads applied. The theoretical buckling re-
sistance (theoretical) is based on a two-branch linear elastic analysis
that is suitable for conventional cylinder shells.
There have been many research studies regarding thin-walled
structures, including many geometric properties and types of loads [1].
Teng and Hu (2007) [2] examined the benefits of the fiber reinforced
polymer (FRP) confinement of hollow steel tubes. Moreover, axial
compression tests on FRP-confined steel tubes were described. Batikha
et al. (2009) [3] investigated a novel method of strengthening
amount of FRP composite used at a critical location can effectively
eliminate the problem and increase the buckling strength. However, the
elephant-foot-type buckling phenomenon is a result of axial load,
especially for thin-walled short cylindrical shells, and is outside the
scope of this work. There have been many studies, both experimental
and theoretical, regarding the calculation of buckling loads for thin-
walled shells [1–19]. As an example, Vakili and Showkati [16] studied
elephant-foot-type buckling and the retrofitting of cylindrical shells
using FRP.
Maali et al. (2012) [4] discussed 14 laboratory specimens in two
groups: shallow conical caps (SCC) and deep conical caps (DCC).
Models were loaded under uniform hydrostatic pressure. The samples
were modified to include either one- or two-line imperfections with
amplitudes of thickness; 1t, 2t, and 3t in depth.
Fatemi et al. (2013) [5] investigated imperfect cylindrical shells
under uniform external pressure. Nilufari et al. (2014) [6] discussed 12
laboratory specimens in three groups, loaded under uniform hydrostatic
pressure. The samples were modified to include circumferential im-
perfections at the junctions between the curved edges of the panels of
the cylindrical and conical shells with amplitudes of 2t, 4t, and 8t in
depth.
Most recently, the effect of longitudinally stiffened cylindrical shells

*
Corresponding author.
E-mail address: acaydin@atauni.edu.tr (A.C. Aydın).

https://doi.org/10.1016/j.tws.2019.106259
Received 31 May 2018; Received in revised form 12 April 2019; Accepted 12 June 2019
Available online 18 June 2019
0263-8231/ © 2019 Elsevier Ltd. All rights reserved.
F.M. Aydın Korucuk, et al. Thin-Walled Structures 143 (2019) 106259

under external pressure was investigated, and researchers showed that Table 1
the buckling capacity increased for the tested models [7–9]. Ad- Tensile Properties of CFRP and epoxy used.
ditionally, Ghanbari Ghazijahani et al. (2014) [10] studied damaged Name and Supplier Type ρ (g/cm3) E (GPa) σ (MPa) ε (%)
cylindrical shells under compression. This study presented an experi-
mental program in which the buckling and failure response of damaged BASF, MasterBrace, Thermoplastic 1.77 227 3800 1.67
300/50 CBS
shell specimens were analysed. Ghanbari Ghazijahani et al. (2014) [11]
BASF, MasterBrace, Low viscosity 0.983 3.034 55.2 3.5
conducted experiments on dented cylindrical shells under peripheral SAT 4500 epoxy
pressure. They reported on an experimental program concerning the
buckling and post-buckling response of thin cylindrical shells with local
dent imperfections under uniform external pressure. The results of this Table 2
study can be used in practical structures with similar geometrical fea- Initial geometries of test models.
tures.
Group Specimen Model Ld hd bd FD di
Ghanbari Ghazijahani et al. (2015) [12] focused on recovering the
fatigue life for larger cutouts by reinforcement with carbon fibre re- Without t-2t-d2 M1 hc/2 t 2t – 2
inforced polymer (CFRP). Fatigue life was not only fully recovered with CFRP 2t-4t-d2 M2 hc/2 2t 4t – 2
t-2t-d4 M3 hc/2 t 2t – 4
the CFRP reinforcement but was also dramatically increased relative to
2t-4t-d4 M4 hc/2 2t 4t – 4
the unreinforced intact specimen. In addition, Ghanbari Ghazijahani t-2t-d6 M5 hc/2 t 2t – 6
et al. (2015) [13] studied a new approach for strengthening in which 2t-4t-d6 M6 hc/2 2t 4t – 6
vertical corrugations were introduced to 11 cylindrical shell specimens Perfect model M7 – – – – –
under uniform external pressure. The results showed a considerable With CFRP-t-2t-d2 M8 hc/2 t 2t 3bd × (ld + 2bd) 2
CFRP CFRP-2t-4t-d2 M9 hc/2 2t 4t 3bd × (ld + 2bd) 2
increase in the buckling capacity of such structures. Ghanbari Ghazi- CFRP-t-2t-d4 M10 hc/2 t 2t 3bd × (ld + 2bd) 4
jahani et al. (2015) [14] studied the effect of large local imperfections, CFRP-2t-4t-d4 M11 hc/2 2t 4t 3bd × (ld + 2bd) 4
known as dents, on the plastic buckling capacity of short steel tubes CFRP-t-2t-d6 M12 hc/2 t 2t 3bd × (ld + 2bd) 6
under axial compression. A total of 11 tests on such short columns CFRP-2t-4t-d6 M13 hc/2 2t 4t 3bd × (ld + 2bd) 6
Perfect model with M14 – – – Entire-surface CFRP –
conducted. Ghanbari Ghazijahani et al. (2015) [15] discussed the
entire-surface
structural behaviour of an innovative composite column through an CFRP
experimental study. The new composite was composed of steel cy-
lindrical hollow sections (CHSs), solid timber infill, and CFRP con- Specimen: hd – bd – di; hd: Dent depth; bd: Dent width; di: dent number (d2, d4,
finements. and d6 as two, four, and six dents, respectively); Ld: Dent length; hc: Height of
The abovementioned literature review may be concluded as follows. cylinder (1250 mm for all models); t: Thickness of cylinder (t = 1 mm for all
The cylindrical shell is a main structural element and is considered to be models); FD: Fibre dimensions (width × height); R: Radius (500 mm).
a basic need in modern industry. The structures of shells are prone to
buckling phenomena owing to their particular shape. The insignificant width of the CFRP strip, and Ld + 2bd is the length of the CFRP strip
thickness of the other dimensions and the emergence of compressive (Ld = dent length, and bd = dent width).
stresses owing to loading are factors that cause the buckling phenom-
enon. In this research, dents are introduced at different depths of t and 2.2. Testing system and manufactured specimen
2t, and with various longitudinal dent numbers (d2, d4, and d6), into
cylindrical tanks. The models are designed to show how the repair of The test machine consists of two plates: upper and bottom plates
dents using CFRP can recover lost capacity as well. with dimensions of 1500 × 1500 mm and a plate thickness of 30 mm.
The models are divided to two groups: the without-CFRP group and The plates were first cut into 1500 × 1500-mm sizes by a CNC cutting
the with-CFRP group. Each group is compared to the individual models device and then ring-cut with a CNC cutting machine. Soldering was
of the group itself, and then both groups are compared together. All applied to minimise the error rate and improve the weld performance
initial and overall buckling and collapse are compared using theoretical via the seam fusing method [4–6,20–22]. To decrease the initial dis-
formulas. Finally, the buckling waves (using theory) and the method of tortion and residual stress, the edges of the cylinder were joined
reversal are described. The purpose of the present work is to investigate through soldering. The steel plates were cut in the exact dimensions of
the effect of dents and/or CFRP on the buckling behaviour of cylinders the models and rolled into cylinders by a rolling machine.
which are not in the range of theoretical formulas. Thus, it will be Fig. 2 shows the test machine used in this research. In one of the
possible to obtain the initial and overall buckling loads according to plates on the upper part, there are two holes in the middle of the pas-
theoretical formulas, without experiments, by using some coefficients. sage of the vacuum hose from the first hole. The load cell is inserted
into the second hole, and four sets of studs on the four sides of the plate
2. Testing process lift the plate using a welding crane. On the sides of the plates, three
holes of the bolt are intended to be six in total, until the configuration
2.1. Test specimens and properties shown in Fig. 2 is achieved. The reason for these six bolts is the in-
stallation of a top plate assembly, which, when tested with the crane,
In this research, 14 laboratory specimens in two groups with ver- prevents these bolts from entering the axial load into the cylinder.
ified dent numbers (d2, d4, and d6) and with dents at different depths After the test machine was manufactured, it was time to build la-
of t and 2t are examined. The first group of specimens (seven speci- boratory samples. Three tensile coupon tests were performed to obtain
mens), labelled ‘without CFRP’, and the second group, ‘with CFRP’, the material properties. The average yield and failure stresses were
were loaded under hydrostatic pressure. Each group contained a perfect found to be 198.8 MPa and 342.4 MPa, respectively. Young's modulus
model and a perfect model with an entire surface of CFRP, with the was calculated to be 210 MPa, and Poisson's ratio was obtained as 0.29.
remaining specimens having a dent with amplitudes of t and 2t To make samples, sheets with a thickness of 1 mm were prepared. Ten
(t = thickness of the cylindrical shell). A perfect model and a perfect sheets were cut to 1250 mm in height and 3140 mm in length with
model with entire-surface CFRP were used for control in each group. scissors. Then, a cylindrical roller was used to create rolled specimens.
The details of the CFRP and epoxy are presented in Table 1, and the During the welding, it was considered that the lowest porosity and in-
details of the specimens are presented in Table 2 and Fig. 1. The CFRP adequacy of welding on the laboratory models were not achieved, and
strip was calculated with formula 3bd × (Ld + 2bd), where 3bd is the the models were prepared without imperfections in welding. In

