Bull Earthquake Eng

DOI 10.1007/s10518-012-9370-y
ORIGINAL RESEARCH PAPER
On the evaluation of the horizontal forces produced
by grain-like material inside silos during earthquakes
Stefano Silvestri Giada Gasparini
Tomaso Trombetti Dora Foti
Received: 29 August 2011 / Accepted: 30 July 2012
Springer Science+Business Media B.V. 2012
Abstract This paper presents analytical developments devoted to the evaluation of the
effective behaviour of grain in at-bottom silos during an earthquake. This research work
starts from all the same basic assumptions of Eurocode 8 except for the one regarding the
horizontal shear forces among consecutive grains. Only this difference leads to a new phys-
ically-based evaluation of the effective mass of the grain which horizontally pushes on the
silo walls. The analyses are developed by simulating the earthquake ground motion with time
constant vertical and horizontal accelerations and are carried out by means of simple dynamic
equilibrium equations that take into consideration the specic mutual actions developing in
the ensiled grain. The ndings indicate that, in case of squat silos (characterized by low, but
usual, height/diameter slenderness ratios), the portion of the grain mass that interacts with
the silo walls turns out to be noticeably smaller than the total mass of the grain in the silo
and the effective mass adopted by Eurocode 8.
Keywords Flat-bottom silos Grain-like material Dynamic equilibrium
Friction coefcient Seismic forces
1 Introduction
In the general issue of the actions of grain-like materials on the walls of at-bottom silos
during an earthquake, the assessment of the horizontal interaction is of particular interest.
This interest is based on the possibility of providing more appropriate design rules closer to
the effective seismic behaviour of silos. A careful evaluation of the forces produced by the
material in the silos makes it possible to safely design silos in a seismic area without waste
of material and excessive redundancy.
S. Silvestri G. Gasparini T. Trombetti
Department DICAM, University of Bologna, Viale Risorgimento 2, 40136 Bologna, Italy
D. Foti (B)
Department ICAR, Technical University of Bari, Via Orabona 4, 70125 Bari, Italy
e-mail: d.foti@poliba.it
1 3
Bull Earthquake Eng
1.1 Scientic background and code provisions
The walls of the silos are typically subjected to both normal and shear stresses (normal pres-
sures, vertical and horizontal friction forces) produced by the particulate material inside the
silo. The magnitude and distribution of normal pressures and shear friction forces through
the height of the wall depend on the properties of the content material and whether the silo is
being lled or discharged. As it is the case for most structures, the effect of lateral loads can
be significant especially on larger silos which contain heavier material since the magnitude
of the horizontal seismic load is directly proportional to the weight of the silo (Dogangun
et al. 2009). Moreover, during an earthquake, the particulate material can oscillate inside
the silo so that the lateral loads due to material ow and the lateral seismic loads must be
considered simultaneously.
The static design of silos (Pozzati and Ceccoli 1972) is usually done in accordance with the
established physical idealization suggested by Janssen (1895) and Koenen (1896), that leads
to the identication of the portion of grain mass which interacts with the silo walls in static
conditions. In more detail, Janssen and Koenen identied the above-mentioned interacting
mass in the conservative case which is represented by the full exploitation of the frictional
vertical stresses along the grain-wall contact surface, which enables a portion of the grain
mass to be completely sustained by the walls. This idealization is conservative since to the full
exploitation of the frictional vertical stresses, it corresponds an upper bound for the actions
induced by the grain-like material on the silo walls. The provisions of Eurocode 1 Part 4
(EN 1991-4 2004) are based on this approach.
Similarly, the seismic design of silos is usually performed on the basis of the identication
of an effective mass which interacts with the silo walls under seismic excitation, i.e. which
horizontally pushes on the silo walls. In this respect, Eurocode 8 Part 4 (EN 1998-4 2006)
provides two methods: (i) a simplied method, as given in point (4) of 3.3, and (ii) a more
accurate one, as given in points (5)(12) of 3.3. Basically, the two methods lead to the same
nal dimensioning of the silo walls.
In the simplied method, the horizontal actions produced by grain-like material inside the
silos are evaluated using the following hypotheses:
1. stiff behaviour of the silo and its content (which means that the silo and its content are
subjected to the input ground accelerations; no amplication is taken into account);
2. the grain mass corresponding to the whole content of the silo except the base cone with
an inclination equal to the internal friction angle of the grain (i.e. the effective mass,
namely the 80% of the total mass), is balanced by the horizontal actions provided by
the walls (supposing that the seismic force corresponding to the mass of the base cone
is balanced by friction and therefore does not push against the walls).
In the accurate method, the effects on the silo of the response of the particulate solid to
the horizontal component of the seismic action are represented through an additional normal
pressure on the walls. The corresponding effective mass can be obtained by integration of
the pressure distribution. The effective mass depends on the slenderness (height/diameter)
ratio of the circular silo, but for common values it is in the range 75 % 95 % of the total
mass (the details will be given in Sect. 8). The walls can be then considered innitely stiff or
not. To take into account the actual stiffness and geometrical properties of the containment
structure, use of nite elements is implicitly suggested.
Although there are many studies developed on the behaviour of liquid silos under earth-
quake ground motion (Hamdan 2000; Nachtigall et al. 2003), there are only a few examples
1 3
Bull Earthquake Eng
of scientic investigations focused on the dynamic behaviour of at-bottomgrain silos during
an earthquake.
A numerical study on silos containing granular material and subjected to seismic exci-
tation has been developed to dene the damage occurring in the silos (Wagner et al. 2002).
Numerical simulations have been carried out which include specic models for (i) the granu-
lar material, (ii) the contact area between the material and the silo wall, and (iii) the silo wall
itself, and the distribution of the damage has been evaluated. The increase in pressure after
shaking due to compaction has been determined and the pressure distribution was shown not
to be the same on opposite sides.
A comprehensive numerical investigation (Holler and Meskouris 2006) was performed
about the seismic behaviour of slender and squat silos and the results had been compared to
the Eurocode 8 prescriptions. The numerical model considers many factors: the nonlinearity
of the granular material itself, the effects due to the varying contact between the granular
material and the silo walls, the soil-structure interaction effects due to the large mass of a full
silo. Also results fromshaking-table tests on a scaled prototype have been taken into account.
In conclusion, slender silos show a very good agreement to the Eurocode provisions, while
for squat silos a reduction of the active mass should be assumed.
1.2 Objective of the research work
Because of research deciency in the scope of this subject in recent publications, the current
methodologies for the seismic design of at-bottom silos lled with grain-like materials can
be summarized in the provisions of the Eurocode 8 Part 4. However, these provisions lead
to an overestimation of the seismic actions induced by the grain to the silo walls, as it was
mentioned before by Holler and Meskouris (2006).
In this research work, a theoretical and analytical study on the dynamic behaviour of grain-
like material inside a circular silo has been developed to explain the discrepancies found in
Holler and Meskouris (2006) with the Eurocode provisions. Some results are similar to those
obtained in Wagner et al. (2002) but with a more analytically dened pressure distribution
along the height of the silo walls.
This research work starts from all the same basic assumptions of Eurocode 8 except for
the one relevant to the horizontal shear forces among consecutive grains. Only this difference
leads to a new physically-based evaluation of the effective mass (not equal to the 80% of
the total mass, assumed by Eurocode 8). The effective mass is here evaluated by means of a
physical idealization which (i) is developed consistently with the one identied by Janssen
and Koenen in the static case and (ii) accounts for the pressure variation in the seismic case.
In more detail, the developments presented here, keeping the validity of hypothesis 1 of the
simplied method provided by Eurocode 8 and removing hypothesis 2 by re-calculating the
effective mass, aim at assessing the effects of the horizontal actions on the silo walls due to
the applied accelerations, on the basis of plain dynamic equilibrium considerations. Time-
history dynamic analyses are not carried out. The results obtained show how these horizontal
actions may be far smaller than those that can be obtained by adopting the Eurocode approach
(i.e. using also hypothesis 2), especially for silos characterized by squat geometrical cong-
uration (this is in accordance with the numerical results presented in Holler and Meskouris
2006). A three-dimensional representation of the results in terms of the effective mass that
acts upon the silo walls is also provided, in order to get a better understanding of the physical
concepts related to the results obtained.
1 3
Bull Earthquake Eng
Fig. 1 Geometry of a at-bottom
grain silo and the reference
system adopted: a vertical view;
b plan view
2 Problem formulation
A silo with radius R and lled with grain-like material up to the height H is considered
(Fig. 1). The free surface of the grain is assumed to be horizontal.
With the objective of obtaining an approximate estimation of the pressures produced by the
grain-like material on the silo walls due to the earthquake acceleration, an idealized system
is studied in idealized conditions.
2.1 Idealized system
The following idealized system is considered as representative of the at-bottom silo lled
with grain-like material:
the grain-like material is assumed to be incompressible and compact, without voids,
as it were composed by a number of innitely stiff and innitely resistant spherical balls,
as depicted in Fig. 2;
the silo walls can be assumed either innitely stiff or exible, with respect to the ensiled
grain. This assumption is relevant to the vertical prole of the horizontal acceleration
along the height of the silo and will be deeply discussed later in Sect. 2.2. It should be
mentioned that hypothesis 1 of the simplied method of Eurocode 8 implies innitely
stiff walls.
It is well known that the grain provides forces on the silo walls (Pozzati and Ceccoli 1972).
Figure 2 illustrates the mutual forces of the schematic idealization adopted:
f
V,,GG
is the vertical normal force, perpendicular to the grain surface, which is
exchanged between two consecutive grains;
f
H,,GG
is the horizontal normal force, perpendicular to the grain surface, which is
exchanged between two consecutive grains;
f
H,,GG
is the horizontal tangential force, parallel to the grain surface, which is
exchanged between two consecutive grains;
f
V,,GG
is the vertical tangential force, parallel to the grain surface, which is exchanged
between two consecutive grains;
f
H,,GW
is the horizontal normal force, perpendicular to the grain surface, which is
exchanged between the grain and the silo wall;
1 3
Bull Earthquake Eng
(a)
(b)
Fig. 2 a Idealized system. b Mutual forces exchanged between two adjacent grains, between the grain and
the silo wall, and between the grain and the silo base
f
V,,GW
is the vertical tangential force, parallel to the grain surface, which is exchanged
between a single grain and the silo wall;
f
H,,GB
is the horizontal tangential force, parallel tothe grainsurface, whichis exchanged
between the grain and the silo base;
f
V,,GB
is the vertical normal force, perpendicular to the grain surface, which is
exchanged between the grain and the silo base.
It should be mentioned that horizontal shear forces which are perpendicular to the sheet
plan are not reported in Fig. 2 and either through the whole section in order to simplify the
graphic representation and also notation, while they exist and should be considered in the
analyses (as it will be done later in Sect. 3.6).
It is here assumed that:
the normal forces ( f
V,,GG
, f
H,,GG
, f
H,,GW
and f
V,,GB
) do not have limitations;
the tangential forces are limited by the friction law of the contact surface considered,
i.e.:
f
H,,GG

