ORAL

SPC-07: SILOS
AND
GRANULAR
MATERIALS
(SIGMA)
C-0143
EFFECTIVE TRANSFER CHUTE DESIGN INCLU-
DING DUST CONTROL FOR HANDLING GRAINS
AND OTHER PRODUCTS
Presenter: Tim Donohue
Australia
Authors: Tim Donohue
1
, Alan Roberts
1
, Craig Wheeler
2
, Dusan
Ilic
1
, Andre Katterfeld
3
1
TUNRA Bulk Solids University of Newcastle
2
University of Newcastle TUNRA Bulk Solids
3
University of Magdeburg Institute of Logistics and Material
Handling

C-0366
VERTICAL LOAD ON DISC EMBEDDED IN STATIC
AND FLOWING GRAIN: EXPERIMENTAL AND
DEM SIMULATIONS
Presenter: Rafal Kobylka
Poland
Authors: Rafal Kobylka
1
, Marek Molenda
2
1
Institute of Agrophysics Polish Academy of Sciences
2
Department of Physical and Technological Properties of
Agromaterials Institute of Agrophysics Polish Academy of
Sciences

C-0371
DUST EXPLOSION RISK DURING GRAPHITE-MO-
DERATED REACTORS DECOMMISSIONING
Presenter: Sophie Trelat
France
Authors: Sophie Trelat
Institute for Radiological Protection and Nuclear Safety
(IRSN), B.P. 17 92262 Fontenay-auxRoses Cedex France

C-0450
AN ARCH-BASED MODEL FOR PREDICTING DY-
NAMIC PRESSURES DURING SILO DISCHARGE
Presenter: Qiang Zhang
Canada
Authors: Xuduo Cheng
1
, Qiang Zhang
2
1
School of Food Science and Engineering Nanjing University
of Finance and Economics
2
Biosystems Engineering, University of Manitoba

C-0461
ANALYSIS OF THE PRESSURE EFECTS IN CORN
DUST EXPLOSION USING COMPUTACIONAL
FLUID DYNAMIC
Presenter: Francisco Carlos Gomes
brazil
Authors: Janana de Andrade Silva
1
, Francisco Carlos Gomes
1
,
Jefferson Luiz Gomes Correa
2
1
Engineering Departament Federal University of Lavras
2
Department of food engineering Federal University of Lavras

C-0488
A NOTE ON LOAD COMBINATIONS FOR SILOS
Presenter: Jorgen Nielsen
Denmark
Authors: Jorgen Nielsen
1
, Jorgen Nielsen
1
, J. Michael Rotter
2
,
John Dalsgaard Sorensen
3
1
Danish Building Research Institute Aalborg University
2
School of Engineering The University of Edinburgh
3
Department of Civil Engineering Aalborg University

C-0823
DETERMINATION OF THE MECHANICAL PRO-
PERTIES OF BIOMASS PELLETS FOR USE IN DEM
SIMULATIONS
Presenter: Francisco Ayuga
Spain
Authors: Carlos Gonzlez-Montellano, lvaro Ramrez-Gmez,
Jos Mara Fuentes, Eutiquio Gallego, Francisco Ayuga
Bipree research group Universidad Politecnica de Madrid

C-1036
TEST SILO FOR DUST EXPLOSIONS
Presenter: lvaro Ramrez-Gmez
Spain
Authors: lvaro Ramrez-Gmez
3
, Alberto Tascn
1
, Pedro Jos
Aguado
2
, Francisco Ayuga
3
3
BIPREE Research Group Universidad Politcnica de Madrid
1
Agricultura y Alimentacin Universidad de la Rioja
2
Departamento de Ingeniera y Ciencias Agrarias Universidad
de Len

C-1065
SIMULATIONS OF DUST EXPLOSIONS IN A 16 M3
SILO
Presenter: Alberto Tascn
Spain
Authors: Alberto Tascn
1
, ngel Ruiz
2
, ngel Couto
2
, Pedro
Jos Aguado
2
1
Agricultura y Alimentacin University of La Rioja
2
Ingeniera y Ciencias Agrarias University of Len

C-1145
A STUDY ON THE DISCHARGE RATE OF PELLETS
STORED IN SILOS: DISCRETE ELEMENT ANALYSIS
Presenter: FRANCISCO AYUGA
Spain
Authors: Carlos Gonzlez-Montellano
1
, Eutiquio Gallego
1
, l-
varo Ramrez-Gmez
1
, Jos Mara Fuentes
1
, Francisco Ayuga
2
1
Bipree Research Group Universidad Politcnica de Madrid
2
Universidad Politcnica de Madrid

C-1406
EXPERIMENTAL STUDY OF THE PRESSURES
EXERTED BY THE WHEAT STORED IN SLENDER
CYLINDRICAL SILOS. COMPARISON WITH EURO-
CODE 1 PART 4.
Presenter: Pedro Aguado Rodrguez
Spain
Authors: Angel Couto
1
, Pedro Aguado Rodrguez
1
, Angel Ruiz
1
,
Alberto Tascn
2
1
ESTI Agraria Av. Portugal 41 24071 LEON (LEON)
2
Agricultura y Alimentacin University of La Rioja Madre de
Dios 51, 26006 Logroo (Spain)

C-1499
DISCRETE ELEMENT MODELING OF COHESIVE
GRAIN MATERIAL
Presenter: Itzhak Shmulevich
Israel
Authors: zvi asaf, Itzhak Shmulevich, Ramy Catav
Civil & Env. Eng. Technion 1 32000 Haifa Civil & Env. Eng.
Technion

C-1578
THE HOUSEHOLD METAL SILO: A HELPFUL TECH-
NOLOGY FOR FOOD SECURITY
Presenter: Danilo Meja Loro
Italy
Authors: Danilo Meja Loro
FAO- Va/Terme Di Caracalla, Rome

C-1629
DISTRETE ELEMENT SIMULATION OF BOTTLE-
NECK FLOW WITH OBSTACLE
Presenter: Fernando Alonso-Marroquin
Australia
Authors: Fernando Alonso-Marroquin
1
, Charmila Sathianan-
dan
2
1
The University of Sydney
2
School of Civil Engineering The University of Sydney

C-1894
EXPLOSION PROTECTION IN GRAIN HANDLING
FACILITIES: FROM COUNT MOROZZO TO COMPU-
TATIONAL FLUID DYNAMICS.
Presenter: Trygve Skjold
Norway
Authors: Trygve Skjold
GexCon AS

C-2002
SAFE STORAGE TIME OF CEREAL GRAIN SUM-
MARIZED IN A MATHEMATICAL MODEL
Presenter: Nils Jonsson Erik
Sweden
Authors: Nils Jonsson Erik
1
, Pernilla Johnsson
2
, Aldo Ritzzo
3
,
Monica Olsen
4
, Leif Gustafsson
5
1
JTI - Swedish Institute of Agricultural and Environmental
Engineering
2
Department of Food Science Swedish University of Agricul-
tural Sciences
3
Department of Chemistry Veterinary and Food Research
Institute
4
Risk Beneft Assessment Department National Food Admi-
nistration
5
Department of Technology ngstrmslaboratoriet, Uppsala
University

C-2113
VIBRATION INDUCED GRANULAR FLOW
THROUGH AN ELBOW
Presenter: Ivn Snchez
Venezuela
Authors: Ivn Snchez
1
, Ramn Darias
2
, Oliver Pozo
3
1
Centro de Fsica, IVIC
2
Fsica Universidad Simn Bolvar
3
Institute for Polymers and Composites/I3N University of
Minho
1

Review of Silo Loadings Associated with the Storage of Bulk
Granular Materials

Alan W. Roberts

Emeritus Professor, TUNRA Bulk Solids Handling Research Associates
The University of Newcastle, NSW, Australia
Email: Alan.Roberts@newcastle.edu.au

Abstract

This paper presents an overview of developments in silo wall load analysis and design over
the period commencing with the 1880s to the present. The pioneering work of Janssen
(1995) in establishing the basic theory of silo wall loads under static filling conditions and the
important contributions by Jamieson (1902-1903) in identifying the overpressures during
symmetrical discharge and the non-symmetry of the wall loads during eccentric discharge
are highlighted. The work of Jenike et al (1961-1973) which commenced in the latter 1950s
is of particular note. The identification of the basic flow patterns of discharging bulk solids
and the linking of these flow patterns to specific bin and silo geometries is reviewed. The
associated bin wall loads that are linked to the flow patterns is discussed. Particular mention
is made of the complex loading problems that are associated with eccentric discharge and
multi outlet bins. The problems of grain swelling due to moisture increases and silo wall
expansion and contraction that accompany temperature variations are outlined. The
application of anti-dynamic or tremmie tubes for controlling wall loads in tall grain silos is
illustrated. The subject of dynamic loads in silos, in particular, the phenomenon of silo
quaking is reviewed. Also discussed is the case of rapid discharge of bulk solids which gives
rise to impact loads which must be considered in the design of bin an its support structure.
Brief mention is made of the loads on structural members that are buried in bins, bulk
storage sheds and stockpiles.

Key Words: Silo loads; flow patterns; grain moisture effects; silo quaking; buried structures

1. Introduction

The design of silos for the storage of bulk granular materials and other bulk products has
been the focus of research for more than 100 years with the early publications appearing in
the literature spanning the period from the latter 1800s to the early 1900s. Notable is the
well known work of the German engineer, H.A. Janssen (1895)[6], but also noteworthy is the
lesser known work of other researchers notably that of the Canadian engineer J.A. Jamieson
(1903-1904)[4,5].

While much of the early work focussed on free-flowing, non-cohesive granular materials, the
recognition of the vast array of cohesive bulk materials handled by industry gave rise to a
much wider focus of the research into silo loads. The research of Jenike et al (1961-73)[7-
12] and Johanson (1064)[13], which began in the late 1950s, represents a major
contribution. The identification of the flow patterns of mass-flow and funnel-flow that are a
function of the silo geometry, and the way these flow modes are influenced by the flow
properties of the bulk material, has had a significant influence on silo design. The
recognition of the role these flow patterns play in controlling the loadings on silo walls has
been the subject of ongoing research over the past 60 years. Despite this, the determination
of silo wall loads remains a subject of considerable complexity as is evidenced by the
number of failures that have occurred.

The complexity of the loadings includes such influences and factors as
2

- Silo geometry in relation to the bulk material flow pattern whether, for example,
mass-flow, funnel-flow or expanded-flow
- Eccentricity of loading and unloading
- Moisture variations in the stored product and, in particular regarding agricultural
grains, the swelling effect due to increases in moisture within contained storage
masses in silos
- Silo expansion and contraction on a night-to-day basis and, in the case of long term
storage, on a seasonal basis summer through to winter.
- Non-uniformity of silo wall loads due to non-uniformity of the wall expansion and
contraction with respect to the aspect of the silo in relation to the position of the sun
- Dynamic effects, such as the natural occurrence of pulsating flows in granular
discharge from silos that give rise to the phenomenon of silo quaking
- The influence of significant draw-down loads that can occur on structural members
buried in stored granular masses

The purpose of this paper is to present a brief overview of the subject of silo wall loads
illustrating some of the foregoing aspects with case study examples. Some emphasis is
given to the subject of silo quaking to explain how and why it occurs and how it can be
predicted and accounted for at the silo design stage so as to ensure it will not be
experienced in practice.

