materials
Article
Impact of Geometrical Imperfections on Estimation of Buckling
and Limit Loads in a Silo Segment Using the Vibration
Correlation Technique
Łukasz Żmuda-Trzebiatowski






Citation: Żmuda-Trzebiatowski, Ł.;
Iwicki, P. Impact of Geometrical
Imperfections on Estimation of
Buckling and Limit Loads in a Silo
Segment Using the Vibration
Correlation Technique. Materials 2021,
14, 567. https://doi.org/10.3390/
ma14030567
Academic Editor: Nicholas Fantuzzi
Received: 12 November 2020
Accepted: 22 January 2021
Published: 26 January 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Materials 2021, 14, 567. https://doi.org/10.3390/ma14030567
and Piotr Iwicki *
Faculty of Civil and Environmental Engineering, Gdansk University of Technology, Narutowicza Str. 11/12,
80-233 Gdańsk, Poland; luktrzeb@pg.edu.pl
* Correspondence: piwicki@pg.edu.pl
Abstract: The paper examines effectiveness of the vibration correlation technique which allows
determining the buckling or limit loads by means of measured natural frequencies of structures. A
steel silo segment with a corrugated wall, stiffened with cold-formed channel section columns was
analysed. The investigations included numerical analyses of: linear buckling, dynamic eigenvalue
and geometrically static non-linear problems. Both perfect and imperfect geometries were considered.
Initial geometrical imperfections included first and second buckling and vibration mode shapes with
three amplitudes. The vibration correlation technique proved to be useful in estimating limit or
buckling loads. It was very efficient in the case of small and medium imperfection magnitudes. The
significant deviations between the predicted and calculated buckling and limit loads occurred when
large imperfections were considered.
Keywords: buckling; vibration correlation technique; thin-walled structures; steel structures; shell
structures; non-destructive testing; finite element method
1. Introduction
Buckling is one of the most common reasons for steel structure failures. Thin-walled
elements subjected to compression are very sensitive to stability loss, therefore it is es-
sential to find a method to monitor the conditions of existing structures. For a long
time, a common buckling test showed a major disadvantage—destructive nature. The
tested loaded structures often buckled with plastic deformations, due to their imperfection
sensitivity. Geometrical imperfections are hard to assess without professional scanning
equipment. The numerical results of limit and buckling loads at a design stage can differ
significantly from their real values. The destructive methods cannot be applied to examine
the existing structures.
The first non-destructive methods employed for steel columns were the Southwell
method [1] and the vibration correlation technique (VCT) [2]. The latter combines vibrations
and stability of investigated structures since they show similarities between buckling and
vibration behaviour (especially columns).
The VCT can be used either to determine actual boundary conditions or to predict the
buckling loads [3].
In the case of axially loaded columns the relation between squared natural frequencies
and applied load is linear [4]. This equation, in terms of relative variables reads:
(ωn /ωn0 )2 = 1 − P/Pn , (1)
where: ωn is the n-th natural frequency of the loaded structure, ωn0 is the n-th natural
frequency of the unloaded structure, P is applied load and Pn is the buckling load cor-
responding to n-th vibration mode. When the n-th natural frequency equals zero, the
https://www.mdpi.com/journal/materials
Materials 2021, 14, 567 2 of 15

compressive load is a corresponding n-th buckling load. Jubb et al. [5] stated that vibra-
tion and buckling modes need to be identical in shape in order for there to be a linear
relationship between the squared natural structural frequency and the buckling load.
While measuring natural frequencies of an existing structure at two load stages it
is possible to draw a line pointing the predicted buckling load. The only condition is
the identity of buckling and vibration modes, thus preliminary numerical calculations to
determine the first buckling mode and a sufficient number of applied sensors to specify
the vibration modes are indispensable.
In the course of the VCT it is also possible to estimate real boundary conditions. It
can be useful in validating numerical models of existing structures. Measuring natural
frequencies of an erected structure and comparing them to the numerical values is the way
to check the stiffness of supports in a numerical model. However, when the geometric
imperfections are not measured, assessment of real boundary conditions by the VCT may
be inaccurate for complex and imperfection-prone structures.
The relationship between the squared natural frequencies and the applied load is
exactly linear only in the case of columns with simple supports [2], where the buckling
and vibration modes are identical. Small deviations from linearity may be observed in
compressed columns and plates or frame systems with different boundary conditions [6].
Nevertheless, the VCT can be still applied successfully. The problems with the VCT
application occur when initial geometrical imperfections are taken into account, here
considerable deviations from linearity are apparent [7].
Massonnet [6] tested uniform beams, plates and cylindrical shells with various sup-
ports and confirmed more or less linear relationship between squared natural frequencies
and applied load of investigated elements.
The VCT has been tested on different structures such as thin rectangular plates [8,9],
shells [10] or steel cylinders [11].
Franzoni et al. [12] presented various modifications to the standard relationship
between vibrations and buckling proposed by researchers throughout the years.
Recent research has focused mainly on shells. Based on a series of experiments,
Arbelo et al. [13] proposed a new modified VCT approach, plotting 1 − (fn /fn0 )2 versus
(1 − P/Pn )2 using best-fit second order curve to find a squared drop of the load-carrying
capacity ξ2 due to the initial imperfections. Next, estimation of the predicted buckling load
(Pimp ) led to the following formula:

Pimp = Pcr (1 − ξ). (2)

