Engineering Structures 33 (2011) 19401954

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct
A simplified analytical procedure for assessing the worst patch load location on
circular steel silos with corrugated walls
E. Gallego

, C. Gonzlez-Montellano, A. Ramrez, F. Ayuga

BIPREE Research Group, Universidad Politcnica de Madrid, Madrid, Spain
a r t i c l e i n f o
Article history:
Received 17 December 2009
Received in revised form
22 February 2011
Accepted 28 February 2011
Available online 31 March 2011
Keywords:
Silo walls
Finite element models
Patch load
Steel structures
a b s t r a c t
Silos are widely used in the food and chemical industries for the storage of granular materials. The
calculation of their wall dimensions is complicated since the interaction between the stored material and
that fromwhich the silo is made is complex, a consequence of their very different mechanical behaviours.
The loads exerted on the silo walls by stored materials must be taken into account in silo design, and the
means for calculating them is contemplated in Eurocode EN 1991-4. The complexity of the phenomena
that occur within silos often leads to the appearance of unexpected and asymmetrically distributed
pressures. This is taken into account in the above Eurocode via the concept of the patch load, which is
asymmetric and can be exerted at any point on the silo wall. A finite element model has been developed
in order to check that the stress resultants derived from the patch load on steel silos with corrugated
walls may be predicted by using the well-known expressions of shell theory. Then, a simplified analytical
procedure has been developed for predicting the worst location of patch loads for all metal silos, but with
special application to corrugated steel silos in Action Assessment Class 3. It has been found that significant
differences may be found for most cases with the worst location for the patch load defined in Eurocode
for welded silos in Action Assessment Class 2. On the other hand, the values obtained for the maximum
meridional membrane stress resultant do not significantly differ, except for high slenderness values in
intermediate slenderness silos.
2011 Elsevier Ltd. All rights reserved.
1. Introduction
Silos have been used to store many kinds of material since the
end of the 19th century. Their use increased greatly as a conse-
quence of industrialisation, and gradually their dimensions be-
came larger. Unfortunately, their exploitation was associated with
a number of accidents, which, on top of their economic conse-
quences, involved the loss of human life. Research on silo design
began soon after. The work of Janssen [1] eventually gave rise to a
theory for the calculationof loads ina static state that is still used in
most modern day standards [25]. During the 20th century exper-
imental work designed to improve our understanding of the pres-
sures exerted on silo walls by stored materials intensified. Work
performed by different authors on concentric silos [611] showed
that, during silo discharge, the pressure exerted is greater than that
associated with filling. The shape of the lateral pressure curve ob-
tained during discharge resembles that produced by Janssen, ex-
cept at certain heights in the silo where there are increases in
pressure.