2. Early Silo Research

While silos have been in existence for many centuries, the known published research into
silo loads was performed over the period of some 30 years commencing in the early 1880s.
A review of this early silo load research has been published by Roberts (1995,1999)[15,18].
The most widely known work in the early period of silo research is that due to the German
Engineer, H.A. Janssen (1895)[6]. This work is significant in that it recognised some
fundamental aspects of internal and boundary friction which limit the magnitude of the loads
on silo floors and walls. Janssens test apparatus is shown in Figure 1.


Figure 1. Model Bin Silo used by Janssen

By comparison, little is known of the work of the Canadian Engineer, J.A. Jamieson (1903,
1904)[4,5] who conducted experiments using the test rig of Figure 2. Jamiesons
contributions are twofold. Firstly, during symmetrical discharge he showed that the wall
pressures increased above the filling pressures. Secondly, and even more significantly, he
examined eccentric discharge and showed that the wall loads on the side nearest the
discharge outlet are lower than those for symmetrical discharge, but greater on the opposite
side. His test results are shown in Figure 3. Thus he demonstrated the non-symmetry of the
3

wall pressures. Had his research been more widely known, some of the silo failures that
occurred 80 or more years later may have been avoided.


Figure 2. Model Bin used by Jamieson Figure 3. Pressures during Fill and
Eccentric Discharge

3. Silo Load Research over the Past 50 Years

Following the foundation work of Jenike, the study of bin wall loads gained new impetus
(Jenike et al (1968-1973)[9-12]. With the flow modes clearly defined and the advantages of
mass-flow being identified, the need for determining the wall loadings in mass-flow bins
became a necessity. The stress fields associated with mass-flow, together with the
corresponding normal wall pressures are shown in Figure 4.


Figure 4. Pressures Acting in Mass-Flow Bins

When a bin is initially filled from the empty condition, a peaked or active stress field occurs
as in Figure 4(a), the major principal stress being almost vertical. When flow occurs, the
support offered by the gate at the outlet is removed and the stress field in the hopper
switches to an arch or passive stress field, the switch travelling up the hopper becoming
locked in at the transition as in Figure 4(b). In the arched or passive stress field, the load is

Peaked or
Active Stress
Field
Arched or Passi ve
Stress Field
o
o
o
o
1
2
1
2
Gate
Cl osed
(a) Initial Filling
(b) Flow (c) Normal Wall Pressure
o
1
Ini ti al Fi ll i ng
Flow
Radial
Stress
Field
Swi tch
Pressure
p
p
p p
nh
nh
n
n
z
z
h
H
h
s
s
h
o
B
D
z
z
v
v
v h
v h
p
p
p
p
y
y
Ef f ectiv e Datum Actual Datum
c
g
z
h
h
h
4

transmitted to the hopper walls with the major principal stress acting more in a horizontal
direction tangential to each arch as in Figure 4(b). Most of the surcharge load due to the
cylinder is carried by the upper section of the hopper giving rise to the high switch stress.
Above the hopper, that is in the cylinder, the peaked or active stress field remains, although
imperfections in the cylinder wall which result in localised flow convergences cause over-
pressures to occur in the cylinder. Imperfections in bin walls, which lead to over-pressures,
may be due to manufacturing and/or constructional details such as weld projections or plate
shrinkages in the case of steel bins, or deformation of formwork in the case of concrete bins.
Jenike et al used strain-energy methods to analyse these over-pressures during flow in the
cylinder. To simplify the computations, design codes generally apply over-pressure factors
to the filling pressures computed by Janssens equation to account for flow conditions in the
cylinder. This is indicated by the upper bound p
n
curve for the cylinder in Figure 4(c). It is to
be noted that when the bin discharges and the flow is stopped, the stress fields and
corresponding pressures shown in Figure 4(c) remain unchanged. The passive stress field
does not revert to that of Figure 4(a). This only occurs if the bin is completely emptied and
then filled again.

A better understanding of the characteristics of funnel-flow and the definition of the effective
transition provided the scope for formalising the computation of wall loads in funnel-flow
bins. There was the realisation that bin wall loads are directly related to the flow pattern
developed during discharge, and this led to the conclusion that, wherever possible, bin
shapes should be kept as simple as possible. While symmetry of the flow channel is seen
as a desirable goal, from a practical point of view, it is virtually impossible to guarantee
symmetrical loading. For instance the filling of the bin needs to be exactly central which,
from a practical point of view, is unlikely. Secondly, the interfacing of the hopper with the
feeder may skew the flow pattern.

The need for ongoing research into bin wall loads had also been encouraged, to a
significant extent, by an increase in the number of reported bin and silo failures. As a result,
there was a pressing need to revise existing bin load codes and to develop new codes in
countries where such codes have not previously existed. The Australian Standard AS3774-
(1996)[1] is one example of the latter. This Standard is quite comprehensive, addressing a
wide range of silo loading conditions including eccentric loads due to non-symmetrical flow
patterns. The new Eurocode (2005)[3] covers the subject of silo loadings in great detail.

4. Loadings Associated with Multi Outlets Bins

Loading conditions in multi outlet bins are quite complex with design misjudgements having
been made on some occasions in the past by assuming uniform, symmetrical loading of the
bin walls. An example of this is the coal bin shown in Figure 5. The bin, constructed of
concrete, is 15m internal diameter with a 25m storage height. The discharge is via seven
vibratory feeders linked to the central reclaim belt conveyor. The walls were designed solely
on the basis of hoop stress with single reinforcing located at the centre of the 250mm wall
thickness. There was no allowance for bending. As Figure 5 shows, the wall pressures in
the region of the outlets are proportional to the flow channel diameters, whereas the
pressures in the stationary regions between the outlets are proportional to the bin diameter.
As a result of the bending in the walls, severe cracking occurred within the first months of
operation. This necessitated the installation of steel cables to strengthen the bin walls as
shown.

The failure of the raw sugar storage bin shown in Figure 6 occurred within the 10 days of
being filled for the first time. The bin was 18m diameter and had 2 rows of 4 outlets on
opposite sides of the centre line. Again, the non-uniformity of the wall loads contributed to
the failure.
5




Figure 5. Cracking Problem in Multi-Outlet Coal Silo




Figure 6. Failure of 8-Outlet Raw Sugar Bin

5. Influence of Moisture and Temperature Variations on Silo Loads

Significant increases in wall loads may occur in grain silos due to increases in grain moisture
content in combination with thermal expansion and contraction (Roberts (1998))[17]. An
example of such loadings have been identified as a contributing factor to the failure and
collapse of the steel wheat storage silo in Australia as shown in Figure 7. Wheat had been
stored in the silo from mid summer to winter, the collapse occurring early morning when the
atmospheric temperature was at its lowest. While the failure may be attributed to the thermal
contraction of the silo walls, there was a period when the grain conditioning aeration fans
were left running during a period of high humidity. This resulted in the moisture content of
the wheat in the lower region of the silo to increase from 9% to 13%. Laboratory tests using
a triaxial cell showed that a 4% increase of moisture content from 9% to 13% resulted in an
increase in the lateral pressure of the wheat of nine times. Grain swelling that accompanies
such a moisture increase results in a reversal of the direction of the wall shear stress. This
leads to an exponential increase in lateral pressure as shown in Figure 8.
Conveyor
Feeders
Low Pressure
High Pressure
Plan View of Bin
Bending of Wall
Flow Channels
18 m
6

Wall loads in storage bins for bulk solids are subject to significant variations in wall loads due
to thermal expansion and contraction. Measurements of wall pressures have shown that
pressures reduce during the day as a bin expands and increase during the night when the
ambient temperature reduces and the bin contracts. However, complete contraction of the
bin may not occur owing to the resistance of the stored bulk solid which has settled during
the expansion phase. Hence, wall loads may not only be influenced by daily temperature
fluctuations, but may also be cumulative over extended periods of undisturbed storage. The
influence of wall load increase due to expansion and contraction is likely to be more
pronounced in the case of grain or relatively free flowing material which can settle more
readily. In the case of grain, the influence can be of considerable significance when coupled
with the effects of moisture increases as discussed above.



Figure 7. Collapsed Grain Silo Figure 8. Wall Pressures under both Normal
and Reverse Friction

When considering expansion and contraction of silos, it is often the case that only the lateral
or radial expansion and contraction are considered. In addition to lateral expansion a bin or
silo will expand upwardly in the longitudinal direction, the amount of longitudinal expansion,
for most bin aspect ratios, being of the same order as the lateral expansion. This is
illustrated in Figure 9. When the bin cools, not only does it contract laterally, but it also
contracts vertically. As the bin walls pull down due to this latter action, there is a strong
likelihood that the shear stress at the wall due to wall friction will change direction from up to
down. This reverse friction is a piston effect which can result in wall pressures significantly
higher than the hydrostatic pressures. Pressures due to reverse friction can significantly add
to those induced by the lateral contraction of the silo.

p
p
t t
n n
Aeration
Ducts
Expansion Contraction

Figure 9. Expansion and Contraction of a Silo

It needs to be noted that the thermal expansion and contraction influences on wall loads are
7

most likely to be non-uniform around the silo wall owing to the relative position of the sun
with respect to the silo. Hence eccentric type loadings will be induced in this way.

6. Use of Anti-Dynamic Tubes for Controlling Wall Loads in Tall Grain Silos

In tall grain silos, the effective transition occurs low down the silo walls. As a result, mass-
flow of grain with flow along the walls occurs over a substantial height of the silo above the
effective transition. The effect is to cause dynamic pressures to be generated, these
pressures being in the order of two to three times the static pressures generated after the
silo is filled from the empty condition.

Figure 10. Model Silo and Test Results Figure 11. Anti-Dynamic
Tube for Eccentric Discharge

One way of controlling the wall loads in existing silos is to employ anti-dynamic, tremmie
tubes which extend slightly less than half the height of the silo. Research using this type of
anti-dynamic tube was conducted by Ooms and Roberts (1985)[14]. Figure 10 shows,
schematically, the 1.2m diameter by 3.5m tall model flat bottom test silo and a sample set of
test results.

The work was initiated in order to provide a simple and low cost solution to controlling the
pressures in a number of badly cracked concrete grain silos approximately ten times the
scale. In effect, the tremmie tube divides the tall funnel-flow silo into two squat silos in
series. The top half of the silo discharges first followed by the bottom part once the level
drops below the top of the tremmie tube and the tube empties. Ports in the bottom of the
tube allow grain to flow laterally to the silo outlet.