Skukis et al. [14] verified this method experimentally with a good result. Recent papers
confirmed effectiveness of the modified VCT by conducting experimental and numerical
analyses on cylindrical shells stiffened with lozenge grid cores [15], pressurised orthotropic
shells with rectangular stiffeners [16], sandwich plates with iso-grid cores [17], composite
unstiffened shells [18] and variable angle tow composite cylindrical shells [19].
Franzoni et al. [12] made an attempt to evaluate the relationship between the applied
load and natural frequencies of a simple supported isotropic unstiffened cylindrical shell an-
alytically. The analytical approach was successfully compared with numerical calculations.
The paper deals with problems of predicting the buckling loads of stiffened shell. The
silo segment strengthened with cold-formed columns is investigated numerically, the VCT
is applied to determine the buckling loads on the basis of calculated natural frequencies.
Various geometrical imperfection shapes and amplitudes are introduced into the model.
The VCT is applied to each numerical model. Distinct structural performance in relation
to the literature-based pattern is observed here. The conducted calculations enhance
the previous analysis of the same silo [20] by introducing geometrical imperfections and
changing boundary conditions.
The research deals with part of a real structure, instead of a small experimental model.
Usually investigation of the issue pertains to actual imperfections which were measured
in experimental models. However, the specific imperfections to emerge in subsequent
 relation to the literature-based pattern is observed here. The conducted calculations en-
hance the previous analysis of the same silo [20] by introducing geometrical imperfections
and changing boundary conditions.
Materials 2021, 14, 567 The research deals with part of a real structure, instead of a small experimental 3 of 15
model. Usually investigation of the issue pertains to actual imperfections which were
measured in experimental models. However, the specific imperfections to emerge in sub-
sequent project stages are unknown. Generally, they are assumed as the first buckling
project stages are unknown. Generally, they are assumed as the first buckling mode or
mode or the first vibration mode, therefore the paper presents the impact of such imper-
the first vibration mode, therefore the paper presents the impact of such imperfections on
fections on relation between squared natural frequencies and compressive force. In the
relation between squared natural frequencies and compressive force. In the introduced
introduced investigation it can be noticed that a zero frequency in the buckling case is not
investigation it can be noticed that a zero frequency in the buckling case is not necessarily
necessarily
the first one,the first be
it may one, it may be
a higher onea as
higher
well.one
Anas well. An innovative
innovative element of element
researchofconcerns
research
concerns the impact of imperfection on limit loads and their determination
the impact of imperfection on limit loads and their determination using the VCT. using the VCT.
TheVCT
The VCTapproach
approachisisan anexperimental
experimentalprocedure.
procedure.The Themain
mainadvantage
advantageofofthethemethod
method
is an accurate non-destructive prediction of the buckling load of
is an accurate non-destructive prediction of the buckling load of a structure. However,a structure. However,
beforeconducting
before conductingthe theexperiments
experimentstheoretical
theoreticalor ornumerical
numericalanalysis
analysisofofthetheinvestigated
investigated
problem should be performed. The main scope of the research was
problem should be performed. The main scope of the research was to perform numerical to perform numerical
simulations of the imperfect silo segment in order to predict relationship
simulations of the imperfect silo segment in order to predict relationship between applied between applied
loadand
load andsquared
squarednatural
naturalfrequencies.
frequencies.As Asaaresult,
result,the
thebuckling
bucklingororlimit
limitload
loadwas
wasfind
findboth
both
by means of the VCT and non-linear static analysis. It can be added that
by means of the VCT and non-linear static analysis. It can be added that some experiments some experiments
aredifficult
are difficulttotoperform
performdue duetotothe
thescale
scaleofofthe
thetested
testedstructures
structuresbut
butthe
thephenomenon
phenomenonmay may
be investigated using numerical
be investigated using numerical simulations.simulations.

2.2.Materials
Materialsand
andMethods
Methods
The
Theinvestigated
investigatedstructure
structurewas wasaastiffened
stiffenedsilo
silosegment.
segment.Based
Basedononresearch
research[21,22],
[21,22],itit
can be concluded that the silo segment model describes well the behaviour
can be concluded that the silo segment model describes well the behaviour of the entire of the entire silo
in theincase
silo of sparsely
the case distributed
of sparsely columns.
distributed columns.The The
silo silo
waswas8.048.04
m inmdiameter
in diameterandand 17.62 m
17.62
high.
m high.TheThewall included
wall includedcorrugated
corrugated steel sheets
steel sheetswith
withthe 7676
the mmmmpitch,
pitch,the
the1818mm mmdepth
depth
and
andthethe0.75 mm
0.75 mmthickness
thickness(Figure 1a). The
(Figure 1a).silo
Thewas
silostrengthened with 18 columns
was strengthened with 18 sparsely
columns
distributed along the along
sparsely distributed circumference. The columns
the circumference. The were
columnsmade of cold-formed
were channel
made of cold-formed
sections
channel of 4 mmofthickness
sections and dimensions
4 mm thickness shownshown
and dimensions in Figure 1b. The
in Figure 1b. entire structure
The entire struc-
material was steel S355 of the following parameters: specific weight 7850 kg/m 3 , elasticity
ture material was steel S355 of the following parameters: specific weight 7850 kg/m3, elas-
modulus 210 GPa
ticity modulus 210and
GPaPoisson’s ratio 0.3.
and Poisson’s ratio 0.3.

16

60
18

76 [mm]
86 [mm]

(a) (b)
Figure1.1.Dimensions
Figure Dimensionsofofsilo
silocross-sections:
cross-sections:(a)
(a)corrugated
corrugatedsheet
sheet(silo
(silowall);
wall);(b)
(b)cold-formed
cold-formedchannel
chan-
nel section (columns).
section (columns).

Theinvestigated
The investigatedstructure
structurewas
wasmodelled
modelledby byaacommercial
commercialABAQUS
ABAQUSpackage,
package,version
version
6.14[23].
6.14 [23].The
Theentire
entiresilo
silowas
wassubstituted
substitutedbybythe
thesegment
segmentincluding
includingthe
thewall
wallsection
sectionofofthe
the
40degrees
40 degreesangle
angle and
and three
three columns (Figure 2a).2a). One
Onecolumn
columnwas wasplaced
placedininthe
themiddle
middleof
the
of thesilo
silopart
partand
andtwo
twocolumn
columnhalves
halves at
at the
the edges. The boundary
boundary conditions
conditions atatthe
theside
side
edgeswere
edges wereintended
intendedto tosimulate
simulatebuckling
bucklingofofthe
theentire
entiresilo
silo[21].
[21].AAsimplified
simplifiedmodel
modelwas was
introducedin
introduced inorder
ordertotolimit
limitthe
thenumber
numberofoffinite
finiteelements,
elements,consequently
consequentlytotoshorten
shortenthethe
computational
computationaltime. time.
Materials 2021, 14, 567 4 of 15
Materials 2021, 14, x FOR PEER REVIEW 4 of 15
Materials 2021, 14, x FOR PEER REVIEW 4 of 15