2
F.M. Aydın Korucuk, et al. Thin-Walled Structures 143 (2019) 106259

Fig. 1. Details of test model.

addition, during rolling, we were careful not to cause any imperfec-
tions.
The edges of the models were welded through metal inert gas
welding by using 0.8-mm electrodes for a 1-mm weld thickness. After
the creation of the laboratory models, it was necessary to create
inadequacies. To create disadvantages in the models, stainless steel
shapes (SPKR) were made in the form of cutting and the wire cut
method. Fig. 2 shows the dent piece for the experimental tests. First, a
dent length was indicated at the centre of the height of the model with a
ruler and a marker, and then a four-base was laid inside the model.

Fig. 2. Details of test machine, crane, dent piece, CFRP detail, and dent implementation in models.

3
F.M. Aydın Korucuk, et al. Thin-Walled Structures 143 (2019) 106259

Table 3
Layout of strain gauges and transducers for all test specimens.
Model Transducers Strain gauges

D1 D2 D3 D4 S1 S2

Height (mm) D1° Height (mm) D2° Height (mm) D3° Height (mm) D4° Height (mm) S1° Height (mm) S2°

M1 650 3 600 91 625 178 610 269 715 93 715 94

M2 615 0 620 102 600 180 650 280 600 110 600 111
M3 650 10 600 112 600 180 640 270 670 119 670 120
M4 650 3 620 90 680 185 650 275 690 118 690 119
M5 650 0 600 90 680 170 600 270 600 110 600 111
M6 600 0 640 90 600 175 650 275 680 127 680 128
M7 650 2 600 85 600 193 635 270 600 128 600 129
M8 600 0 660 90 600 185 600 290 680 110 680 111
M9 620 2 600 85 600 170 600 260 630 105 630 106
M10 650 10 600 100 620 180 670 280 640 110 640 111
M11 650 0 650 100 600 182 650 270 650 115 650 116
M12 600 0 670 90 620 180 600 280 600 124 600 125
M13 620 10 600 90 650 180 670 255 650 120 650 121
M14 650 11 600 80 600 175 660 278 630 107 630 108

D1o = Degree from start point of cylinder (0–360°); Height is from ground of the cylinder.
D: Displacement transducer; S: Strain gauge.