The design of the bottom ports and tube sizing in relation to the silo outlet dimension are
important in order to promote automatic choking of the lateral flow at the bottom until the
tube empties. No valves are necessary. The arrangement ensures that at no time does the
effective transition intersect the walls of the silo and hence the pressures never exceed the
values corresponding to the static or initial filling condition. This is illustrated in the test
results of Figure 10. The anti-dynamic tube has also been successfully used to control the
flow in silos having eccentric discharge, as in Figure 11, and multiple discharge points.
Without the tube in place, the walls of such silos are subject to significant bending stresses
in addition to the hoop stresses. It is to be noted that the anti-dynamic tube described here is
suitable only for free flowing, cohesionless bulk solids such as grain. They should not be
used for cohesive bulk solids. It is also important to note that the cumulative effect of shear
stresses causes high draw-down forces which must be accounted for in their design and
installation in the silo.
Measured
Pressures
Mean Static Pressure
Mean Flow Pressure
(Tube Fitted)
Mean Flow Pressure
(No Tube)
Pressure
Model
Silo
Anti-
Dynamic
Tube
H
Ht =
0.41H
Lowest Observed
Effective Transition
u = 45
o
u
1
2
8


7. Dynamic Effects The Silo Quaking Problem

A recurring problem in bin and silo loadings is that due to the phenomenon of silo quaking.
Gravity flow in bins and silos, characteristically, is a cyclic or pulsating type flow which
arises as a result of changes in density during flow and by varying degrees of mobilisation of
the internal friction and flow channel boundary surface friction. Quaking is known to occur in
bins of all types, namely mass-flow, funnel-flow, expanded-flow and bins with single and
multi-outlets. In general, the pulsating load problem is more pronounced at low flow rates
where the period of pulsing may be from a few seconds to many seconds or even minutes.
The consequence of the pulsating loads may range from nuisance value arising from the
transmission of shock waves through the ground to disturb neighbouring areas, to structural
fatigue failure when the natural frequencies of the silo and structure are excited by the flow
pulses. An overview of the silo quaking problem, based on the work of Roberts and
Wensrich (1996-2003)[16-20,24,25], is now presented.

7.1 Experimental Studies

Quaking in tall mass-flow bins may be experienced when the height of fill is above a critical
height H
cr
where H
cr
D, as depicted in Figure 12. Figure 12(a) shows the test bin of
dimensions 1.2m diameter by 3.5m high, the bin being fitted with a stainless steel hopper.
Load cells which measure normal and shear stress simultaneously are fitted into the silo
walls. Above the height H
cr
, plug type flow occurs with the velocity profile substantially
uniform over the cross-section. Below the critical level in the region of the transition, the
flow starts to converge due to the influence of the hopper and the velocity profile is no
longer uniform. The velocity profile is further developed in the hopper as shown. As the flow
pressures form in the hopper, dilation of the bulk solid occurs. As a result of this dilation, it
is possible that the vertical supporting pressures decrease slightly thus reducing the support
given to the plug of bulk solid in the cylinder. This causes the plug to drop momentarily
giving rise to a load pulse. The cycle is then repeated.


(a) Test bin showing load cell locations (b) Velocity profiles

Figure 12. Mass-flow Test Bin
H
H
D
cr
Critical Level
for Mass Flow
Velocity
Profiles
9

Examples of wall pressure and shear stress records at locations 5 and 14 are shown in
Figure 13. The pulsing was quite pronounced at location 14 well up the cylinder, but was
less evident further down the cylinder. Virtually no pulsing was shown to occur at location 5
near the transition.

(a) Results for location 5 (b) Results for location 14

Figure 13. Sample of Normal Pressures and Shear Stresses at Two Locations

7.2 Dynamic Load Model

The dynamic load model proposed initially by Roberts (1996)[16] was later modified in the
light of the work of Wensrich (2002)[24,25]. Wensrich showed that in tall mass-flow silos,
shock waves travel upwards through the silo from the transition giving rise to an exponential
increase in the amplitude of the dynamic wall pressures as illustrated in Figure 14. The
amplitude of the pulse pressure Ap
ws
is given by




wo ws
y/R K -
e p p A = A

(1)

















Figure 14. Dynamic Loads Induced in Silo

D
W
H
h
S
sh
y
Plane of
shock
A
p
ws
h
s
H
s
D
W
H
h
S
y
Plane of
shock
A
p
h
s
H
s
A
p
A
p
ws
wo
Shock pressure
on walls grows
exponentially
Increment
due to shock
(amplitude)
A
p
wo
t
t
T = Shock period
Tp
A
p
A
p
ws
Shock pressure
on walls
Increment
due to shock
(amplitude)
A
p
wo
t
t
T = Shock period
Tp
D
W
H
h
S
sh
y
Plane of
shock
A
p
ws
h
s
H
s
D
W
H
h
S
y
Plane of
shock
A
p
h
s
H
s
A
p
A
p
ws
wo
Shock pressure
on walls grows
exponentially
Increment
due to shock
(amplitude)
A
p
wo
t
t
T = Shock period
Tp
A
p
A
p
ws
Shock pressure
on walls
Increment
due to shock
(amplitude)
A
p
wo
t
t
T = Shock period
Tp
10


As y h, it is assumed that Ap
ws
'tails off' to approach the flow pressure at the top surface in
the cylinder. It is noted that Ap
ws
is the additional wall pressure applied to the initial or static
pressure. Figure15 shows the measured values (plotted points) and calculated (full lines)
wall pressure amplitudes for the wheat silo of Figure 12(a). The calculated values are based
on the following values:
H
sh
= 2.14 m (location 14 of Figure 12(a)); | = 30
o
; = 0.85 t/m
3
; D = 1.2 m:
Overall fill height above transition, H
sh
+ h = 3.0 m: K = 0.4 ;k
r
= 0.29
k
r
= ratio of shear to normal stress amplitude.

The agreement between the measured and calculated normal and shear stresses is
considered very satisfactory.


Figure 15. Computed and Measured Figure16. Pulse Period Versus Velocity
Normal and Shear Stress Amplitudes

7.3 Pulse Period
The following equation, proposed by Roberts (1996)[16], has been shown to be a good
predictor of the pulse period.
)
a
v
t 2 (
a
v
)
a
v
(t T
o o
+ + + = (2)
where
) h (h
p
a
s
vo
+
A
=
A
Q
v =
v

t
y
o
A
=
a = acceleration of upper mass during pulse motion such that a g
v = average velocity of bulk solid in the cylinder during discharge,
Q = discharge rate
t
o
= time for motion of upper mass to be initiated
Ac
y
= dynamic displacement of consolidated mass in vertical direction
0 0.5 1 1.5 2
0
0.5
1
1.5
2
2.5
Normal Stress Amplitude
Shear Stress Amplitude
Predicted Normal Stress Amplitude
Predicted Shear Stress Amplitude
STRESS AMPLITUDE (kPa)
H
E
I
G
H
T

A
B
O
V
E

T
R
A
N
S
I
T
I
O
N


(
m
)
0
5
10
15
20
25
30
0 0.5 1 1.5 2 2.5 3
Ac
Ac
Ac
Acy = 5 mm
y = 2.5 mm
y =1 mm
y = 0.5 mm
Velocity (mm/s)
P
u
l
k
s
e

P
e
r
i
o
d


(
s
e
c
.
)
11

Since velocities in the upper cylindrical section are usually quite low, for most practical
cases, the ratio
v
a
0. Hence T = t
o
. That is,

v

T
y
A
= (3)
The parameter Ac
y
is dependent the average particle size, void ratio and properties of the
boundary surface of the flow channel. A plot of pulse period versus average discharge
velocity in the cylindrical section of a silo is presented in Figure 16. As shown by Roberts
(2003)[20], Acy is related to the contact stiffness at the boundary of the channel as follows:


K

T
n
y
= A
(4)
where K
T
= contact stiffness (
mm
kPa
) and o
n
= normal stress (kPa).

7.4 Controlling Shock Loads in Tall Mass-Flow Silo Case Study

In order to attenuate the growth in shock wave amplitude up the silo, Roberts and Wensrich
conducted experiments using a hopper insert which, in effect, divides a tall mass-flow silo
into two squatter mass-flow silos. Laboratory experiments on a small scale silo showed that
the acceleration amplitude above the insert is reduced by approximately 50%. These results
formed the solution to the quaking problem being experienced by the 200 tonne wheat
conditioning silo shown in Figure 17.

Figure 17. 200 tonne Wheat Conditioning Silo
Wheat is dosed with water prior to being fed into the top of the silo, the objective being to
bring the wheat to a uniform moisture content of around 16% prior to discharge for the
milling process. The magnitude of the shock loads were quite severe, particularly when the
silo was full or near full. Observation showed the shock loads were transmitted through the
ground into the neighbouring concrete silo structure leading to some cracking.

12

Calculations showed that the vertical dynamic load for the full silo when discharging grain
amounted to 30% of the full static load, thereby increasing the total vertical load to 130% of
the static load, The load analysis indicated that if the silo is operated at reduced head or fill,
the shock loads are significantly reduced. For instance, at 60% full, the dynamic load is 46%
of its maximum value, whereas at 50% full, the dynamic load is only 21% of its maximum
value. Obviously, operating at 50% full would be the preferred option provided the residence
time of the wheat in the silo is still sufficient to achieve good conditioning. The vibration
frequency of the silo structure due to the vertical motion and swaying varies with the degree
of fill. The fundamental natural frequency of the silo when 100% full is estimated to be 40%
of the corresponding frequency when the silo is empty.

8. Quaking in Other Types of Bins and Silos
A similar action to that described for mass-flow bins may occur in tall funnel-flow bins where
the effective transition intersects the wall in the lower region of the silo. As a result, there is
flow along the walls of a substantial mass of bulk solid above the effective transition.

8.1 Squat Funnel-Flow, Expanded-Flow and Intermediate-Flow Bins

The flow-patterns in squat funnel-flow and expanded-flow bins are illustrated in Figure 18.
During funnelflow with no flow along the walls, as depicted in Figure 18(a), dilation of the
bulk solid occurs as it expands in the flow channel. As a result some reduction in the radial
support given to the stationary material may occur. If the hopper is fairly steeply sloped, say
|u > o)], then the stationary mass may slip momentarily causing the pressure in the flow
channel to increase as a result of the 'squeezing' action. The cycle then repeats. A similar
behaviour may occur in expanded flow bins, such as the bin depicted in Figure 18(b).

Figure 18. Funnel-flow and expanded-flow
A similar flow pattern is that of intermediate-flow as described by Benink (1989)[2]. This
mode of flow, illustrated in Figure 19, is a special form of mass-flow which can occur in
squat mass-flow bins in which the surcharge head H < H
cr
. Such flow is defined by a rapid
flow in the central region of the hopper, and a slower flow in the outer regions as illustrated
in Figure 19(a). Funnel-flow is really the special case of intermediate-flow in which the outer
region is stationary.