GG
f
V,,GG
f
V,,GG

GG
f
H,,GG
f
H,,GB

GB
f
V,,GB
f
V,,GW

GW
f
H,,GW
where
GG
,
GB
and
GW
are the friction coefcients related to the contact surfaces of
grain-grain, grain-base and grain-wall, respectively.
It should be mentioned that, in this research work, f
H,,GG
, even if limited by the friction
law, is assumed different from zero. On the contrary, when giving the effective mass equal
to 80% (simplied method) or providing the analytical expression for the additional nor-
mal pressure due to the horizontal seismic action (accurate method), Eurocode 8 implicitly
assumes f
H,,GG
= 0. This implies a lateral sliding behaviour of each grain layer upon the
one below during an earthquake, so that the lateral equilibrium is ensured by the reaction of
the walls which take all the weight of the grain. This is the only basic difference between
the assumptions of the analytical developments reported in this study and the assumptions
of Eurocode 8. This difference will lead to a new evaluation of the effective mass.
1 3
Bull Earthquake Eng
Fig. 3 Representation of the mutual actions which are exchanged between consecutive grains, between the
grain and the silo wall, and between the grain and the silo base
In order to perform an integral evaluation of the global forces that the grain produces on
the silo walls, the grain-like material is treated as a set of overlapped layers of innitesimal
height dz (transition from the discrete approach to the continuous approach). z represents
the distance of a single horizontal layer of grain from the free surface. Within this con-
tinuous approach, the above-mentioned concentrated normal and tangential forces become
distributed normal pressures p and tangential stresses , respectively:
f
V,,GG
becomes p
v,GG
(z);
f
H,,GG
becomes p
h,GG
(z);
f
H,,GG
becomes
h,GG
(z);
f
V,,GG
becomes
v,GG
(z);
f
H,,GW
becomes p
h,GW
(z);
f
V,,GW
becomes
v,GW
(z);
f
H,,GB
becomes
h,GB
(z);
f
V,,GB
becomes p
v,GB
(z).
This notation is illustrated in Fig. 3 and, as far as normal forces are concerned, it is the one
adopted by Eurocode 8.
It is reasonable to assume that the vertical pressures, p
v,GG
(z), tend to diminish from
the core of the grain towards the silo walls where their value is equal to zero (Fig. 4a). The
vertical pressures should be necessarily equal to zero next to the walls, due to the fact that
some grain is sustained by the walls through friction, and not sustained by the underlying
layers of grain. If it were not so, the silo walls could be designed only for their self weight,
without considering any load coming from the grain mass, in both static and seismic cases.
A rst idealize model of the actual distribution of these vertical pressures was proposed
by Janssen (1895) and Koenen (1896). With the purpose of evaluating the effective mass of
grain which leans against the walls and also providing conservative design indications for the
static case, they assumed that the vertical pressures, p
v,GG
(z), at the base of a grain portion at
a generic height z, are equally distributed over the whole surface (Fig. 4b). This model leads
to a conservative estimation of the forces on the walls in that the frictional vertical stresses
along the grain-wall contact surface are fully exploited, whilst the actual frictional stresses
are likely to be lower.
Another idealize model that will be useful for the assessment of the actions induced to
the silo walls by the horizontal accelerations in the seismic case and that lead to conservative
results (with the same concept of the one proposed by Janssen and Koenen) is here intro-
duced. This idealization (Fig. 4c) considers each grain layer as divided into two equivalent
portions composed of (i) grain completely leaning against the layers below (central portion)
and (ii) grain completely sustained by the walls (and therefore characterized by a null verti-
cal pressure between one grain and another). This schematization implies the existence of a
1 3
Bull Earthquake Eng
(a) (b)
(c) (d)
Fig. 4 a Actual distribution of p
v,GG
(z), b the schematization by Janssen and Koenen, c the schematization
proposed in this study, d subdivision of each grain layer: grain completely leaning against the layers below
(internal disk in blue colour) and grain completely sustained by the walls (external torus in red colour)
specic distance s fromthe silo wall, in correspondence of which
v,GG
= 0 so that p
v,GG
=
0 for all the points with a distance less than s from the silo wall (Fig. 4d).
Therefore, a horizontal layer of grain characterized by a height dz and placed at a generic
distance z measured from the free surface is assumed to be divided into two portions:
an internal disk with an diameter of 2r (corresponding to the grain leaning on the layers
below), highlighted in Fig. 5 with blue hatching;
an external torus with an unknown thickness s (corresponding to the grain sustained
by the walls), highlighted in Fig. 5 with red hatching.
The dimensions of the internal disk and the external torus vary by the distance z measured
from the free surface of the grain, as the thickness s of the external torus varies by z.
The internal disk D is characterized by a height dz and a radius r(z) = R s(z) and is
placed at a depth z measured fromthe free surface of the grain, namely at a height h = Hz
from the ground.
The external torus is subdivided into sectors (of circular annulus), each one of them, with
a central angle d, is identied by the central angle measured clockwise from the negative
semi-axis of x, as indicated in Fig. 5. Each sector of the external torus (which in this research
study will be referred to as element E) is characterized by a height dz and a thickness s(z).
A system of auxiliary coordinates (O, ) on the horizontal plane is also dened, where
represents the radial direction (perpendicular to the lateral surface of the silo) and represents
the direction perpendicular to as indicated in Fig. 5.
2.2 Idealized conditions
The above-described idealized system is studied in the following idealized conditions.
The earthquake ground motion is simulated with vertical and horizontal accelerations,
which, in general, are functions of the time t and the height h (or, equivalently, the dis-
tance z from the free surface of grain, as depicted in Fig. 5). As far as the vertical direction
1 3
Bull Earthquake Eng