40°
40°

(a) (b)
(a) (b)
Figure 2. (a) Scheme of boundary conditions in simplified silo segment; (b) Visualisation of nu-
Figure 2. (a) Scheme of boundary conditions in simplified silo segment; (b) Visualisation of
merical model.
Figure 2. (a) Scheme of boundary conditions in simplified silo segment; (b) Visualisation of nu-
numerical model.
merical model.
The lower edges of the silo segment were fixed, the upper edges restrained transla-
The lower edges of the silo segment were fixed, the upper edges restrained translations
tions The lower edges
in radial of the silo segment
and circumferential were fixed,
directions, the upper
the lateral edges
edges restrained
of side columns transla-
con-
in radial and circumferential directions, the lateral edges of side columns constrained
tions in translations
strained radial and circumferential
in a circumferential directions,
directiontheand
lateral edges
rotations of side
around columns
radial con-
and verti-
translations in a circumferential direction and rotations around radial and vertical axes.
strained
cal axes. translations
The were
columns in a circumferential
were toattached direction and rotations around radial and verti-
The columns attached the walltoby themeans
wall by
of means
point tieof connectors
point tie connectors
simulating simulat-
rivets
cal
ing axes.
rivets
or screws. The
or columns
screws. were attached to the wall by means of point tie connectors simulat-
ing rivets
Thesilo
The or wall,
silo screws.
wall,as aswell
wellasasthe
thecold-formed
cold-formedcolumns,
columns,wereweremodelled
modelledwithwith4-node
4-nodeshell
shell
The
elements silo
with wall, as
reduced well as the
integration. cold-formed
A simplifiedcolumns,
numericalwere modelled
model with
consisted
elements with reduced integration. A simplified numerical model consisted of 200,340 finite 4-node
of 200,340shell
fi-
elements
nite with
elements. Areduced
single integration.
corrugation A simplified
wave was numerical
approximated
elements. A single corrugation wave was approximated by 8 elements. model
by 8 consisted
elements. of 200,340 fi-
nite The
elements.
The load
load wasA single
was imposed corrugation
imposed in inthe wave
theform was approximated
formofofsupport
supportdisplacement by
displacement 8 elements.
applied
applied vertically to the
vertically to
nodes The load
located was
on imposed
the upper in the
edges form
of the of support
columns. displacement
The silo wall
the nodes located on the upper edges of the columns. The silo wall was not loaded in applied
was not vertically
loaded in to the
order
nodes
to
order tolocated
avoid aonnumber
a number
avoid theofupper
local edges
buckling
of local of the
modes
buckling columns. ofThe
of the
modes the silo wall sheets.
corrugation
corrugationwassheets.
notAloaded
pattern inoforder
A pattern dis-
of
to avoid
placement a number
load of
subjected local
to buckling
one of the modes
columns of the
is corrugation
shown
displacement load subjected to one of the columns is shown in Figure 3. in sheets.
Figure 3. A pattern of dis-
placement load subjected to one of the columns is shown in Figure 3.

Figure 3. A pattern of displacement load subjected to the upper edges of silo columns.
Figure3.3.AApattern
Figure patternof
ofdisplacement
displacementload
loadsubjected
subjectedtotothe
theupper
upperedges
edgesofofsilo
silocolumns.
columns.
Both perfect and imperfect silo geometry cases were considered. Two primary free
Bothperfect
vibration
Both perfect
(F1, F2)and imperfect(B1,
andimperfect
buckling siloB2)
silo geometry caseswere
mode shapes
geometry cases were considered.
wereconsidered.
implemented Two
Two primary
byprimary free
the software
free
vibration
inner
vibration (F1,F2)
procedure
(F1, F2)to
and
and the buckling
numerical
buckling (B1,
(B1, B2)mode
model
B2) mode
in the shapes were
form were
shapes of implemented
initial geometrical
implemented byimperfections
by thesoftware
the software
innerprocedure
(Figure
inner procedure tothe
4). Relevant
to the
linearnumerical
buckling
numerical model ininthe
analysis
model theform
was form ofofinitial
performed initial geometrical
togeometrical
copy the nodal imperfections
coordinates
imperfections
(Figure
of
(Figure 4).Relevant
buckling
4). Relevant
modes and linear
linear bucklinganalysis
introducing
buckling analysis
such a newwas
was performed
geometry
performed intotocopy
copy
order the
tothe nodalfurther
conduct
nodal coordinates
anal-
coordinates
of buckling
of buckling
yses. modes amplitudes
Imperfection and
andintroducing
introducing such
in each sucha new
case awere
new geometry
set as 2 mm
geometry in order 5to
(a),order
in mm conduct
to(b) andfurther
conduct anal-
20 further
mm (c).
yses.total
Imperfection
analyses.
The ofamplitudes
Imperfection
number in each
amplitudes
the investigated in case
eachwere
structurescase set
was as 2set
were
equal mm (a),
toas13.2 mm5 mm(a),(b)5andmm20(b)mm (c).
and
20
Themm (c).number
total The total
ofnumber of the investigated
the investigated structures structures
was equal was to 13.equal to 13.
Materials 2021, 14, 567 5 of 15
Materials 2021, 14, x FOR PEER REVIEW 5 of 15

Figure 4. Initial geometrical imperfections: (a) first vibration mode; (b) second vibration mode; (c)
Figure 4. Initial geometrical imperfections: (a) first vibration mode; (b) second vibration mode;
first buckling mode; (d) second buckling mode.
(c) first buckling mode; (d) second buckling mode.

