Stability of cylindrical steel silos composed of corrugated sheets and columns

based on FE analyses versus Eurocode 3 approach

P. Iwicki, K. Rejowski, J. Tejchman

PII: S1350-6307(15)30059-5
DOI: doi: 10.1016/j.engfailanal.2015.08.017
Reference: EFA 2660

To appear in:

Received date: 27 February 2015

Accepted date: 11 August 2015

Please cite this article as: Iwicki P, Rejowski K, Tejchman J, Stability of cylindrical steel
silos composed of corrugated sheets and columns based on FE analyses versus Eurocode
3 approach, (2015), doi: 10.1016/j.engfailanal.2015.08.017

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Stability of cylindrical steel silos composed

of corrugated sheets and columns based on
FE analyses versus Eurocode 3 approach

PT
P. Iwicki, K. Rejowski and J. Tejchman

RI
Faculty of Civil and Environmental Engineering, Gdańsk University of Technology
Narutowicza 11/12, 80-233 Gdańsk-Wrzeszcz, Poland

SC
piwicki@pg.gda.pl, rejowskikarol@wp.pl, tejchmk@pg.gda.pl

NU
Abstract
MA
Comprehensive static and dynamic 3D stability finite element analyses of a cylindrical steel silo
composed of corrugated sheets and open thin-walled vertical stiffener profiles were carried out by
taking the geometric and material non-linearity into account. The silo was subjected to
ED

axisymmetric and non-axisymmetric loads imposed by a bulk solid following Eurocode 1.

Theoretical and measured initial geometric imperfections were considered in calculations. The
PT

results of non-linear analyses were compared to those according to Eurocode 3. Modifications of

the Eurocode formulae were proposed. Some recommendations for the silo engineering design were
CE

elaborated.
AC

Keywords: silo stability, finite element method, geometric imperfections, non-linear dynamic
analyses, non-linear static analyses, corrugated walls, columns

1. Introduction

Thin-walled metal cylindrical silo shells are vulnerable to buckling failure caused by the wall
friction force due to shearing of bulk solids along silo walls. The buckling failures occur
particularly during silo eccentric filling or discharge which are unavoidable because of a non-
homogeneous nature of bulk solids [1], [2]. As a consequence, non-uniform horizontal wall
pressures develop which contribute to undesired meridional bending and a non-symmetric
distribution of wall compressive forces. The non-symmetry of pressures distribution can also result
from local deformations of silo structural elements and global geometric inaccuracies during a silo

1
 ACCEPTED MANUSCRIPT
assembly. The buckling strength of silos depends on many different factors as: the aspect ratio of
the structure, discharge flow pattern, form and amplitude of geometric, loading and material
imperfections, type of joints, boundary conditions at ends, level of internal pressurization and
stiffness of the stored bulk solid [3]-[7]. For thin-walled cylindrical silos made from isotropic plain
rolled sheets with a different aspect ratio, a summary of various failure modes was presented in [8]

PT
and [9].

RI
Metal silos are frequently built of thin-walled horizontally corrugated curved sheets strengthened by
vertical stiffeners (columns) distributed uniformly around the silo circumference and connected

SC
with screws due to an economical steel consumption and a small silo weight. In those silos,
horizontally corrugated wall sheets carry circumferential tensile forces caused by horizontal wall

NU
pressure of a bulk solid and vertical columns carry vertical compressive forces exerted by wall
friction from a bulk solid. Thus, their behaviour is completely different from this for silos made
MA
from isotropic rolled thin-walled walls. In contrast to many stability analyses performed for silos
with isotropic rolled thin-walled walls ([8]-[16]), comprehensive analyses of silos consisting of
horizontally corrugated sheets and vertical stiffeners are rare. Our initial 3D FE calculation results
ED

presented in [17]-[19] evidently showed that the Eurocode approach [20] (see Eqs.1 and 2) provided
significantly too conservative outcomes for silos with corrugated sheets and columns.
PT

The silo stability is usually investigated by non-linear static studies based on the equilibrium path
CE

between the load and displacement. In analyses, the geometric non-linearity or both the geometric
and material non-linearity may be taken into account. In order to capture the unstable post-buckling
AC

behaviour, the arc-length method by Riks is usually used [21] that is the most fundamental
procedure for a stable analysis process under the global load control. It works well in snap-through
problems, where the equilibrium path in a load-displacement space is smooth and does not branch.
However, the solution is sometimes impossible to be achieved due to the convergence loss caused
by a localized instability (e.g. surface wrinkling, local buckling or material instability) resulting in
stiffness lack at the limit point. In order to overcome these problems and to reach a solution,
dynamic analyses may be applied [22]-[28], where the time history of a structure response is traced
during growing load. The advantage of dynamic analyses as compared to static analysis concerns
the ensured convergence of the load-displacement path equilibrium in the post-peak regime, when a
local transfer of strain energy from one part of the structure to its neighbouring parts occurs (local
instability) and a global solution method may not work. The application of static stability analyses
is certainly more realistic in a pre-peak regime, however in a post-peak regime this assumption is
not unique, since the stability loss connected usually with structural stability jumping modes has a
2
 ACCEPTED MANUSCRIPT
dynamic character. A dynamic stability analysis is also physically justified in silo shells under
compression since: a) dynamic effects occur when silo walls buckle during both filling and
emptying [29], [30] and b) the wall friction force rapidly grows at the onset of solid discharge
which is a dynamic process [31], [32]. The capability of coupled static and dynamic 3D FE
calculations for buckling investigations of cylindrical metal silo shells was shown in [28].

PT
The aim of our study is to predict the stability strength and critical regions of real cylindrical metal

RI
silos composed of horizontally corrugated sheets strengthened by vertical columns under
compressive loading by means of dynamic non-linear finite element analyses and to compare the

SC
strength with the Eurocode 3 approach [20]. First, a usual static analysis was performed in order to
compare the calculated stresses with those estimated based on measured displacements during silo

NU
filling and discharge. Next comprehensive static and dynamic analyses with different initial
geometric imperfections were carried out. Both global and local geometric imperfections were
MA
taken into account. The initial geometric imperfections were assumed based on the solution results
of a linear buckling problem (LBA) (global and local imperfections) and dynamic eigen-value
problem (global imperfections). In addition, real geometric imperfections measured on a similar silo
ED

type were taken into account. Non-linear static and dynamic FE analyses with both the geometric
and material non-linearity were carried out with an imperfect silo by taking into account loads
PT

exerted by a bulk solid during eccentric discharge (defined by Eurocode 1 [33]). The three-
dimensional numerical finite element (FE) calculations were carried out with the commercial code
CE

ABAQUS [21]. Attention was paid to the both force-displacement curves and deformation modes.
The choice of the mesh refinement, element type, time increment and arc-length increment were
AC

based on experience of our previous similar stability calculations for silos [17]-[19], [28]. The
calculated results on the stability of the silos were compared with the recommendations given by
Eurocode 3 [20]. The improvement of the standard formulae [20] was proposed based on the
calculations for 3 different silos.

The innovative elements of the paper are: a) a comprehensive comparative study of a dynamic and
static buckling process of a real metal silo composed of horizontally corrugated sheets strengthened
by vertical columns by taking into account different (theoretical and measured) initial geometric
imperfections that has not been performed to our knowledge yet and b) an elaboration of some
recommendations for the silo engineering design as compared to the requirements by Eurocode 3
[20]. The dynamic analyses were carried out by taking into account the different contribution of
inertial forces on the buckling process. At the high load action time, the effect of inertial forces was

3
 ACCEPTED MANUSCRIPT
almost eliminated based on the evolution of kinetic energy of the silo. Note that the existing
stability dynamic FE simulations are mostly devoted to shells [22]-[27].