Fig. 5 External torus (red
hatching) and internal disk (blue
hatching) of the grain. a Vertical
section, b plan view
Fig. 6 Accelerated conditions:
silo subjected to g +a
ev
g and
to a
eh
g
is concerned, the absolute acceleration experienced by the silo and its content is given by
a
v
(t, z) = g + a
ev
(t, z) g, where g is the gravity acceleration and a
ev
(t, z) g is the
additional vertical acceleration due to the earthquake (Fig. 6). In this study, the term addi-
tional means additional with respect to the acceleration of gravity. As far as the horizontal
direction is concerned, the absolute acceleration is given by a
h
(t, z) = a
eh
(t, z) g, where
a
eh
(t, z) g is the horizontal acceleration due to the earthquake (Fig. 6). Both a
ev
and a
eh
are
expressed as fractions of g. For comparative purposes, Eurocode 8 Part 4 3.3 (EN 1998-4
2006) refers to a
eh
(t, z) as parameter (z), which is dened as the ratio of the response
horizontal acceleration of the silo at a vertical distance z from the equivalent surface of the
stored contents, to the acceleration of gravity.
As far as the time variation of the earthquake input is concerned, the earthquake ground
motion is simulated with time constant vertical and horizontal accelerations: a
ev
(t, z) =
a
ev
(z) and a
eh
(t, z) = a
eh
(z). It is clear that this assumption leads to a conservative sim-
plication, given that it is representative of a ctitious single instant of time in which both
the vertical and the horizontal accelerations are supposed to reach contemporarily their peak
values. The same assumption is also adopted by Eurocode 8 Part 4 3.3 (EN 1998-4 2006)
by means of parameter (z), which is not a function of time.
As far as the space variation of the earthquake input (i.e. the acceleration prole along the
height of the silo) is concerned, in general, different trends, such as constant, linear, parabolic
or more complex proles, can be considered, e.g.:
1 3
Bull Earthquake Eng
for the vertical earthquake acceleration:
constant : a
ev
(z) = a
ev0
linear : a
ev
(z) = a
ev0
+a
ev1
(H z)
parabolic : a
ev
(z) = a
ev0
+a
ev1
(H z) +a
ev2
(H z)
2
for the horizontal earthquake acceleration:
constant : a
eh
(z) = a
eh0
linear : a
eh
(z) = a
eh0
+a
eh1
(H z)
parabolic : a
eh
(z) = a
eh0
+a
eh1
(H z) +a
eh2
(H z)
2
where a
ev0
, a
ev1
, a
ev2
, a
eh0
, a
eh1
, and a
eh2
are constant coefcients. By introducing param-
eter (z) as a function of the vertical distance z, Eurocode 8 Part 4 3.3 (EN 1998-4 2006),
accounts for but does not specify anything about the variation of the horizontal acceleration
along the height of the silo. The choice of the vertical proles of both the vertical and the
horizontal earthquake accelerations is strictly related to the dynamic behaviour (i.e. mass
and stiffness) of the system composed by the silo walls and the grain material. If the silo is
assumed to be innitely stiff, no amplication is to be considered and thus spectral acceler-
ations coincide with ground accelerations (i.e. the response acceleration of the silo does not
vary along the height of the silo). If the silo is assumed to be exible, variation along the
height of the silo should be considered for the earthquake accelerations.
In next Sect. 3, a general methodology will be developed in which the most suitable
proles can be introduced. Thus, the analytical developments will be presented in the most
general case of a
ev
(z) and a
eh
(z), considered as generic functions of z.
From a practical point of view, the structural designer can then assume the most suitable
prole on the basis of his/her judgement on the specic problem at hand. As illustrative
examples: (i) if the silo turns out to be very stiff with respect to the horizontal actions, then a
constant prole for the horizontal earthquake acceleration can be reasonably assumed; (ii) if
the silo turns out to be characterized by a predominant shear-type lateral deformation, then
a linear prole for the horizontal earthquake acceleration can be reasonably assumed; (iii) if
the silo turns out to be characterized by a predominant exure-type lateral deformation, then
a parabolic prole for the horizontal earthquake acceleration can be reasonably assumed.
The choice of the most suitable prole of the earthquake accelerations is beyond the goals of
this research work. However, in rst approximation, with specic reference to the horizontal
earthquake acceleration, the following procedure can be suggested:
Step 1: a rst-tentative prole for the horizontal acceleration can be assumed (e.g. a constant
one);
Step 2: the resultant pressure distribution is calculated;
Step 3: the pressure distribution (i.e. the set of horizontal forces) is introduced into a struc-
tural (namely, a nite-element) model of the silo;
Step 4: the displacement prole is obtained;
Step 5: byassumingthat the silodynamic behaviour is basicallygovernedbythe rst mode of
vibration and that the accelerations can be approximately considered proportional to
the displacements, the tentative prole for the horizontal acceleration can be updated
on the basis of the obtained displacement prole;
Step 6: again from Step 2.
This iterative procedure can be stopped as soon as the displacement prole is sufciently
close to the acceleration prole.
1 3
Bull Earthquake Eng
It is also clear that a more accurate evaluation of the proles of the earthquake accelera-
tions and, in general, of the dynamic interaction between the grain and the silo walls can be
obtained only by means of: (i) numerical simulations which account for collisions and con-
tacts of elements (Discrete Element Methods, e.g. see the work by Dvornik and Lazarevic
1999) or (ii) experimental shaking-table tests (as the ones that the authors are currently
developing at the EQUALS centre of Bristol).
2.3 Objective of the analytical developments
The objective of the analytical developments presented in the following sections is to deter-
mine the measure of the thickness of the external torus, in accelerated conditions, by means
of plain equilibrium equations, in order to quantitatively identify the portion of the grain
that leans against the layers below and the one that pushes on the walls (i.e. the effective
mass). Consequently, as derivative results that are basic information regarding the pressure
distributions (namely, normal pressures and tangential stresses) will be also achieved.