13


Figure 19. Intermediate-flow
8.2 Expanded-Flow Surge Bin Case Study

The example of an expanded-flow surge bin handling Potash is now considered. The bin
constructed of stainless steel, is shown in Figure 20. The relevant properties of the Potash
are:
Bulk Density = 1.2t/m
3
; Flow Rate q
m
= 4000tph; Effective Angle of Internal friction o = 45
o
;
Hopper Half-Angle o = 25
o
; Average Particle Size of Potash = 7.5mm.

8.2.1 Analysis of Loads due to Central Channel
Referring Figure 21, it is assumed that during discharge the central channel or plug acts as a
dynamic mass moving under accelerated flow. The driving force is the weight of the material
in the channel, the downward movement being resisted by the components of the shear and
normal forces around the boundaries of the flow channel. The slope of the central inner
moving plug is defined by the angle c.

Figure 20. Expanded-Flow Surge Bin Figure 21. Loads Acting in Flow Channel
Slow Flow in
Outer Region
Pulsating Flow in
Central Region
Hopper
(a) Flow Patterns (b) Effective Mass
12000 OD
5697 OD
4
5
0
0
3
2
6
4
1
6
0
0
3
5
0
0
44
o
25
o
z
Inclined Slide
Gates
1500 OD
c
z
h
Di
Do
D
Dy
y
pn pn
Fs
W
o
c
t t
Inner
Moving
Plug
ps
vs vm
vav
Velocity Profile
14

Since shear at the boundary is due to internal friction, it is assumed that slip and hence
failure corresponds to the maximum shear stress, it follows from Mohr circle analysis that
= sin o (5)
The corresponding boundary friction angle is
|
w
= tan
-1
(sin o) (6)
The total support force provided by the flow channel boundary is

(7)
The solution of equation (7) leads to

] (8)
where

(9)

(10)


(16)
{

(
|

) }

q

(11)
[

)] (12)
The weight of the moving central core is
(

(13)
where

and

The net dynamic force is

(14)
The dynamic load factor depends on the elasticity of the impact during the deceleration
phase. For a suddenly applied load, k
D
= 1. The acceleration is given by
a =
W
g F
D
(15)
where
= g = bulk specific weight, kN/m
3
= bulk density, t/m
3
y = height of moving mass, m k
h
= pressure ratio defined by equation (6)
p
v
= average vertical pressure p
n
= normal pressure
p
s
= surcharge pressure, kPa m = 1
c = half-angle of flow channel. o = half-angle of hopper
|
w
= friction angle at boundary = friction coefficient
15


8.2.2 Predicted Performance Results

Figure 22. Predicted Dynamic Parameters for the Potash Surge Bin
Referring to the potash surge bin of Figure 20, the slide gates are partially closed to control
the discharge flow rate through the split chute to 4000tph. Based on the above analysis, the
dynamic forces, accelerations, velocities and pulse frequencies have been computed for the
for various central flow channel heights 0 to 8.5m, the results being presented in Figure 22.
Based on the effective angle of internal friction o = 45
o
, the half-angle of the flow channel,
c = 6
o
(Roberts (2005)([21]. The lower frequencies for the full flow channel are most likely to
be the critical values for the bin design.
9. 6000 Tonne Multi-Outlet Coal Bin Case Study

Silo-quaking problems have been known to occur in bins with multiple outlets. By way of
illustration, a case study concerning a 6000 tonne coal bin is reviewed. The bin, illustrated in
Figure 23, has seven outlets as shown. Coal is discharged by means of seven vibratory
feeders onto a centrally located conveyor belt. When the bin was full or near full, severe
shock loads were observed at approximately 3 second intervals during discharge. The
discharge rate from each feeder was in the order of 300 t/h. When the level in the bin had
dropped to approximately half the height, the shock loads had diminished significantly. With
all the outlets operating, the effective transition was well down towards the bottom of the bin
walls. Substantial flow occurred along the walls, and since the reclaim hoppers were at a
critical slope for mass and funnel-flow as determined by flow property tests, the conditions
were right for severe 'silo quaking' to occur.

Confirmation of the mechanism of silo quaking was obtained in field trials conducted on the
bin. A set of dynamic strain results is shown in Figure 24. In one series of tests the three
feeders along the centre line parallel with the reclaim conveyor were operated, while the
four outer feeders were not operated. This induced funnel-flow in a wedged-shaped pattern
as indicated in Figure 23, with the effective transition occurring well up the bin walls, that is
H
m
< H
cr
. The same was true when only the central feeder (Fdr. 1) was operated. In this
latter case the stationary material in the bin formed a conical shape. Under these conditions,
the motion down the walls was greatly restricted and, as a result, the load pulsations were
barely perceptible
0.1
0.2
0.3
0.4
0.5
0.6
10
20
30
40
50
60
70
0 2 4 6 8 10
Velocity (m/s)
Frequency (hz)
V
e
l
o
c
i
t
y


(
m
/
s
)
F
r
e
q
u
e
n
c
y


(
h
z
)
Flow Channel Height z
0
5
10
15
20
0
2
4
6
8
10
0 2 4 6 8 10
FD (kN)
Acceleration (m/s^2)
D
y
n
a
m
i
c

F
o
r
c
e

F
D


(
k
N
)
A
c
c
e
l
e
r
a
t
i
o
n

(
m
/
s
^
2
)
Flow Channel Height z (m)
(a) Dynamic Force and Acceleration
(b) Velocity and Frequency
16


Figure 23. 6000 Tonne Multi-Outlet Coal Bin

In a second set of trials, the three central feeders were left stationary, while the four outer
feeders were operated. This gave rise to the triangular prism shaped dead region in the
central region, with substantial mass-flow along the walls. The load pulsations were just as
severe in this case as was the case with all feeders operating.



Figure 24. Variation in Dynamic Micro-Strain in Bin Column (peak to peak measurement)

A critical factor in the operation of quaking silos is the dynamic stability of the overall
structure. The silo in question is supported on columns from a concrete base which, in turn,
is supported on piles as illustrated in Figure 25. In view of the significant decrease in total
mass of the silo from the full to empty condition, there will be a corresponding increase in
the natural frequencies of the silo during the emptying process. The modes of vibration
involve combinations of vertical, swaying and twisting modes which are induced as a result
of non-symmetrical loading of the silo, the pulsating flow during discharge and variations in
the ground stiffness in the zone of the supporting piles.



H
H
D
D
Eff ective transition
for central feeders
operationg
Eff ective transition
for all feeders or
outside feeders
operating
H
Feeders
Conveyor
m
cr
m
F
Conveyor
Feeders
Plan View of Bin
1
2
3
4
5
6
7
0
2
4
6
8
10
12
14
20 30 40 50 60 70 80 90 100
All Feeders
Feeders 4.5.6.7
Feeders 1,2,3
Feeder 1
BIN LEVEL %
D
Y
N
A
M
I
C

M
I
C
R
O
-
S
T
R
A
I
N
17

k
k
1
2
Sw aying
Mode
Columns
Piles
Vertical
Mode
Concrete
Base
Ground
M

Figure 25. Silo Dynamic Mode
10. Dynamic Loads Due to High Load-Out Rates

Dynamic loads also occur during flood type loading of mineral ores into rail wagons. As an
illustration, the case of an iron ore train loading bin, illustrated in Figure 26, is considered
(Roberts (2008))[22].
















(a) Train Loading Bin (b) Bin Flow Patterns and Loads















(a) Wagon Load Rates and Total Loads (d) Rail Wagon Load Patterns

Figure 26 . Load-Out Bin for Filling Iron Ore Rail Wagons


Vex
Vey
Vo
u = 50
o
Fix
Fiy
FH
FV
Fic
Tic
TD Tvi
Fvi FD
Pivot Point for Swing Chute
Pivot Point for Clamshell
Top of Wagon
Trim Level
12000 dia
8
0
0
0
1
0
9
0
0
1770 sq
Vex
Vey
Vo
u = 50
o
Fix
Fiy
FH
FV
Fic
Tic
TD Tvi
Fvi FD
Pivot Point for Swing Chute
Pivot Point for Clamshell
Top of Wagon
Trim Level
Vex
Vey
Vo
u = 50
o
Fix
Fiy
Vex
Vey
Vo
u = 50
o
Fix
Fiy
FH
FV
Fic
Tic
TD Tvi
Fvi FD
Pivot Point for Swing Chute
Pivot Point for Clamshell
Top of Wagon
Trim Level
12000 dia
8
0
0
0
1
0
9
0
0
1770 sq

0
10
20
30
40
50
60
70
0
20
40
60
80
100
120
140
0 10 20 30 40 50
Load-Out Rate (t/h)
Total Load (tonne)
TIME (sec)
L
O
A
D
-
O
U
T

R
A
T
E


(
t
/
h

x

1
0
^
3
)
T
O
T
A
L

D
I
S
C
H
A
R
G
E


(
t
o
n
n
e
)
Qaverage = 8.64 x 10^3 t/h

VT
Zone 1
30 t
3.5 sec
Zone 2
16.5 t
6.5 sec
Zone 3
57 t
27 sec
Zone 4
16.5 t
13 sec
18

Each wagon holds 120 tonne of ore, the filling time per wagon being approximately 50 sec.
The load out is controlled by a clam shell gate operating on a swinging chute as depicted.
As an empty wagon moves under the bin load-out chute, there is an initial surge in the flow
rate peaking around 60,000 t/h. This causes high vertical and lateral impact loads. Once the
chute chokes, the remainder of the wagon is loaded at a rate of approximately 7000 t/h, with
the flow rate reduced to zero as the gate closes with the wagon full. The shock loads on the
bin and structure need to be taken into account in the design.

10. Loads on Buried Structural Members

Large bulk storage facilities such as stockpiles, bulk storage sheds and bins often contain
structural elements which become partly or totally buried by the bulk solid during the filling
process. Typical examples include trestle support legs on stockpile load-out conveyors as in
Figure 26 and support columns in bulk storage sheds as illustrated in Figure 27.

The loads on these structural elements are quite complex and depend on a number of
factors such as the methods of filling and reclaim, load settlement, the flow properties of the
bulk solid and any lateral deflection of the structural element. Furthermore, the cumulative,
in-plane type compressive loads due to the shear stresses acting along the structural
members can contribute to the buckling loads, thus exacerbating the lateral bending.
Depending on the method of reclaim employed, the lateral loads may increase further during
the unloading process. These various loading conditions have been described and design
equations presented by Roberts (2007)[23]. He examined the case of the bulk phosphate
storage shed of Figure 28.



Figure 27. Conveyor Support Legs Buried in Iron Ore Stockpile

Load-Out Load - Out
Trestle Legs
Conveyor
Stockpile
Trestle Legs
Conveyor
19



Figure 28. Bulk Phosphate Storage Shed

The loads on the buried columns can be quite variable depending on the following
conditions:

- Variations in the flow properties of the material
- Loading and unloading history and mechanism for loading and unloading
- Length of time the bulk material remains undisturbed in the stockpile
- Rigidity of the stockpile floor - whether any settlement has occurred
- Variations in stress fields from active to passive

The models for the active and passive states of stress on the upper loading face of the
column are shown in Figure 29. In view of the uncertainty as to how the stockpile will be
filled and reclaimed, it is necessary to consider both stress states. The degree of support
provided by the bulk solid on the rear face of the column depends on the degree of
cohesiveness of the bulk solid. Observations have shown that for free flowing granular
materials, flow of the material to the back face is quite rapid and the rear support provided
by the material is quite strong. On the other hand, cohesive materials do not back-fill so
readily so that the degree of support is much weaker.