2. Silo description

PT
The detailed numerical calculations were carried out with a real cylindrical metal silo belonging to a
silo battery built in Poland (Fig.1). The height of a silo was H=17.62 m and its diameter D=8.02 m

RI
(H/D=2.2). The cross-section area was 50.52 m2 and the perimeter 25.2 m. The silo mantle

SC
consisted of 21 horizontally corrugated sheets 890 mm×2940 mm×0.75 mm based on a foundation
slab. The silo was strengthened by 18 vertical columns composed of open thin-walled profiles with

NU
a varying cross-section in the form of the 'C' and 'V' - shape (Table 1) and thickness (t=1.5-4 mm)
along the column height (Table 1) uniformly placed along the silo circumference at the constant
distance of ds=1.4 m. The columns were connected to the wall sheets by screws. The corrugation
MA
had 76 mm pitch and 18 mm depth. The outflow took place through a hopper separated from the
rest of the silo. The silo roof was made from metal sheets inclined under an angle of 25o to the
horizontal and stiffened by 36 radial beams. The silo contained corn and was designed for
ED

concentric filling and funnel flow emptying with the constant outlet velocity of 50 m3/h.
PT

The silo wall loads induced by maize were calculated according to Eurocode 1 [33]. During
axisymmetric emptying, the standard maximum wall normal and shear stress in the bin were
CE

ph=30.52 kPa and pw=19.38 kPa, respectively. When considering a possible non-symmetric
emptying, they increased up to ph=36.26 kPa and pw =26.68 kPa [33] (Fig.2). In numerical analyses
AC

it was assumed that the load level according Eurocode 1 [33] was the reference value, i.e. the load
factor λ was always related to the wall shear stress of pw=26.68 kPa. The maximum characteristic
circumferential tensile normal stress in corrugated sheets was equal to σt=169 MPa and was
significantly smaller than the permissible steel normal stress equal to 355 MPa.

For the case of the column distance ds=1.4 m in the silo of Fig.1 (ds>ds,max), the characteristic
buckling load bearing capacity N of single vertical column in a metal silo with corrugated walls of
Fig.1 is given by the following formula in Eurocode 3 [20]

N = 2 EJ × K (1)
with

4
 ACCEPTED MANUSCRIPT
D
K = ks and D = 0.13Etd 3 , (2)
ds3

where EJ is the bending stiffness of columns in the plane perpendicular to the wall, the foundation
stiffness K denotes the bending stiffness of corrugated sheets between vertical columns, ݇௦ =6 is the

PT
coefficient, D denotes the wall sheet bending stiffness, E is the modulus of elasticity, t denote the
sheet thickness and d is the sheet height. The following assumptions were met to lay down Eqs.1

RI
and 2:
a) the 2D behaviour of the column is considered only,

SC
b) the number of buckling half-waves along the circumference is equal to the half of the
columns number, if the number of column is a multiple of 4 (buckled silo shape is axially anti-

NU
symmetric),
c) the column is loaded by vertical forces prescribed at both ends only (horizontal pressure is
MA
not considered),
d) the column is supported at one side by elastic springs simulating the presence of corrugated
sheets,
ED

e) the column has hinges at its ends.

PT

For the small column distance ds<ds,max, where

CE

0,25
 r 2 Dy 
= kdx 
 C y 
d s ,max , (3)
 
AC

there exists the approach in Eurocode [20] for the silo buckling based on an orthotropic shell theory.
The critical buckling resultant force nx,Rcr per the unit circumference of an orthotropic shell should
be evaluated at each appropriate silo level by minimizing Eq.4 with respect to the critical
circumference wave number j and the buckling height li as

 1  A 
nx , Rcr = min  2 2  A1 + 2   , (4)
 jω  A3  

where: Dy - the flexural rigidity parallel to the corrugation, Cy - the stretching stiffness parallel to the
corrugation and r - the cylinder radius and kdx – the coefficient recommended to be taken as 7.4, j - the
circumference wave number, ω - the parameter including buckling height li and A1,2,3 - the

5
 ACCEPTED MANUSCRIPT
parameters including the flexural and stretching stiffness in orthogonal directions of the equivalent
orthotropic shell.

The load bearing capacity N of a single vertical column in metal cylindrical silos should be
apparently always smaller than the plastic force [20]

PT
Nb , Rk = Aeff f y ,

RI
(5)

SC
where Aeff is the effective cross-sectional area of the column and fy denotes the yield stress.

NU
Due to the column distance in the silo of Fig.1 ds=1.4 m > ds,max=1.16 m (Eq.3), the appropriate
standard formula for the silo buckling strength is described by Eqs.1 and 2. Using Eqs.1 and 2, the
MA
characteristic buckling strength was exceeded in 1 column profile twice for symmetric emptying
and in 2 profiles about 2.5 times for non-symmetric emptying (Table 1). The column buckled for
the characteristic wall shear stress equal to pw=10.67 kPa (the limit load factor λ=0.4). In turn, the
ED

buckling strength calculated by Eq.4 was almost 8 times higher than this by Eqs.1 and 2.
PT

3. Input FE data
CE

The 3D stability finite element calculations were carried out with the entire cylindrical bin (curved
corrugated walls and 18 vertical thin-walled columns with a varying cross-section). The hopper and
AC

silo roof were neglected in non-linear calculations. Only in a static analysis for the design load level
and in a dynamic eigen-value analysis, the roof was taken into account. The corrugated sheets were
fixed to the columns at the sheet wave top at a constant distance of 0.076 m (thus, screw
connections were not modelled). The 4-node thin shell elements with a reduced integration point
were employed [21]. The total amount of finite elements was 884 880 (for the model without a roof)
(Fig.3a). The minimum 6 shell elements were assumed to describe the curvature of corrugated
sheets (Fig.3b). The element size was 8.6×93.6 mm2 (540 elements were taken along the silo
circumference). For columns, the elements size was 10×36 mm2 (265 elements were taken along the
silo height). The bin columns were fixed at the bottom. The columns were modelled by elements of
the size 14.9×32 mm2 (550 elements were taken along the silo height) (Fig.3b). The choice of the
element type and mesh refinement were based on own experiences from similar static analyses for a
silo with corrugated walls [17]-[19]. Along the top, radial and circumferential displacements were

6
 ACCEPTED MANUSCRIPT
restrained; only vertical displacements were allowed there. The following numerical FE analyses
were carried out:
1. LBA – a linear buckling (eigen-value) analysis with a perfect silo, performed in order to obtain
initial geometric imperfections.
2. DEVA - a dynamic eigen-value analysis of a perfect silo performed in order to obtain initial

PT
geometric imperfections and to find silo natural frequencies.
3. GNA – a geometrically non-linear analysis of a perfect and an imperfect silo in order to

RI
calculate the limit load factor.
4. GMNA – a geometrically and materially non-linear analysis of a perfect and an imperfect silo in

SC
order to calculate the limit load factor.

NU
The steel was assumed to be elastic (LBA, DEVA, GNA) or elastic-perfectly plastic (GMNA) with
the following properties: modulus of elasticity E=210 GPa, Poisson’s ratio v=0.3 and yield stress
MA
fy=355 MPa. In non-linear analyses, the automatic Riks method and implicit dynamic method by
Newmark [21] was used to determine a load-deflection relationship and limit load factor. The
minimum arc length increment was 1×10-6. In order to predict the effect of inertial forces, dynamic
ED

analyses were performed for the different action time ta of the prescribed vertical load: 2 s, 10 s,
20 s i 40 s mainly in the form of a linear function (i.e. the load linearly grew with time up to ta) or in
PT

the form of a bilinear form (the load linearly grew with time up to the maximum load factor and
then was kept constant during ta) (Fig.3C). The time increment in the automatic time increment
CE

approach was ∆t=2.5×10-5. The static analyses were about 1-2 times less expensive (in the CPU
time) than the dynamic ones for the time action time of ta=2 s and about 5 times less expensive than
AC

the dynamic ones for the time action time of ta=20 s.