3 Dynamic equilibrium in accelerated conditions
The following assumptions are considered:
presence of gravity acceleration g;
presence of time constant vertical acceleration a
ev
(z) towards z (positive upwards);
presence of time constant horizontal acceleration a
eh
(z) towards x (positive to the right);
the inertial forces acting on the internal disk D, due to the horizontal acceleration, are
completely balanced by the resultant of the shear (tangential) stresses produced on the
lower surface of the disk (absence of horizontal sliding of the grain). This assumption
represents the fundamental difference from the Eurocode 8 approach, which implicitly
assumes horizontal sliding of the grain;
the probable negative variation (depression) of the horizontal pressure between element
E and the silo wall, due to the effects of the horizontal acceleration, is to such an extent
that it cannot completely neutralize the horizontal pressure, p
h,GW
(z), between element
E and the silo wall (hypothesis of not neutralizing the pressure). This assumption, which
is also consistent with the Eurocode 8 provisions, implies that at any point on the silo
wall, the sum of the static pressure of the particulate material on the wall and the seismic
action effect must not be less than zero, which physically means that the grain cannot
pull the walls. Also, this assumption is strongly conrmed by the numerical values of
the seismic pressures with respect to the static ones (see section 45).
The direction of the horizontal acceleration (towards x) is rotated by an angle on the hor-
izontal plane compared to the direction (towards ) perpendicular to the external vertical
surface of element E.
As it was mentioned in Sect. 2.1, the analyses are not carried out on the basis of the Janssen
and Koenens hypothesis (Pozzati and Ceccoli 1972), and they are performed by dividing
the grain material into disks and elements, which implies the generation of additional forces
between the silo wall and the grain, in presence of horizontal and vertical earthquake accel-
erations.
With reference to Figs. 7 and 8, the unknown quantities of the problem are:
1. p
v,GG
(z) = vertical pressures acting on disk D;
2. p
h,GG
(z) = horizontal pressures which are exchanged between disk D and element E;
1 3
Bull Earthquake Eng
(a)
(b)
Fig. 7 Vertical longitudinal section: a schematic trend of s(z); b vertical and horizontal actions operating on
disk D and on the symmetrical elements E
Fig. 8 Horizontal cross-section: horizontal actions operating on the symmetrical elements E
1 3
Bull Earthquake Eng
3.
h,GG
(z) = horizontal tangential stresses acting on the surfaces of disk D;
4.
v,GW
(z) = vertical tangential stresses acting on the silo wall;
5. s(z) = thickness of element E;
6. p
h,GW
(z) = horizontal pressures which are exchanged between element E and the silo
wall;
7.
h,GW
(z) = horizontal tangential stresses acting on the silo wall.
The mutual actions exchanged between the grain and the silo walls are assessed in this
study (as it is usually done in seismic analyses where the effects of horizontal accelerations
are evaluated), by the study of free-body diagrams which are representative of the dynamic
equilibrium conditions.
The equations at disposal are:
1. Vertical forces equilibrium of disk D;
2. Pressure ratio relationship between vertical and horizontal pressures in the grain;
3. Horizontal (radial) forces equilibrium of disk D;
4. Friction law for the boundary between element E and the silo wall;
5. Vertical forces equilibrium of element E;
6. Horizontal (radial) forces equilibrium of element E;
7. Horizontal (tangential) forces equilibrium of element E.
Figures 7 and 8 show the mutual actions that disk D, elements E and the external walls of
the silo exchange. It must be noticed that, in addition to the normal pressures and the shear
stresses which are exchanged between the portions of grain and between the grain and the
silo wall, there are the vertical and horizontal forces, which are detailed as follows:
V
D
= self-weight of disk D acting towards z due to the effect of the gravity
acceleration ( is the density of the grain-like material);
a
ev
(z) V
D
= inertial force coming from the centre of the mass of disk D and acting
towards z due to the effect of the vertical acceleration a
ev
(z) (the iner-
tial force is downward, as the acceleration a
ev
has been assumed positive
upwards);
a
eh
(z) V
D
= inertial force coming from the centre of the mass of disk D and acting
towards x, due to the effect of the horizontal acceleration a
eh
(z) (iner-
tial force towards the left as the acceleration a
eh
has been assumed to be
positive towards the right);
V
E
= self-weight of element E acting towards z due to the effect of the gravity
acceleration ( is the density of the grain-like material);
a
ev
(z) V
E
= inertial force coming from the centre of the mass of element E and acting
towards z due to the effect of the vertical acceleration a
ev
(z) (the iner-
tial force is downward, as the acceleration a
ev
has been assumed positive
upwards);
a
eh
(z) V
E
= inertial force coming from the centre of the mass of element E and act-
ing towards x, due to the effect of the horizontal acceleration a
eh
(z)
(the inertial force is towards the left as the acceleration a
eh
has been
assumed to be positive towards the right).
3.1 Vertical forces equilibrium of disk D
Vertical forces equilibrium of disk D provides:
p
v,GG
(z +dz) A
D
= p
v,GG
(z) A
D
+ V
D
+a
ev
(z) V
D
(1)
1 3
Bull Earthquake Eng
where A
D
= r
2
is the area of the top and bottom surfaces of the disk, V
D
= A
D
dz is the
volume of the disk, and clearly p
v,GG
(z +dz) = p
v,GG
(z) + dp
v,GG
. Thus, Eq. (1) leads
to:
dp
v,GG
= (1 +a
ev
(z)) dz (2)
Integrating Eq. (2) gives:
p
v,GG
(z) =
_
(1 +a
ev
(z)) dz +C
1
(3)
where C
1
is a constant of integration that can be obtained imposing the boundary condition
[on the top surface of the grain the vertical pressure is null, i.e. p
v,GG
(z = 0) = 0].
3.2 Pressure ratio relationship between vertical and horizontal pressures in the grain
If is the pressure ratio of the grain-like material, the following relationship holds between
vertical and horizontal pressures inside the grain:
p
h,GG
(z) = p
v,GG
(z) (4)
3.3 Horizontal (radial) forces equilibrium of disk D
Horizontal (radial) forces equilibrium of disk D provides:

h,GG
(z +dz) A
D
= a
eh
(z) V
D
+
h,GG
(z) A
D
(5)
where
h,GG
(z +dz) =
h,GG
(z) +d
h,GG
. Thus, Eq. (5) leads to:
d
h,GG
= a
eh
(z) dz (6)
Integrating Eq. (6) gives:

h,GG
(z) =
_
a
eh
(z) dz +C
2
(7)
where C
2
is a constant of integration that can be obtained imposing the boundary condition
(on the top surface of the grain the shear stress is null, i.e.
h,GG
(z = 0) = 0).
3.4 Friction law for the boundary between element E and the silo wall
If
GW
is the friction coefcient of the grain-wall contact surface, the following relationship
holds between the normal pressures and the vertical shear stresses along the contact surface
between the grain of element E and the silo wall:

v,GW
(z) =
GW
p
h,GW
(z) (8)
3.5 Vertical and horizontal (radial) forces equilibrium of element E
Vertical and horizontal forces equilibriumequations of element E are coupled in the following
system of equations:
_
V
E
(1 +a
ev
(z)) =
v,GW
(z) A
E
p
h,GW
(z) A
E
= a
eh
(z) V
E
+ p
h,GG
(z) A
E
(9)
where:
1 3
Bull Earthquake Eng
A
E
= R d dz is the area of the vertical lateral surface of element E;
V
E
= s(z)
_
R
s(z)
2
_
d dz is the volume of element E;
a
eh
(z) represents the component of the horizontal acceleration a
eh
(z) perpendicular to
the external vertical surface (towards ) of element E (see Fig. 8): a
eh
(z) = a
eh
(z) cos .
Equation (9) may be rewritten as follows:
_
_
_
s(z)
_
R
s(z)
2
_
(1 +a
ev
(z)) =
v,GW
(z) R
_
p
h,GW
(z) p
h,GG
(z)
_
R = a
eh
(z) s(z)
_
R
s(z)
2
_
(10)
Substituting Eq. (8) into the rst of Eq. (10):
_
_
_
s(z)
_
R
s(z)
2
_
(1 +a
ev
(z)) =
GW
p
h,GW
(z) R
_
p
h,GW
(z) p
h,GG
(z)
_
R = a
eh
(z) s(z)
_
R
s(z)
2
_
(11)
After some calculations, this system of equations provides the closed-form expressions of
p
H,GW
(z) and s(z).
As far as the horizontal pressures exerted by the grain on the silo wall are concerned, it is
possible to obtain:
p
h,GW
(z) =
p
h,GG
(z)
1 (z) a
eh
(z) cos
GW
(12)
where (z) =
1
1+a
ev
(z)
. Equation (12) gives the pressures between the grain and silo walls.
As far as the thickness of the portion of grain which is sustained entirely by the silo wall
is concerned, the following quadratic equation in s(z) is obtained:
s(z)
_
1
s(z)
2R
_
=
p
hGG
(z) (z)
GW
(1 (z) a
eh
(z) cos
GW
)
(13)
Assuming (z) =
2 p
hGG
(z)(z)
GW
z (1(z)a
eh
(z) cos
GW
)
, Eq. (13) can be rewritten as follows:
s
2
(z) 2R s(z) + R (z) z = 0 (14)
The two solutions of the last equations are:
s(z) = R
_
R
2
R (z) z (15)
Clearly, the thickness s(z) cannot be larger than the radius R of the silo, so that the only
solution that has a physical meaning is the following (with sign -):
s(z) = R
_
R
2
R (z) z (16)
i.e.
s(z) = R
_
R
2
R
2p
hGG
(z) (z)
GW
(1 (z) a
eh
(z) cos
GW
)
(17)
Equation (19) gives the thickness of the grain layer that leans against the walls.
1 3
Bull Earthquake Eng
3.6 Horizontal (tangential) forces equilibrium of element E
Horizontal (tangential) forces equilibrium of element E provides:

h,GW
(z) A
E
= a
eh//
(z) V
E
(18)
where a
eh//
(z) = a
eh
(z) sin represents the component of the horizontal acceleration a
eh
parallel to the external vertical surface (towards ) of element E (see Fig. 8). Thus, Eq. (18)
leads to:

h,GW
(z) = a
eh
(z) sin s(z)
_
1
s(z)
2R
_
(19)
4 Specialisation to the case of constant vertical proles of both the vertical
and the horizontal earthquake accelerations
In this and all the next sections, the following assumptions are made:
constant vertical acceleration along the height of the silo, i.e. a
ev
(z) = a
ev0
;
constant horizontal acceleration along the height of the silo, i.e. a
eh
(z) = a
eh0
.
Equation (3) specialized as follows:
p
v,GG
(z) = (1 +a
ev0
) z (20)
Equation (4) specialized as follows:
p
h,GG
(z) = (1 +a
ev0
) z (21)
Equation (7) specialized as follows:

h,GG
(z) = a
eh0
z (22)
Equation (12) specialized as follows:
p
h,GW
(z) =
z

0
(1
0
a
eh0
cos
GW
)
(23)
where
0
=
1
1+a
ev0
.
Equation (17) specialized as follows:
s(z) = R
_
R
2
R
2
GW
1
0
a
eh0
cos
GW
z (24)
Equation (19) specialized as follows:

h,GW
(z) =
_
a
eh0
sin
GW
1
0
a
eh0
cos
GW
_
z (25)
5 Graphic representations of the horizontal pressures between the grain
and the silo wall
Equation (12) represents the rst fundamental result of this work. According to the assump-
tions made in the previous sections, the equation provides the horizontal pressures which are
exchanged between the grain material and the silo wall in seismic conditions.
1 3
Bull Earthquake Eng
In case of absence of the horizontal acceleration [a
eh
(z) = a
eh0
= 0], Eq. (12) simplies
as follows:
p
h0
(z) =
z

0
= (1 +a
ev0
) z (26)
The overpressure (or depression) between the grain and the silo wall due to the effects of the
horizontal acceleration can be dened as follows:
p
h
(z, ) = p
h,GW
(z, ) p
h0
(z) (27)
For illustrative purposes, by considering a typical silo dened by the following dimensions
and characteristics: R = 10 m, H = 20 m, = 900 kg/m
3
,
GW
= 0.40, = 0.50, a
eh
=
a
eh0
= 0.40, a
ev
= a
ev0
= 0.15 and in order to visualize the results given by Eqs.(26) and
(27), Figs. 9a, 10a and 11a show the p
h0
plots and Figs. 9b, 10b and 11b show the p
h
plots for significant horizontal sections of the silo at different heights z starting from the free
surface of the grain. In the plots:
the black curve represents the external circumference of the silo with radius R (unit of
measurement [m]),
the blue plot (Figs. 9a, 10a, 11a) represents the diagram of pressures p
h0
for a xed
height z and 0

360

(accelerated conditions with vertical acceleration only, unit

of measurement [N/cm
2
]),
the red plot (Figs. 9b, 10b, 11b) represents the diagram of pressure variations p
h
for a
xed height z and 0

360

(due to horizontal acceleration only, unit of measure-

ment [N/cm
2
]).
It is interesting to notice that, at any height z:
the plot of the horizontal pressures p
h0
is axial-symmetrical, while the plot of the pressure
variations p
h
due to horizontal earthquake acceleration only is not axial-symmetrical;
pressure variations p
h
due to horizontal earthquake acceleration only are very small
(10 %30 % of p
h0
) with respect to horizontal pressures p
h0
under accelerated condi-
tions with vertical acceleration only.
6 Portions of grain relative to the behaviour under accelerated conditions
Equation (17) represents the second fundamental result of this research work. s (z, ) repre-
sents the thickness of the portion of grain which actually pushes on the silo walls in seismic
conditions.
In order to give physical insight into this result and to facilitate the understanding of
the mutual actions which develop between the grain and the silo walls under accelerated con-
ditions, in this section three-dimensional graphic representations are provided of the volumes
of the grain portions which are basically (i) supported by the grain below and (ii) sustained
by the silo walls. Figure 12 provides these graphic representations for a silo characterized by
the same parameters reported in the previous section.
Portion A1 is the amount of grain leaning against the lower portion of the material up
to the silo foundation. This portion of grain does not interact with the silo walls. From a
geometrical point of view, it overlaps with the vertical-axis overturned truncated cone solid
(Fig. 12, blue colour), in which the minor base is the one obtained by drawing the curve
r (z, ) = R s (z, ) for 0

360

on the plane z = H (at the silo foundation) and

the major one is the one conned by the silo circumference (at the top of the grain, i.e. z = 0).
1 3
Bull Earthquake Eng
Fig. 9 Horizontal cross-section of the considered silo at height z = 0.50H
Portion A2 is the amount of grain that is completely sustained by the lateral walls of the
silo. This is the portion of the grain that interacts with the silo walls. From a geometrical
point of view, it coincides with the vertical-axis cylindrical annulus (Fig. 12, red colour),
with a thickness s (z, ) which is variable according to the height z and the angle on the
horizontal plane.
1 3
Bull Earthquake Eng
Fig. 10 Horizontal cross-section of the considered silo at height z = 0.75H
Also, the volumes of portions A1 and A2 (V
A1
and V
A2
) can be calculated:
V
A1
= RH
2
_
_
1

GW
_
1
2
0
a
2
eh0

2
GW
_
_
(28)
1 3
Bull Earthquake Eng
Fig. 11 Horizontal cross-section of the considered silo at height z = 1.00H
V
A2
= RH
2
_
_