Figure 29. Stress States for Buried Column
Load -Out
Conveyor
Stockpile
Columns
Load -Out
Conveyor
Stockpile
Active Case
o
1
Settlement
During Loading &
Flow During
Unloading
Column
z
p
nl
t
t
l
u
p
nu
v
p
|
u
R
Direction of Major
Consolidation Stress
Passive Case
|
z
z
g
h
Column
o
1
o
1
t
s
t
c
p
ns
p
v
q
s
c
s
q
c
p
nc
c
c
20

As an example of the loads that may occur, the case of the bulk phosphate storage shed of
Figure 28 is illustrated. The support columns are 250mm square steel tubing buried to a
depth of 13.5m. Figure 30 shows the loads acting on the leading side of the columns for
both the active and passive states. The full lines depict the load conditions with some back
support, whereas the dotted lines are the load conditions without rear support. In addition
the compressive loads due to shear draw-down need to be considered since these
contribute to the compressive buckling loads. In view of the slenderness of the support
column in this case study, it is not surprising that some columns were deformed due to
bending.





















Figure 30. Lateral loads on Support Columns of Bulk Phosphate Storage Shed

11. Concluding Remarks

The aim of this review paper has been to highlight some salient aspects in the research into
silo load analysis that have occurred since the early work of Janssen and others
commencing in the latter part of the 19th century until the present. It is quite evident that
significant progress has been made. Yet, at the same time, it is also evident that the subject
of silo loads remains a subject of considerable complexity. Apart from the analysis and
design of the bin or silo and its support structure, the complexity is greatly magnified by the
difficulty of fully understanding the properties of the bulk material being stored and handled
and the way these bulk materials are influenced by consolidation stresses particularly under
such variable conditions of prolonged undisturbed storage coupled with moisture and
temperature variations. While the loading of silo and bin walls may be confidently predicted
in the case symmetrical, single centrally located outlets, the case of multi-outlet bins and
bins with eccentric discharge points create significant design challenges. Other areas for
research include the important subjects of dynamic loads and bulk material flow induced
vibrations. Invariably, modern industrial development in the agricultural and process
industries demands increased tonnage rates, coupled with more efficient, economic and
trouble-free operation. This provides the impetus for more exacting analysis and design
procedures. It is the driving force for more in-depth research into this important subject, that
of silo load analysis. There exists a wealth of knowledge which provides the datum for future
research directions. This coupled with modern computer technology and keen young
engineering and science based intellect, advances in the important subject of silo design
and related areas of bulk handling are assured.
0 10 20 30 40 50 60 70 80
0
2
4
6
8
10
12
14
FT - Active (kN/m)
FT - Passive (kN/m)
FT1 - Active (kN/m)
FT1 - Passive (kN/m)
LATERAL FORCE ON COLUMN (kN/m)
D
E
P
T
H


z

(
m
)
Bulk Density = 2.1 t/m^3
21


12. References

1. Australian Standard (1996) AS3774, Loads on Bulk Solids Containers. Standards
Association of Australia
2. Benink, E.J. (1989). Flow and Stress Analysis of Cohesionless Bulk Materials in Silos
Related to Codes. Doctoral Thesis, The University of Twente, Enschede, The
Netherlands.
3. EN 1991-4, (2005) Eurocode 1 Actions on Structures. Part 4. Silos and Tanks. Final
Draft
4. Jamieson, H.A. (1903). Grain Pressures in Deep Bins. Trans. Canadian Society of
Civil Engineers, Vol. XVII.

5. Jamieson, J.A. (1904). Grain Pressures in Deep Bins. Engineering News. Vol. LI,
No.10, 236-243.
6. Janssen, H.A (1895). Versuche u ber Getreidedruck in Silozellen (On the
Measurement of Pressures in Grain Silos). Zeitschrift des Vereines deutscher
Ingenieure, 1045-1049.
7. Jenike, A.W. (1961). Gravity Flow of Bulk Solids. Bul. 108, The Univ. of Utah,Engng.
Exp. Station, USA.
8. Jenike, A.W. (1964). Storage and Flow of Solids. Bul. 123, The Univ. of Utah,Engng.
Exp. Station, USA.

9. Jenike, A.W. and Johanson, J.R. (1969). On the Theory of Bin Loads. Trans. ASME.,
Jnl. of Engng. for Industry. Series B. Vol.91, No.2. 339.
10. Jenike, A.W., Johanson,J.R. and Carson, J.W. (1973). Bin Loads - Part 2: Concepts.
Trans. ASME., Jnl. of Engng. for Industry. Series B. Vol.95, No.1.
11. Jenike, A.W., Johanson, J.R. and Carson, J.W. (1973). Bin Loads - Part 3: Mass-Flow
Bins. Trans. ASME., Jnl. of Engng. for Industry. Series B. Vol.95, No.1, 6.
12. Jenike, A.W., Johanson, J.R. and Carson, J.W. (1973). Bin Loads - Part 4: Funnel-
Flow Bins. Trans. ASME., Jnl. of Engng. for Industry. Series B. Vol.95, No.1, 13.

13. Johanson, J.R. (1964). Stress and Velocity Fields in the Gravity Flow of Bulk Solids.
ASME, Jnl. of Appl. Mechanics, Vol. 131, Ser. E, No. 3. 499-506.
14. Ooms, M. and Roberts. A.W. (1985. The Reduction and Control of Flow Pressures in
Cracked Grain Silos. Bulk Solids Handling, Vol. 5. No.5. 1009-1016.
15. Roberts A.W. (1995) One Hundred Years of Janssen. Bulk Solids Handling 15(3)
369-383.
16. Roberts, A.W. (1996). Shock Loads in Silos - The 'Silo Quaking' Problem., Bulk Solids
Handling, Vol 16, No. 2. 59-73.
17. Roberts, A.W. (1998), Basic Principles of Bulk Solids, Storage, Flow and Handling.
The University of Newcastle Research Associates Ltd. (TUNRA)
22

18. Roberts, A.W. (1999). Particle Technology Reflection and Horizons: An Engineering
Perspective, Transactions, Institution of Chemical Engineering, Part A, Vol 76, No A7,
1999, 775- 796.
19. Roberts, A.W and Wensrich,C.M. (2002). Flow Dynamics or Quaking in Gravity
Discharge from Bins. Chemical Engineering Science, Vol. 57.295-305.
20. Roberts, A.W. (2003). Review of the Silo Quaking Problem in Bins of Various
Geometrical Shapes and Flow Pattern. Task Force Quarterly, Academic Computer
Centre in Gdansk, Poland, Vol 7. 623-641.
21. Roberts, A.W. (2005). Characterisation for Hopper and Stockpile Design, Chapter 3,
Characterisation of Bulk Solids, Ed D. McGlinchey, Blackwell Publishing. 85-131.
22. Roberts, A.W. (2008). Shock Loads in an Iron Ore train Loading Bin. Proc. Structures
and Granular Solids Conference, Royal Society of Edinburgh. CRC Press, Taylor and
Francis Group, London. 67-76.
23. Roberts, A.W. (2007). Loads on Support Structural Elements Buried in Stockpiles,
Particle and Particle Systems Characterisation, 2 Vol 24, Issue 4-5, 352-359.
24. Wensrich, C.M. (2002). Analytical and Numerical Modelling of Quaking in Tall Silos.
PhD Thesis, The University of Newcastle, Australia.
25. Wensrich, C.M. (2002), Experimental Behaviour of Quaking in Tall Silos. Powder
Technology, Vol. 127, 87-94.




Effective Transfer Chute Design including Dust Control for Handling
Grains and other Products
Timothy J . Donohue
1
*, Alan W. Roberts
1
, Craig Wheeler
1
, Dusan Ilic
1
, Andre
Katterfeld
2

1
TUNRA Bulk Solids, University of Newcastle, Callaghan, 2308, Australia
2
Institute of Logistics and Material Handlings, University of Magdeburg, Germany
*Corresponding author. E-mail: Timothy.Donohue@Newcastle.edu.au
Abstract
The effective design of transfer chutes is of utmost importance in the materials handling
chain to ensure reliable flow of the product being transported. In addition to this, the control
of fugitive dust is needed to meet occupational health and safety standards as well as
environmental concerns. In the design of transfer chutes classical continuum methods [1]
(Roberts, 2003) are available to utilise as well as more recent computational techniques. This
paper presents the well-established continuum method as well as two computational
techniques available to analyse flow through transfer chutes. Discrete Element Modelling
(DEM) is an excellent tool to model and visualise the granular flow through a transfer chute,
while Computational Fluid Dynamics (CFD) is more focussed on the flow of air (and dust)
through transfer chutes. Details of these modelling techniques will be presented in this paper
as well as case studies to illustrate the application of these methods. By way of example, one
case study will demonstrate the advantages of fast, controlled density, accelerated flow in a
grain ship loading chute in which the dust emissions during grain loading were reduced by up
to 60%. This problem was solved by pilot scale testing in combination with computational
simulation.
Key words: transfer chutes, dust, modelling, grain
1. Introduction
In many bulk material handling operations the major dust problems result from the loading,
unloading and transfer operations of belt conveying systems. The magnitude of this problem
is identified by the U.S. Department of Labor (1987) [2] which cite these areas of belt
conveyor operations as the most frequent sources of dust emissions. For example, a
particular mine in Australia uses 85x10
6
litres of water to control dust generation in only
seven conveyor transfer stations. In terms of water usage alone, the annual use for dust
control would be very significant indeed, but in addition to this other external energy sources
such as dust extraction fans are also used. Given that the use of water or other external
energy costs increase running costs, there is a great need for passive dust control systems.
As the tools for effective transfer chute design are discussed in this paper the theme of
passive dust control is also present.

The modelling of granular flow is itself a very complex task, and is affected by a number of
factors including the implicit material properties, the plant geometry and the interaction of
these two (wall friction). In the past when analysing the flow of material through a conveyor
transfer station the well-established continuum method [1] (Roberts, 2003) has been used in
which the stream of material is treated as a continuum and some other relevant assumptions
are made. However, as the continuum model does not deal with individual particles, there is
no facility available to study the flow of dust through a transfer station. Another of the other
main tools available for studying granular flow is Discrete Element Modelling (DEM), which
models individual particles and makes calculations based on Newtons laws of motion at very
small time steps. While it is not possible to model dust directly using DEM, it can serve as a
valuable design tool for transfer chutes in a similar manner as the continuum method does.
The last design tool discussed in this paper is Computational Fluid Dynamics (CFD), which
can model the dust directly through a multiphase simulation. The purpose of this paper is to
present an overview of these design tools and illustrate how they can be used effectively in
the design of transfer chutes, with a particular focus on dust control.