A static load pump was employed with a maximum load of 900 kN,
stroke of 300 mm, and constant speed of 0.016 mm/s. The pump was
used to create dents in the models [23–30]. In Fig. 2, the dent im-
plementation in the models is shown. All dents were manufactured to
be inward and equidistant from each end for all of the models. Fur-
thermore, the dents were manufactured with equal angles (3600/dent
number). For the group with CFRP models that were to be used, two-
component epoxy-coated adhesive was stuck onto the specified parts.
Table 2 lists the data for the experimental models.
To install the laboratory models, the following was accomplished.
Beginning at the bottom of the laboratory model, after fitting the model
on the bottom plate of the silicone silencer model, the edge of the model
was filled to the plate inside and outside the cylinder to prevent air
from escaping from the inside of the cylinder. After that, to fit the other
side of the cylinder, the bolts around the device were opened, and the
upper plate was slowly raised and guided by a crane to the cylinder.
Prior to guiding the top plate, the bolts were adjusted to the height of
the cylinder to prevent a sudden collision of the plate with the cy-
lindrical model.
Finally, after inserting the top plate, the bolt shown in Fig. 2 was
closed to prevent gravity loading on the cylindrical model. Alternately,
the crane transported the top plate during the test to maintain a gravity-
load-avoidance sample. After the plate around the cylinder was filled
with silicon glue, silicon was added to an argon boil in order to prevent
air from escaping.
After performing the above steps, four metal support rings, two on
the bottom and two on the top of the cylinder, were installed so that all
laboratory models with a simple support were closed only in the radial
direction.
To carry out a vacuum pump test, a vacuum pump manufactured by
Zinisan with a hydrostatic pressure capacity of 600 kPa was used. The
literature regarding cylindrical shells, most of which are mentioned
here, was designed for decreased thickness and height. Furthermore,
the vacuum systems of their experimental setups were manual and had
limited capability. However, the setup developed for this work has an
automatic vacuum system. Thus, cylindrical models with 1 mm of
thickness and 1250 mm of height could be tested. Fig. 2 shows the
vacuum pump used in the experimental tests.
To collect information, a data logger was used: the INSTURUNET
1420 model with 120 channels and 13 inputs. Displacement transducers
were used to measure the deformations under external pressure. Four
linear variable displacement transducers with a maximum displacement
of 300 mm of the Japanese SDP-300 type, labelled as D1, D2, D3, and
D4, were used to record the displacements at different locations along
the height and circumference of the specimens (locations of LVDTs are
presented in Table 3). The error of the LVDT was 0.001 mm. During the
test, we connected all four LVDTs to the base, to prevent magnetization.
Fig. 2 shows the LVDTs. In all of the experiments, two strain gauges
(S1 = vertical and S2 = horizontal) were installed. Table 3 lists the
locations of the strain gauges. Before installing the strain gauges, we
placed the strain gauges thoroughly with a flap and then, after cleaning,
with a special adhesive on the desired area. The strain gauges were of
type YEFLA-5. Their resistance was 121 Ω ± 0.5%, the coefficient of
sensitivity was 2.1 ± 2%, and their size was 5 mm × 3 mm. This is
shown in Fig. 3. Furthermore, two cameras recorded the experiments in
order to obtain the initial, overall, and collapse buckling loads ac-
cording to time stamps. The first buckling wave creation time signified
the initial buckling, and so on.
3. Results
3.1. Models without CFRP
Fig. 4 shows some of the load-total displacement and load-strain
curves of the models without CFRP. The initial buckling, overall
buckling, and collapse of the without-CFRP group are summarised in
Table 4 and presented in Fig. 5 for all models. Before the initial buck-
ling, the cylindrical shell was steady and had no buckling behaviour.
Each of these breaks represents one or more than one wave in these
load-total displacement curves. There is also a correlation between the
load-total displacement and load-strain plots of these two completely
different tools, which can be taken as further proof of the installation
and measurement precision.
Vertical strains were more common than horizontal strains. This
may be explained by the effect of external pressure. Table 4 shows a
comparison of the initial buckling, overall buckling, and collapse of the
without-CFRP group. Table 4 shows that, after the initial buckling load,
the models, themselves resisted for post-buckling behaviour. Then, the
models entered the stage of post-buckling. A phenomenon observed in
all models was an increase in the buckling capacity after the initial
buckling. In general, the overall buckling load was approximately
20–60% of the initial buckling load, and the collapse buckling load was
approximately 25–70% of the initial buckling load. Furthermore, the
initial buckling load decreased with an increased number of dents from
2 to 6.
The initial buckling load decreased for the dented models when