GW
_
1
2
0
a
2
eh0

2
GW
_
_
(29)
1 3
Bull Earthquake Eng
Fig. 12 Three-dimensional views of portion A1 (in blue) and of portion A2 (in red) of the at-bottom grain
silo: a overview and b sectioned view
As an illustrative example, for a silo (i) characterized by a height equal to diameter, containing
granular material with a pressure ratio = 0.5, a grain-wall friction coefcient
GW
= 0.37,
and (ii) subjected to earthquake accelerations characterized by a
eh0
= 0.30 and a
ev0
= 0.10,
the portion of the grain mass interacting with the silo walls proves to be equal to the 19% of
the total grain mass. This amount is far smaller than the 80% considered by the simplied
method of Eurocode 8 and than the amount (75 % 95 %) that can be obtained using the
accurate method of Eurocode 8.
1 3
Bull Earthquake Eng
7 Limits of validity of the proposed theory
The proposed theory has some limitations which are related to the mathematical definition
of some physical quantities:
it is necessary that portion A1 exists. This physical condition is rendered into the follow-
ing mathematical limitation:
s (, z = H) < R, (30)
which, taking into account Eq. (16), leads to:
_
R
2
R (, z = H) H > 0, (31)
and then to:
R > (, z = H) H, (32)
The maximum value of (, z = H) occurs for = 0. Thus, the condition given by
Eq. (32) requires that the slenderness ratio, =
H
2R
, should be:
H
2R
<
1
2 ( = 0, z = H)
=
1
0
a
eh0

GW
4
GW
(33)
otherwise portion A1 cannot exist and the theory proposed in this study cannot be applied;
it is also necessary that portion A2 exists. This physical condition translates into the
following mathematical limitation:
s (, z) = R
_
R
2
R (, z) z > 0, , z (34)
that requires:
R >
_
R
2
R (, z) z, (35)
After some calculations, the condition expressed by Eq. (35) becomes:
a
eh0
<
1

GW
; (36)
it is also fundamental that the square root in Eq. (16) exists. This leads to the following
condition:
a
eh0
<
_
1
2
GW
H
R
_
1

0

GW
; (37)
by referring to Eq. (23), in order not to have innite values of the horizontal normal
pressures, it is necessary that:
1
0
a
eh0
cos
GW
= 0, (38)
which gives:
a
eh0
=
1

0
cos
GW
, (39)
However, the condition expressed by Eq. (39) is already encompassed in the condition
expressed by Eq. (36);
1 3
Bull Earthquake Eng
nally, in order to prevent horizontal sliding of the grain on the foundation and to guar-
antee that the inertial forces acting on internal disk D due to the horizontal acceleration
are completely balanced by the resultant of the shear stresses developing at the disk
foundation, it is necessary that the horizontal acceleration is lower than the following
limiting value:
a
eh0
(1 |a
ev0
|)
GB
. (40)
It should be noted that all these limitations identify a class of circular at-bottom silos,
for which the theory proposed in this paper can be applied.
8 Methodology proposed for the assessment of the seismic action on at-bottom grain
silos
It can be reasonably assumed that structures characterized by high values of the vertical
and horizontal stiffnesses, such as silos, do not amplify or diminish the acceleration induced
by the earthquake ground motion at their base. They are therefore subjected to the stresses
caused by the inertial forces that arise due to the accelerations at the base that, during the
seismic events, vary continuously by time. It is clear that the highest stresses induced by the
seismic action are those that derive from the peak ground acceleration.
Therefore, the seismic design of silos can be based on the assumption that the action
induced by an earthquake ground motion is modelled as a couple of horizontal and vertical
accelerations, a
eh0
g and a
ev0
g, representative of the maximum shaking of the ground.
By approximating the seismic action on the silos as two time constant, horizontal and
vertical accelerations equal to a
eh0
g and a
ev0
g, their effect can be evaluated by referring
to the results obtained from the analytical elaborations described in the previous sections.
It must be mentioned that the actions developing on the silo walls are not axial-symmetrical.
The base shear (at the bottom of the silo walls) is given by the integral, on the lateral
surface of the grain, of the projection of the additional pressures towards x (namely, along
the direction of the horizontal acceleration). Skipping all the analytical passages, the base
shear is given by:
T = a
eh0
RH
2
_

GW
1
2
0
a
2
eh0

2
GW
_
(41)
The simplied method of Eurocode 8 for circular at-bottom silos would give:
T
EC8,si mpli f i ed
= a
eh0
R
2
H 0.8 (42)
The accurate method of Eurocode 8 for circular at-bottom silos would give:
T
EC8
= a
eh0
R
2
H
_
1
R
6H
_
(43)
By eliminating the acceleration term, from the base shear formula, then the rest is the effec-
tive mass. Equation (43) indicates that, in the accurate method of Eurocode 8, the effective
mass is represented by the factor
_
1
R
6H
_
as percentage of the total mass of the grain.
As illustrative examples, if H = 4R (slenderness ratio = 2) then the effective mass is
96% of the total mass; if H = 2R ( = 1) then the effective mass is 92% of the total mass,
and if H = R( = 0.5) then the effective mass is 83% of the total mass. These values are
similar to the 80% value adopted in the simplied method of Eurocode 8.
1 3
Bull Earthquake Eng
The bending moment at the bottom of the silo walls is given by the integral, on the lateral
surface of the silo, of the projection of the additional pressures towards x (namely, along the
direction of the horizontal acceleration) multiplied by their height from the silo foundation.
Skipping all the analytical passages, the base bending moment is given by:
M =
1
3
a
eh0
RH
3
_

GW
1
2
0
a
2
eh0

2
GW
_
(44)
The simplied method of Eurocode 8 would give:
M
EC8,si mpli f i ed
= a
eh0
R
2
H 0.8
H
2
(45)
The accurate method of Eurocode 8 would give:
M
EC8
= a
eh0

R
2
2

_
H
2

R
2
27
_
(46)
For the immediate assessment of the benets that the methodology applied in this study
engenders with respect to Eurocode 8 calculations, then it is appropriate to dene the follow-
ing ratios between: (i) the base shear obtained from the formulation presented and the one
obtained by using the accurate method of Eurocode 8, (ii) the bending moment obtained from
the formulation presented and the one obtained by using the accurate method of Eurocode 8:

T
=
T
T
EC8
=
H
2
_

GW
1
2
0
a
2
eh0

2
GW
_
R
_
H
R
6
_ =
2
_

GW
1
2
0
a
2
eh0

2
GW
_
_
1
1
12
_ (47)

M
=
M
M
EC8
=
2
3
H
3
_

GW
1
2
0
a
2
eh0

2
GW
_
R
_
H
2

R
2
27
_ =
4
3

_

GW
1
2
0
a
2
eh0

2
GW
_
_
1
1
108
2
_ (48)
where =
H
2R
is the slenderness ratio. Note that the
T
ratio can also be seen as the ratio
between the effective mass obtained from the formulation presented and the one obtained
using the accurate method of Eurocode 8.
For illustrative purposes, Fig. 13 reports the plots of
T
and
M
ratios as functions of the
slenderness ratio =
H
2R
, for some specic values of the parameters:
GB
= 0.40, =
0.50, a
eh0
= 0.15, 0.25, 0.35, a
ev0
= 0.10 and
GW
= 0.25, 0.40, 0.50, 0.60.
A meticulous exam and inspection of the plots reported in Fig. 13 indicate that the differ-
ence between the results of the proposed methodology and those of the Eurocode 8 method is
maximum (i.e. low values of the ratios) for squat silos, characterized by lower values of the
slenderness ratio. It is possible to notice that both
T
and
M
ratios increase linearly, except
for very small and not feasible values of =
H
2R
close to either
1
12
or
1

108
[coming out
from the analytical expressions reported in Eqs. (47) and (48)]. This also conrms the results
obtained by Holler and Meskouris (2006). As illustrative example, for squat silos character-
ized by height smaller than diameter ( < 1), considerable reductions of the effective mass
can be obtained: 30 %70 % (mainly depending on the value of the friction coefcient

GW
). Also:
the advantages of the theory presented here are larger for the bending moment, rather
than for the shear force;
1 3
Bull Earthquake Eng
Fig. 13
T
and
M
ratios as function of the silo slenderness ratio = H/(2R) for different values of the
parameters
GW
and for a, b a
eh0
= 0.15; c, d a
eh0
= 0.25; e, f a
eh0
= 0.35
1 3
Bull Earthquake Eng
among the parameters which govern the behaviour of silos, the most important one seems
to be the grain-wall friction coefcient
GW
. The benecial effect in terms of reductions
of base shear (i.e. effective mass) and base bending moment increases if
GW
decreases;
again, the benecial effect in terms of reductions of base shear and base bending moment
increases with decreasing values of ;
the values of the earthquake accelerations (both horizontal and vertical) do not inuence
so much the trend of the
T
and
M
ratios.
9 Conclusions
In this paper, the actions of grain on the walls of circular at-bottom silos during earthquake
ground motions have been studied analytically.
This research work starts from all the same basic assumptions of Eurocode 8 except the
one regarding the horizontal tangential forces between consecutive grains. This difference
leads to a newphysically-based evaluation of the effective mass of grain, which pushes on the
silo walls. The effective mass is evaluated in this study by means of a physical idealization
which (i) is developed consistently with the conservative one identied by Janssen (1895)
and Koenen (1896) in the static case and (ii) accounts for the pressure variation in the seismic
case.
The results obtained show how, even if still evaluated in a conservative way, the effective
mass may be far less than the one given by Eurocode 8. Thus the horizontal actions may
be far less with respect to those that can be obtained by applying the Eurocode provisions,
especially for silos characterized by squat geometrical conguration. This is in accordance
with the few numerical results presented in the scientic literature (Holler and Meskouris
2006). In more detail, the results indicate that, in case of squat silos characterized by low but
common height/diameter slenderness ratios, the portion of grain mass interacting with the
silo walls proves to be noticeably lower than the one obtained using Eurocode 8. Basically, for
squat silos characterized by height smaller than diameter ( < 1), considerable reductions
of the effective mass can be obtained: 30 % 70 % (mainly depending on the friction
coefcient
GW
).
The limits of validity of the analytical developments allow for the identication of a class
of circular at-bottom grain silos, for which the theory proposed in this paper can be applied.
Acknowledgments Dr. Luca Celeghini of Petkus Technologie GmbH (Eisenacher Str. 42, D-99848 Wutha-
Farnroda, Germany) is acknowledged for the idea which inspired most of the analytical developments and for
providing useful material. The authors are also grateful to Dr. Francesco Fabbri and to Dr. Giulia Lavatura for
their contributions relevant to the analytical and numerical developments focused on the design of a shaking-
table test for the experimental verication of the methodology proposed, which is currently under progress at
the EQUALS centre of Bristol, within a European SERIES project.
References
Dogangun A, Karaca Z, Durmus A, Sezen H (2009) Cause of damage and failures in silo structures. J Perform
Constr Facil ASCE 23(2):6571
Dvornik J, Lazarevic D(1999) Loading of granular material on silo walls. In: Proceedings of the 1st conference
on applied mathematics and computation, Dubrovnik, Croatia, 1318 Sept 1999, pp 215222
EN 1991-4 (2004) Eurocode 1. Basis of design and action on structures. Part 4. Actions in silos and tanks.
CEN, Brussels
EN1998-4 (2006) Eurocode 8. Design of structures for earthquake resistance. Part 4. Silos, tanks and pipelines.
CEN, Brussels
1 3
Bull Earthquake Eng
Hamdan FH (2000) Seismic behaviour of cylindrical steel liquid storage tanks. J Constr Steel Res 53:307333
Holler S, Meskouris K (2006) Granular material silos under dynamic excitation: numerical simulation and
experimental validation. J Struct Eng ASCE 132(10):15731579
Janssen HA (1895) Versuche ber Getreidedruck in Silozellen. Zeitschrift des Vereines deutcher Ingenieure.
39:10451049 (in German)
Koenen M (1896) Berechnung des Seiten und Bodendrucks in Silozellen. Centralblatt der Bauverwaltung
16:446449
Nachtigall I, Gebbeken N, Urrutia-Galicia JL (2003) On the analysis of vertical circular cylindrical tanks
under earthquake excitation at its base. Eng Struct 25:201213
Pozzati P, Ceccoli C (1972) Teoria e Tecnica delle Strutture. Volume Primo, Preliminari e Fondamenti. UTET,
Torino
Wagner R, Noh S-Y, Butenweg C, Meskouris K (2002) Seismic excited granular material silos. In:
Grundmann H, Schueller GI (eds) Proceedings of Eurodyn 2002, pp 253258. ISBN:90 5809 510X
1 3