2. Continuum Method

The continuum method, developed by Roberts (1969) [3], is a design tool developed for the
analysis of streamlined flow of the bulk solid through a transfer chute. It has been shown
that streamlined flow is a very effective flow mode for passive dust control. The continuum
method primarily deals with hood sections (Figure 1a) and spoon chutes (Figure 1b) and is
based on a lumped parameter model. The basis for the continuum method is shown in Figure
1. The application of the continuum method to an industrial case study is discussed at a later
stage in this paper, and the reader is referred to the work of Roberts [1,3] for further details
on the specifics of the continuum model.


Figure 1: Chute Flow Models (Roberts [1, 3])

3. Discrete Element Modelling

Discrete Element Modelling (DEM) is a numerical method for simulating the flow of granular
bulk solids. The basic principle of DEM is to model each individual particle as a separate
entity that can undergo a range of forces as in reality. These forces typically include gravity,
contact forces with other particles and walls, with cohesive and adhesive forces being
typically applied as well if the bulk solid is cohesive in nature. Calculations for the forces, and
resulting displacements, are made for every particle at very small time steps throughout the
simulation. As a result, DEM simulations can be very computationally intensive. The reader is
directed to [4] (Grger and Katterfeld, 2007) and [5] (Katterfeld and Grger, 2007) for more
fundamental information about DEM and its uses in bulk solids handling.

For a first evaluation of the dust emission in a transfer chute, it is often sufficient to analyse
the material flow in the system, which can be done using DEM. Generally, it can be said that
dust is generated where there is an abrupt change of the particle speed or direction. These
changes in direction cause a change in the porosity of the bulk solid which has a twofold
effect;
A bulk solid dispersing or dilating, for example during free fall, will tend to induce air
into the falling stream of material
A bulk solid that is being compacted, for example in an impact zone, will cause an
exhaustion of air
Within a transfer chute, there will always be a certain amount of fine particles suspended in
the air. These suspended dust particles become important in the impact zones where there
will be a large exhaustion of air. Therefore, it is necessary to ensure a smooth redirection of
the material flow with a minimum loss of momentum. Furthermore, it is necessary to ensure a
compact material flow as any dispersion of the flow will increase the porosity, thus leading to
an increase in the air flow induced into the stream. This dispersion can potentially create
more dust as it will need to be exhausted from the granular stream upon impact. With the
help of these two guidelines for smooth redirection and compact material stream, it is
possible to evaluate the general dust emission behaviour of many transfer chutes.

As an example of this type of approach, consider the two designs of a transfer chute
presented in Figure 2. The figure shows a comparison of the steady state material flow in a
transfer chute with the original design (left) and the optimised design (right). The design
criteria discussed previously can be easily evaluated via the simulated particle flow and the
particle velocity (particle colour).


Figure 2: (a) Original Design (b) Optimised Design

The above example shows the benefits of implementing the basic guidelines of ensuring a
smooth redirection of flow and a compact material stream. The main advantage of this type
of approach over a method such as the previously discussed continuum method is the ability
to visualise the effect that the change in geometry has on the flowing granular stream. While
the use of DEM on its own, or in conjunction with the continuum method, has its benefits, air
flow patterns can only be visualised through the use of Computational Fluid Dynamics (CFD).
This is discussed in the following section.

4. Computational Fluid Dynamics

Computational Fluid Dynamics (CFD) is a numerical modelling technique that can be used
for a wide variety of applications. A great deal of literature exists covering the fundamentals
of CFD and the associated set of equations that are solved. The focus of this paper is on the
use of CFD and how it can assist in analysing transfer chute flows, so the reader is referred
elsewhere to the great wealth of information that exists for a detailed explanation on the
fundamentals of CFD.

For transfer chute analysis using CFD, a multiphase simulation is typically used, of which
there are two main approaches. The first approach is known as the Eulerian-Lagrangian
method where the fluid phase is treated as a continuum and the dispersed phase is solved
by tracking a large number of particles. This method is limited to small volume fractions, as
computationally the process for large volume fractions is too intensive. However, this method
models the real behaviour of the particles where a coefficient of restitution is defined
between particle to particle and also particle to wall. The other method is the Eulerian-
Eulerian approach where all of the phases are treated as interpenetrating continua. This
approach requires the solid dispersed phase to have a low value of dynamic viscosity so that
it can be modelled as a fluid. This method can be used to solve flow fields for densely
packed solids. However there is no direct accounting for the friction between the solid
phased and the walls.

To illustrate the combined approach of continuum and CFD modelling, a case study is
presented in which a grain ship loading chute is analysed. A pilot scale test rig was built to
investigate a range of chutes, with each chute being evaluated in terms of dust generation.
The pilot scale test rig can be seen in Figure 3.

Figure 3: Pilot scale test rig
Using the previously described continuum method and the design guideline of keeping the
stream concentrated, a constant radius spoon was designed to replace the original chute.
Photos of the chutes used in the scale modelling can be seen in Figure 4. Using a DustTrak
aerosol monitor, the average dust concentration level for the series of tests showed more
than a 50% reduction in dust compared to the original loading spoon. The reduction in dust
levels were attributed to the concentrated product stream entraining the dust. A complete
overview of this pilot scale testing can be found in the work by Wheeler et al (2007) [6].

Using CFD, a further optimisation of this chute was found through the inclusion of a
converging section leading into the constant radius spoon. The two geometries for the
loading spoon chute that were used in the CFD modelling can be seen in Figure 5. For this
CFD modelling, a three phase Eulerian-Eulerian simulation was used which included an air
phase, a solid dispersed phase of 1mm particles, representing the grains, and also a solid
dispersed phase of 0.1mm particles representing the dust. A full discussion of the CFD
simulation can be found in Donohue et al (2009) [7].


Figure 4: (a) Original design with spring loaded gates (b) Constant radius spoon revised design

Figure 5: CFD modelling results for two spoon chute geometries (a) Standard chute (b)
Converging chute
Results for the volume solid fraction for the grains as they flow through the two chutes can be
seen in Figure 5. As the figure shows, the converging section has the effect of streamlining
the flow and concentrating the flow towards the back wall of the spoon chute. The impact
angle with the back wall is also smaller in the converging chute, which leads to faster
discharge velocities from the spoon chute (evident by the smaller burden depth of material
discharging from the chute). Comparisons of the dust profiles can also be made, with Figure
6 showing the volume fraction of dust through the cross-section at the discharge point. As
the figure shows, the top layer of material at the discharge point for the standard chute
contains a significant proportion of dust. In comparison, the top layer of the material for the
converging chute contains almost no dust. The principal reason for this is due to the dust
becoming better encapsulated in the material stream due to the converging section. This is
advantageous as when the material travels through the chute and discharges on to a
stockpile the encapsulated dust within the material stream will be embedded into the
stockpile thus reducing the amount of dust that becomes airborne. Experimental testing in
the pilot scale test rig supported these CFD findings in that a further 20% reduction in dust
was found through the use of the converging chute compared to the standard chute.


Figure 6: Comparison of dust volume fraction through the cross-section at the discharge poin
for (a) Standard chute, and (b) Converging chute
5. Concluding Remarks

A well-established continuum method in addition to two numerical modelling techniques has
been presented as possible options to approach the design of transfer chutes. In particular,
the design of transfer chutes with respect to dust control was discussed. Methods such as
the continuum method and DEM can be useful as a first approach in designing transfer
chutes for optimum dust control performance, and when used the basic principles of fast,
controlled density, accelerated flow should be maintained. However, to investigate the flow of
air (and dust) directly in transfer chutes then the numerical tool of CFD is necessary. A case
study was presented to highlight the use of CFD in approaching the problem of controlling
dust in transfer chutes.

6. References

[1] Roberts, A.W., 'Chute Performance and Design for Rapid Flow Conditions', Chemical
Engineering Technology, Vol 26, 2003.

[2] U.S. Department of Labor, Occupational Safety and Health Administration, Dust Control
Handbook for Minerals Processing, 1987.

[3] Roberts, A. W.; An Investigation of the Gravity Flow of Non-cohesive Granular Materials
through Discharge Chutes, Transactions of A.S.M.E., J nl. of Eng. in Industry, Vol. 91 (2),
Series B, 1969, pp 373-381.

[4] Grger, T.; Katterfeld, A.; Application of the Discrete Element Method in Materials
Handling: Basics and Calibration; Bulk Solids Handling Vol. 27 (1), 2007, pp 17-23

[5] Katterfeld, A; Grger, T.; Application of the Discrete Element Method in Materials
Handling: Transfer Stations; Bulk Solids Handling Vol. 27 (3), 2007, pp 158-167

[6] Wheeler, C.A.; Krull, T.; Roberts, A.W.; Wiche, S.J .; Design of Ship Loading Chutes to
Reduce Dust Emissions, Process Safety Progress, Vol. 26 (2) 2007, pp 229-234

[7] Donohue, T.J .; Roberts, A.W.; Wheeler, C.A.; McBride, W.; Computer Simulations as a
Tool for Investigating Dust Generation in Bulk Solids Handling Operations, Particle and
Particle Systems Characterisation, Vol 26, 2009, pp 265-274.
Vertical Load on Disc Embedded in Static and Flowing Grain:
Experimental and DEM Simulations
Rafa Kobyka, Marek Molenda
Institute of Agrophysics PAS, Doswiadczalna 4, Lublin, 20-290, Poland
*Corresponding author. E-mail: rkobylka@ipan.lublin.pl

Abstract
Vertical loads on discs embedded in grain were measured in a model silo 0.4 m in diameter and
1.5 m high. Five disk inserts (with ratios of disc diameter to silo diameter d/D of: 0.188, 0.253,
0.375, 0.55 and 0.675) were axially suspended at height h of h/D = 0.5 in the cylindrical
container and vertical loads on discs during storage and discharge were estimated. DEM
simulations of a laboratory experiment were performed with an assembly of 40000 spherical
particles filling (0.1 m in diameter) cylindrical container. Insert loads during silo discharge were
found in good agreement (up to certain disc diameter) with results of laboratory testing. The ratio
of the mass of material in laboratory and numerical experiments was found a good scaling
factor.
Key words: Granular mechanics, inserts load, DEM

1. Introduction
Inserts are commonly used as construction members in storage silos during filling (to improve
the mixing of material) or discharge (e.g. to modify flow pattern or decrease dynamic loads shift).
Even though the first experiments with such objects are dated over 40 years ago (Johanson,
1966; Johanson, 1967), questions of estimation of loads on objects embedded in granular
material are still not fully covered by standard design codes (Wojcik & Tejchman, 2008).
Numerous attempts have been undertaken to measure or/and calculate insert loads based on
continuum approach (e.g. Johanson, 1966; Walker, 1966; Tsunakawa & Aoki, 1975; Strusch &
Schwedes, 1994; Ding et. al., 2011). However the continuum approach, successfully used for
calculation of the stress distribution inside grain silo, fails to describe the phenomena involving
several particle dimensions, anisotropy of load or disturbed discharge. For such purposes, the
family of numerical methods (as FEM or DEM) is considered a promising alternative (e.g.
Schuricht et. al., 2008).