Corresponding address: EUIT Agrcola. Avda. Complutense s/n, 28040 Madrid,

Spain. Tel.: +34 91 336 3705.
E-mail address: eutiquio.gallego@upm.es (E. Gallego).
The existence of load eccentricities causes asymmetrical load
distributions on the silo wall [12]. A number of authors [1315]
have reported the lateral pressures on the silo wall closest to the
outlet to be smaller thanthose onthe opposite wall. It has also been
constantly confirmed that the pressures exerted during discharge
are greater than those seen during filling. The work of Pieper [16]
was of great impact because it showed that square silos with outlet
eccentricities of 50% and 100% behave in a manner opposite to that
described by the above authors. Thus, they report lateral pressure
to increase on the wall closest to the outlet, rather than on the
opposite wall.
Nielsen [17] demonstrated the existence of asymmetrical
pressure distributions in concentric silos both during filling and
discharge. These results came about via work on the possible
influence of the location of sensors placed on silo walls on the very
measurements they took [18]. In these experiments a geometrical
imperfection to the interior surface was introduced on the silo
wall near the place where the sensors were located and it was
demonstrated that the distribution of pressures on the wall was
altered by this. His work led to numerous findings [1921],
including evidence that geometrical asymmetries of the silo wall
could significantly contribute towards structural failure through
the notable local increase in lateral loads.
Stiglat [22] proposed the inclusion of two rectangular zones of
additional pressure on the silo walls to account for the pressures
0141-0296/$ see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2011.02.032
E. Gallego et al. / Engineering Structures 33 (2011) 19401954 1941
Fig. 1. Distribution of pressures exerted by the stored material.
that must have arisen to cause the damage he observed in concrete
silos. This finding was taken into account by the DINstandard [23],
and it could be considered the origin of the patch load concept.
The European Commission put into operation work [24] that led to
the current silo design standard EN 1991-4 [2], the responsibility
of which fell to the European Committee for Standardisation
(CEN). CEN established a project team (CEN TC250/SC1/PT4) to
develop this standard on actions in silos and tanks. This project
team adopted a proposal from [25] to treat the patch load as a
cosine function in the draft ENV 1991-4 (1996) standard [26]. The
concept of the patch load included in the Eurocode is a simplified
procedure that intends to take into account the appearance of
unexpected asymmetrical pressure distributions. The basis of
Rotter and Nielsen [25] proposal for metal silos was to introduce
an acceptable increase in the axial compression developed at the
base of the silo. This rule was based on the concept that the stress
resultants in the silo wall should all be simple cosine and sine
functions of the circumferential coordinate, so the patch load was
chosen to produce this effect.
The pressures measured in a full scale, operational silo at
Karpalund [2729] were used by Ooi et al. [30] to obtain a pressure
pattern that could explain the pressure asymmetries measured.
Further works by Ooi and Rotter [31], Ooi et al. [32] could not
provide more evidence of pressure patterns anything like the
cosine patch load include in the EN 1991-4 standard, so the
magnitude of the pressure on this patch load is still effectively the
one devised by Stiglat [22].
Eurocode EN 1991-4 [2] determines the loads exerted by the
stored material on the silo walls during silo filling and discharge.
These loads are taken into account via pressure distributions,
which may be global and symmetrical, or asymmetric and local.
Symmetrical pressure distributions come about throughhorizontal
pressures (p
h
) and frictional tractions (p
w
) on all points of the
silo wall, and show axial symmetry. In contrast, local asymmetric
distributions (the patch load) come about through horizontal
pressures, and their area of exertion is restricted to a localised band
on the silo wall of height S that can appear at any depth (z
p
) in
the silo (Fig. 1). The pressure values that characterise the patch
load (p
ps
) are constant for any meridional line, while continuous
variation is seen over the circumferential coordinates (). The
inclusion of this asymmetrical distribution allows the possible
distribution of asymmetrical loads caused by eccentricities during
filling anddischarge, or to imperfections inthe silo wall, to be taken
into account [3335].
The stress resultants caused by symmetrical pressure distri-
butions are easy to calculate using classical analytical equations
[1,36]. However, the stress resultants derived from asymmetrical
distributions are not so easily determined [37,38], a consequence
of their nature and arbitrary position. These effects are more im-
portant the larger the silo. Eurocode EN1991-4 [2] therefore allows
one to ignore or simplify the way in which the patch load is taken
into account in some cases. However, it is obligatory to take it into
full account in slender and intermediate slender silos in Action As-
sessment Class 3 (AAC3), and this requires the use of numerical
calculation methods and thus appropriate software. Finite element
models have been widely used to try to capture the complex pres-
sure patterns in silos [3944].
In addition, the patch load must be placed at any location on the
silo wall. At this regard, Eurocode EN1991-4 [2] makes a distinction
between welded silos and bolted silos. In the case of welded silos
Eurocode EN1991-4 [2] proposes a simply defined location for the
patch load that effectively maximises the axial compression at
the base. This approximation is valid because these silos are not
susceptible to bursting failures and the wall thickness is governed
by axial compression requirements, producing thicker walls than
are needed to resist bursting. This approximated location of the
patch load is still valid for bolted silos when axial compression is
considered. However, bolted silos must have adequate bolting at
every level to ensure that the peak pressure that might occur at
each point is supported. For this reason, Eurocode EN1991-4 [2]
still requires that the patch load be placed at any and all locations
on a bolted silo wall, in order to account for the increase in the
horizontal pressures at any level due to the patch load.
Large metallic silos are often made from corrugated steel.
Some authors have produced numerical models [34,45,46] or
have proposed analytical equations [47] for the analysis of the
effects of the patch load in metallic silos, particularly for silos
in which vertical compression is the only critical loading. The
present work examines the stress resultants produced by the patch
load on cylindrical silos made of corrugated steel, focusing on
axisymmetric silos in which the patch load must be taken into
account, i.e., slender silos or silos of intermediate slenderness
belonging to AAC3 [2].
For this, a finite element model [48] was developed that
simulates a corrugated wall, introducing the loads exerted by the
stored material and the patch load at different positions.
The development of numerical models for designing large silos
takes time and is costly. When the patch load needs to be taken
into account this can be even more expensive and time consuming
since the patch load can act at any point on the silo wall; there
is no single point where the appearance of the patch load is
clearly more unfavourable. The results obtained were used to
validate an alternative analytical procedure that calculates, for any
position on the wall, the worst location of the patch load for axial
compression. This analytical procedure avoids the need to use a
complex numerical calculation as well as the need to take into
account the different positions where this load might act.
2. Methodology
An analysis was made for 64 different models (Table 1) using
16 different silos that are flat-bottomed, ground-supported and
bolted. The patch load was taken into account each time using
the Finite Element Method (FEM), contemplating four possible
1942 E. Gallego et al. / Engineering Structures 33 (2011) 19401954
Fig. 2. Wavelength characteristic of the corrugated silo wall (left). Distinction between internal and external wall sections (right).
Table 1
Shape characteristics of the examined silos.
Silo (S) d
c
(m) H (m) (h
c
/d
c
) Capacity (t)
S1 21.83 29.28 1.34 10062.8
S2 21.83 30.39 1.39 10447.0
S3 21.83 31.51 1.44 10831.1
S4 23.65 25.01 1.06 10089.0
S5 23.65 26.13 1.10 10539.8
S6 23.65 27.25 1.15 10990.7
S7 23.65 28.36 1.20 11441.5
S8 23.65 29.48 1.25 11892.3
S9 23.65 30.60 1.29 12343.2
S10 23.65 31.72 1.34 12794.0
S11 27.29 27.65 1.01 14852.2
S12 27.29 28.77 1.05 15452.4
S13 27.29 29.89 1.10 16052.6
S14 27.29 31.00 1.14 16652.8
S15 27.29 32.12 1.18 17253.1
S16 27.29 32.64 1.02 23857.1
positions for this load. Each model was analysed for two different
loading states. In the first, only the symmetrical component of
the horizontal pressures was taken into account; in the second
the asymmetrical horizontal pressures (the patch load) were taken
into account as well. Frictional tractions were not considered in the
model since they are assumed to be taken by stringer stiffeners,
although the latter ones were not simulated in the numerical
model. The comparison of the stress resultants associated with
these loading states allowed those due to the patch load alone to
be isolated, and permitted the validation of an analytical procedure
for estimating the effects of the patch load.
2.1. Silo shape
The overall shape of each silo, S, was taken to be defined by
the diameter (d
c
) and height (H) of the highest stored solid to
wall contact. True silo shapes were taken into account (Table 1),
with the proviso that these were AAC3 silos of intermediate
slenderness [2]. The slenderness of the silos (Eq. (1)), , was
obtained from the equivalent height of the silo, h
c
, and the silo
diameter, d
c
. The equivalent height of the silo, h
c
, is measured from
the base to the equivalent surface. Thus, its value is obtained from
the sum of the height of the highest stored solid to wall contact,
H, and the height of the equivalent surface of the base of the top
pile, h
o
, the value of which was calculated (Eq. (2)) from the silo
diameter and the angle of repose of the stored material (