Numerical
Numericalcalculations
calculationscovered
coveredthe thefollowing
followinganalyses
analysesfor foreach
eachsilo
silogeometry
geometrymodel:model:
•• Linear buckling, resulting in 5 primary buckling modes and
Linear buckling, resulting in 5 primary buckling modes and eigenvalues. The buckling eigenvalues. The buck-
ling modes were bound to correspond to the vibration modes,
modes were bound to correspond to the vibration modes, therefore it was essential therefore it was essen-
tial to compare
to compare themthem in order
in order to draw
to draw rightright conclusions.
conclusions. The buckling
The buckling eigenvalues
eigenvalues were
were
used used to assess
to assess the primary
the primary buckling
buckling loads.loads.
•• Geometrically
Geometricallynon-linearnon-linearanalysis,
analysis,leading
leadingto to aa full
full equilibrium
equilibrium path:path: the
the segment
segment
displacement load vs. the average vertical displacement
displacement load vs. the average vertical displacement of the columns top edges. of the columns top edges. A
static, general method based on Full Newton solution
A static, general method based on Full Newton solution technique was applied [23].technique was applied [23].
The
The initial,
initial, minimum
minimum and and maximum
maximum increment
increment sizessizes were
were equal
equal to
to 0.001,
0.001, 1010−−15
15 and
and
0.01,
0.01,respectively.
respectively. In Inorder
order to to avoid
avoid problems
problems with with solution
solution convergence
convergence automatic
automatic
stabilization
stabilizationwas wasintroduced
introducedin inthe
the models
models by by numerical
numerical damping
damping factor
factor (equal
(equal to to
10−9−).9 ).The
10 Theanalytical
analyticalaimaimwas wastotodetermine
determinethe thelimit
limitload,
load,i.e.,
i.e.,the
theload
loadcorresponding
corresponding
to
tothe
theload-displacement
load-displacementpath pathmaximum
maximum(P (Plim).).
lim
•• Combined
Combined geometrically non-linear and dynamic eigenvalue
geometrically non-linear and dynamic eigenvalue analysis.
analysis. Non-linear
Non-linear
analysis
analysiswas wasperformed
performedrepeatedly—each
repeatedly—eachtime timethe thedisplacement
displacement load
loadincreased
increased to to
a
certain
a certain value in order
value in orderto pre-stress
to pre-stressthe structure. After each
the structure. Afterprocess a dynamic
each process eigen-
a dynamic
value
eigenvalue problem was solved,
problem resulting
was solved, in 5 vibration
resulting modes and
in 5 vibration modes corresponding natural
and corresponding
natural frequencies.
frequencies. The vibration
The vibration modes weremodesnecessary
were necessary
to tracetotheir
traceshape
their shape variation
variation dur-
during
ing the loading
the loading course course
[20]. It[20].
was It wastoeasy
easy confuseto confuse the frequencies
the natural natural frequencies
of various of
various orders as the pre-load increased. The primary
orders as the pre-load increased. The primary natural frequencies allowed applying natural frequencies allowed
applying
the vibration the correlation
vibration correlation
techniquetechnique
and drawing and drawing the relationships
the relationships between between
the ap-
the applied
plied load and load
theand the squared
squared naturalnatural frequencies.
frequencies.

3.3.Results
Resultsand
andDiscussion
Discussion
Theaim
The aimof
ofthe
theperformed
performedanalyses
analysesisisto
toverify
verifythe
thevibration
vibrationcorrelation
correlationtechnique
techniqueinin
its standard form, with itis ability to predict buckling loads of the silo segment stiffened
its standard form, with itis ability to predict buckling loads of the silo segment stiffened
withthin-walled
with thin-walledcolumns.
columns.The Thecharacteristic
characteristiccurves
curvesofofthe
theperfect
perfectand
andimperfect
imperfectstructure
structure
aretraced
are tracedand
andanalysed.
analysed.
Four primary vibration and buckling modes of the silo segment without initial im-
Four primary vibration and buckling modes of the silo segment without initial im-
perfections are shown in Figure 5. It may be observed that the buckling modes shapes
perfections are shown in Figure 5. It may be observed that the buckling modes shapes
differ from the vibration modes completely. However, during structural loading process
differ from the vibration modes completely. However, during structural loading process
selected vibration modes change their shapes. While the first and fourth vibration modes
selected vibration modes change their shapes. While the first and fourth vibration modes
remain unaffected by the applied load, the second and third modes continuously alter their
remain unaffected by the applied load, the second and third modes continuously alter
forms. In our case, at first, the half-wave along the silo height of the second vibration mode
their forms. In our case, at first, the half-wave along the silo height of the second vibration
narrowed making it possible for two new smaller waves to occur, forming three semi-waves
mode narrowed making it possible for two new smaller waves to occur, forming three
along the silo height. This shape resembled the second buckling mode. Similarly, the third
 Materials 2021, 14, x FOR PEER REVIEW 6 of 15
Materials 2021, 14, x FOR PEER REVIEW 6 of 15
Materials 2021, 14, 567 6 of 15

semi-waves along the silo height. This shape resembled the second buckling mode. Simi-
semi-waves
larly, the thirdalong the silo
vibration height.
mode This shape
consisting of tworesembled
semi-wavesthe second
along the buckling mode.
silo height Simi-
changed
vibration
larly, the mode
third consisting
vibration of
mode two semi-waves
consisting of two along the
semi-waves silo height
along thechanged
its form in the same way, narrowing and allowing two new semi-waves to occur, creating silo its
height form
changedin
the
its same way,
formofinfour
a total narrowing
the same and allowing
way, narrowing
semi-waves. This shape, two
andin new
allowing semi-waves to occur,
two new semi-waves
turn, resembled creating a
to occur,
the first buckling total of
creating
mode. The
four semi-waves.
achanges
total ofinfour Thismodes
shape,This
semi-waves.
vibration in turn, resembled
shape,
during the in turn, the
loading first are
resembled
course buckling
the mode.
first
presented in The
buckling changes
mode.
Figure in
The
6. When
vibration
changes
the appliedmodes during
in vibration the loading
modesthe
load exceeded during course are
the loading
maximum presented in
courseit are
limit point, Figure
waspresented 6.
impossible When
in to the
Figure
followapplied
6. the
Whenvi-
load
the exceeded
applied
bration modes the
load maximum
in exceeded
some cases limit
the thepoint,
of maximum it limit
was impossible
investigated point, to
it was
structures follow thegeometrical
impossible
with initial vibration
to followmodes
the vi-
imper-
inbration
some cases
fections.modes ofin
the investigated
some cases of thestructures with initial
investigated geometrical
structures imperfections.
with initial geometrical imper-
fections.

Pcr,1 = 1360 kN f1 = 8.48 Hz

Pcr,1 = 1360 kN f1 = 8.48 Hz

Pcr,2 = 1360 kN f2 = 15.12 Hz

Pcr,2 = 1360 kN f2 = 15.12 Hz

Pcr,3 = 1528 kN f3 = 17.26 Hz

Pcr,3 = 1528 kN f3 = 17.26 Hz

Pcr,4 = 1529 kN f4 = 17.64 Hz

Pcr,4 = 1529 kN f4 = 17.64 Hz
(a) (b)
(a) (b)
Figure 5. Primary one-to-four modes: (a) buckling; (b) vibration.
Figure 5. Primary one-to-four modes: (a) buckling; (b) vibration.
Figure 5. Primary one-to-four modes: (a) buckling; (b) vibration.