4. Measurements in silo
4.1 Natural vibration frequencies

The natural frequencies of the silo structure were performed by means of 10 PCB accelerometers
356B18 in 10 points along the silo mantle at the distance of 2.5 m and the portable system for sound
and vibration measurements LMS SCADAS [34]. The silo vibrations enforced with the help of a
modal hammer were registered in a direction perpendicular to the silo mantle in the empty silo. The
following natural frequencies of the empty silo were identified in the measurements: 6.9 Hz, 7.5 Hz,
8.0 Hz, 11.6 Hz, 13.3 Hz, 13.7 Hz, 18.7 Hz and 19.4 Hz. The measured accelerations during silo

7
 ACCEPTED MANUSCRIPT
emptying were negligible due to funnel flow. The measured first eigen-frequency of the silo,
6.90 Hz, was almost the same as the calculated value of 6.89 Hz by FEM.

4.2 Initial geometric displacements in empty silo

PT
The initial geometric displacements were precisely measured in a similar silo type located in Poland
(H=25 m, D=15.4 m) (Fig.4A). The silo consisted of horizontally corrugated curved sheets

RI
strengthened by 34 vertical columns at the distance of ds=1 m. The corrugation had 76 mm pitch

SC
and 18 mm depth. The geodetic silo displacements were solely measured along column joints (12
points) at the vertical distance of 2.5 m with the accuracy of 3 mm in an empty silo directly after its
assembly before the first filling process. The column displacements were induced by the assembly

NU
method which did not require a crane. Initially, the thin-walled columns with the height of 2.5m
were erected along the entire silo circumference. Then the corrugated sheets were attached to the
MA
columns and to the each others. The silo ring was next lifted by about 2.5 m (distance equal to the
column height) by means of chains and new columns with sheets were fixed below. The columns
and sheets in 2 rows were connected by screws. The assembly process was repeated until the silo
ED

top edge was reached. After the silo assembly, the columns had the form of a zigzag line. The
column displacements induced obviously some sheet inaccuracies. The maximum radial column
PT

displacement to the outside was 34 mm and to the inside 50 mm (at the connections points). The
radial displacements measured of Fig.4A were next transferred to the analyzed silo (H=17.62 m,
CE

D=8.02 m). The grid of measured displacement at 12 points along 34 columns was proportionally
transformed to the scaled silo diameter into the grid of 7 points along 18 columns (Figs.4B and 4C).
AC

The column displacements in the analysed silo were assumed equal to the displacement of
neighbouring points in the measured silo with the maximum absolute value (Fig.4C). The maximum
column radial displacement to the outside was in the calculations as in the real silo, i.e. 34-50 mm.

5. Initial geometric wall imperfections in FE analyses

The different types of initial geometric wall imperfections were assumed in FE calculations (Fig.5):

1) the first eigen-mode of free vibration of the perfect silo (Fig.5A). It was assumed due to various
dynamic loads occurring in silos (e.g. wind, filling and emptying) which might contribute to
dynamic vibrations. The eigen-mode takes the shape of a half-wave in a vertical direction and 14
half-waves in a circumferential direction (Fig.5A).

8
 ACCEPTED MANUSCRIPT

2) the first eigen-mode from a linear buckling analysis (LBA) of a perfect silo (Fig.5B). In order to
eliminate a possible local stability loss in both columns and sheets, the wall thickness of the thin-
walled columns was increased up to 20 mm and the compressive vertical load due to wall friction
only was applied to the columns only. The buckled mode corresponded to 13 half-waves occurring

PT
along the part of the silo circumference and was non-symmetric (Fig.5B). There were solely two
half-waves in a vertical direction.

RI
3) the thirty fifth eigen-mode from a linear buckling analysis (LBA) of a perfect silo (Fig.5C) which

SC
was characterized by a local deformation mode of thin-walled column walls in the most compressed
lower region (Fig.5C).

NU
4) The measured displacements according to measurements in an empty similar silo of Fig.4B
MA
(Fig.5D). Their shape differed from the theoretical ones: slightly in a circumferential direction
(Fig.5E) and strongly in a vertical direction (the measured displacements uniformly covered the
entire silo in contrast to a theoretical distribution).
ED

6. FE results for silo

PT

6.1 Stability of perfect silo

CE

The calculated buckling load factor from a linear buckling analysis (LBA) for non-symmetric wall
shear and normal stresses was λ=5.5, i.e. the wall shear stresses at buckling were almost 6 times
AC

higher than the non-symmetric wall stresses by Eurocode 1 [12] (pw=5.5×26.68 kPa and
ph=5.5×36.26 kPa). The first 200 buckling modes (LBA) indicated solely local buckling of silo
column walls. The limit load factors from non-linear analyses were: λ = 3.18 (GNA) and λ=1.45
(GMNA) (Fig.6A). In static cases, the equilibrium path in a post-buckling range could not be
determined due to a convergence loss in spite of a comprehensive change of the parameters in the
Riks method (Fig.6A). It was found with the aid of dynamic calculations wherein the load increased
linearly in time (Fig.3C) several times faster than the period of natural vibrations (the first-eigen-
frequency was f=9 Hz for the silo without the roof and the period of natural vibrations
T=1/f=0.11 s). For the load action time ta=20 s, the dynamic load factor λ= 3.264 was similar as the
static solution result. The relationship between the load factor and silo top edge vertical
displacement was approximately linear up to the peak, and after the peak it indicated a pronounced
fall for GNA and a mild decrease for GMNA. The decrease of λ in a post-critical regime was

9
 ACCEPTED MANUSCRIPT
caused by the presence of inertial forces (the load factor λ was calculated as the sum of the vertical
reactions at the silo column supports). The failure deformation mechanism was characterized by a
local stability loss of the silo column walls (GNA) (Figs.6Ba and 6Bc) or by reaching the yield
stress at the wall connectors between 2 different column profiles (GMNA) (Figs.6Bb and 6Bd).

PT
6.2 Stability of imperfect silo
Initial modal geometric imperfection (1st eigen-mode)

RI
For the silo with modal imperfections from DEVA (Fig.5A) with the initial horizontal amplitude of

SC
w=25 mm, a decrease of the load action time ta caused an increase of the silo buckling strength by
20% due to the presence of inertial forces which act against the buckling (Fig.7A). The dynamic

NU
limit load factor was in the range of λ=3.66 for ta=2 s and λ=3.03 for ta=20 s. For ta≥20 s there was
no a significant difference in the load-displacement paths (Fig.7B), hence ta=20 s was assumed for
MA
further calculations. The results for ta=20 s (linear load function of Fig.3C) and for ta=2 s (bilinear
load function of Fig.3C) were similar.
ED

The silo dynamic deformation at the limit point was similar for the different load action time
(Fig.7C). A small difference between the deformed modes concerned the number of buckled
PT

columns. For the low load action time, most of single locally buckled columns in the lower silo
region under the hopper top edge were displaced to the silo inside. For the larger load action time
CE

(ta≥10), some columns (6-8) also buckled to the silo outside.

AC

Up to the limit load factor (ta=20 s), the silo behaved in a quasi-static way - the kinetic energy of the
system was close to zero (Fig.7D). Later, the silo behaviour became slightly dynamic due to an
increase of the kinetic energy. At the end of the analyses, the kinetic energy was 10% of the total
energy. For the bilinear load function and ta=2 s, the kinetic energy was 20% of the total energy.