r
). The
material stored was taken to be wheat (
r
= 34). The slenderness
of all the silos considered was in the range of 1 < < 2; thus, they
were all of intermediate slenderness [2].
=
h
c
d
c
=
H +h
0
d
c
(1)
h
0
=
d
c
6
tan
r
. (2)
The wall of the silos was assumed to be made of corrugated steel,
the corrugations running circumferentially (Fig. 2(b)). The authors
useda finite element consisting of flat plate elements, sothe curved
surface was divided into small straight sections; for this reason the
shape of the silos reflected approximately real rather than truly
real conditions (Fig. 2(a)). This approximate shape facilitated the
applicationof loads to the walls. The metal wall of a real silo usually
differs slightly with height in terms of its thickness; this allows
the resistance required of the wall to withstand the pressures it
faces at each height to be obtained, optimising the cost of silo
manufacture. The silos examined were assumed to have walls of
varying thickness covering 0.8, 1, 1.25, 1.5, 1.7, 2, 2.5, 3, 3.5, 4, 4.5,
5, 5.5, 6, 6.5, 7, 7.5, 8, 8.5 and 9 mm.
Analytical methods were used to determine the wall thickness
required to resist the circumferential tensile membrane stress re-
sultant caused by the maximum horizontal symmetrical pressure
at each height. The walls were assumed to have stringer stiffeners
designedto resist the total vertical loadonthe wallcommonly the
case in the design of large silos [40]. With the aimof taking into ac-
count the patch load on the wall thickness values, the values of the
symmetrical pressures were increased in line with the patch load
following the simplified procedure outlined in Eurocode 1991-4.
The necessary wall thickness at each height was obtained fromthe
structural verifications defined in the different parts of Eurocode 3
for S355 quality steel [4952].
2.2. Loads
The 64 different models were analysed solely applying the
loads associated with the horizontal pressure component (Fig. 1)
under two sets of loading state conditions. The first (T1) took
into account the symmetrical distribution of the pressure exerted
by the stored wheat against the silo wall, while the second (T2)
E. Gallego et al. / Engineering Structures 33 (2011) 19401954 1943
Table 2
Mechanical properties of the stored material (according to Annex E in [2]).
Mechanical property Mean value Modification coefficient a Upper value Lower value Units
Maximum bulk unit weight 9 7.5 kN/m
3
Angle of repose (
r
) 34
Angle of internal friction (
i
) 30 1.12 26.79 33.6
Lateral pressure ratio (K
m
) 0.54 1.11 0.49 0.60
Coefficient of friction with the wall () 0.50 1.16 0.43 0.58
Reference coefficient for the patch load (C
op
) 0.50
took into account the symmetrical distribution of these pressures
plus the asymmetric distribution (patch load). The pressure values
applied were calculated using the procedures documented in Table
3.1 of EN 1991-4 [2] for the mechanical properties of wheat found
in annex E of Eurocode EN 1991-4 [2] (Table 2). The wall friction
coefficient was determined by using the expressions given in
annex D of Eurocode EN 1991-4 [2] for wall surface category D4.
The patch load can theoretically be exerted at any position for
AAC3 silos [2]. To analyse the possible influence of its position, four
patch load positions were taken into account for each silo: H
p
=
H/5 (P1), H
p
= 2H/5 (P2), H
p
= 3H/5 (P3) and H
p
= 4H/5 (P4)
(Fig. 1). Eurocode EN 1991-4 [2] provides different equations for
calculating the symmetrical horizontal pressures (p
h
) at eachdepth
z on the silo wall, measured from the equivalent surface (Fig. 1),
for both filling and discharge, depending on the slenderness of
the silo. Eq. (3) allows the horizontal pressures in intermediate
slenderness silos (1 < < 2) to be calculated.
p
h
(z) = C
h
p
ho
_
1
_
z h
0
z
0
h
0
+1
_
n
_
. (3)
In Eq. (3), p
ho
is the asymptotic horizontal pressure at great depth
due to stored particulate solid (Eq. (4)), whose value is a function of
the Janssen characteristic depth z
o
(Eq. (5)), C
h
is a coefficient that
takes into account filling or discharge (Eq. (6)) and n is an exponent
whose value is given by Eq. (7).
p
h0
= Kz
0
(4)
z
0
=
A
KU
(5)
C
h
=
_
1 Filling
1 +0.15 ( 1) Discharge
(6)
n = (1 +tan
r
) (1 z
0
/h
0
) . (7)
In Eqs. (3)(5), A is the cross-sectional area of the silo, U is the
circumference of this cross section, , K and are the values of
the material properties adopted, according to Table 3.1 of Eurocode
1991-4, for calculating the maximum normal pressures due to the
stored material: upper unit weight, upper lateral pressure ratio
and lower coefficient of wall friction respectively (Table 2). When
the wall frictional loading component is considered, the discharge
factor C
h
is substituted by C
w
(Eq. (8)):
C
w
=
_
1 Filling
1 +0.1 ( 1) Discharge.
(8)
The patch load is a distribution of asymmetrical pressures exerted
over a concentric band of the silo wall with a height of S = 0.2 d
c
,
the centre of which (Fig. 1) can be at any height, H
p
, on the silo wall.
The value of the pressures, p
ps
, that characterise the patch load
(Eq. (9)) is a function of the depth z
p
(z
p
= h
c
H
p
) at which
it is exerted and the circumferential coordinate () involved. C
p
is a coefficient (Eq. (10)) that takes into account the slenderness
of the silo and whether filling or discharge is underway. Its value
Fig. 3. Geometrical symmetry of the exerted loads.
is a function of C
op
, a reference coefficient for the patch load
characteristic of each particular stored material (Table 2).
p
ps
(z
p
, ) = p
p
(z
p
, ) cos = C
p
p
h
(z
p
) cos (9)
C
p
=
_
_
_
0.21C
op
_
1 e
1.5[1]
_
Filling
0.42C
op
_
1 e
1.5[1]
_
Discharge < 1.2
0.272C
op
( 1) Discharge > 1.2.
(10)
2.3. Finite element models
2.3.1. Finite element mesh
Only half of a silo was simulated since the cylindrical shape
and the form of the patch load allow the meridional symmetry
of the silo to be established (Fig. 3). The influence of the non-
simulated half was taken into account by adopting symmetry
boundary conditions on the vertical edges of the model.
The numerical model was employed using commercial ANSYS
software [53]. The silo was meshed with a shell planar quadrilat-
eral element with the capacity to simulate bending and membrane
effects. The above element had four nodes, each with six degrees
of freedom. The steel wall of the silo was assumed to show linear
elastic and isotropic behaviour, which is defined by its modulus
of elasticity and the Poisson coefficient (E = 210, 000, 000 kPa
and = 0.3). Fig. 4 shows the result of rotation around the silos
vertical axis of a generic, straight section with the simplified shape
shown in Fig. 2(b). This rotation gives rise to a circumferential band
that was meshed with a single row of 68 elements. Fig. 4(b) shows
a detail of the mesh of finite elements generated for the set of cir-
cumferential bands.
2.3.2. Application of loads
The loading states T1 and T2 assume the exertion of horizontal
pressures on the silo wall. The different models were simulated
assuming pressures during discharge. To simplify matters, these
loads were applied discretely over the elements belonging to
the circumferential band associated with the vertical straight
section representing half a corrugation wavelength (Fig. 2(b)). This
assumes that the horizontal loads are applied discontinuously,
with no loads on the elements in the inclined, intermediate
sections of a corrugation wavelength. To take this into account, the
pressure applied on the elements of each straight, vertical section
was bigger than that at the height z where the pressure acts.
1944 E. Gallego et al. / Engineering Structures 33 (2011) 19401954
Fig. 4. Circumferential band associated with one of the straight sections of the wavelength shape (a). Detail of the finite element mesh (b).
Fig. 5. Model constraints.
2.3.3. Constraints
The nodes at the base of the silo (line 2, Fig. 5) correspond to
points for anchoring to the foundation, which was treated as fixed.
Both the circumferential displacements and rotations about the
vertical axis of the nodes on the vertical edges of the model were
restrained to simulate the half of the silo that was not modelled
(see line 1, Fig. 5).
3. Results
3.1. Nomenclature
Each silo, S, was analysed under the conditions of the two
loading states (T1 and T2) indicated in Section 2.2, and with four
possible positions (P) for the patch load in T2. The results were
represented in five generatrices (Fig. 6) for the simulated half of the
silo, L, all represented with a particular circumferential coordinate
(
L
).
The development of all the R stress resultants shown in Fig. 7
were monitored for all the possible combinations of the variables
STP with respect to the circumferential () and meridional
(x) coordinates of the silo wall (Fig. 6). The representation of
these results generates a three-dimensional function of R =
f (x, )/(STP), where the STP variables are fixed. If T = T1,
the function R is independent of the variable P in the absence of
the patch load (R = f (x, )/(ST1)). For the scenario SP, the
result of R
T2
= f (x, )/(ST2P) minus R
T1
= f (x, )/(ST1)
Fig. 7. Stress resultants (R) exerted on the silo wall (positive values).
allows the distribution of the pressures on the silo wall caused
solely by the patch load to be determined. The function R
P
=
R
T2
R
T1
is three-dimensional (R
P
= f (x, )/(SP)) for fixed
values of SP. The three-dimensional values of R
P
= f (x, )/(SP)
were calculated for the circumferential coordinate
L
associated
with each generatrix L in order to study the patch load in detail
and facilitate the validation of the analytical expressions proposed
in this paper. This calculation generates the two-dimensional
function R
P,L
= f (x)/(SP
L
) that shows the development of
the stress resultants (R) on the wall at the meridional coordinate
x (Fig. 6) of each generatrix analysed.
3.2. Stress resultants due to the patch load
The stress resultants caused by the patch load were inves-
tigated using the scenario S3P4, since the shape of the graph
obtained would be the same for any scenario. Fig. 8 shows the
results obtained under the T2 loading state conditions (R
T2
=
f (x, )/(S3T2P4)). For each stress resultant, R, the influence of
the patch load was obtained via comparison with the results ob-
tained under the T1 conditions.
3.2.1. Circumferential membrane stress resultant
Fig. 9 shows the development of the circumferential membrane
stress resultant, n