2nd mode
2nd mode

3rd mode
3rd mode

Figure 6. Changes of second and third vibration modes during loading course.
Figure 6.6.Changes
Figure Changesofofsecond
secondand
andthird
thirdvibration
vibrationmodes
modesduring
duringloading
loadingcourse.
course.
Materials 2021, 14, x FOR PEER REVIEW 7 of 15

Materials 2021, 14, 567 7 of 15

The full equilibrium paths between the average vertical displacements of the upper
edges of the columns and the applied compressive loads are presented in Figure 7. The
diagrams include the silo parts with various imperfection shapes and amplitudes. Deno-
tations of Thethe full equilibrium paths
abbreviations used between
in the the average
figure vertical displacements
description are following: of the upper perfect
B0—the
edges of the columns and the applied compressive loads are presented in Figure 7. The dia-
geometry, B1—the initial imperfections as the first buckling mode, B2—the initial imper-
grams include the silo parts with various imperfection shapes and amplitudes. Denotations
fections
of theas the secondused
abbreviations buckling mode,
in the figure F1—theare
description initial imperfections
following: as the
B0—the perfect first vibration
geometry,
mode, F2—the
B1—the initialinitial imperfections
imperfections as the firstasbuckling
the second
mode,vibration
B2—the initial mode. The terms
imperfections ‘2 mm’, ‘5
as the
mm’secondand ’20 mm’mode,
buckling denote imperfection
F1—the amplitudes.
initial imperfections as theThefirstimperfect geometry
vibration mode, F2—thewith the
initialassumed
greatest imperfections as the second
imperfection vibration mode.
magnitude reducesThethe
terms ‘2 mm’,
limit load ‘5over
mm’three
and ‘20 mm’ Unsur-
times.
denote imperfection amplitudes. The imperfect geometry with
prisingly, the buckling modes set as initial imperfections were more disadvantageousthe greatest assumed im-
perfection magnitude reduces the limit load over three times. Unsurprisingly, the buckling
thanmodes
the vibration modes (about twice). Moreover, the first buckling mode influenced the
set as initial imperfections were more disadvantageous than the vibration modes
limit(about
load slightly more
twice). Moreover, than
thethe
firstsecond
buckling one.
modeNote that thethe
influenced first and
limit second
load buckling
slightly more loads
are close,
than thehowever their
second one. corresponding
Note that the first andbuckling modes diverge.
second buckling On the
loads are close, contrary,
however their the first
vibration mode showed
corresponding buckling amodes
muchdiverge.
smaller Onimpact on the
the contrary, thestructural load-carrying
first vibration mode showedcapacity
(Plima) than
much thesmaller impact
second on the structural load-carrying capacity (Plim ) than the second one.
one.

2000 2000
a a
b 1500 b
1500
c c
d d
P [kN]
P [kN]

1000 e 1000 e
f f
g g
500 500

0 0
0 20 40 60 80 0 20 40 60 80
d [mm] d [mm]

(A) (B)
Figure 7.Figure
Evolution of totalofvertical
7. Evolution reactions
total vertical and and
reactions average vertical
average displacement
vertical displacement of of columns topedges
columns top edgesobtained
obtained from geo-
from
geometrically
metrically non-linearnon-linear
analysesanalyses of the
of the silo silo segment
segment withwith initial
initial imperfectionsin
imperfections inthe
the form
form of:
of:(A)
(A)buckling
bucklingmodes: (a) B0,
modes: (a) B0, (b)
B1_2 mm,(b) (c)
B1_2 mm,mm,
B1_5 (c) B1_5
(d) mm,
B1_20 (d)mm,
B1_20(e)mm,
B2_2(e)mm,
B2_2 (f)
mm, (f) B2_5
B2_5 mm,mm, (g) (g) B2_20
B2_20 mm;mm;(B)
(B)vibration
vibration modes:
modes:(a)(a)B0,
B0,(b)(b)
F1_2
F1_2 mm,
(c) F1_5 mm,
mm,(c)(d)F1_5 mm,mm,
F1_20 (d) F1_20 mm, mm,
(e) F2_2 (e) F2_2
(f)mm,
F2_5(f)mm,
F2_5(g)
mm, (g) F2_20
F2_20 mm.mm.
Figures 8 and 9 show the vibration correlation technique results. In each case four
Figuressquared
primary 8 and 9natural
showfrequencies
the vibration werecorrelation
followed. The technique
diagramsresults. In each
show various case four
rela-
primary
tionships between natural frequencies and the applied load. Depending on geometricalrelation-
squared natural frequencies were followed. The diagrams show various
ships between natural
imperfection shapes andfrequencies
amplitudes, and
thethe
firstapplied load. Depending
natural frequency reaching the onzero
geometrical
value im-
was different. In a geometrically perfect structure the second and third
perfection shapes and amplitudes, the first natural frequency reaching the zero value wasfrequencies reached
zero almost
different. In a at the same timeperfect
geometrically (Figure 8A). Their corresponding
structure the second and vibration
thirdmodes at thesereached
frequencies
points resembled the first and second buckling modes. The predicted buckling load (Ppred )
zero almost at the same time (Figure 8A). Their corresponding vibration modes at these
was equal to 1383 kN, whereas the calculated buckling load was equal to 1360 kN. The rela-
points
tiveresembled
difference wasthe1.7%.
first The
andlimit
second
loadbuckling
was equal modes. Thei.e.,
to 1734 kN, predicted
1.275 times buckling load (Ppred)
greater than
wasthe
equal to 1383
buckling load.kN,No whereas the calculated
initial imperfections buckling
introduced to theload was equal
numerical model to 1360 kN. The
exceeded
relative difference
the outcome. was 1.7%.
Moreover, The and
the first limitfourth
loadmodes
was equal
were to 1734 kN,
straight, while i.e.,
the1.275
secondtimes
and greater
third ones (exhibiting changes during the loading course) non-linear.
than the buckling load. No initial imperfections introduced to the numerical model ex- It should be noted
that the
ceeded predicted
outcome.buckling load is defined
Moreover, as applied
the first load value
and fourth modes either
were when naturalwhile
straight, frequency
the second
equals zero or when it reaches minimum value (if it does not intersect with load axis).
and third ones (exhibiting changes during the loading course) non-linear. It should be
noted that predicted buckling load is defined as applied load value either when natura
frequency equals zero or when it reaches minimum value (if it does not intersect with load
axis).
Materials 2021, 14, 567 8 of 15

Materials 2021, 14, x FOR PEER REVIEW 8 of 15

Figure 8. Relationship between applied load and squared natural frequencies ((a) first, (b) second, (c) third, (d) fourth) of
Figure 8. Relationship between applied load and squared natural frequencies ((a) first, (b) second, (c)
the silo segment with initial geometrical imperfections: (A) none, (B) B1_2 mm, (C) B1_5 mm, (D) B1_20 mm, (E) B2_2 mm,
(F) B2_5third, (d)B2_20
mm, (G) fourth)
mm.of the silo segment with initial geometrical imperfections: (A) none, (B) B1_2 mm,
(C) B1_5 mm, (D) B1_20 mm, (E) B2_2 mm, (F) B2_5 mm, (G) B2_20 mm.
Materials 2021, 14, 567 9 of 15
Materials 2021, 14, x FOR PEER REVIEW 9 of 15

Figure
Figure 9.9. Relationship
Relationship between
between applied
applied load
load and
and squared
squared natural
natural frequencies
frequencies ((a)
((a) first,
first, (b)
(b) second,
second, (c)
(c) third,
third, (d)
(d) fourth)
fourth) of
of
the
the silo segment with initial geometrical imperfections: (A) F1_2 mm, (B) F1_5 mm, (C) F1_20 mm, (D) F2_2 mm, (E)
silo segment with initial geometrical imperfections: (A) F1_2 mm, (B) F1_5 mm, (C) F1_20 mm, (D) F2_2 mm, (E) F2_5
F2_5
mm, (F) F2_20 mm.
mm, (F) F2_20 mm.