The results of a geometrically non-linear static and dynamic analyses GNA for the different initial
horizontal amplitude w of the initial imperfection shape of Fig.5A are in Figs.8 and 9. The static and
dynamic limit load factor λ diminished with increasing initial imperfection amplitude w: the static
from λ=3.20 at w=0 mm down to λ=2.60 at w=75 mm and the dynamic: from λ=3.26 at w=0 mm
down to λ=2.80 at w=75 mm (lower by 0-20% than for the perfect silo). The static analysis results
differed from dynamic ones; there was a clear separation of static and dynamic paths at the limit
load. The differences were at the limit load of 5-10% (2.933<λ<3.264) and in the post-buckling

10
 ACCEPTED MANUSCRIPT
range of 20-30% (Figs.8 and 9). The dynamic limit load could be grater that the static one due to the
fact that in the static studies the real limit load was not reached. For a small imperfection amplitude
(w<25 mm), local buckling of column walls occurred and when an imperfection amplitude was
higher, the silo tended to buckle in a global mode similarly to the assumed initial imperfection
shape (Figs.8C and 9C).

PT
Figures 13 and 14 include the static and dynamic FE results for GMNA. The limit load factor (static

RI
and dynamic) varied between λ=1.22 (w=75 mm) and λ=1.45-1.47 (w=0 mm) (Figs.10 and 11) and

SC
was smaller than in GNA by about 50%. The static and dynamic non-linear load-displacement paths
were similar but in the case of static computations (Figs.10 and 11), the convergence was again lost
in a post-peak region in several simulations. Some dynamic fluctuations were obtained for the silo

NU
with w>75 mm (Fig.11). The failure mode in static and dynamic analyses was similar as for GNA
characterized by local buckling of silo column walls at connectors (w<50 mm) and global buckling
MA
of columns and corrugated walls (w≥50 mm) according to the assumed initial imperfection shape
(Figs.10C and 11C).
ED

Initial global geometric imperfection according to LBA (1st eigen-mode) with increased
column wall stiffness
PT

The FE results with the initial imperfection shape of first LBA eigen-mode of the silo with
CE

increased column wall thickness of Fig.5B are shown in Figs.12 and 13 (GNA) and Figs.14 and 15
(GMNA). In static analyses, the limit load factor λ was again not always determined due to a
AC

convergence loss (Figs.12 and 14). The dynamic limit load factor changed between λ=2.62
(w=100 mm) and λ=3.26 (w=0 mm) for GNA and between λ=1.19 (w=100 mm) and λ=1.47
(w=0 mm) for GMNA depending mainly upon the imperfection amplitude w (Figs.13 and 15). The
limit load factor λ was reduced by 20% (GNA) and 20% (GMNA) with respect to a perfect silo.
The silo deformation at the limit load factor was characterized by local column plastification at
column connectors for small imperfection amplitudes. The deformed columns moved into the silo
inside. For larger imperfection amplitudes, the columns were again plastified at connectors;
columns displaced into silo inside and outside (Figs.13 and 15) for both GNA and GMNA.

Initial local geometric imperfection according to LBA (35th eigen-mode)

11
 ACCEPTED MANUSCRIPT
The dynamic FE results with the initial imperfection of Fig.5C are presented in Figs.16 (GNA) and
20 (GMNA). The dynamic limit load factor was reduced from λ=3.26 (w=0 mm) down to λ=3.21
(w=10 mm) for GNA and from λ=1.45 (w=0 mm) down to λ=1.45 (w=10 mm) for GMNA. For the
limit load factor, a local stability loss of columns happened with small imperfection amplitudes. For
larger imperfection amplitudes, most of locally buckled columns moved into the silo inside and

PT
only a few to the outside (Fig.16). The columns were buckled near the hopper top edge and near
connection points of different profiles ('V' and 'C'). At the limit state for GMNA, 3 columns with

RI
the largest imperfection amplitude were plasticized.

SC
Initial global geometric imperfections based on measurements

NU
The dynamic FE results with the initial imperfection of Fig.5D are presented in Figs.18 (GNA) and
22 (GMNA). The dynamic limit load factor was reduced to λ=3.2 (w=35 mm) for GNA and to
MA
λ=1.44 (w=50 mm) for GMNA. The limit load factor λ was lower than for a perfect silo by 0.5%
(GNA) and by 2% (GMNA). At the limit load factor (GNA), a local stability loss of columns
occurred; the buckled columns were pushed into the silo inside at the column connectors (Fig.18C).
ED

The silo deformation at the limit state for GMNA was characterized by plastic column deformation
at connectors and by plastic deformation in the silo mantle and columns in the case of large
PT

imperfection amplitude (Fig.19)

CE

In the silo with initial imperfections based on measurements, the influence of the patch load was
also analyzed (the patch load was prescribed at the silo mid-height) (Figs.19Ad and 19Cd). The
AC

limit load factor was additionally reduced by 20% to λ=1.44 for w=50 mm. The deformation for the
limit load factor became non-symmetric around the silo circumference (Fig.19Cd).

The smallest silo limit load factor was obviously obtained for the silo with initial imperfections
according to the first eigen-mode from LBA with the increased column wall stiffness. It was smaller
by 10% than this according to the case with the measured initial imperfections and 3 times higher
than the silo buckling strength following Eurocode 3 [20].

The summary of calculation results of calculated limit load factor λ as compared to Eurocode 3
(Eq.1, 2 and 5) is presented in Fig.20 (GMNA). The calculated limit load is higher by 5-30% than in
Eurocode 3 depending upon the assumed initial imperfections.

12
 ACCEPTED MANUSCRIPT
7. Comparison with stability calculation results for similar silos

We compared our stability results in the silo of Fig.1 (H=17.62 m, D=8.2 m, H/D=2.2, 18 columns
for ds=1.4 m, corrugation: 76 mm pitch and 18 mm depth) with our previous ones for 2 silos with
corrugated walls and thin-walled columns: silo '1' (H=21.48 m, D=5.25 m, H/D=4.1, 18 columns for

PT
ds=0.93 m, corrugation: 76 mm pitch and 18 mm depth) [17] and silo '2' (H=20.11 m, D=12.5 m,
H/D=1.6, 28 columns for ds=1.4 m, corrugation: 119 mm pitch and 10 mm depth) [18]. All three

RI
silos had a similar height (18-22 m) but a different diameter (5-12.5 m), column spacing (0.9-1.4 m)

SC
and corrugation profile.

NU
The maximum load factor was compared for the standard wall loads during non-symmetric
emptying using GMNA with the initial geometric imperfections based on LBA for the maximum
horizontal imperfection amplitude w=50-100 mm (Table 2). This factor was always the smallest in
MA
the silo [18] for w=50-100 mm and the highest in the silo [17] for w=50 mm and in the silo of Fig.1
for w=100 mm. The maximum load factor increased with increasing corrugation depth by 20% for
w=50 mm and by 180% for w=100 mm for the similar slenderness ratio H/D (H/D=2.2-1.6) and
ED

decreased by 27% with increasing H/D (H/D=2.2-4.1) for the large horizontal imperfection
amplitude of w=100 mm. For the lower horizontal imperfection amplitude of w=50 mm, it was
PT

higher with increasing slenderness ratio H/D by 17% in contrast to metal cylindrical silos made
from plain isotropic sheets. The effect of the imperfection amplitude became stronger if H/D was
CE

higher. As compared to the performed numerical analyses (GMNA), the silo buckling resistance
according to Eurocode 3 [20] was lower by the factor: 1.3-2.2 [17], 1.3-3.1 [18] and 3.1 (silo of
AC

Fig.1).