, on the silo wall under the T1 loading state

conditions (n
,T1
= f (x, )/(S3T1)). This stress resultant is the
only one whose value is not zero during this loading state, and
it corresponds to a circumferential tension value that is constant
Fig. 6. Position of the generatrices L.
E. Gallego et al. / Engineering Structures 33 (2011) 19401954 1945
(a) n
,T2
/(S3P4T2). (b) n
x,T2
/(S3P4T2).
(c) n
x,T2
/(S3P4T2). (d) m
x,T2
/(S3P4T2).
(e) m
,T2
/(S3P4T2). (f) m
x,T2
/(S3P4T2).
Fig. 8. Stress resultants R
T2
= f (x, )/(S3T2P4) on the silo wall.
for a height x. This value is proportional to the horizontal pressure
at each depth z considered, p
h
(z), the constant of proportionality
being the radius of the silo (Eq. (11)).
n
,T1
= p
h
(z) r
c
(11)
Fig. 8(a) shows the development of the stress resultant n

on the
wall of the silo under the T2 loading state conditions (n
,T2
=
f (x, )/(S3P4T2)). Modifications of the initial value, n
,T1
,
only occurred in the band of application of the patch load as
a consequence of the pressure caused by the latter load. These
1946 E. Gallego et al. / Engineering Structures 33 (2011) 19401954
Fig. 9. Circumferential membrane stress resultant on the silo wall in the scenario
S3T1.
Fig. 10. Circumferential membrane stress resultant on the silo wall for scenario
S3P4, exclusively caused by the patch load.
modifications increased the value of the circumferential tension
over the interval 0 < < 90, and a reduction over that of 90
< < 180. The absolute value of these variations was greater
the closer the coordinate was to the limit values = 0 and
= 180. In the case of = 90, no change in the value of the
circumferential tension was seen.
The distribution of the patch load generates circumferential
tensionor compressionmembrane stress resultant where there are
pressures exerted inwards or outwards on the silo respectively.
The absolute value of this tensile membrane stress resultant or
compression depends on the angle , following the pressure
values of the patch load. Fig. 10 shows the effect of the patch
load on the membrane stress resultant, n