The
The limit
limit loads
loads of
of structures
structures with
with various
various initial
initial imperfections
imperfections are are frequently
frequently similar
similar
or
or much smaller than the numerical value of the lowest buckling load. The buckling
much smaller than the numerical value of the lowest buckling load. The buckling loads
loads
predicted
predicted by by means
means of of the
the VCT
VCT represent,
represent, in contrary to
in contrary the perfect
to the perfect structure,
structure, the
the limit
limit
load.
load. The
The VCT
VCT seems
seems to to indicate
indicate the
the minimum
minimum of of the
the lowest
lowest buckling
buckling load
load and
and the limit
the limit
load
load as
as the
the predicted buckling load
predicted buckling load (P
(Ppred
pred),),not
notjust
justthe
the first
first of
of the
the two.
two. In
In the
the case
case of
of the
the
investigated shell the
investigated shell the buckling
bucklingload
loadwas
wassmaller
smallerthanthan the
the limit
limit load
load forfor
thethe perfect
perfect struc-
structure
ture and greater than the limit load for the structure with imperfections (Table 1). This
relation is also illustrated in Figure 10.
Materials
Materials 2021,
2021, 14,
14, x567
FOR PEER REVIEW 10
10 of
of 15
15

Table 1. Summary
and greater than of
thethe GNA,
limit LBA
load forand
theVCT results.with imperfections (Table 1). This relation
structure
isThe
alsoNumerical
illustrated in Figure 10. The Difference
Plim [kN] (GNA) Pcr [kN] (LBA) Ppred [kN] (VCT)
Model [%]
Table 1. Summary of the GNA, LBA and VCT results.
B0 1734 1360 1383 1.7
B1_2 mm 1120
Plim [kN] 1357
Pcr [kN] P1126
pred [kN] The 0.5
Difference
The Numerical Model
B1_5 mm 938(GNA) (LBA)
1340 (VCT)
952 [%]
1.5
B1_20 mmB0 517 1734 11291360 1383
571 1.7
10.4
B1_2 mm 1120 1357 1126 0.5
B2_2B1_5
mm mm 1134 938 13571340 1142
952 0.7
1.5
B2_5 mmmm
B1_20 958 517 13441129 982571 2.5
10.4
B2_2 mm 1134 1357 1142 0.7
B2_20 mm
B2_5 mm
544 958 11561344
594982
9.2
2.5
F1_2 mmmm
B2_20 1379 544 13601156 1372
594 −0.9
9.2
F1_5F1_2
mm mm 1361 1379 13591360 1372
1356 −0.9
−0.2
F1_5 mm 1361 1359 1356 −0.2
F1_20 mm
F1_20 mm 1186 1186 13461346 1184
1184 −0.2
−0.2
F2_2F2_2
mm mm 1336 1336 13591359 1332
1332 −0.3
−0.3
F2_5 mm 1284 1353 1281 −0.2
F2_5 mmmm
F2_20 1284 997 13531245 1281
1000 −0.2
0.3
F2_20 mm 997 1245 1000 0.3

2000 2000
GNA GNA
LBA LBA
1500 VCT 1500 VCT
P [kN]
P [kN]

1000 1000

500 500

0 0
0 2 5 20 0 2 5 20
amplitude [mm] amplitude [mm]
(a) (b)
2000 2000
GNA GNA
LBA LBA
1500 VCT 1500 VCT
P [kN]

P [kN]

1000 1000

500 500

0 0
0 2 5 20 0 2 5 20
amplitude [mm] amplitude [mm]
(c) (d)
Figure
Figure 10.
10. GNA,
GNA, LBA
LBA andand VCT
VCT loads
loads versus
versusimperfection
imperfectionamplitude
amplitudeofofthethesilo
silosegment
segmentwith
withinitial
initial geometrical
geometrical imper-
imperfec-
fections as modes: (a) first buckling (B1), (b) second buckling (B2), (c) first vibration (F1), (d) second vibration (F2).
tions as modes: (a) first buckling (B1), (b) second buckling (B2), (c) first vibration (F1), (d) second vibration (F2).

Equation
Equation (1) (1) states
states that
that natural
natural frequencies
frequencies should
should tend
tend to
to zero
zero when
when the
the applied
applied
compressive load becomes a subsequent buckling load. However, numerical
compressive load becomes a subsequent buckling load. However, numerical computations computa-
tions discard
discard this statement.
this statement. The squared
The squared natural
natural frequencies
frequencies followed
followed one
one of of ways:
two two ways:they
they dropped
dropped to the tozero
the zero
valuevalue or they
or they decreased
decreased to thetominimum
the minimum valuevalue at some
at some level,level,
next
next
increasing. In almost all VCT diagrams one of the primary natural frequencies fell to
increasing. In almost all VCT diagrams one of the primary natural frequencies fell to
zero;
zero; only
only in
in the
the structure
structure with
with the
the second
second buckling
buckling mode
mode asas initial
initial imperfection
imperfection of
of 22 mm
mm
amplitude (Figure 8E)
amplitude (Figure 8E)three
threelowest
lowestfrequencies
frequencies reached
reached thethe minimum
minimum values
values and and
thenthen
rose
Materials 2021, 14, 567 11 of 15