The objective of the reported project was to measure vertical loads on disk embedded in wheat
stored in a model silo, to perform numerical simulation of the laboratory testing and to compare
the results of testing and simulations.



2. Methods and Materials
2.1. Laboratory testing
The experiment was carried out in a cylindrical model silo 1.5 m high, 0.4 m in diameter (Fig. 1).
The silo wall (1) was supported on three adjustable stands (3) resting on steel rack (2) at an
angular distance of 120. The smooth galvanized steel wall was covered with abrasive cloth of
100 grit, to increase wall friction and to reflect the conditions observed in a full scale silo made of
concrete or corrugated steel. A flat silo bottom (4) was placed on a steel tripod (5) of adjustable
height. Under the silo bottom, moveable container (6) was placed to hold discharging material.
Wheat was transported using vacuum feeder (7), from which the model silo was filled. Plate (8)
was used to fix three load cells (9) used to measure insert loads. The load cells were located at
the same distance from the silo axis as discs fixings and connected with them with steel strings.
Data acquisition unit was used to supply excitation voltage of load cells and to measure vertical
load. Signal from the unit was transferred and recorded at a PC. Load cells type KMM30
(Omega Engng.) of nominal load of 500 N +/- 2% were used and their readings were recorded
every 0.8 s until discharge was complete.

Figure. 1. Laboratory set-up to measure vertical load on insert immersed in wheat grain during
discharge: a) real view, b) scheme of the axial cross-section of the bin. Numbers are described
in text.
Test material was wheat cv. Nawra of moisture content of 11.5% (w.b.) and bulk density of
826.7 kg/m
3
. Effective angle of internal friction (=29) was measured using direct shear test
following Eurocode 1 (Eurocode, 2006) procedure. Coefficients of friction between grain and the
insert (galvanized steel),
c
=0.16, and between grain and silo walls, =0.43 were taken from
earlier measurements (Molenda et. al, 2001).
a) b)
Tests were conducted for five different diameters of the discs, chosen in a way that its surface
area as compared to the middle-sized disc is changing as follows; 0.25, 0.5, 1, 2, 3. The five disk
inserts, with ratios of disc diameter to silo diameter d/D of: 0.188, 0.253, 0.375, 0.55 and 0.675,
were axially suspended at height h of h/D = 0.5 (see Fig. 2b). The model silo was axially filled
with approximately 150 kg of wheat using vacuum feeder. After filling the grain was allowed to
equilibrate during a rest period of 20 minutes. Static load was taken as an average of 20
subsequent measurement cycles. Next, the discharge orifice (0.025 m in diameter) was opened
and discharge commenced with discharge rate of 9.3 kg/min that resulted in the movement of
grain along the wall with the velocity of approximately 0.0015 m/s. Testing for each disc diameter
was performed in three replications, each time rotating the insert with suspension by angular
distance of 120 to minimize an influence of possibly non identical conditions of the suspension
(e.g. unequal length or elasticity of strings, human error etc.).
2.2 Simulation Setup
Numerical simulations to estimate vertical load of grain on the disc were carried out using open
source code PAPA (Parallel Algorithm for PArticle flow) (Schwarzer, 2004). Numerical
experiment was performed in a model silo 0.1 m in diameter and 0.33 m high (Fig. 2). The silo
was divided into a two parts: lower chamber 0.25 m high was used as the one in which inserts
were placed and for that essential computations were performed, while the upper chamber was
used as a preparatory chamber, where particles were randomly generated and settled due to
gravity. Chambers were separated by a conical hopper with an angle between walls and
horizontal plane of approximately 8. Loading orifice had diameter D
o
=0.036 m and a height
H
s
=0.015 m.

Figure 2. Scheme of the silo used in numerical experiment
Mono-sized spherical particles (with diameter d
p
=0.0038 m calculated based on volume
equivalent to volume of the wheat kernel) with properties comparable to wheat kernels were
used. Particle density was
p
of 1290 kg/m
3
and silo wall density
s
of 7800 kg/m
3
(for steel),
Young modulus for particles E
p
=868 MPa and for walls E
s
=210
5
MPa and Poisson ratio
c
=0.18
and
s
=0.3 respectively for particles and walls were taken after J. Wiacek Ph.D. thesis (Wiacek,
2008). Kuwabara-Kono contact model was chosen with damping coefficients and B
responsible for energy dissipation during particle rebound were assessed comparing maximal
heights which particle reached after rebounding in case of real tests and DEM simulations. Vales
of those parameters are assumed as follows: =0.15 for both materials and B=310
5
for steel,
while B=910
3
for wheat kernels. Dimensions of the silo (see Fig. 2) and the insert sizes were
taken to correspond to experimental ones. The diameter of the discharge/loading orifice
(equivalent to approximately 9.5 particle diameters) was minimum to ensure undisturbed flow.
The silo diameter equivalent to approximately 26.3 particle diameters was large enough to
reduce disturbance from the wall on simulated grain inside the silo. The smallest, basic diameter
of the disc was chosen to be bigger than discharge gate and the gap between insert and wall
(approx. 8.2 particle diameters) was wide enough to minimize flow disturbances.

3. Results
Figures 3-5 present loads on inserts immersed in grain silo during discharge measured in
laboratory experiment (black line, values on left vertical axis) and obtained from DEM
simulations (points, values on right vertical axis). The ratio of mass of the grain above the insert
from laboratory testing to corresponding mass in numerical experiment was used to compare
experimental and numerical results quantitatively.
In case of the smallest discs, results obtained using DEM were in very good qualitative and
quantitative agreement with an experimental data (see Figure 3).
For the medium disc sizes up to the height of grain column point z
0
1.2 h/D, DEM results
follow fairly close experimental data, but above that point the fit is distinctly worse (Fig. 4).
With further increasing size of the disc, DEM results clearly overestimate experimental data. Gap
between the insert and silo walls became too narrow to allow for reliable flow of the material
(Fig. 5).

Figure 3. Vertical load on disc insert immersed in wheat grain during discharge phase
(diameter 0.1875 d/D)

Figure 4. Vertical load on disc insert immersed in wheat grain during discharge phase
(diameter 0.375 d/D)

Figure 5. Vertical load on disc insert immersed in wheat grain during discharge phase
(diameter 0.675 d/D)
4. Conclusions
Values of estimated vertical loads on discs during discharge corresponded fairly well to the
experimental results. Ratio of the mass of material in laboratory and numerical experiments was
found to be a good scaling factor.
Results of simulations and laboratory testing were in fairly good agreement in the case of two
smallest discs, while in case of the largest discs (d/D=0.55 and higher) numerical values
overestimated experimental results over two times. This is probably an effect of arching due to
low distance between the disc and the wall (less than ten particle diameters in simulations) that
precluded fluid-like behavior of flowing bedding.
Kuwabara-Kono contact model used for numerical simulations was adopted as the one which
was proven working very well in case of two colliding spheres as well as in the case of sphere
colliding with flat surface. Performed computations showed that contact model did not cope with
increasing vibrations at lower level of the bed caused by mass of the upper layers of the grain.
Damping at low relative speed between particles appeared to be too weak
Acknowledgement
This research was partially supported by the Ministry of Science and Higher Education, Poland,
Grant No. N N310 305739.
References
Ding S., Dyroy A., Karlsen M., Enstad G. G., Jecmenica M. (2011). Experimental Investigation of
Load Exerted on a Double-Cone Insert and Effect of the Insert on Pressure Along Walls of a
Large-Scale Axisymmetrical Silo. Particulate Science and Technology, 29(2), 127-138.
Eurocode 1, Actions on structures. Part 4. Silos and tanks. EN 1991-4, 2006.
Johanson J. R. (1966). The use of flow-corrective inserts in bins. J. Engng Ind., Trans. ASME,
88, 224-230.
Johanson J. R. (1967/68). The placement of inserts to correct flow in bins. Powder Technol., 1,
328-333.
Kuwabara G., Kono K. Restitution Coefficient in a Collision between Two Spheres. Jpn. J. Appl.
Phys. 26, 1230-1233 (1987).
Molenda M., Horabik J., Ross I. J. (2001). Comparison of loads on smooth- and corrugated-wall
model grain bins. Int. Agrophisics, 15, 95-100.
Schuricht T., Furll C., Enstad G. G. Experimental and calculated loads on ,,cone in cone'' inserts.
Proceedings RELPOWFLOIV, 10-12 June 2008.
Schwarzer S. PAPA - Parallel Algorithm for PArticle flow problems, Stuttgart, 2006.
Strusch J., Schwedes J. (1994). The use of Slice Element Methods for Calculating Insert Loads.
Bulk Solids Handling 14(3): 505-512.
Tsunakawa H., Aoki R. (1975). The Vertical Force of Bulk Solids on Objects in Bins. Powder
Technology, 11, 237-243.
Walker D. M. (1966). An approximate theory for pressures and arching in hoppers. Chem.
Engng Sci. 21, 975-997.
Wiacek J. Discrete element modeling of quasi-static effects in grain assemblies, Ph.D. thesis.
Institute of Agrophysics PAS , Lublin, 2008.

Dust explosion risk during graphite-moderated reactors
decommissioning
Sophie Trlat
1*
1
Institut de Radioprotection et de Sret Nuclaire (IRSN), B.P. 17 92262 Fontenay-aux-Roses
Cedex France
* Corresponding author. Email: sophie.trelat@irsn.fr

Abstract

This document discusses about dust explosion hazard concerning graphite-moderated reactors
decommissioning. Concerns about initiation and effects of a dust explosion scenario lead to a
request for more data concerning parameters such as: graphite dust particle size and its
chemical composition, ignition energy during decommissioning

Keywords: dust explosion, graphite, graphite-moderated reactors decommissioning

1. Introduction
The UNGGs (Natural Uranium Graphite Gas) or graphite-moderated reactors were the first
generation of French nuclear power reactors. They were developed in the 1950-60s in France.
Such reactors used natural uranium fuel, graphite as a neutron moderator and carbon dioxide as
a coolant. The moderator components consist of large graphite blocks (several thousands of
tons of graphite per reactor). All French reactors of this type have been shut down, but they
have not been fully decommissioned yet: indeed, fuel assemblies were taken out but graphite
moderators of these old reactors are still in place, waiting for the reactor dismantling, planned in
a near future. One of the graphite-moderated reactors decommissioning safety problems is the
explosion hazard of graphite dusts. A safe decommissioning is a lengthy process that requires
careful planning, close attention to technical details and a complete risk assessment of the plant.
Thus, nuclear graphite dust explosion hazard data are requested to perform a risk assessment.
In France and in other countries (the United Kingdom, Germany, Italy, Japan, etc), some studies
were carried out on decommissioning issues and on graphite dust explosion in order to examine
if graphite dust should present significant hazards for reactor decommissioning (Roubineau,
2006; IAEA, 2006). Some results are given in various publications mostly focused on the
determination of explosion sensitivity parameters (minimum ignition energy and minimum
explosible concentration) and explosion severity characteristics (maximum explosion pressure
and maximum rate of pressure rise) of graphite dusts. However, some works are still in progress
in France in order to enhance knowledge in this field. For example, concerns about the initiation
and effects of a dust explosion event in a graphite-moderated reactor during decommissioning
lead to a request for more data concerning parameters such as: particle size of graphite dust
encountered in UNGGs and its chemical composition, ignition energy during decommissioning,
etc.
This paper provides an overview of the knowledge about dust explosion hazard during graphite-
moderated reactors decommissioning. It aims to highlight the main findings of such studies and
to suggest research axes.