, via the function n

,P
arising fromn
,T2
n
,T1
. n
,P
membrane stress resultants are only
produced in the patch loads band of application, the tension and
compression values depending on the value of the circumferential
coordinate .
The effect of the patch load on the circumferential membrane
stress resultant can be examined analytically using Eq. (12),
employing the value of a patchloadat a depthz
p
withrespect to the
equivalent surface (Fig. 1) and for a point with the coordinates z,
(Fig. 6). If Eq. (9) is introduced into Eq. (12), and bearing in mind
the existence of n
,P
in the patch loads application band, then the
definitive value of n
,P
can be derived (Eq. (13)).
n
,P
(z, ) = p
p
(z
p
, )r
c
(12)
n
,P
(z, ) = r
c
C
p
p
h
(z
p
) cos z
p
s/2 z z
p
+s/2. (13)
The analytical expression provided by Eq. (13) predicted the value
obtained by the numerical model well in all the cases analysed, as
shown by Fig. 11 for line L5 in the scenario S3P4.
Fig. 11. Circumferential membrane stress resultant for line L5 in scenario S3P4,
obtained numerically and analytically.
3.2.2. Meridional membrane stress resultant
Under the conditions of T1, the development of circumferential
membrane stress resultants, n

, is sufficient to balance the sym-

metrical distribution of the pressures. The meridional membrane
stress resultants, n
x,T1
, therefore adopt null values throughout the
silo wall. Under the T2 conditions, the additional presence of the
patch load induces the appearance of meridional membrane stress
resultants in the silo wall below the point where the patch load
acts. Fig. 8(b) shows the absolute meridional membrane stress re-
sultants to increase linearly with the depth below the point where
the patch load is applied for any coordinate . The membrane
stress resultant, n
x,T2
, which is equivalent to n
x,P
since n
x,T1
= 0,
adopts compression values when 0 < < 90 and tensile values
when 90 < < 180. The magnitude of the meridional membrane
stress resultant increases the closer this coordinate is to the lim-
its = 0 and = 180. When = 90 the value of n
x,T2
is null.
The development of the meridional membrane stress resultant is
entirely due to the patch load since the latters distribution has a
non-null resultant. The resultant of the patch load (Eq. (14)) is a
horizontal force F at a depth z
p
(Fig. 1) with the same orientation
as the pressures of the patch load [25,26].
F(z
p
) = 0, 5 d
c
C
p
p
h
(z
p
). (14)
The force F resulting from the action of the patch load on the
band of the silo wall considered induces the appearance of global
bending moments (Fig. 12) for any cross section of the silo at a
depth z (Fig. 1) below the point of application of the patch load,
with a magnitude of F (z z
p
).
The global bending moment, M, generated by the resultant of
the patch load is compensated for by the appearance of membrane
stress resultants of compression and tension on opposite sides
of the bending axis. Both forces are of the same magnitude
and are situated symmetrically with respect to the bending axis.
The analytical determination of the meridional membrane stress
resultant, n
x,P
, at a point on the wall of coordinates (z, ), is
performed assuming that a cross section of the silo at a point below
where the patch load acts, z
p
, would behave like a tubular member
under bending. The meridional membrane stress resultant can be
directly expressed [47] as a function of the patch load (Eq. (15)).
n
x,P
(, z, z
p
) = 2C
p
p
h
(z
p
)s(z z
p
) cos () /d
c
. (15)
E. Gallego et al. / Engineering Structures 33 (2011) 19401954 1947
Fig. 12. Analytical determination of n
x
.
Fig. 13. Meridional membrane stress resultant on the line L5 for silo S3, obtained
analytically and numerically, for the four patch load positions (P) considered.
Eq. (15) predicted very well the numerical value of the meridional
membrane stress resultant obtained in all the cases analysed; this
is seen in Fig. 13 which shows the analytical and numerical results
for line L5 of silo S3 for the four patch load positions (P) taken into
account.
Several factors exist that prevent one knowing at the outset
the position of the patch load that will generate the greatest
meridional membrane stress resultant, n
x,P
at any determined
coordinates (z, ). Firstly, the value of n
x,P
depends on the
horizontal pressure, p
h
, at the position, z
p
(the centre of the patch
load position), which increases with the depth of the patch load.
n
x,P
also depends on the value of z z
p
, which is greater for higher
patch load positions and z coordinates close to the silo base. Fig. 13
shows that, of the positions studied here, position P3 is the least
favourable for a silo height (x) of <5.5 m, while P4 is the least
favourable for all other silo heights.
3.2.3. Meridional bending moment
Meridional shell bending moments m
x
arise due to the alternat-
ing eccentricity of the corrugated wall when it is subjected to axial
compression n
x
. These bending moments do not appear in the body
of the shell under T1 loading conditions because wall friction has
been omitted from this load case. Small local meridional moments
do occur near the base due to the compatibility requirement under
internal pressure, but these are not shown because the focus here
is on the effect of the corrugations.
Fig. 14. Appearance of the moment m
x,P
as a consequence of the meridional
membrane stress resultant n
x,P
.
Fig. 15. Meridional bending moment for the line L5 in silo S3 obtained analytically
and numerically for four patch load positions considered.
Fig. 8(f) shows the appearance of meridional bending moments
under T2 loading state conditions at all points on the wall under
the point where the patch load acts. The value of these moments
increases linearly with depth. The meridional membrane stress
resultant n
x
acts in the midplane of the corrugated wall, which lies
at a distance b from the vertical plane sections of the corrugation
wave (Fig. 14). Thus, Eq. (16) can be used to analytically calculate
the meridional moment at each point on the silo wall (x, ) due to
the patch load, m
x,P
, as the product of the meridional membrane
stress resultant and the distance b.
m
x,P
(x, ) = n
x,P
(x, ) b (16)
Eq. (16) predicted very well the numerical value of the meridional
bending moment in all the cases analysed, as shown by Fig. 15
for line L5 in the scenario S3P4. Note that, since the sign of n
x,P
is constant for L5, the value of the moment m
x,P
changes sign
depending on whether the vertical plane section is interior or
exterior to the midplane of the wall.
For unstiffened corrugated walls, the meridional bending
moments may lead to the failure mode known as roll-down in
which corrugations collapse [54]. Special design provisions are
presented in Eurocode EN 1993-4-1 [52] to cover this effect.
3.2.4. Circumferential bending moment
The circumferential bending moments, m

are a direct Poisson

consequence of the meridional bending moments and so are
quantitatively linked to them (m

= m
x
). The stresses derived
from the appearance of the circumferential bending moments are
irrelevant in silo design.
1948 E. Gallego et al. / Engineering Structures 33 (2011) 19401954
3.2.5. Membrane shear stress resultant
Under the T1 loading state conditions, membrane shear
stress resultants (n
x
) are not produced; neither are meridional
membrane stress resultants (n
x
), meridional bending moments
(m
x
) nor circumferential bending moments (m