In turn, the silo buckling strength calculated by the alternative approach in Eurocode 3 [20] (Eqs.3
and 4) was almost equal to the limit load for a perfect silo using GNA (λ=3.1) for the silo of Fig.1.
The buckling strength by Eqs.3 and 4 was λ=4.61 for the silo in [17] and λ=0.63 for the silo in [18].

In addition, a comparison between the numerical and Eurocode results (Eqs.1-4) was performed for
10, 19, 22. columns in the silo of Fig.1 applying GNA (Fig.21). With 22 columns, the calculations
were calculated according Eqs.3 and 4 instead of Eqs.1 and 2. For this silo, the limit load factor by
Eqs.3 and 4 was equal λ=4.22 while the limit load factor for the initial imperfections in the form of
the first eigen-frequency was very similar (λ=4.26). For the silo with 18 columns of Fig.1, Eqs.3
and 4 were also more realistic (the difference was solely 10%) while when using Eqs.1 and 2, the

13
 ACCEPTED MANUSCRIPT
differences were huge (λ=3 against λ=0.39) (Fig.21). Eqs.1 and 2 are more realistic when elastic
buckling takes place in spite of the limit condition (Eq.5).

In order to improve the Eurocode formulae and to be on the safe side, we propose to apply always
Eqs.3 and 4 with the reduction factor d s d s ,max within the range ds>ds,max instead of the

PT
recommended Eqs.1 and 2 (see curve 'd' for the silo of Fig.1 in Fig.22). The proposed formula is
also on the safe side for the silos described in [17] and [18]. For the silo [17], the limit load factor

RI
by FEM was higher by 25% and for the silo in [18], the limit load by FEM was higher by 200%

SC
than the Eqs.3 and 4 with ds>ds,max). The latter results indicate that Eqs.3 and 4 are sometimes also
too conservative (thus they also require the improvement due to the presence of the bulk solid).

NU
Note that Eqs.1 and 2 may also be modified in the range ds<ds,max to obtain better agreement with
FE results. The stiffness of the silo column elastic foundation provided by corrugated sheets has to
take into account both the circumferential wall curvature and tensile force and smaller half-wave
MA
length of the buckled wall in the circumferential direction (equal to the distance between the
columns).
ED

Our studied will be continued in order to optimize the Eurocode approach to the silo problem [20].
The stiffness of the stored bulk solid will also be taken into account [35] by describing the bulk
PT

solid behaviour with a hypoplastic constitutive model [1], [36], [37].

CE

8. Conclusions
AC

Some conclusions can be derived from our comprehensive dynamic and static stability FE studies
for silos (Fig.1, [17], [18]) made from corrugated curved sheets strengthened by vertical columns
with different initial geometric imperfections:

• The silo design according to the orthotropic shell theory is in Eurocode more realistic (Eqs.3 and
4) than the design based on the method considering the column stability on the elastic foundation
(Eqs.1 and 2). The latter formulae are largely too conservative as compared to the FE results. We
propose to always apply Eqs.3 and 4 independently of ds with the reduction factor d s d s ,max in the

range ds>ds,max (instead of Eqs.1 and 2).

• Non-linear implicit dynamic analyses were effective to obtain the full load-displacement
relationship during the entire buckling process. In static cases, the equilibrium path might not be
14
 ACCEPTED MANUSCRIPT
always possible to be determined due to a convergence loss (in particular, for geometrically non-
linear studies at small imperfection amplitudes). In non-linear dynamic analyses, the effect of
inertial forces was almost negligible for the load action time higher than 20 s (the static and
dynamic non-linear load displacement paths were similar).

PT
• The pattern of measured geometric initial imperfections significantly differed from the eigen-
mode from LBA. Based on measurements of initial imperfections, the stability strength (GMNA)

RI
was higher by 10% (for the initial horizontal imperfection amplitude w=50 mm) than this calculated
with initial imperfections based on the 1st eigen-mode with the increased column wall stiffness

SC
from LBA. The silo imperfection in the shape of the first eigen-mode (with the increased column
wall stiffness) always resulted in the lowest limit load. It is recommend to simulate the assembly

NU
process of silo in order to obtain the more realistic initial geometric imperfections.
MA
• For perfect silos, the maximum limit load was 1.3-3.5 larger than by Eurocode 3 for GMNA. For
imperfect silos with the initial horizontal imperfection amplitude of 50 mm, the maximum limit
load was smaller by 1-20% than by Eurocode 3 (GMNA).
ED

• For geometrically non-linear analyses, the effect of initial imperfections caused a decrease of the
PT

maximum limit load by about 20% with respect to perfect silos that is in agreement with the
standard approach. The material non-linearity decreased the buckling strength (even by 55%) as
CE

compared to geometrically non-linear calculations. For perfect silos (GNA), the maximum limit
load was 8 times larger than according to Eurocode 3.
AC

• The silo limit load factor depended upon the magnitude and shape of initial imperfections. The
lowest limit load factor was calculated for a global initial geometric imperfection from a linear
buckling analysis with an artificially increased thickness of column walls.

• For imperfect silos with small initial imperfection amplitudes, their deformation was characterized
by local buckling or plastification of columns at connectors depending on the performed analysis
(GNA or GMNA). When the initial imperfection amplitude was high (>5 cm), the silos always
tended to buckle in a global mode similar to the assumed initial imperfection shape.

15
 ACCEPTED MANUSCRIPT
• The corrugation pitch depth of wall sheets should be as large as possible. The columns profiles
should be restrained against local buckling by means of additional steel plates, in particular in
regions of a cross-section column profile change due to the additional bending.

Acknowledgements

PT
The financial support of the Polish National Research Centre NCN in the frame of the Grant No
2011/01/B/ST8/07492 "Safety and optimization of cylindrical metal silos containing bulk solids

RI
with respect to global stability " and of the Polish National Centre for Research and Development
NCBR in the frame of the Grant POIG.01.03.01-00-099/12 "Innovative method of dimensioning and

SC
construction of large industrial silos made from corrugated sheets" is gratefully acknowledged. The
numerical calculations were performed using the computing resources of CI TASK at Gdańsk

NU
University of Technology.
MA
References
[1] Tejchman J.. Confined granular flow in silos-experiments and numerical investigations. Berlin-
Heidelberg: Springer (eds. W. Wu and R. I. Borja); 2013.
ED

[2] Wójcik M, Tejchman J. Modeling of shear localization during confined granular flow in silos
within non-local hypoplasticity. Powder Technology, 2009; 192(3): 298–310.
PT

[3] Brown CJ, Nielsen J. Silos: Fundamentals of theory, behavior and design. London: E and FN
Spon; 1998.
CE

[4] Fayer N, Otten L. Handbook of Powder Science and Techmology. New York: Chapmann and
Hall; 1984.
AC

[5] Rotter JM. Guide for the Economic Design of Circular Metal Silos. London: Spon Press, 2001.