4
F.M. Aydın Korucuk, et al.
Fig. 3. S1 and S2 strain gauges of M2 model.
compared with the perfect model (M7). The dent number, depth, and
width are the main parameters. However, the dent numbers showed the
greatest decreasing effect in the d6 models (M5 and M6). The de-
creasing effects of the dent number became apparent after the number
of dents was greater than four, which was lower than half that of the
perfect model. The decrease in the initial buckling load was approxi-
mately 14–28% for models M1, M2, M3, and M4 (Table 4).
The gradual decrease in the initial buckling load was the result of
increasing dent numbers. However, the proportional decreasing of the
initial buckling is increased with decreasing angles of the dents, and the
possibility of wave formation increased as the dent numbers increased.
Furthermore, the dent width is the second important parameter for the
initial buckling load. The decrease in the initial buckling load was as
low as 17.51% for the M3 model but was 26.26% for the M2 model.
The initial buckling load decreased with increasing amplitude when
the dent number was the same as the dent depth. With an increasing
dent number from 2 to 6, the initial buckling load was −1.59 and
−92.73% for M3 and from M1 and M5 to M1, respectively. Thus, the
initial buckling load decreased with increasing amplitude for dent
numbers with different dent depths. Table 4 presents models with the
same dent number and varying dent depths (M2 compared to M1, M4
compared to M3, and M6 relative to M5). The initial buckling load
decreased by 10.16%, 9.15%, and 13.96% for M2 and from M1 and M4
and from M3 and M6 to M5, respectively. It can be concluded that the
models with the same dent number and depth were different with in-
creasing dent depth, and the model was weakened. However, the dent
depth was limited by doubling the size, as the thickness of the cylinder,
throughout this work.
The overall buckling load is the second checkpoint for the models.
However, the dents and all of the related parameters decreased the
initial buckling load, and the overall buckling increased by approxi-
mately 7–12% for models M2 and M3. Furthermore, the range between
the initial and overall buckling was similar. The dents behaved as
stiffeners after the initial buckling for models M2 and M3. In this work,
the dent depths and dent widths were planned according to the thick-
ness of the cylinder. However, Table 4 indicates that there must be
transition zones between M1 to M2 and M3 to M4 for the overall
buckling. The initial overall buckling range increased for all models.
For the M5 model, the range was three times greater than that for the
perfect model.
The collapse loads of the models were greater than that of the
perfect model at approximately 10%, except for M4 and M1. For M6,
5
Thin-Walled Structures 143 (2019) 106259
the collapse range (the range between the overall to collapse load) was
approximately four times greater than that of the perfect model. Thus,
the plastic behaviour of the model appeared at this stage. When the
dent numbers reached four, the collapse range reached the smallest
value; however, increasing the dent depth, width, and number widened
the range significantly. Thus, an exploration of dent implementation
according to the dent cross-section geometry and depth-width ratio can
be performed in a future work.
Fig. 6 shows the before-and-after tests of the models without CFRP.
The wave numbers of the models without CFRP are equal (6); never-
theless, the wave lengths are not. The failures in all of the samples are
V-shaped, that is, raising the edges in the form of a V causes the model
to fail. In all of the without-CFRP group models, the first buckling wave
was observed from the locations of the dents.
3.2. Models with CFRP
The CFRP reinforcement geometry was designed for one section
throughout this study. The CFRP dimensions were designed according
to the dent properties. However, the dimensions around the dent were
fixed as one dent width. Thus, the reinforcement effect of the CFRP is
concluded according to this idea.
Fig. 7 shows some of the load-total displacement and load-strain
curves of the models with CFRP. The initial buckling, overall buckling,
and collapse of the with-CFRP group are summarised in Table 4 and
presented for all models in Fig. 5.
Table 5 shows a comparison of the initial buckling, overall buckling,
and collapse of the with-CFRP group. Table 4 shows that after the initial
buckling load, the models, resisted themselves for later stages of the
buckling. Then, the models entered the stage of post-buckling. A phe-
nomenon observed in all specimens was an increase in buckling capa-
city after the initial buckling. For the perfect models, the initial, overall,
and collapse buckling loads increased by 55.6%, 24.3%, and 18.5%,
respectively. These increments are the greatest values that can be ex-
pected for the CFRP coverage. However, six dents are not expected in a
model; the lowest initial buckling loads were observed according to
theoretical considerations. The dent area, defined with the dent depth
and height, is the main reason for buckling capacity loss, and the dent
area is a more important parameter than the dent number, as seen in
Fig. 5. Furthermore, the decrease in the capacity loss was not as high as
for the models with six dents.
However, the initial buckling of the models decreased for all dented
F.M. Aydın Korucuk, et al. Thin-Walled Structures 143 (2019) 106259

Fig. 4. Load-total displacement and load-strain curves of models without CFRP.

6
F.M. Aydın Korucuk, et al. Thin-Walled Structures 143 (2019) 106259

Table 4 according to the fully CFRP-covered model M14. In addition, an in-

Comparison of initial buckling, overall buckling, and collapse of models. creasing dent area presented a greater decreasing initial-buckling-ca-
Model P1 (kPa) P2 (kPa) P3 (kPa) (P2 - (P3 - (P1 - M7 (P1 - M14 pacity loss of 44–65% for models M9, M11, and M13. Nevertheless, the
P1)/P2 P1)/P3 P1)/P1 P1)/P1 greater values of capacity loss decreased to approximately 24% for the
(%) (%) (%) (%) overall buckling and collapse capacity. The gradual decrease was best
observed in the collapse loads of the models with CFRP. However, the
M1 161.40 216.85 225.84 25.57 28.53 −14.62 −78.30
M2 146.51 251.50 276.36 41.75 46.99 −26.26 −96.42
overall-to-collapse buckling range was not as expected. This range was
M3 157.42 265.29 267.51 40.66 41.15 −17.51 −82.81 the smallest for M12.
M4 144.22 197.83 208.96 27.1 30.98 −28.27 −99.54 It can be concluded that models with the same dent number and
M5 86.63 227.39 267.24 61.90 67.58 −113.54 −232.19 dent depth are different with increasing dent depth, and the model is
M6 76.02 194.21 254.03 60.86 70.07 −143.34 −278.56
weakened. The dent depth is a weaker element of the model. Fig. 8
M7 184.99 233.52 248.56 20.78 25.57 – −55.57
M8 168.92 225.03 283.66 24.93 40.45 −9.51 −70.36 shows the before-and-after tests of the models with CFRP. It is well seen
M9 154.77 190.82 225.82 18.89 31.46 −19.53 −85.94 that the number of waves obtained from the experimental test of the
M10 167.08 224.15 231.45 25.46 27.81 −10.72 −72.24 with-CFRP group was 5, but the lengths of the waves were not equal.
M11 160.92 219.80 221.49 26.79 27.35 −14.96 −78.83
The failures in all of the samples were in the form of a V shape, that
M12 105.73 201.29 201.44 47.47 47.51 −74.96 −172.18
M13 98.36 207.41 222.75 52.58 55.84 −88.07 −192.58
is, raising the edge in the form of a V shape caused the model to fail. In
M14 287.78 290.12 294.49 0.81 2.28 35.72 – the case of the entire-surface perfect CFRP model, it can be said that
after modelling the buckling wave in the experimental model after it
P1: Initial buckling; P2: Overall buckling; P3: Collapse buckling; M: Model; left the CFRP device, the waves returned to a relatively basic state. This
Experimental waves were six and five for models without and with CFRP, re- means that in the fully enclosed model with CFPR, the resulting waves
spectively. can be elastoplastic.