). However, under
the T2 loading state conditions these membrane shear stress
resultants are produced below the point where the patch load acts
(Fig. 8(c)). Their value is constant for the same circumferential
coordinate independent of the silo height, with = 0 and
= 180 providing null values, and maximum values obtained
for = 90. The stress resultant, n
x,P
(x, ), can be determined
analytically using the membrane theory of shells, which leads to
Eq. (17). It was checked that Eq. (17) analytically predicted very
well the value for the numerically obtained membrane shear stress
resultants.
n
x,P
=
F
r
sin . (17)
However, stress resultants n
x
can be ignored in silo design
because the patch load, at the code defined magnitude, induces
very small stress resultants compared with those induced by
symmetrical wall friction.
3.2.6. Twisting shear moment
The twisting shear moments m
x
, adopt null values throughout
the silo wall under the T1 loading state conditions, while under
the T2 conditions they are not null (Fig. 8(d)). However, twisting
shear moments m
x,P
never play a significant role in the behaviour
of cylindrical shell structures [55]. Therefore, these moments are
not relevant for silo design.
3.3. A simplified analytical procedure for taking into account the
patch load in Action Assessment Class 3 (AAC3) slender and
intermediate slenderness silos
AAC3 slender and intermediate slenderness silos must be de-
signed following Eurocode EN 1991-4 [2], taking into account the
patch load. This requires numerous calculations to be performed
since the patch load can appear at any depth in a silo. Therefore,
numerical methods are required for this calculation. However, nu-
merical methods are time consuming, expensive to perform and
require special training. The analytical procedure proposed in this
work allows, for any point on a silo wall, the determination of the
most unfavourable stress resultants associated with a patch load
acting at any point. This obviates the needfor numerical techniques
andthe individual assessment of the stress resultants causedby the
patch load acting at different positions.
The equations that define the patch load in Eurocode were
derived from a cosine model, as proposed by Rotter and
Nielsen [25]. The pattern of pressures is sufficiently smooth to
produce a match between membrane theory and the bending
theory of shells used in the Finite Element analysis. Thus, this is
the reason why numerical and analytical results agree, as it has
beenshowninSection3.2. The proposedsimplifiedprocedure rests
on these expressions, which allow obtaining the stress resultants
caused by the patch load at a particular position in an analytical
way. Furthermore, these expressions in Section 3.2 are valid for
silos of intermediate slenderness and slender silos, and during
filling or discharge, as long as the values of p
h
and C
h
are calculated
properly. For silos of intermediate slenderness, these values are
given by Eqs. (3) and (6). In the case of slender silos ( > 2) the
values of p
h
, C
h
and C
w
are calculated using Eqs. (18)(20):
p
h
(z) = C
h
p
ho
_
1 e

z
z
0
_
(18)
C
h
=
_
1 Filling
1.15 Discharge
(19)
C
w
=
_
1 Filling
1.1 Discharge.
(20)
The values of p
ho
and z
o
for the case of slender silos are calculated
in the same way as for intermediate slenderness silos (Eqs. (4)
and (5)). Similarly, the value of the patch load pressures, p
ps
for
slender silos can be calculated using Eq. (8) provided that the
corresponding value of p
h
is used.
3.3.1. Starting hypothesis for the development of the proposed
simplified analytical procedure
The proposed procedure is based on the following conventional
design practices:
The silo in question is a AAC3 silo, either slender or of
intermediate slenderness [2]. The cross section is circular
and the wall is made from corrugated steel sheets reinforced
with vertical stiffeners around the silo. The distance between
these vertical stiffeners, according to Eurocode EN 1993-4-1
(5.3.4.3.4) [52], permits the assumption that the vertical forces
are entirely assumed by the stiffeners. The wall is therefore
assumed to have a zero meridional membrane stress resultant.
The sheets of corrugated steel forming the silo wall assume only
those forces that generate circumferential stresses due to the
horizontal symmetrical pressure and the patch load. The patch
load may act at any point, thus, the steel sheeting is made such
that any cross section at depth z can withstand circumferential
tensile membrane stress resultant equivalent to the sum of:
the circumferential membrane stress resultant caused by the
symmetrical horizontal pressure at depth z (n
,sym
).
The maximum circumferential membrane stress caused by
the patch load n
,asym
at depth z.
The vertical stiffeners assume the meridional membrane stress
resultants acting on the silo wallforces caused by the
symmetrical frictional traction and the meridional effects of the
patch load (n
x
). The uniform compression load assumed by the
vertical stiffeners at a depth z is the sum of:
The meridional stress resultant n
x,sym
caused by the symmet-
rical frictional traction between the top of the silo and a band
around the wall at depth z.
The maximum meridional stress resultant n
x,asym
at depth z
caused by any possible distribution of the patch load.
The value of n
,sym
can be determined for the symmetrical
horizontal pressure and radius of the silo using Eq. (11). The value
of n
x,sym
corresponds to the values of n
z,Sk
defined in Eurocode EN
1991-4, andis the result of the integrationof the frictional tractions
present at smaller values of the depth than z considered [47].
Since the symmetrical pressure values are well known, the use of
the proposed method only requires the definition of the values of
n
,asym
and n
x,asym
to be added to them.
3.3.2. Calculation of n
x,asym
The values of n
x,sym
are always compressive; the calculation of
n
x,asym
therefore always requires that one take into account the
position in the silo where the greatest meridional stress resultants
of compression due to the patch load arise, i.e., line L5 where
L5
=
0. At this position, and for a patch load k at a depth z
p
(Fig. 1), the
value of n
x,P,k
at a point at depth z (z > z
p
) is provided by Eq. (21)
for intermediate slenderness silos, and by Eq. (22) for slender silos.
n
x,P,k
= 0.4C
p
C
h
p
ho
z
o
_
1
_
1 +
p
2
0
1
0
_
n
_

z

p
_
(21)
n
x,P,k
= 0.4C
p
C
h
p
ho
z
o
_
1 e

p
_

z

p
_
. (22)
E. Gallego et al. / Engineering Structures 33 (2011) 19401954 1949
Fig. 16. Representation of the n
x,P,k
lines for 10 patch load positions. Comparison
of the strict and simplified linear envelope curves.
These equations have been obtained by substituting in Eq. (15) the
values of p
h
given in Eqs. (3) and (18) for intermediate slenderness
silos and slender silos respectively, and considering a value of s =
0.2d
c
. In addition, dimensionless variables (
p,

0
,
z
) have been
introduced (Eqs. (23)(25)).

p
= z
p
/z
0
(23)

0
= h
0
/z
0
(24)

z
= z/z
0
. (25)
If the depth below the surface z at any point is smaller than z
p
,
then n
x,P,k
automatically adopts a null value. Eqs. (21) and (22)
represent a straight line that provides the values of n
x,P,k
for all
the heights z of the silo, and for the position of the patch load k.
n
x,P,k
adopts a null value when z = z
p
, and a maximum value
when z = h
c
. The depth at which the patch load acts can be any;
thus, infinite n
x,P,k
lines corresponding to the infinite number of
possible positions of the patch load k would be obtained. Fig. 16
shows, for silo S3, 10 n
x,P,k
lines for 10 positions of the patch load
on the silo wall. Each of the infinite number of n
x,P,k
lines provides
a meridional membrane stress resultant different at each point
on the wall. However, the design of the silo wall is based on the
maximum value of the meridional membrane stress resultant. The
curve that for each height (h
c
z) provides the maximum value
is here given the name enveloping curve (n
x,p,max
); this is shown in
Fig. 16.
The strict envelope curve n
x,P,max
does not show tangents with
the entire set of n
x,P,k
lines. The analytical calculation of this
envelope curve is therefore complicated. However, if some points
(n
x,P,max i
, z
i
) of this strict envelope curve are determined, then a
simplified envelope curve (n

x,P,max
= f (z)) may be obtained by
using regressiontechniques. The number of points tobe considered
depends on the accuracy and the order of the function used to
approximate the strict envelope curve. In order to obtain the value
of n
x,P,max i
most unfavourable value of n
x,p
for a particular depth
it is required to determine the position of the patch load (z
p,max i
)
for which the load n
x,P,max i
is obtained. This position may be found
by solving Eq. (26), which maximises the function n
x,P,k
= f (
p
)
for a given z
i
position of the silo wall.
d
_
n
x,P,k
(
p
)
_
d
p
= 0. (26)
The use of a Taylor series (Eq. (27)) is required to obtain an
analytical solution of Eq. (26) to calculate an approximate solution
for some exponential and power terms that appears when deriving
the generic expression of n
x,P,k
= f (z
p
). Below, the detailed
procedure for solving Eq. (26) is presented for both the case of
intermediate slenderness silos and slender silos.
y = ln(1 +x) x
x
2
2
+
x
3
2

x
4
2
+ +(1)
n1
x
n
n
1 x 1. (27)
Intermediate slenderness silos.
For the particular case of intermediate slenderness silos,
Eq. (26) leads to:
d
d
p
_
0.4C
p
C
h
p
ho
z
o
_
1
_
1 +
p
2
0
1
0
_
n
_
_

z

p
_
_
= 0.
(28)
Thus, operating in Eq. (28) provides the following:
_
1 +

p

0
1
0
_
n1
_
1 2
0
n
z
1
0
_

_
1 +

p
(n +1)
1 2
0
n
z
_
= 1. (29)
Using Naperian logarithms for both terms in Eq. (29) provides the
expression:
(n 1) ln
_
1 +

p

0
1
0
_
+ln
_
1 2
0
n
z
1
0
_
+ ln
_
1 +

p
(n +1)
1 2
0
n
z
_
= 0. (30)
Seeking a closed formsolution for
p
it is necessary to approximate
the logarithmic terms in
p
using a series. The first and third terms
of Eq. (30) can be substituted for the first two terms in the Taylor
series (Eq. (27)) to finally obtain:
(n 1)
_