[6] Safarian SS, Harris EC. Design and Construction of Silos and Bunkers. New York: Van
Nostrand Reinhold Company, 1985.
[7] Tejchman J. The influence of bulk solids on buckling strength of silos. Powder Handling and
Processing 2004; 16 (4/5) 1-6.
[8] Sadowski AJ, Rotter JM. Steel silos with different aspect ratios: I — Behaviour under
concentric discharge. J. Constr. Steel Res. 2011; 67 (10): 1537-44.
[9] Sadowski AJ. Rotter JM. Steel silos with different aspect ratios: II — behaviour under eccentric
discharge. J. Constr. Steel Res. 2011; 67 (10): 1545-53.
[10] Sadowski AJ, Rotter JM. Buckling of very slender metal silos under eccentric discharge. Eng.
Struct. 2011; 33(4): 1187–94

16
 ACCEPTED MANUSCRIPT
[11] Song CY, Teng JG. Buckling of circular steel silos subjected to code-specified eccentric
discharge pressures. Eng. Struct. 2003; 25 (11):1397-417.
[12] Song CY. Effects of patch loads on structural behaviour of circular flat-bottomed steel silos.
Thin-Walled Struct. 2004; 42 (11): 1519-42
[13] Knebel K, Schweizerhof K. Buckling of cylindrical shells containing granular solids. Thin-

PT
Walled Struct. 1995; 23(1-4): 295–312.
[14] Cao Q, Zhao Y. Buckling strength of cylindrical steel tanks under harmonic settlement. Thin-

RI
Walled Struct 2010; 48 (6): 391-400.
[15] Chen L, Rotter JM, Doerich C. Buckling of cylindrical shells with stepwise variable wall

SC
thickness under uniform external pressure. Eng. Struct. 2011; 33 (12): 3570-8.
[16] Khelil A. Buckling of steel shells subjected to nonuniform axial and pressure loading. Thin-

NU
Walled Struct. 2002; 40 (11): 955–70.
[17] Wójcik M, Iwicki P, Tejchman J. 3D buckling analysis of a cylindrical metal bin composed of
MA
corrugated sheets strengthened by vertical stiffeners. Thin-Walled Struct. 2011; 49 (8): 947-963.
[18] Iwicki P, Wójcik M, Tejchman J. Failure of cylindrical steel silos composed of corrugated
sheets and columns and repair methods using a sensitivity analysis. Eng Fail Anal 2011; 18 (8):
ED

2064-83.
[19] Wójcik M, Sondej M, Tejchman J. Buckling analyses of cylindrical metal silos containing bulk
PT

solids. Proc. Int. Conf. Advances In Civil, Structural and Mechanical Engineering, Hongkong 2013.
[20] EN1993-4-1. Eurocode 3: Design of steel structures. Part 4–1: Silos, tanks and pipelines -
CE

Silos. Brussels: CEN; 2007.

[21] ABAQUS, Theory Manual, Version 6.8, Hibbit, Karlsson & Sorensen Inc; 2008.
AC

[22] Kleiber M, Kotula W, Saran M. Numerical analysis of dynamic quasi-bifurcation. Eng.

Comput. 1987; 4 (1): 48-52.
[23] Choong KK, Ramm E. Simulation of buckling process of shells by using the finite element
method. Thin-Walled Struct. 1998; 31 (1-3): 39-72.
[24] Kuhl D, Ramm E. Generalized energy-momentum method for non-linear adaptive shell
dynamics. Comput. Methods Appl. Mech. Eng. 1999; 178 (3-4): 343-66.
[25] Botasso CL, Bauchau, OA, Choi JY. An energy decaying scheme for nonlinear dynamics of
shells. Comput. Methods Appl. Mech. Eng. 2002; 191 (27-28): 3099-121.
[26] Chróścielewski J, Lubowiecka I, Witkowski W. Dynamics based on six-field theory of shells
in the context of energy-conserving scheme. Proc. Int. Conference SSTA, Gdańsk, Poland, 2005;
303-7.
[27] Kobayashi T, Mihara Y, Fujii F. Path-tracing analysis for post buckling process of elastic
cylindrical shells under axial compression. Thin-Walled Struct. 2012; 61: 180-187.
17
 ACCEPTED MANUSCRIPT
[28] Iwicki P, Tejchman J, Chróścielewski J. Dynamic FE simulations of buckling process in thin-
walled cylindrical metal silos. Thin-Walled Structures 2014; 84, 344–359.
[29] Knödel P. Stabilitatsuntersuchungen an kreiszylindrischen stahlernen Siloschussen.
Dissertation, University of Karlsruhe 1995.
[30] Ummenhofer T, Stabilitatsverhalten imperfekter zylindrischer Stahlsiloschalen -

PT
experimentelle und numerische Untersuchungen. Dissertation, University of Karlsruhe 1996.
[31] Gudehus G, Tejchman J, Silo music and silo quake - experiments and a numerical Cosserat

RI
approach. Powder Technology 1993; 76 (2): 201-12.
[32] Tejchman J. Technical concept to prevent the silo honking. Powder Technology 1999; 106 (1-2):

SC
7-22.
[33] EN1991-1-4. Eurocode 1: Basis of design and action on structures. Part 4: Actions in silos and

NU
tanks. Brussels: CEN, 1995.
[34] Wilde K, Rucka M, Rejowski K, Kuczyńska N, Wójcik M, Tejchman J. Internal report on
MA
measured accelerations, displacements and temperature of the silo in Radomice near Czerniejewo
(in polish). Faculty of Civil and Environmental Engineering, Gdansk University of Technology
2013.
ED

[35] Wójcik, M., Tejchman, J. Simulation of buckling process of cylindrical metal silos with flat
sheets containing bulk solids. Thin-Walled-Struct., doi: 10.1016/j.tws.2015.02.025, 2015.
PT

[36] Tejchman J. FE modeling of shear localization in granular bodies with micro-polar

hypoplasticity. Berlin-Heidelberg: Springer Series in Geomechanics and Geoengineering (eds.
CE

W. Wu and R. Borja); 2008.

[37] Tejchman J, Górski J. Computations of size effects in granular bodies within micro-polar
AC

hypoplasticity during plane strain compression. Int. J. for Solids and Structures 2008; 45 (6) 1546-
1569.

18
 ACCEPTED MANUSCRIPT

LIST OF FIGURES

Fig.1: Metal cylindrical silos with corrugated walls and 18 vertical columns at distance of 1.4 m
(height H=17.62 m and diameter D=8.02 m)

PT
Fig.2: Vertical wall frictional traction and horizontal wall pressure in silo during emptying

RI
according to Eurocode 1 [15] (pw – vertical frictional traction, ph – horizontal normal pressure)

SC
Fig.3. FE mesh for silo: A) entire bin and B) columns and C) relationship between load and time in
dynamic analyses: linear function (line 'a'), bilinear function (line 'b') and limit load factor for

NU
ta=20 s (line 'c')
MA
Fig.4: Initial horizontal wall geometric displacements of columns in cylindrical metal silo with
corrugated sheets: A) measured along perimeter (H=25 m, D=15.4 m, 34 columns at distance of
1.4 m), B) transferred to analyzed silo (H=17.62 m, D=8.02 m, 18 columns at distance of 1.4 m)
ED

and C) measured (a) and assumed (b) along silo perimeter at height of 2.5 m.
PT

Fig.5: Different initial wall geometric displacements assumed for FE analyses: A) 1st dynamic
eigen-mode of free vibration for perfect silo, B) 1st eigen-mode of perfect silo with column wall
CE

thickness increased up to 20 mm, C) 35th eigen-mode of perfect silo and D) from measurements in
empty silo of Fig.4B (a) 3D view, b) horizontal cross-section, c) side view), E) silo cross-section at
AC

height of 2.5 m normalized with respect to maximum imperfection amplitude: a) from

measurements, b) 1st dynamic eigen-mode of free vibration for perfect silo, c) 1st eigen-mode of
perfect silo with increased column wall thickness

Fig.6: Non-linear static and dynamic FE analyses of perfect silo: A) vertical load factor λ against
vertical displacement u of silo top edge and B) silo deformation for limit load factor with given
stresses by von Mises [Pa] and scale factor 6 (a) static analysis GNA, b) static analysis GMNA,
c) dynamic analysis GNA and d) dynamic analysis GMNA for load action time ta=20 s