again. In some cases it was difficult to match frequencies with correct vibration modes
when the applied load was nearly equal to the maximum load capacity, therefore the curves
corresponding to squared natural frequencies had stopped before they reached the load
axis level (Figures 8E and 9C,E).
Moreover, it can be observed that the greater the imperfection amplitude the more
curvilinear the first squared frequency relation. The relationship between squares of natural
frequencies and the compressive load was linear when the applied load was low or medium.
While the limit load tends to its limit value the curve shapes become non-linear in a higher
extent. When the curves frequently reached zero values rapidly, the VCT effectiveness
was reduced.
In most cases of the conducted analyses the first squared frequency to reach the zero
value was the second or the third one in a row. The second and third vibration modes
resembled the first and second buckling modes. While the corresponding buckling loads
were very close, the curves representing the second and third squared natural frequencies
often dropped to zero at the same or nearly the same time. In the structure with the second
buckling mode as initial imperfection of 2 mm amplitude the first and second curves
reached the minimum values and no natural frequency was equal to zero (Figure 8E). When
the initial imperfection took the buckling mode shape (first or second) with the greatest
assumed amplitude (20 mm), the first natural frequency curve reached zero (Figure 8D,G).
In both cases the second curve reached the minimum value, next it increased. When one
of the three lowest natural frequencies were equal to zero, the other ones reached their
minimum values and rose again. The exception was the perfect silo segment (Figure 8A),
where the first and fourth curves were unaffected by the loading process. It may be
incorrect in the case of vibration mode shapes assumed as initial geometrical imperfections
because full analyses could not be finished due to the problems with determining the
correct numbering of vibration modes and, in consequence, natural frequencies.
The predicted by the VCT buckling or limit loads are shown in Table 1. Denotations
are as follows: Plim is the maximum load-carrying capacity (limit load) based on non-linear
analysis, Pcr is the lowest buckling load obtained from the linear buckling analysis, Ppred is
the predicted load received in the VCT and the difference is assessed between the Ppred and
the minimum of Plim and Pcr . If the zero value was not achieved by any curve, the predicted
buckling load was assumed equal to the load when the squared natural frequency reached
its minimum value. It can be noted that the difference was less than 2.5% in a majority of
structural analyses. Only ‘B1_20 mm’ and ‘B2_20 mm’ structures, where large imperfection
amplitudes were introduced, showed lower effectiveness of the VCT. It should be also
highlighted that the predicted maximum load in the structures, where the buckling modes
were introduced as initial imperfections, was always greater than the Plim or Pcr , it leads to
underestimation of structural load-carrying capacity.
Figures 11 and 12 illustrate the relationship between 1 − (fn /fn0 )2 and (1 − P/Pn )2
investigated in each structure. A squared drop of the load-carrying capacity ξ2 was deter-
mined according to the modified VCT proposed by Arbelo et al. [13]. Next, Equation (2)
was employed to assess the predicted buckling load. The results are collected in Table 2.
The buckling loads are equal or close to the results of the standard VCT. The maximum
relative differences of both methods are 1.7%.
 Materials 2021, 14, 567 12 of 15
aterials 2021, 14, x FOR PEER REVIEW 12 of 15

Figure 11. Relationship between relative load and natural frequencies ((a) first, (b) second, (c)
Figure 11. Relationship between relative load and natural frequencies ((a) first, (b) second, (c) third, (d) fourth) of the silo
third, (d) fourth) of the silo segment with initial geometrical imperfections: (A) none, (B) B1_2 mm,
segment with initial geometrical
(C) B1_5 imperfections:
mm, (D) B1_20 (A) none,
mm, (E) B2_2 mm, (F)(B)B2_5
B1_2mm,
mm,(G)
(C)B2_20
B1_5 mm,
mm,according
(D) B1_20tomm,the (E) B2_2 mm, (F) B2_5
modi-
mm, (G) B2_20fied
mm,VCTaccording to the modified VCT by Arbelo.
by Arbelo.
 Materials 2021, 14, 567 13 of 15
Materials 2021, 14, x FOR PEER REVIEW 13 of 15

Figure 12.12.
Figure Relationship between
Relationship betweenrelative
relativeload
loadand
andnatural
natural frequencies ((a) first,
frequencies ((a) first,(b)
(b)second,
second,(c)
(c)third,
third,(d)(d) fourth)
fourth) of of
thethe
silosilo
segment with initial geometrical imperfections: (A) F1_2 mm, (B) F1_5 mm, (C) F1_20 mm, (D) F2_2 mm, (E) F2_5 mm, (F)
segment with initial geometrical imperfections: (A) F1_2 mm, (B) F1_5 mm, (C) F1_20 mm, (D) F2_2 mm, (E) F2_5 mm, (F)
F2_20 mm, according to the modified VCT by Arbelo.
F2_20 mm, according to the modified VCT by Arbelo.

Table 2. Arbelo method results.

The Numerical The Difference

ξ2 Pcr [kN] Pimp [kN]
Model [%]
B0 0 1360 1360 0
B1_2mm 0.02513 1357 1142 2.0
B1_5mm 0.08468 1340 950 1.3
B1_20mm 0.2463 1129 569 10.0
Materials 2021, 14, 567 14 of 15

Table 2. Arbelo method results.

The Numerical Model ξ2 Pcr [kN] Pimp [kN] The Difference [%]
B0 0 1360 1360 0
B1_2mm 0.02513 1357 1142 2.0
B1_5mm 0.08468 1340 950 1.3
B1_20mm 0.2463 1129 569 10.0
B2_2 mm 0.02731 1357 1133 −0.1
B2_5 mm 0.07254 1344 982 2.5
B2_20 mm 0.2365 1156 594 9.2
F1_2 mm 0 1360 1360 0
F1_5 mm 0 1359 1359 0
F1_20 mm 0.01467 1346 1183 −0.3
F2_2 mm 0 1359 1359 1.7
F2_5 mm 0.002816 1353 1281 −0.2
F2_20 mm 0.03863 1245 1000 0.3

4. Conclusions
The following conclusions stem from the performed numerical analysis only.
• The vibration correlation technique seems to be very effective in estimating buck-
ling or limit loads of silos strengthened with columns sparsely distributed along
their circumference.
• The VCT allows to predict the silo segment buckling load for the perfect structure and
the limit load in the case of the imperfect structure.
• The increase of geometrical imperfection magnitude reduces the VCT precision in the
case of the structures with buckling loads as initial imperfections.
• The loading course in numerical simulations affects the shape of selected vibra-
tion modes.
• The relationship between squared natural frequencies and the applied load is linear
in small load cases only. Considerable non-linearity occurs when the applied load
becomes close to the minimum buckling load or the limit load.
The future research will be focused on sensitivity analysis application in the VCT in
order to improve the estimation of real buckling loads by means of experimental tests. A
number of experiments is planned to be conducted in laboratory conditions to compare
numerical and experimental results.