0
1
0

1
2
_

0
1
0
_
2
_
+ln
_
1 2
0
n
z
1
0
_
+

p
(n +1)
1 2
0
n
z

1
2
_

p
(n +1)
1 2
0
n
z
_
2
= 0. (31)
Thus, resolving Eq. (31), and grouping the terms in function of the
exponent order that affects
p
, provides the following:
_
1 n
2 (1
0
)
2

1
2
_
(n +1)
1 2
0
n
z
_
2
_

2
p
+
_
(n 1) (1 2
0
)
(1
0
)
2
+
(n +1)
1 2
0
n
z
_

p
+
_
(1 n)
0
(2
0
)
2 (1
0
)
2
+ln
_
1 2
0
n
z
1
0
__
= 0. (32)
The solution of the second degree equation shown in Eq. (32)
provides the value of
p,i
associated with z
p,max i
, which is defined
to be the position of the patch load that gives rise to the maximum
meridional membrane stress resultant n
x,P,max i
at the depth below
the surface z
i
. The value of this membrane stress resultant is
obtained using Eq. (33).
n
x,P,max i
= 0.4C
p
C
h
p
ho
z
o
_
1
_
1 +
p,i
2
0
1
0
_
n
_

z,i

p,i
_
. (33)
1950 E. Gallego et al. / Engineering Structures 33 (2011) 19401954
(a) Intermediate slenderness silo
(d
c
= 21.83 m, h
c
= 31.51 m; = 1.44).
(b) Slender silo
(d
c
= 18.5 m, h
c
= 42.08 m; = 2.27).
Fig. 17. Variation of the value of n
x,p,k
at different positions z
i
with the value of z
p
.
Since a Taylor series was used to obtain an approximate solution,
the range of validity of Eqs. (32) and (33) is conditioned by meeting
the conditions set out in Eqs. (34) and (35).
1

p

0
1
0
1 (34)
1

p
(n +1)
1 2
0
n
z
1. (35)
Slender silos.
For the particular case of slender silos, Eq. (26) leads to:
d
dz
p
_
0.4C
p
C
h
p
ho
z
o
_
1 e

p
_

z

p
__
= 0. (36)
Thus, operating in Eq. (36) provides the following:
e

p
_
(
z
+1)
_
1

p

z
+1
__
= 1. (37)
Using Naperian logarithms for both terms in Eq. (37) provides the
expression:

p
+ln (
z
+1) +ln
_
1

p

z
+1
_
= 0. (38)
The third term of Eq. (40) can be substituted by the first two terms
in the Taylor series (Eq. (27)) to obtain:

p
+ln (
z
+1)
_

p

z
+1
+
1
2
_

p

z
+1
_
2
_
= 0. (39)
Thus, solving Eq. (39) and grouping the terms depending on the
exponent order of
p
would provide:
_
1
2
1
(
z
+1)
2
_

2
p
+
_

z
+2

z
+1
_

p
ln (
z
+1) = 0. (40)
The solution of the second degree equation shown in Eq. (40)
provides the value of
p,i
associated with z
p,max i
, which is defined
to be the position of the patch load that gives rise to the maximum
meridional membrane stress resultant n
x,P,max i
at the depth below
the surface z
i
. The value of this membrane stress resultant is
obtained using Eq. (41).
n
x,P,k
= 0.4C
p
C
h
p
ho
z
o
_
1 e

p,i
_

z,i

p,i
_
. (41)
Since a Taylor series was used to obtain an approximate solution,
the range of validity of Eqs. (40) and (41) is conditioned by the
meeting of the conditions set out in Eq. (42).
1

p

z
+1
1. (42)
Calculation of the simplified envelope curve.
The simplified envelope curve may be approximated to a
particular function type if some points (n
x,P,max i
, z
i
) of the strict
envelope curve are known. A linear approximation has been
considered (Fig. 17) to exemplify this procedure. So, two points
of the envelope curve were needed to calculate the simplified
envelope curve. These were adopted those with z
i
= 0 and
z
i
= h
c
, and the corresponding values of the maximummeridional
membrane stress resultant due to the patch load, n
x,P,max i
=
0, and n
x,P,max,hc
, respectively. The latter value is obtained for
a position z
i
= h
c
from Eqs. (32) or (40), depending on the
slenderness of the silo. Thus, the expression of the simplified
envelope curve (n