19
 ACCEPTED MANUSCRIPT
Fig.7: Non-linear static and dynamic FE analyses (GNA) of imperfect silo for initial horizontal
imperfection in form of first eigen-mode of free vibration (Fig.5A) with initial horizontal amplitude
of w=25 mm: A) vertical load factor λ against vertical displacement u of silo top edge from
dynamic calculation with linear load function of Fig.3C for ta: a) 2 s, b) 10 s, c) 20 s, d) 40 s,
e) static calculation and f) dynamic calculation for ta=2 s with bilinear load function of Fig.3C, B)

PT
vertical load factor λ against load action time ta, C) silo deformation for limit load factor (with
stresses by von Mises [Pa] and scale factor 6) for ta: a) 2 s, b) 10 s, c) 20 s and d) 40 s and D)

RI
system energy evolution against time t and vertical top displacement u for ta=20 s (linear load

SC
function): a) total energy, b) elastic energy and c) kinetic energy

Fig.8: Non-linear static FE analyses (GNA) of imperfect silo with imperfection in form of first

NU
eigen-mode of free vibration of Fig.5A: A) evolution of load factor λ against vertical displacement
of silo top edge u for initial horizontal imperfection amplitude w: a) 0 mm, b) 25 mm and c) 75 mm,
MA
B) limit load factor λ against initial horizontal imperfection amplitude w and C) silo deformation at
limit load factor (with stresses by von Mises [Pa] and scale factor 6) for w: a) 0 mm, b) 25 mm and
c) 75 mm
ED

Fig.9: Non-linear dynamic FE analyses of imperfect silo (GNA) with imperfection in form of first
PT

eigen-mode of free vibration of Fig.5A for load action time ta=20 s: A) evolution of vertical load
factor λ against vertical displacement u of silo top edge for initial horizontal imperfection amplitude
CE

w: a) 0 mm, b) 25 mm and c) 75 mm, B) limit load factor λ against initial horizontal imperfection
amplitude w, C) silo deformation at limit load factor with stresses by von Mises [Pa] and scale
AC

factor 6 for w: a) 0 mm, b) 25 mm and c) 75 mm.

Fig.10: Non-linear static FE analyses (GMNA) of imperfect silo with imperfection in form of first
eigen-mode of free vibration (Fig.5A): A) evolution of load factor λ against vertical displacement
of silo top edge u for initial horizontal imperfection amplitude w: a) 0 mm, b) 25 mm and c) 75 mm,
B) limit load factor λ against initial horizontal imperfection amplitude w and C) silo deformation at
limit load factor with stresses by von Mises [Pa] and scale factor 6) for w: a) 0 mm, b) 25 mm and
c) 75 mm

20
 ACCEPTED MANUSCRIPT
Fig.11: Non-linear dynamic FE analyses of imperfect silo (GNA) with imperfection in form of first
eigen-mode of free vibration (Fig.5A) for load action time ta=20 s: A) evolution of vertical load
factor λ against vertical displacement u of silo top edge for initial imperfection amplitude w: a)
0 mm, b) 25 mm and c) 75 mm, B) limit load factor λ against initial horizontal imperfection
amplitude w, C) silo deformation at limit load factor with stresses by von Mises [Pa] and scale

PT
factor 6 for initial imperfection amplitude w: a) 0 mm, b) 25 mm, c) 50 mm, d) 75 mm and e)
100 mm

RI
Fig.12: Evolution of vertical load factor λ against vertical displacement u of silo top edge from

SC
non-linear static FE analyses of imperfect silo (GNA) with imperfection in form of first-eigen mode
from LBA with increased column wall stiffness (Fig.5B) for initial imperfection amplitude w: a)

NU
0 mm, b) 25 mm, c) 75 mm, d) 100 mm
MA
Fig.13: Non-linear dynamic FE analyses of imperfect silo (GNA) with initial geometric
imperfection in form of first eigen-mode from LBA with increased column wall stiffness (Fig.5B)
for load action time ta=20 s: A) evolution of vertical load factor λ against vertical displacement u of
ED

silo top edge for imperfection amplitude w: a) 0 mm, b) 25 mm, c) 50 mm and d) 75 mm and B)
load factor λ against initial w
PT

Fig.14: Non-linear dynamic FE analyses of imperfect silo (GMNA) with initial geometric
CE

imperfection in form of first eigen-mode from LBA with increased column wall stiffness (Fig.5B)
for load action time ta=20 s: A) evolution of vertical load factor λ against vertical displacement u of
AC

silo top edge for imperfection amplitude w: a) 0 mm, b) 75 mm and c) 100 mm, B) load factor λ
against initial amplitude w, C) silo deformation at limit load factor with stresses by von Mises [Pa]
and scale factor 6 for w: a) 0 mm, b) 75 mm and c) 100 mm

Fig.15: Non-linear dynamic FE analyses of imperfect silo (GMNA) with initial geometric
imperfection in form of first eigen-mode from LBA with increased wall thickness of columns
(Fig.5B) for load action time ta=20 s: A) evolution of vertical load factor λ against vertical
displacement u of silo top edge for initial imperfection amplitude w: a) 0 mm, b) 25 mm c) 50 mm,
d) 75 mm and e) 100 mm, B) load factor λ against w and C) silo deformation at limit load factor
with stresses by von Mises [Pa] and scale factor 6 for w: a) 0 mm, b) 25 mm and c) 75 mm

21
 ACCEPTED MANUSCRIPT
Fig.16: Non-linear dynamic FE analysis of imperfect silo (GNA) with initial geometric
imperfection in form of first eigen-mode from LBA (Fig.5C) for load action time ta=20 s: A)
evolution of vertical load factor λ against vertical displacement u of silo top edge for imperfection
amplitude w: a) 0 mm and b) 10 mm, B) load factor λ against w and C) silo deformation at limit
load factor with stresses by von Mises [Pa] and scale factor 6 for w: a) 0 mm and b) 10 mm

PT
Fig.17: Non-linear dynamic FE analysis of imperfect silo (GMNA) with initial geometric

RI
imperfection in form of first eigen-mode from LBA (Fig.5C) for load action time ta=20 s: A)
evolution of vertical load factor λ against vertical displacement u of silo top edge for imperfection

SC
amplitude w: a) 0 mm and b) 10 mm, B) load factor λ against w and C) silo deformation at limit
load factor with stresses by von Mises [Pa] and scale factor 6 for w: a) 0 mm and b) 10 mm

NU
Fig.18: Non-linear dynamic FE analysis of imperfect silo (GNA) with initial geometric
MA
imperfection according to measurements in empty similar silo (Fig.5D) for load action time ta=20 s:
A) evolution of vertical load factor λ against vertical displacement u of silo top edge for
imperfection amplitude w: a) 0 mm and b) 35 mm, B) load factor λ against w and C) silo
ED

deformation at limit load factor with stresses by von Mises [Pa] and scale factor 6 for w: a) 0 mm
and b) 35 mm
PT

Fig.19: Non-linear dynamic FE analysis of imperfect silo (GMNA) with initial geometric
CE

imperfection according to measurements in empty similar silo (Fig.5D) for load action time ta=20 s:
A) evolution of vertical load factor λ against vertical displacement u of silo top edge for
AC

imperfection amplitude w: a) 0 mm and b) 35 mm c) 50 mm d) 50 mm with patch-load, B) load

factor λ against w and C) silo deformation at limit load factor with stresses by von Mises [Pa] and
scale factor 6 for w: a) 0 mm, b) 35 mm c) 50 mm and d) 50 mm with patch-load