models with CFRP. The decrease for model M9 was particularly inter-
3.3. Buckling loads of all models
esting. While there was a gradual decrease for models M8, M10, M11,
M12, and M13, the initial buckling of M9 was approximately 86% of the
In comparing the with-CFRP group to the without-CFRP group, each
fully covered model M14. This is thought to be the result of a dent area
of the models in the without-CFRP group was compared with one an-
increment effect. Furthermore, the increment of the dent area is not as
other in the with-CFRP group. Thus, the initial buckling load is very
effective for M9 as for the models with four and six dents.
important in the thin-walled structure when compared to the initial
The overall buckling loads were observed as the initial ones.
buckling load of all of the models. The results are listed in Table 5. The
However, the initial-to-overall buckling range was very small for the
initial buckling load of the without-CFRP group increased by
perfect model M14, but model M13 showed the greatest increment. The
4.45%–35.72% compared to the without-CFRP group. Alternately, be-
striking point is that the dent area is the main parameter that increases
cause the initial buckling load is important in this type of structure, the
the range between the initial and overall buckling. Furthermore, the
use of CFRP at the dent sites can have a significant effect on the initial
increment of the initial-to-overall buckling range is lower than that for
buckling load. Tables 3 and 4 show that for experimental waves of the
the models without CFRP. Thus, the CFRP approached the values of the
without-CFRP and with-CFRP groups, the wave number of the with-
initial and overall buckling loads. This means less plastic behaviour at
CFRP group was lower than that of the without-CFRP group. Thus, the
this stage for the models without CFRP. While the overall buckling
use of CFRP in cylindrical thin-walled shells can have a significant ef-
improvement was approximately 24% for the perfect model, only M8,
fect on the waves and plastic range of the models.
M11, and M13 experienced greater overall buckling of approximately
When models M1, M3, and M5 were compared with M8, M10, and
3–10% when compared to similar models without CFRP.
M12, the initial buckling load increased by 4.45%, 5.78%, and 18.06%,
The initial and overall buckling loads of M8 and M10 were similar,
respectively. When models M2, M4, and M6 were compared with M9,
and the collapse load decrease was the lowest for the model M8
M11, and M13, the initial buckling load increased by 5.33%, 10.38%,

Fig. 5. Comparison of buckling loads of all models.

7
F.M. Aydın Korucuk, et al.
Fig. 6. Models without CFRP (M1–M7).
and 22.71%, respectively. Furthermore, as the dent number increased,
the capacity loss decreased to the maximum value for the models
without CFRP. Increasing the dent number also increased the CFRP
area. Thus, the capacity loss resulted in a decrease with the usage of
CFRP. However, a similar variation in the overall buckling and collapse
loads was not observed.
When models M8, M9, M10, and M11 were compared with M1, M2,
M3, and M4, respectively, the CFRP increased the initial buckling by
approximately 5–10%, which is relatively little. However, the initial
buckling increased by approximately 18–22% for M12 and M13 in ac-
cordance with M5 and M6. Thus, the improvement of CFRP became
8
Thin-Walled Structures 143 (2019) 106259
more visible with increasing dent numbers and areas. Furthermore, the
CFRP dimensions can be the main parameter of future work. The per-
fect models presented a substantial improvement in the initial buckling
by approximately 56% (Table 4).
The plastic range from the initial to collapse load was the greatest for
similar models M6 and M13. This is the result of increasing the dent area,
which increased the plastic deformability of the models. In particular, si-
milar models M2 and M9 emphasised the dent cross-section geometry and
depth-width ratio to be searched for the future work, as mentioned earlier.
Furthermore, the dent area also affects the CFRP area. Thus, the CFRP
geometry, area, type, etc., may be used as parameters.
F.M. Aydın Korucuk, et al. Thin-Walled Structures 143 (2019) 106259

Fig. 7. Load-total displacement and load-strain curves of models with CFRP.

Table 5
Initial buckling load comparison of models.
(M8P1 - M1P1)/M8P1 (M9P1 - M2P1)/M9P1 (M10P1 - M3P1)/M10 P1 (M11P1 - M4P1)/M11 P1 (M12P1 - M5P1)/M12P1 (M13P1 - M6P1)/M13P1 (M14P1 - M7P1)/M14P1

4.45 5.33 5.78 10.38 18.06 22.71 35.72

9
F.M. Aydın Korucuk, et al. Thin-Walled Structures 143 (2019) 106259

Fig. 8. Models with CFRP (M8–M14).

4. Comparison of theory with experiments
The study of shell structures and their stability is important. The
history of applying stability equations to shell structures dates back to
the 18th century. In 1795, Euler established the theory of stability of
beams [31]. Then, the theory of the problems of shell-structure
mechanics and many methods for solving them were introduced, and a
large number of researchers have since dealt extensively with this
subject. A general theory of shells was presented for the first time in the
second half of the nineteenth century by Love (1959) [32]. Experi-
mental tests were evaluated to observe the initial buckling loads of
cylindrical shells.