x,P,max
= f (z)) will be given by Eq. (43) for the
linear approximation used.
n
x,P,max
= n
x,asym
= n
x,P,k,h
c
(1 (h
c
z) / (h
c
h
0
)) . (43)
Variation of the value of n
x,p,k
at a position z
i
with the value of z
p
.
Fig. 17 shows the variation of n
x,p,k
with respect to z
p
for
different fixed positions z
i
in two silos of different slenderness.
Each curve in Fig. 17 represents one of the functions given by
Eqs. (21) or (22) (depending on the slenderness) for a particular
value of z
i
. It can be observed that the peak value of n
x,p,k
for a given
depth belowthe surface z
i
is found to be at a position z
p,max
, which
is always placed close to half of the distance between the top of the
silo and the considered depth, z
i
. This position is slightly closer to
the top of the silo and tends to become constant for large values
of z
i
. This pattern is a consequence of the asymptotic trend of the
symmetrical normal pressure distributionfor deeppositions. Inthe
asymptotic area, the value of p
h
remains almost constant and then
(z
i
z
p
) is the only factor that affects n
x,p,k.
according to Eqs. (21) or
(22). The thick black line has been obtained in Fig. 18 by joining up
the points of maximum n
x,p,k
for each position z
i
considered, and
its asymptotic trend for the value of z
p,max
can be clearly seen.
Comparison between the exact and approximate solutions for
z
p,max
and the values proposed by EN 1991-4 [2]. Influence of these
solutions on the value of n
x,p,max
.
The solution obtained for intermediate slenderness silos and
slender silos through Eqs. (29) and (37), respectively, provides
E. Gallego et al. / Engineering Structures 33 (2011) 19401954 1951
Fig. 18. Comparison between the value of z
p,max
proposed by EN 1991-4 [2] and the exact and approximate solutions obtained for a position z
i
= h
c
. Influence of these
solutions on n
x,p,max
.
the value of the most unfavourable position of the patch load
(z
p,max
) for a given position z on the silo wall. The calculation of the
exact solution for those equations can only be obtained by using
different iterative procedures. However, an approximate solution
may be obtained if Taylor series (Eq. (27)) are used, which led to
Eqs. (32) and (40) for intermediate slenderness silos and slender
silos, respectively. In addition, these approximate solutions are
valid only if the conditions set out in Eqs. (34)(35) (for
intermediate slenderness silos) and (42) (for slender silos) are
satisfied.
Following the simple proposal of Rotter and Nielsen [25],
Eurocode EN 1991-4 [2] indicates for welded silos that the
most unfavourable position of the patch load (z
p,max
) may be
approximated by the lesser of z
0
and 0.5h
c
, regardless the position
z
i
considered on the silo wall. The rules provided in Eurocode EN
1991-4 [2] for determining the most unfavourable location of the
patch load in welded silos have been used as a reference value
for the case of bolted silos studied in this work. The patch load
must be placed at any and all locations in bolted silos due to the
bursting failures, in order to ensure that the peak pressure that
might occur at each point is supported. However, the reference
value defined for welded silos is still valid for bolted silos if axial
compression is being considered. For this reason, the value of z
p,max
and its influence on n
x,p,max
have been calculated by using three
methods: the exact solution, the approximate solution obtained
by using the Taylor series and the calculation obtained according
to the criterion defined in EN 1991-4 [2]. Fig. 18 shows the values
for z
p,max
and n
x,p,max
for a position z
i
= h
c
in slender silos and
intermediate slenderness silos as a function of the slenderness of
the silo.
It can be seen that the value defined in EN1991-4 [2] for z
p,max
is
often significantly different from the one obtained from the exact
or the approximate solutions for all possible slenderness values.
The approximate solution provides results very close to those
obtainedwiththe exact solutionfor slender silos but are somewhat
different for intermediate slenderness silos at high slenderness
values. The reason may be found in the fact that the precision of the
Taylor approximationis lower for highslenderness values since the
conditions set out in Eqs. (34)(35) approach to the limiting values
1. In fact, this also causes that no approximate solution can be
obtained from Eq. (32) for z
i
= h
c
and slenderness values greater
than 1.65. However, the maximum difference between the exact
value and the approximate one for n
x,p,max
is less than 7% for all
slenderness values.
However, the resulting values of n
x,p,max
derived from these
positions do not significantly differ between all methods. It is
important to bear in mind that the value of n
x,p,max
is used for
the design of the silo. Significant differences appear only for high
slenderness in intermediate slenderness silos between the exact
value of n
x,p,max
and those obtained from the approximate value
of z
p,max
or the criterion defined in the Eurocode. The shape of
curves plotted in Fig. 18 may explain these small differences in
n
x,p,max
, despite the great differences of position obtained between
the Eurocode method and the approximate and exact solutions for
the localisation of the worst position of the patch load. If the curve
associated to a position z
i
/h
c
= 0.92 is observed in Fig. 18(a), it
can be found that the patch load position z
p
/h
c
= 0.38 produces
1952 E. Gallego et al. / Engineering Structures 33 (2011) 19401954
Fig. 19. Comparison between the value of z
p,max
proposed by EN 1991-4 [2] and the exact and approximate solutions obtained for this value at position z
i
= 0.75 h
c
.
Influence of these solutions on the n
x,p,max
value.
the peak value of n
x,p,max
. However, almost the same values of
n
x,p,k
may be obtained for patch load positions, z
p
/h
c
, in the range
0.30.45.
The values of z
p,max
for the exact solution remain almost
unchanged regardless of the slenderness considered. So, the worst
position of the patch load is independent of the slenderness of the
silo for a given position z
i
. However, the value of z
p,max
may change
considerably when the depth below the surface z
i
is different, as it
will be shown later.
The worst location for the patch load, as well as the maximum
n
x,p,max
are plotted in Fig. 19 for a depth below the surface z
i
=
0.75h
c
. The value of z
p,max
permitted by EN 1991-4 [2] is again
greater than the one obtained from the exact or the approximate
solutions for all possible values of slenderness. In this case, if a
point z
i
different from h
c
has been considered, the approximate
value predicted for intermediate slenderness silos is much closer
to the exact value. It indicates that the conditions set out in Eqs.
(34)(35) are better achieved than for z
i
= h
c
. The approximate
and the exact values are almost identical for slender silos, as it
occurred for z
i
= h
c
.
In this case, the differences existing for the worst position of the
patch load between the values predicted by Eurocode and those
obtained with the exact solution have now a greater influence in
n
x,p,max
than for z
i
= h
c
. This difference is particularly evident for
high slenderness values in intermediate slenderness silos. In this
case, the values of n
x,p,max
obtained from the exact or approximate
solutions are almost the same but the value of n
x,p,max
obtained
from the proposed z
p,max
value in EN 1991-4 [2] is up to a 25%
smaller. Small differences of n
x,p,max
between all methods are
obtained for low slenderness values in intermediate slenderness
silos and for any slenderness in slender silos.
3.3.3. Determination of the n
,asym
value
The value of n
,sym
always reflects tension. Thus, n
,asym
= f (z)
can be obtained analytically for a circumferential coordinate = 0
(generatrix L5) (Eq. (44)) via the use of Eq. (13). The patch load
can act at any depth in the silo; Eq. (44) therefore directly provides
the maximummeridional membrane stress resultant curve at each
depth below the surface z (n
,asym
= f (z)).
n
,P,L5
(z) = C
p
p
h
(z)r. (44)
4. Conclusions
This work reports the results of a numerical and analytical study
of the stress resultants caused by the patch load, as defined in
Eurocode EN 1991-4 [2], in cylindrical, flat-bottomed silos with
walls made of corrugated steel sheeting, and whose vertical loads
are carried on wall vertical stiffeners.
The most important stress resultants to consider in silo design
are:
The circumferential membrane stress resultant in the band
of application of the patch load. Circumferential tension from
axisymmetric loads must be increased by the magnitude arising
from the tensile side of the patch load.
E. Gallego et al. / Engineering Structures 33 (2011) 19401954 1953
Meridional membrane stress resultant (n
x
) acting on all points
of the wall below the point where the patch load acts. This is
due to the global bending moment acting on the cross section
of the silo at these positions, which in turn is due to the non-
null resultant force of the distribution of the patch load. For
silo design, the most important effect caused by this global
bending moment is the increase in axial compression on the
compressive side.
Meridional moment (m
x
) acting on all points of the wall
below where the patch load acts. This is a consequence of the
eccentricity betweenthe meridional membrane stress resultant
(n
x
) and the position of the vertical straight sections in the
corrugation wave.
The knowledge of these effects led to the validation of analytical
expressions that allow the designer to determine without the
need for numerical methods the stress resultants derived from
the patchload. Intest examples, the agreement betweenthe results
obtained by this method and those obtained numerically was very
good. Using these new expressions a simplified procedure was
designed to determine the stress resultants produced by the patch
load in slender silos and silos of intermediate slenderness (all AAC3
3 [2]). This procedure allows the stress resultants to be determined
in an entirely analytical manner without the need to take into
account multiple alternatives for the point where the patch load
acts.
Acknowledgements
The authors wish to express their gratitude for partially financ-
ing this work to PRADOvia the researchproject Automatizacin del
clculo de silos de acuerdo a la normativa internacional and Spanish
Plan for Research, Development and Innovation via the research
project Study on Buckling and Dust Explosions in Agricultural Steel
Silos. Simulation with Numerical methods and Experimental Valida-
tion AGL2009-13181-C02-01.
The authors would also like to thank the reviewers for their
helpful and positive comments and, in particular, to one of them,
whose suggestions and amendments have greatly improved this
paper.
References
[1] Janssen H. Versuchber getreidedruck in silozallen (research on pressure in
silos). Z Ver Dtsch Ing 1885;39:10459.
[2] EN 19914. Actions on structures. Silos and tanks. Brussels: CEN. 2006.
[3] DIN 1055. Design loads for buildings and loads in silo bins. DIN 1055 part 6.
Berlin: Deutsches Institut fr Normung. 2005.
[4] ACI 313-97. Standard practice for design and construction of concrete silos and
stacking tubes for storing granular materials. 1997.
[5] ISO DIS 11697. Basis for design of structures: loads due to bulk materials. ISO
draft. International Standard, ISO. 1995.
[6] Platanov PN, Kovtun AP. Davlenie Zerna na stenki silosov elevatorov (grain
pressures over the silowalls). MukomolnoElevatornaia Promyshlennost 1959;
25(12):224.
[7] Pieper K, Wenzel F. Comments on DIN1055: design loads for buildings loads
in silo bins. BetonStahlbetonbau 1963;611.
[8] Sugita M. Flow and pressures of non-cohesive granular materials in funnel
flow bin. Trans ASME 1972;72-MH-20:18.
[9] Reimbert M, Reimbert A. Pressures and overpressures in vertical and
horizontal silos. In: Proc. international conference on design of silos for
strength and flow. Powder Advisory Centre. 1980. p. 4960.
[10] Ramrez A, Nielsen J, Ayuga F. On the use of plate-type normal pressure cells
in silos: part 1: calibration and evaluation. Comput Electron Agric 2010;71(1):
716.
[11] Ramrez A, Nielsen J, Ayuga F. On the use of plate-type normal pressure cells
in silos: part 2: validation for pressure measurements. Comput Electron Agric
2010;71(1):6470.
[12] Ramrez A, Nielsen J, Ayuga F. Pressure measurements in steel silos with
eccentric hoppers. Powder Technol 2010;201(1):720.
[13] Prante V. Messungen des getreidedruckes gegen silowandungen. Z Ver Dtsch
Ing 1896;40:11225.
[14] Jamieson JA. Grain pressures in deep bins. Eng News 1904;51(10):23643.
[15] Reimbert M, Reimbert A. Recherches novelles sur les efforts exercs par le
matieres pulverulentes ensiles sur les parois des silos. Ann Inst Tech Bat
Travaux Publi I 1943;(11):148.
[16] Pieper K, Wagner K. Der Einfluss Verschiedener Auslaufarten auf die
Seitendrucke in Silozellen. Weiesbaden (Germany): Aufbereitungs-Technik;
1969.
[17] Nielsen J. Silotryk (pressures in silos). Ph.D. thesis. Lyngby: Department of
Structural Engineering, Technical University of Denmark; 1972.
[18] Askegaard V, Bergholdt M, Nielsen J. Problems in connection with pressure cell
measurements in silos. Bygningsstatiske Med 1971;42(2):3374.
[19] Munch-Andersen J, Nielsen J. Size effects in large grain silos. Bulk Solids Handl
1986;6:8859.
[20] Munch-Andersen J. The boundary layer in rough silos. In: Proc. of the
2nd international conference on bulk materials, storage, handling and
transportation inst. eng. aust. 1986.
[21] Nielsen J. Load distribution in silos influenced by anistropic grain behaviour.
In: Proc. of the international conference on bulk materials, storage, handling
and transportation, inst. eng. aust. 1983.
[22] Stiglat K. Schden an silos in mannheim. Arbeitstagung Lagern von
Sojaschrot. Lehrstuhl fr Hochbaustatik. Germany: Technische Universitt
Braunschweig; 1976.
[23] DIN 1055-6. Design loads for buildings and loads in silo bins. DIN 1055 part 6.
Berlin: Deutsches Institut fr Normung. 1977.
[24] European comission. Silo and tanks. Outline for part 11. Report for the
eurocode task group on actions on structures. 1987.
[25] Rotter JM, Nielsen J. Steel silo patch pressures in the draft European standard.
Paper presented to CEN TC250/SC1/PT4 actions in Silos and Tanks. Karlsruhe;
September 1996. p. 8.
[26] CEN ENV 19914. Actions on structures. Silos and tanks. Brussels: CEN. 1996.
[27] Nielsen J, Andersen EY. Loads in grain silos. Bygningsstatiske Med 1982;53(4):
12335.
[28] Nielsen J, Kristiansen NO. Related measurements of pressure conditions in full
scale barley silo and in a model silo. In: Proc. of the international conference
on design of silos for strength and flow, University of Lancaster. 1980.
[29] Nielsen J, Kristiansen NO. Pressure measurements on a silo in Karpalund.
Nordic group for silo research. Report no. 5. Department of Structural
Engineering, Technical University of Denmark. 1979 (in Danish).
[30] Ooi J, Pham L, Rotter JM. Systematic and random features of measured
pressures on full-scale silo walls. Eng Struct 1990;12(2):7487.
[31] Ooi JY, Rotter JM. Measured pressures in full scale silos: a new understanding.
In: Proc., International conference: bulk materialstowards the year 2000.
Institution of Mechanical Engineers. 1991. p. 195200.
[32] Ooi JY, Rotter JM, Zhong Z. A statistical evaluation of filling and discharging
pressures on silo walls. In: Proc., 7th world congress on chemical engineering.
2005, p. 10. Abstract p. 337.
[33] Rotter JM. Silo and hopper design for strength. In: McGlinchey D, editor. Bulk
solids handlingequipment selection and operation. Oxford: Blackwell; 2008.
p. 99134 (Chapter 3).
[34] Gillie M, Rotter JM. The effects of patch loads on thin-walled steel silos. Thin-
Walled Struct 2002;40(10):83552.
[35] Song CY, Teng JG. Buckling of circular steel silos subject to code-specified
eccentric discharge pressures. Eng Struct 2003;25(11):1397417.
[36] Reimbert A, Reimbert M. Silos thorie et pratique. France: Editions Eyrolles;
1959.
[37] Ayuga F. Some unresolved problems in the design of steel cylindrical silos.
In: Chen JF, Ooi JY, Teng JG, editors. Structures and granular solids. From
scientific principles to engineering applications. London: Taylor & Francis
Group; 2008.
[38] Rotter JM. Cylindrical shells: unsymmetrical loadings and supports.
In: Brown CJ, Nielsen J, editors. Silos: fundamentals of theory, behaviour
and design. London: E & FN Spon; 1998. p. 36777 (Chapter 16).
[39] Vidal P, Gallego E, Guaita M, Ayuga F. Finite element analysis under different
boundary conditions of the filling of cylindrical steel silos having an eccentric
hopper. J Constr Steel Res 2008;64:48092.
[40] Vidal P, Gallego E, Guaita M, Ayuga F. Simulation of the filling pressures
of cylindrical steel silos with concentric and eccentric hoppers using 3-
dimensional finite element models. Trans ASABE 2006;49(6):188195.
[41] Gallego E, Rombach GA, Neumann F, Ayuga F. Simulations of granular flow
in silos with different finite element programs: ANSYS vs. SILO. Trans ASABE
2010;53(3):81929.
[42] Ooi JY, She KM. Finite element analysis of wall pressure in imperfect silos.
Internat J Solids Struct 1997;34(16):206172.
[43] Keiter TWR, RombachGA. Numerical aspects of FE simulations of granular flow
in silos. J Eng Mech 2001;127(10):104450.
[44] Vidal P, Guaita M, Ayuga F. Analysis of dynamic discharge pressures in
cylindrical slender silos with a flat bottom or with a hopper: comparison with
Eurocode 1. Biosyst Eng 2005;91(3):33548.
[45] Briassoulis D. Finite element analysis of a cylindrical silo shell under
unsymmetrical pressure distributions. Comput Struct 2000;78(13):27181.
[46] Song CY. Effects of patch loads on structural behavior of circular flat-bottomed
steel silos. Thin-Walled Struct 2004;42(11):151942.
[47] Rotter JM. Guide for the economic design of circular metal silos. London: Spon
Press; 2001.
1954 E. Gallego et al. / Engineering Structures 33 (2011) 19401954
[48] Gonzlez-Montellano C, Ramrez A, Gallego E, Ayuga F. On the steel cost of
circular flat-bottomed silos designed using the Eurocodes. Struct Eng Mech
2009;33(5):56172.
[49] EN 1993-1-1. Design of steel structures: general rules and rules for buildings.
Brussels: CEN. 2005.
[50] EN1993-1-3. Design of steel structures: cold-formed thin gauge members and
sheeting. Brussels: CEN. 2006.
[51] EN 1993-1-8. Design of steel structures: design of joints. Brussels: CEN. 2005.
[52] EN 1993-4-1. Design of steel structures: silos. Brussels: CEN. 2007.
[53] ANSYS 10.0. ANSYS users manual. ver. 10.0. Houston: Swanson Analysis
Systems Inc.; 2008.
[54] Rotter JM, Zhang Q, Teng JG. Corrugation collapse in circumferentially
corrugated steel cylinders. In: First national structural engineering conference,
Melbourne, Institution of Engineers. 1987. p. 37783.
[55] Seide P. Small elastic deformations of thin elastic shells. Leyden (Holland):
Noordhoff; 1975.