Fig.20: Summary of FE results: evolution of load factor λ against vertical displacement u of silo top
edge and limit load factor λ for for imperfection amplitude w=50 mm (GMNA) for: a) perfect silo,
b) imperfect silo with measured geometric imperfection, c) imperfect silo with imperfection in form
of first eigen-mode of free vibration, d) imperfect silo with geometric imperfection in form of first
eigen-mode from LBA with increased column wall thickness and e) based on Eurocode 3 (Eqs.1, 2
and 5)

22
 ACCEPTED MANUSCRIPT
Fig.21: Maximum load factor λ for silo of Fig.1 with initial imperfection in form of first eigen-
mode of free vibration with amplitude w=25 mm (GNA) against column number (10, 19, 22) as
compared to Eurocode formulae (Eqs.1-4)

Fig.22: Maximum load factors λ versus number of columns: for silo of Fig.1: a) Eurocode 3 (Eqs.3

PT
and 4), b) Eurocode 3 (Eqs.1 and 2), c) FE results and d) our proposition (Eqs.3 and 4 with
reduction factor d s d s ,max ), silo in [17]: e) Eurocode 3 (Eqs.3 and 4), f) Eurocode 3 (Eqs.1 and 2),

RI
g) FE results and h) our proposition (Eqs.3 and 4 with reduction factor d s d s ,max ) and silo in [18]:

SC
i) Eurocode 3 (Eqs.3 and 4), j) Eurocode 3 (Eqs.1 and 2), k) FE results and l) our proposition (Eqs.3
and 4 with reduction factor d s d s ,max ) (vertical lines 'm-p' denote column number for silo of Fig.1

NU
(line 'm'), silo in [17] (line 'n') and silo in [18] (line 'o'))
MA
ED
PT
CE
AC

23
 ACCEPTED MANUSCRIPT

LIST OF TABLES

Table 1: Silo column buckling capacity according to Eurocode 3 [20] during symmetric and non-
symmetric emptying

PT
Normal wall Normal wall Max. load Max. load

RI
Silo Critical
Profile force force (non- factor λ factor λ
height compressive

SC
cross- (symetric symetric (non-
level force [13] (symetric
section empting) [26] empting) [26] symetric
[m] empting)
[kN] [kN] [kN] empting)

NU
0,84 C1.5 0 0 51,32 - -
1,67 C1.5 1,82 2,51 51,32 28,2 20,5
MA
2,51 C1.5 8,02 11,03 51,32 6,4 4,7
3,34 C2.0 17,49 24,07 59,87 3,4 2,5
4,18 C2.0 29,49 40,58 59,87 2,0 1,5
ED

5,02 C2.0 43,49 59,86 59,87 1,4 1,0

5,85 C3.0 59,11 81,36 75,05 1,3 0,9
PT

6,69 C3.0 76,07 104,70 75,05 1,0 0,7

7,52 C3.0 94,13 129,56 75,05 0,8 0,6
CE

8,36 C4.0 113,14 155,72 83,21 0,7 0,5

9,20 C4.0 132,94 182,97 83,21 0,6 0,5
AC

10,03 C4.0 153,43 211,18 83,21 0,5 0,4

10,87 V04 174,52 240,21 165,91 1,0 0,7
11,70 V04 196,14 269,96 165,91 0,8 0,6
12,54 V04 218,22 300,35 165,91 0,8 0,6
13,38 V04 240,71 331,31 165,91 0,7 0,5
14,21 V04 240,71 331,31 165,91 0,7 0,5
15,05 V04 240,71 331,31 165,91 0,7 0,5
15,88 V05 240,71 331,31 187,57 0,8 0,6
16,72 V05 240,71 331,31 187,57 0,8 0,6
17,62 V05 240,71 331,31 187,57 0,8 0,6

24
 ACCEPTED MANUSCRIPT

Table 2: The maximum load factor λ from FE stability calculations for 3 different silos with
corrugated walls and columns for different horizontal imperfection amplitude w (Eqs.1 and 2)

PT
amplitude silo [17] silo [18] silo (present study)

RI
w [mm] λ λ λ

SC
50 1.45 1.03 1.24
100 0.87 0.42 1.20

NU
MA
ED
PT
CE
AC

25
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
ED
PT
CE
AC

FIGURE 1

26
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
ED
PT
CE
AC

FIGURE 2

27
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
ED
PT
CE
AC

A) B)

28
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
ED
PT

C)
CE
AC

FIGURE 3

29
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
ED

A)
PT
CE
AC

B)

30
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
ED

C)
PT

porównać z innymi imperfekcjami

CE
AC

FIGURE 4

31
 ACCEPTED MANUSCRIPT

PT
A) B) C)

RI
SC
NU
MA
ED
PT
CE
AC

a) b)

32
 ACCEPTED MANUSCRIPT

D)

PT
RI
SC
NU
MA
a) b) c)
ED

E)
PT
CE
AC

E)

FIGURE 5

33
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
A)
MA
ED
PT
CE
AC

a) b)

B)
c) d)

FIGURE 6
34
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
A)
ED
PT
CE
AC

B)

35
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
a) b) c) d)
MA C)
ED
PT
CE
AC

D)

FIGURE 7

36
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA A)
ED
PT

B)
CE
AC

C)
a) b) c)

FIGURE 8

37
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA A)
ED
PT
CE

B)
AC

C)
a) b) c)

FIGURE 9

38
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
A)
MA
ED

B)
PT
CE
AC

C)
a b) c)

FIGURE 10

39
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
A)
MA
ED

B)
PT
CE
AC

C)
a) b) d)

FIGURE 11

40
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
ED
PT
CE
AC

FIGURE 12

41
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
ED

A)
PT
CE
AC

B)

FIGURE 13

42
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
A)
MA
ED
PT

B)
CE
AC

C)

a) b) c)

FIGURE 14

43
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
A)
MA
ED
PT

B)
CE
AC

a) b) d)

FIGURE 15

44
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
A)
MA
ED

B)
PT
CE
AC

C)

a) b)

FIGURE 16

45
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA A)
ED
PT

B)
CE
AC

a)) b)
C)

FIGURE 17
46
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
A)

3.28
MA
3.26
3.24
λ

ED

3.22
3.2
0 10 20 30
PT

w [mm]
B)
CE
AC

C)
a) b)

FIGURE 18
47
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
A)
MA
ED
PT

B)
CE
AC

C)

a) b) c) d)

FIGURE 19

48
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
ED
PT
CE
AC

FIGURE 20

49
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
C)
ED
PT
CE
AC

FIGURE 21

50
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
ED
PT
CE
AC

FIGURE 22

51
 ACCEPTED MANUSCRIPT
Graphical Abstract

PT
RI
SC
NU
MA
ED
PT

Fig.1: Metal cylindrical silos with corrugated walls and 18 vertical columns at distance of 1.4 m
(height H=17.62 m and diameter D=8.02 m)
CE

Comprehensive static and dynamic 3D stability FE analyses of a cylindrical steel silo composed of
AC

corrugated sheets and vertical stiffeners were carried out by taking the geometric and material non-
linearity. The silo was subjected to axisymmetric and non-axisymmetric loads imposed by a bulk
solid following Eurocode 1. Theoretical and measured initial geometric imperfections were
considered in calculations. The differences between static and dynamic analysis results were
discussed. The results of non-linear analyses were compared to those according to Eurocode 3.
A modification of the Eurocode formulae was proposed. Some recommendations for the silo
engineering design were elaborated.

52
 ACCEPTED MANUSCRIPT
Research Highlights
• We deal with stability of large cylindrical metal silos composed of corrugated sheets.
• We carry out 3D FE stability calculations with the finite element method.
• We compare our FE results with the Eurocode approach.
• A modification of the Eurocode formulae was proposed.

PT
RI
SC
NU
MA
ED
PT
CE
AC

53