10
F.M. Aydın Korucuk, et al. Thin-Walled Structures 143 (2019) 106259

Subsequently, these results were compared with the relevant codes Table 7
and standards as per Jawad (1994) [33], Venstel and Krauthaer (2001) Comparison of buckling loads obtained from experiments and theory.
[34], Ross (2007) [35], the British Standards Institution Code (2009) Model P1/PJ P2/PJ P1/PR P2/PR P1/PBSI P2/PBSI P1/PECCS P2/PECCS
[34], and the European Recommendations Code (1988) [33]. Jawad
(1994) [33], Venstel and Krauthaer (2001) [34], and Ross (2007) [35] M1 1.16 1.56 1.15 1.55 1.22 1.64 1.12 1.51
M2 1.06 1.81 1.05 1.80 1.11 1.90 1.01 1.74
obtained the critical load of the external pressure in cylindrical shells.
M3 1.13 1.91 1.12 1.89 1.19 2.01 1.09 1.84
Cylindrical shells are assumed to be simply supported with axial re- M4 1.04 1.43 1.03 1.41 1.09 1.50 1.002 1.37
straints in these equations: M5 0.62 1.64 0.62 1.62 0.65 1.72 0.60 1.58
t M6 0.55 1.40 0.54 1.39 0.57 1.47 0.52 1.34
0.92 E ( R )2.5 M7 1.34 1.68 1.32 1.67 1.40 1.77 1.28 1.62
Jawad theory: Pj hc M8 1.22 1.63 1.20 1.61 1.28 1.70 1.17 1.56
R (1) M9 1.12 1.38 1.10 1.36 1.17 1.44 1.075 1.32
M10 1.21 1.62 1.19 1.60 1.26 1.70 1.16 1.55
t
2.6E ( 2R ) 2.5 M11 1.16 1.59 1.15 1.57 1.22 1.66 1.11 1.52
Ross theory:PR = hc t M12 0.76 1.45 0.75 1.44 0.80 1.52 0.73 1.39
2R
0.45( 2R )0.5 (2) M13 0.71 1.50 0.70 1.48 0.75 1.57 0.68 1.44
M14 2.08 2.09 2.06 2.07 2.18 2.20 1.99 2.01
The British Standards Institution code (BSI) (2009) [36] in the
fourth edition of the Specification for Unfired Fusion Welded Pressure
Vessels is also considered in this study. The code predicts the elastic For example, Table 7 shows a comparison between the initial and
instability pressure (Pmc) for shell structures. In this theory, ε is a overall buckling values obtained from theory with those from the ex-
parameter that is a function of 2Rm/t and L/2Rm. Rm is the radius of periments. For perfect model M7, the initial buckling from the theories
the cylindrical shell. is lower than the experimental one by approximately 29–34%, as seen
in Fig. 9. The differences between the theoretical and test values are
Et
British Standards Institution code (BSI) theory:Pmc =
(3) related to the material and geometrical properties, material behaviour,
R
considerations and assumptions in the development of the equations,
Finally, the design formula provided by the European test conditions, etc. The CFRP effect was visualised best for the M14
Recommendations ECCS-CECM-EKS for the Buckling of Steel Shells model. The initial buckling load increment was approximately
(ECCS, 1988) [37] is considered in this research and is used to predict 100–108% for the CFRP-covered perfect model. Thus, hc/R = 2.5 and
the buckling pressure of simply supported elastic shell structures. In this tc/R = 0.002 were the effective parameters for the initial buckling load
formula, βmin is a parameter that is a function of t/R and hc/R, where R capacity.
is the radius of the cylindrical shell: While the dent numbers were 2 and 4 for models M1 and M3, the
t increments were 16 and 13%, respectively. However, the initial buck-
European Recommendations code (ECCS) theory:PECCS = E min ling loads were similar to the theoretical ones for M2 and M4. The dent
R (4)
width and depth were 2tc and 4tc for models M2 and M4, respectively,
The theoretical predictions of Jawad, Ross, the British Standards while they were half that for M1 and M3. Parallel models M9 and M11
Institution Code, and the European Recommendations Code for the increased the initial buckling by approximately 12–16%, which may be
initial buckling pressure of a perfect model are summarised in Table 7. the observation of the CFRP within the theories. Furthermore, the in-
The number of circumferential buckling waves is calculated through cremental effect of CFRP is observed at approximately 21–22% for
approximations by Teng et al. (2001) [38,39]: models M8 and M10, while the dent width and depth were tc and 2tc.
These observations of the initial buckling loads are not only related to
R R
n= 2.74 the theoretical ranges but also to the dent areas. In particular, the initial
hc t (5)
buckling load was as low as 50% for models M5 and M6, while it was
The numbers for the circumferential buckling waves theory are approximately 30% for M12 and M13.
listed in Table 6. For the models without CFRP, the initial buckling load obtained
PJ: Initial buckling load from Jawad theory; PVK: Initial buckling from the experiments is approximately 0.52–1.40 times the buckling
load from Venstel and Krauthaer theory; PR: Initial buckling load from load obtained from the Jawad, Ross, BSI, and ECCS theories.
Ross theory; PBSI: Initial buckling load from British Standard [31]; Furthermore, the overall load is approximately 0.68–2.01 times that of
PECCS: Initial buckling load from European Standard [33]. the mentioned theories. In the case of the with-CFRP group, the initial
Table 6 shows the buckling loads obtained from the theoretical buckling obtained from the experiments is approximately 0.70–2.18
formulas, and Table 7 and Fig. 9 show a comparison of those obtained times the buckling load, while the overall buckling load is approxi-
from experiments and theory. However, the theoretical buckling loads mately 1.32–2.20 times that obtained from the relevant theories
are for perfect models; the models with dents and/or CFRP are pre- (Table 7).
sented for all, throughout this work. Thus, the theoretical formulations In general, the buckling loads obtained from the theories are far
can be evaluated for the nonperfect models. The present work examines smaller than the overall buckling loads and even those at the collapse of
the models for hc/R = 2.5 and t/R = 0.002, which are not in the range the experiments. To obtain the initial and overall buckling loads ac-
of the Jawad theory [9,33–39]. The latter steps of this work can be cording to theoretical formulas, the presented coefficients in Table 7
extended with theoretical evaluations for the dented models. Following have to be multiplied by the theoretical ones in order to obtain the
a comparison of the tests with the theoretical predictions for the initial initial, overall, and collapse buckling. However, the theoretical for-
buckling of the perfect models, the effects of dents and/or CFRP on the mulations designed for perfect models are only estimation steps to
thin-walled cylindrical shells are evaluated. evaluate the buckling loads of nonperfect models.