Author Contributions: Conceptualization, P.I.; methodology and validation, Ł.Ż.-T.; numerical

investigation, Ł.Ż.-T.; formal analysis, Ł.Ż.-T. and P.I.; visualization, Ł.Ż.-T.; writing—original draft
preparation, Ł.Ż.-T.; writing—review and editing, P.I.; supervision, P.I. All authors have read and
agreed to the published version of the manuscript.
Funding: This research received no external funding.
Data Availability Statement: Data is contained within the article.
Acknowledgments: Abaqus calculations were carried out at the Academic Computer Centre in Gdańsk.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Southwell, R.V. On the analysis of experimental observations in problems of elastic stability. Proc. R. Soc. Lond. Ser. A 1932, 135,
601–606. [CrossRef]
2. Singer, J.; Arbocz, J.; Weller, T. Buckling Experiments: Experimental Methods in Buckling of Thin-Walled Structures; John Wiley & Sons:
New York, NY, USA, 2002; Volume 2.
3. Arbelo, M.A.; Kalnins, K.; Ozolins, O.; Skukis, E.; Castro, S.G.P.; Degenhardt, R. Experimental and numerical estimation of buck-
ling load on unstiffened cylindrical shells using a vibration correlation technique. Thin. Wall Struct. 2015, 97, 273–279. [CrossRef]
4. Lurie, H. Lateral Vibrations as Related to Structural Stability. Ph.D. Thesis, California Institute of Technology, Pasadena,
CA, USA, 1950.
Materials 2021, 14, 567 15 of 15

5. Jubb, J.E.M.; Phillips, I.G.; Becker, H. Interrelation of structural stability, stiffness, residual stress and natural frequency. J. Sound
Vib. 1975, 39, 121–134. [CrossRef]
6. Massonnet, C. Les Relations Entre Les Modes Normaux de Vibration et la Stabilite des Systemes Elastiques. In Bulletin des Cours et
des Laboratoires D’essais des Constructions du Genie Civil et D’hydraulique Fluvial; Goemaere, J., Ed.; Imp. Du Roi: Brussels, Belgium,
1940; Volume 1.
7. Chailleux, A.; Hans, Y.; Verchery, G. Experimental study of the buckling of laminated composite columns and plates. Int. J. Mech.
Sci. 1975, 17, 489–498. [CrossRef]
8. Singhatanadgid, P.; Sukajit, P. Experimental determination of the buckling load of rectangular plates using vibration correlation
technique. Struct. Eng. Mech. 2011, 37, 331–349. [CrossRef]
9. Abramovich, H.; Govich, D.; Grunwald, A. Buckling prediction of panels using the vibration correlation technique. Prog. Aerosp.
Sci. 2015, 78, 62–73. [CrossRef]
10. Souza, M.A.; Assaid, L.M.B. A new technique for the prediction of buckling loads from non-destructive vibration tests. Exp. Mech.
1991, 31, 93–97. [CrossRef]
11. Skukis, E.; Kalnins, K.; Chate, A. Preliminary Assessment of Correlation between Vibrations and Buckling Load of Stainless Steel
Cylinders. In Shell Structures: Theory and Applications; CRC Press/Balkema: London, UK, 2014; Volume 3, pp. 325–328.
12. Franzoni, F.; Degenhardt, R.; Albus, J.; Arbelo, M.A. Vibration correlation technique for predicting the buckling load of
imperfection-sensitive isotropic cylindrical shells: An analytical and numerical verification. Thin. Wall Struct. 2019, 140,
236–247. [CrossRef]
13. Arbelo, M.A.; de Almeida, S.F.M.; Donadon, M.V.; Rett, S.R.; Degenhardt, R.; Castro, S.G.P.; Kalnins, K.; Ozolins, O. Vibration
correlation technique for the estimation of real boundary conditions and buckling load of unstiffened plates and cylindrical shells.
Thin. Wall Struct. 2014, 79, 119–128. [CrossRef]
14. Skukis, E.; Ozolins, O.; Kalnins, K.; Arbelo, M.A. Experimental test for estimation of buckling load on unstiffened cylindrical
shells by vibration correlation technique. Procedia Eng. 2017, 172, 1023–1030. [CrossRef]
15. Shahgholian-Ghahfarokhi, D.; Gholamhossein, R. Buckling load prediction of grid-stiffened composite cylindrical shells using
the vibration correlation technique. Compos Sci. Technol. 2018, 167, 470–481. [CrossRef]
16. Franzoni, F.; Odermann, F.; Wilckens, D.; Skukis, E.; Kalnins, K.; Arbelo, M.A.; Degenhardt, R. Assessing the axial buckling
load of a pressurized orthotropic cylindrical shell through vibration correlation technique. Thin. Wall Struct. 2019, 137,
353–366. [CrossRef]
17. Shahgholian-Ghahfarokhi, D.; Aghaei-Ruzbahani, M.; Gholamhossein, R. Vibration correlation technique for the buckling load
prediction of composite sandwich plates with iso-grid cores. Thin. Wall Struct. 2019, 142, 392–404. [CrossRef]
18. Franzoni, F.; Odermann, F.; Lanbans, E.; Bisagni, C.; Arbelo, M.A. Experimental validation of the vibration correlation technique
robustness to predict buckling of unstiffened composite cylindrical shells. Compos. Struct. 2019, 224, 1–9. [CrossRef]
19. Labans, E.; Abramovich, H.; Bisagni, C. An experimental vibration-buckling investigation on classical and variable angle tow
composite shells under axial compression. J. Sound Vib. 2019, 449, 315–329. [CrossRef]
20. Żmuda-Trzebiatowski, Ł.; Iwicki, P. Correlation between Natural Frequencies and Buckling Load in A Stiffened Shell. In Shell
Structures: Theory and Applications; CRC Press/Balkema: London, UK, 2018; Volume 4, pp. 481–484.
21. Iwicki, P.; Rejowski, K.; Tejchman, J. Determination of buckling strength of silos composed of corrugated walls and thin-walled
columns using simplified wall segment models. Thin. Wall Struct. 2019, 135, 414–436. [CrossRef]
22. Iwicki, P.; Rejowski, K.; Tejchman, J. Simplified numerical model for global stability of corrugated silos with vertical stiffeners. J.
Constr. Steel Res. 2017, 138, 93–116. [CrossRef]
23. Dassault Systèmes Simulia Corp. Abaqus 6.14 Documentation; Dassault Systèmes: Providence, RI, USA, 2014.