Table 6
Theoretical predictions for relevant theories.
PJ (kPa) PR (kPa) PBSI (kPa) PECCS (kPa) Approximate Number of Waves [35]

Perfect model 138.2 139.7 131.8 143.9 8

11
F.M. Aydın Korucuk, et al.
Fig. 9. Buckling loads from theories and experimental models.
Table 6 shows the number of waves obtained from the experiments
and the theory of Donnell. It is well seen that the number of waves
obtained from the theory in the without-CFRP group is two more than
the number of waves obtained from the experiments. The value in the
with-CFRP group is 3. The present work examines the models in order
to predict the buckling loads of thin-walled cylindrical shells, which are
out of the range of relevant theories. Thus, the wave propagation and
the wave numbers may not be observed within this experimental plan.
5. Conclusions
This study discussed the buckling behaviour of thin-walled cylind-
rical shells. Fourteen simply supported specimens with different dent
depths (2t and 4t) and various longitudinal dent numbers (d2, d4, and
d6) were loaded under hydrostatic pressure. The following results were
noted:
After the models were exposed to an initial buckling load, they
entered the stage of post-buckling, and the resistance of the models was
increased. A phenomenon observed in all models was an increase in the
buckling capacity after the initial buckling.
1. The initial buckling load decreased with increased number of dents
from 2 to 6 for all groups of models.
2. For all models, the initial buckling load decreased with an in-
creasing amplitude for the dent numbers with the same dent depth.
Thus, the initial buckling load decreased with increasing amplitude
for the dent numbers with different dent depths.
3. The dent depth is a weakening element for the models.
4. When the dent number reached 4, the collapse range became the
smallest, but increasing the dent depth, width, and dent number
widened the range significantly. Thus, an exploration of dent im-
plementation according to the dent cross-section geometry and
depth-width ratio is a future work.
5. The dent area is the main parameter that increases the range from
the initial to overall buckling. Furthermore, the increment of the
initial to overall buckling range was lower than for the models
without CFRP. Thus, the CFRP was closer to the values of the initial
and overall buckling loads. This indicates less plastic behaviour at
this stage according to the models without CFRP.
6. The failures in all samples were in the form of a V shape; that is,
raising the edges in this form caused the model to fail. However, for
the CFRP-covered perfect model, it can be said that after modelling
the buckling wave in the experimental model after leaving the
CFRP device, the waves returned to a relatively basic state. This
12
Thin-Walled Structures 143 (2019) 106259
means that in the fully enclosed model with CFRP, the resulting
waves can be elastoplastic.
7. The initial buckling load of the models without CFRP increased by
4.45%–35.72% compared to the without-CFRP group. Alternately,
because the initial buckling load is important in this type of
structure, the use of CFRP at the dent sites can have a significant
effect on the initial buckling load.
8. For the experimental waves of all models, the wave number of the
with-CFRP group was lower than that of the without-CFRP group.
Thus, the use of CFRP for cylindrical thin-walled shells can have a
significant effect on the waves and the plastic range of the models.
9. In general, the buckling loads obtained from the theories are far
smaller than the overall buckling loads and even those at the col-
lapse of the experiments. To obtain the initial and overall buckling
loads according to theoretical formulas, the presented coefficients
have to be multiplied by the theoretical ones in order to obtain the
initial, overall, and collapse buckling. However, the theoretical
formulations, designed for perfect models, are only estimation steps
to evaluate the buckling loads of nonperfect models.
10. It is clear that the number of waves obtained from theory in the
without-CFRP group was two more than the number of waves ob-
tained from the experiments. The value in the with-CFRP group was
3. The present work examined the models in order to predict the
buckling loads of thin-walled cylindrical shells, which are out of the
range of relevant theories. However, the theoretical formulations
designed for perfect models are only estimation steps to evaluate
the buckling loads of nonperfect models. The later steps of this
work can be extended with theoretic evaluations of the dented
models.
Abbreviations
ρ Density
σ Tensile Strength
ε Parameter that is a function of 2R/t and hc /2R
Ld = Dent length
bd = Dent width
t Thickness of cylinder
hc Height of cylinder
di Number of dents (d2, d4, d6)
FD Fibre dimensions
R Radius
D Displacement transducer (Di is coded to describe the number
of the displacement transducer mentioned)
F.M. Aydın Korucuk, et al.
S Strain gauge (Si is coded to describe the number of the strain
gauge mentioned)
Di° Angle of the displacement transducer number i, from the start
point of the cylinder (welding)
P1 Initial buckling load
P2 Overall buckling load
P3 Collapse buckling load
Mi Model i (M1, M2, .., M14)
MiPj Relevant buckling load of model i (for example, M5P1)
PJ Critical load of external pressure from Jawad theory
PVK Critical load of external pressure from Venstel and Krauthaer
theory
PR Critical load of external pressure from Ross theory
PBSI Critical load of external pressure from British Standards
Institution code
PECCS Critical load of external pressure from European
Recommendations code
N Number of circumferential buckling waves
Pmc Elastic instability pressure
βmin Parameter that is a function of t /R and hc /R
E= Young's modulus
References
[1] C.D. Miller, Shell Structure, CRC press LIC, 1999.
[2] J.G. Teng, Y.M. Hu, Behaviour of FRP-jacketed circular steel tubes and cylindrical
shells under axial compression, Constr. Build. Mater. 21 (2007) 827–838.
[3] M. Batikha, J.F. Chen, J.M. Rotter, J.G. Teng, Strengthening metallic cylindrical
shells against elephant's foot buckling with FRP, Thin-Walled Struct. 47 (2009)
1078–1091.
[4] M. Maali, H. Showkati, S. Mahdi Fatemi, Investigation of the buckling behavior of
conical shells under weld-induced imperfections, Thin-Walled Struct. 57 (2012)
13–24.
[5] S.M. Fatemi, H. Showkati, M. Maali, Experiments on imperfect cylindrical shells
under uniform external pressure, Thin-Walled Struct. 65 (2013) 14–25.
[6] A. Niloufari, H. Showkati, MaaliM, S. Mahdi Fatemi, Experimental investigation on
the effect of geometric imperfections on the buckling and post-buckling behavior of
steel tanks under hydrostatic pressure, Thin-Walled Struct. 74 (2014) 59–69.
[7] T. Ghanbari Ghazijahani, H. Jiao, D. Holloway, Influence of a cutout on circular
steel hollow sections under cyclic loading, J. Constr. Steel Res. 100 (2014) 12–20.
[8] T. Ghanbari Ghazijahani, H. Jiao, D. Holloway, Structural behavior of shells with
different cutouts under compression: an experimental study, J. Constr. Steel Res.
105 (2015) 129–137.
[9] T. Ghanbari Ghazijahani, H. Sadighi Dizaji, J. Nozohor, T. Zirakian, Experiments on
corrugated thin cylindrical shells under uniform external pressure, Ocean Eng. 106
(2015) 68–76.
[10] T. Ghanbari Ghazijahani, H. Jiao, D. Holloway, Experimental study on damaged
cylindrical shells under compression, Thin-Walled Struct. 80 (2014) 13–21.
[11] T. Ghanbari Ghazijahani, H. Jiao, D. Holloway, Experiments on dented cylindrical
shells under peripheral pressure, Thin-Walled Struct. 84 (2014) 50–58.
[12] T. Ghanbari Ghazijahani, H. Jiao, D. Holloway, Fatigue experiments on circular
hollow sections with CFRP reinforced cutouts, J. Constr. Steel Res. 106 (2015)
322–328.
[13] T. Ghanbari Ghazijahani, H. Jiao, D. Holloway, Longitudinally stiffened corrugated
cylindrical Shells under uniform external pressure, J. Constr. Steel Res. 110 (2015)
191–199.
[14] T. Ghanbari Ghazijahani, H. Jiao, D. Holloway, Plastic buckling of dented steel
13
Thin-Walled Structures 143 (2019) 106259
circular tubes under axial compression: an experimental study, Thin-Walled Struct.
92 (2015) 48–54.
[15] T. Ghanbari Ghazijahani, H. Jiao, D. Holloway, Timber filled CFRP jacketed circular
steel tubes under axial compression, Constr. Build. Mater. 94 (2015) 791–799.
[16] M. Vakili, H. Showkati, Experimental and numerical investigation of elephant foot
buckling and retrofitting of cylindrical shells by FRP, J. Compos. Constr. August 20
(4) (2016).
[17] A. Orifici, C. Bisagni, Perturbation-based imperfection analysis for composite cy-
lindrical shells buckling in compression, Compos. Struct. 106 (2013) 520–528.
[18] C. Bisagni, Composite cylindrical shells under static and dynamic axial loading: an
experimental campaign, Prog. Aero. Sci. 78 (2015) 107–115.
[19] S.C. Whitea, P.M. Weavera, K.C. Wub, Post-buckling analyses of variable-stiffness
composite cylinders in axial compression, Compos. Struct. 123 (2015) 190–203.
[20] C.D. Babcock, Experiments in shell buckling, in: Y.C. Fung, E.E. Sechler (Eds.), Thin-
shell Structures: Theory, Experiment and Design, Prentice-Hall, Englewood Cliffs,
NJ, 1974, pp. 345–369.
[21] J. Singer, Buckling experiments on shells—a review of recent developments, Solid
Mech. Arch. 7 (1982) 213–313.
[22] M. Maali, A.C. Aydin, H. Showkati, M. Sagiroglu, M. Kilic, The effect of longitudinal
imperfections on thin-walled conical shells, Journal Of Building Engineering 20
(Nov.2018) 424–441, https://doi.org/10.1016/j.jobe.2018.08.005.
[23] M. Maali, M. Sagiroglu, M.S. Solak, Experimental behaviour of screwed beam-to-
column connections in cold-formed steel frames”, Arabian Journal of Geosciences
11 (2018) 205.
[24] M. Sagiroglu, M. Maali, M. Kılıç, A.C. Aydın, A novel approach for bolted T-stub
connections”, International Journal of Steel Structures (2018), https://doi.org/10.
1007/s13296-018-0034-x.
[25] M. Maali, M. Kılıç, M. Sagiroglu, A.C. Aydın, Experimental model for predicting the
semi-rigid connections' behaviour with angles and stiffeners, Adv. Struct. Eng. 20
(6) (2017) 884–895.
[26] M. Maali, M. Kılıç, A.C. Aydın, Experimental model of the behaviour of bolted
angles connections with stiffeners, Int. Journal of Steel Structures 16 (3) (2016)
1–15.
[27] M. Maali, A.C. Aydin, M. Sagiroglu, Investigation of innovative steel runway beam
in industrial building, Microsilica SADHANA Academy Proceedings in Engineering
Sciences 40 (October 2015) 2239–2251.
[28] A.C. Aydın, M. Maali, M. Kılıç, M. Sagiroglu, Experimental investigation of sinus
beams with end-plate connections, Thin-Walled Struct. 97 (December) (2015)
35–43.
[29] A.C. Aydın, M. Kılıç, M. Maali, M. Sagiroglu, Experimental assessment of the semi-
rigid connections behavior with angles and stiffeners, J. Constr. Steel Res. 114
(November 2015) 338–348.
[30] M. Sagiroglu, Experimental evaluation of the post-fire behavior of steel T-compo-
nent in the beam-to-column connection, Fire Saf. J. 96 (2018) 153–164.
[31] Z.E. Mazurkiewicz, R.T. Nagorski, Shells of Revolutions, Elsevier, 1991.
[32] A.E.H. Love, A Treatise on the Mathematical Theory of Elasticity, fourth ed., Dover
Publisher, New York, 1959.
[33] M.H. Jawad, Theory and Design of Plate and Shell Structures, Chapman & Hall,
1994.
[34] E. Venstel, T. Krauthaer, Thin Plates and Shells: Theory: Analysis, and Applications,
Marcel Dekker, Inc., New York, 2001.
[35] C.T.F. Ross, A proposed design chart to predict the inelastic buckling pressures for
conical shells under uniform external pressure, Mar. Technol. 44 (2) (2007) 77–81.
[36] BSI, Specification for Unfired Fusion Welded Pressure Vessels, fourth ed., British
Standards Institution, UK PD5500, 2009.
[37] ECCS, Buckling of Steel Shells – European Recommendations, ECCSCECM-EKS,
European Convention for Constructional Steelwork, Belgium, 1988.
[38] J.G. Teng, Y. Zhao, L. Lam, Techniques for buckling experiments on steel silo
transition junction, Thin-Walled Struct. 39 (2001) 685–707.
[39] M. Maali, M. Kılıç, Z. Yaman, E. Ağcakoca, A.C. Aydın, Buckling and post-buckling
behavior of various dented cylindrical shells using CFRP strips subjected to uniform
external pressure: comparison of theoretical and experimental data, Thin-Walled
Struct. 137 (2019) 29–39.