2. Characteristics of French UNGG reactors
French type UNGG reactors were mainly constructed in France, with one export sale: an UNGG
in Spain (Vandellos). Nine UNGG plants were built in the 1950-60s in France; all of them are
now shut down (Table 1), but they have not been fully decommissioned yet. Indeed, the fuel
assemblies were taken out. However, graphite moderators are still in place.

The earlier units were Chinon and Marcoule, with heat exchangers outside the main pressure
vessel. Later units (Saint-Laurent-des-Eaux and Bugey) moved these heat exchangers inside
the pressure vessel. Fig. 1 illustrates these two kinds of reactors. Power reactors Chinon,
Saint-Laurent-des-Eaux and Bugey were operated by EdF, whereas Marcoule units were
operated by the French atomic energy organization Commissariat lEnergie Atomique (CEA).

TABLE 1: French UNGG reactors characteristics (Bonal et al., 2002)
Unit Owner Power (MWe) Construction start Construction achievement Shut down
Marcoule G1 CEA 7 1952 1956 1968
Marcoule G2 CEA 38 1955 1958 1980
Marcoule G3 CEA 38 1956 1959 1984
Chinon A1 (CHA1) EdF 70 1957 1963 1973
Chinon A2 (CHA2) EdF 180 1959 1965 1985
Chinon A3 (CHA3) EdF 360 1961 1966 1990
Saint Laurent A1 (SLA1) EdF 390 1963 1969 1990
Saint Laurent A2 (SLA2) EdF 450 1966 1971 1992
Bugey EdF 540 1965 1972 1994

Nota: the general designs of the Marcoule reactors are quite different of those of the power
reactors. So, for this paper, examples of graphite dust explosion risks are related to these power
reactors.



FIGURE 1: View of an UNGG reactor. In the left, a reactor with heat exchangers outside the
vessel; in the right, a reactor with heat exchangers inside the vessel

Decommissioning planning can be divided into two main periods: one first stationary period
which takes place now, before the beginning of the dismantling of graphite blocks, and a second
period which will consist in removing graphite blocks from the core. Graphite dust explosion may
potentially occur in both periods.
In the first one, it could happen as follows for reactors with heat exchangers inside the vessel:
the large graphite blocks which constitute the moderator components (several thousands of
tons of graphite per reactor) may fall in case of an earthquake of high intensity or due to a
large corrosion of the internal metallic structures,
this event could possibly generate an explosible dust cloud and bring enough energy to ignite
it.
In the second period, there will be dismantling operations to remove graphite blocks from the
core, and in case of accident, there is a risk of dust explosion. In fact, the methodology for
graphite removal strongly depends on the design and supporting facilities available. The option
of flooding the vessel to provide a radiological barrier has been proposed for Bugey, SLA2,
SLA1 and CHA3, providing the pre-stressed concrete containments can be shown to resist the

water pressure. However, in the case of the other reactors, such as CHA2 and CHA1, which
have a different design, adapting the present structure to ensure water tightness is considered to
be technically impossible. Then, in the first period, dust explosion risk remains for all six French
power UNGG reactors; in the second period, the risk only remains for CHA2 and CHA1 because
the others will be dismantled under water. Therefore, according to IRSN, it is worth looking at all
possible scenarios of graphite dust cloud explosion within French reactors undergoing
decommissioning and examining precisely if the critical conditions for such an explosion to occur
can be achieved. The following conditions must be satisfied simultaneously for a dust explosion
to occur (Eckhoff, 2003): the dust must be combustible; the dust must be airborne; a sufficient
energy ignition source is requested; the dust concentration must lie within the explosible range;
the particle-size distribution must be sufficient for flame propagation; there must be confinement.
Thus, it is necessary to determine: explosibility of graphite dust; ignition characteristics of
graphite dust (minimum inflammability energy); energy that might be brought by falling graphite
blocks; chemical composition and particle-size distribution of the dust likely to be found on
surfaces within reactor vessels; flame speeds and temperatures in case of a graphite dust
explosion; overpressures generated by a graphite dust explosion.

3. Main findings from graphite dust deflagration studies
The main source of international data with particular reference to the decommissioning of
graphite moderated reactors has been given by Wood (2007), the EPRI Project Manager. In this
report, the relevant work conducted in each participating country (France, Italy, Japan and the
United Kingdom) is presented in detail. In order to answer the questions which arose in section
2, the relevant findings are summarized and discussed in the following paragraphs, along with
some additional data from another author, Denkevits et al, (2006). The test programs presented
in the EPRI documents (Wood, 2007) have been mainly conducted by France (Carson, 2000;
Roubineau, 2004; Roubineau, 2006), Italy, Japan and the United Kingdom into the behaviour
of nuclear-grade graphite dusts in the presence of an ignition source with a variety of different
graphite grades. So, one of the main differences is the granulometry. Table 2 presents a
comparison of the particles sizes of the dusts used in the different studies (supposed to
represent typical distributions of material within the reactor vessels) reported in (Wood, 2007). It
can be seen that the UK dust and the Japanese fine dust, are much finer than the samples
used in the other studies from (Wood, 2007), the Italian dusts seem to be the most coarse.

TABLE 2. Comparison of the granulometries: estimation of the particle size at the peak of the
distribution and of the proportion of the dust lying below 5 m (Wood, 2007)
Italy unirradiated graphite dust peak > 100 m; 1.5% < 5 m
Italy reactor dust peak ~ 20 m; 4% < 5 m
Japan coarse dust defined as < 300 m
Japan fine dust defined as < 25 m
UK peak 6 - 7 m; 33% < 5 m
France (INERIS) peak ~ 60 m; 7% < 5 m
France (CNPP) peak 23 30 m; ~ 10 % < 5 m

In order to make an overall summary of the explosion parameters of pure graphite dusts, the
differences in experimental utilities used in the studies reported in EPRI document (Wood, 2007)
have to be commented:
Italian work was conducted by Stazione Sperimentale per I Combustibili using an ASTM
standard 20-litres vessel (ASTM E1515-03a; ASTM E1226-05) with two 5kJ igniters
providing a total 10 kJ ignition energy,
Japanese tests utilized a Hartmann-style apparatus of 30-litres capacity (Wood, 2007), they
used ignition energies from 10 J up to 10 kJ,

In the United Kingdom, some tests were realized by the UK Fire Research Station (FRS) in a
20-litres spherical vessel, with a 100 J capacitive discharge. The University of Leeds (United
Kingdom) employed a standard 1-m
3
test vessel (constructed to the general specifications of
the ISO standard (1992) with chemical igniters of energy between 5 and 15 kJ and a
methane driver system for explosion parameters determination; some tests were also
conducted with a Hartmann apparatus using an arc of energy 4 J (Wood, 2007),
In France, INERIS (Wood, 2007; Carson, 2000) carried out some experiments in a standard
20-litres vessel using a pair of chemical igniters of total energy 10 kJ (each igniter has
individual energy of 5 kJ). CNNP works were then conducted in a 742-litres and 4 meter-
length horizontal plexiglass tube (50 cm diameter) with energy values from 60 to 180 kJ, this
apparatus aimed to study flame propagation.
Basic explosion parameters for nuclear-grade graphite dusts employed in the studies presented
in (Wood, 2007) are:
minimum explosible concentration: in the range 60 - 125 g.m
-3
,
minimum ignition energy: around 2 kJ over a range of practical dust concentrations (in the
range 60 - 1500 g.m
-3
),
maximum explosible concentration: definition of the maximum value is difficult from the data
available in (Wood, 2007), but French works (Carson, 2000) gave a positive result at 1500
g.m
-3
. So the maximum explosible concentration was supposed to be over 1500 g.m
-3
,
maximum rate of pressure rise in standard test: in the range 118 - 250 bar.s
-1
,
deflagration index Kst: in the range 14 - 76 bar.m.s
-1
,
maximum overpressure: in the range 5.43 6.7 bar.

The effects of other particules or additives in graphite dust were also investigated by Nuclear
Power Engineering Corporation (NUPEC) in Japan and reported in (Wood, 2007). According to
Japanese experiments, two additives iron powder and iron oxide (FeO), are able of acting as
inerters at the lower ignition energies and dust concentrations. From these results, EPRI report
(Wood, 2007) concluded that both Fe and FeO cause an inerting effect, in that the threshold
ignition energy for a deflagration is approximately doubled from 2 to 4 kJ. However, in these
reported tests, the precise concentration of Fe powder in the graphite dust and particle-size
distributions for the additives are not available.
Series of tests were also carried out (Denkevits et al., 2006; Denkevits & Dorofeev, 2003) in a
20-litres sphere facility (called DUSTEX facility) using two pyrotechnical igniters of 10 kJ total
release energy, to evaluate the explosion properties of pure graphite dusts (carbon content
higher than 99.8 % wt). The dust explosion parameters, as above, were measured. The effect of
the dust particle size on the explosion parameters was also investigated. The maximum
overpressures (Fig. 2) and pressure rise rates were measured as a function of characteristic
dust particle size, the size being tested were 4 m, 25-32 m and 40-45 m. From this graph,
we can see that the highest overpressure value (6.6 bar) characterizes the 4 m dust. This value
is in good correlation with the EPRI results from (Wood, 2007). Denkevits et al., (2006) also
identified that graphite dusts with particle sizes from 4 to 40 m were able to explode in a wide
range of concentrations. His criterion for dust explosion is that the maximum overpressure
should exceed 0.2 bar. This kind of criterion was not specified in any studies from EPRI
document (Wood, 2007). Then, as expected, the finest dust (4 m) has the lowest minimum
explosion concentration (70 g.m
-3
) and the lowest minimum ignition energy (1kJ), which is lower
than in EPRI report (2 kJ). Indeed, in EPRI report (Wood, 2007), Italian studies confirmed
positive results for an explosion with 350, 500 and 700 g.m
-3
concentrations and with 2 kJ
ignition energy. And Japanese results (Wood, 2007) identified a graphite dust explosion with
concentration below 250 g.m
-3
, 2 kJ ignition energy and particle size less than 25 m.


FIGURE 2: Maximum overpressure versus dust concentration for 4, 2532, and 4045 m
graphite dusts (Denkevits et al., 2006)
According to all these experiments, we can draw the following conclusions:
dust particle size is very influent on explosion properties. That is the reason why it is
important to bear in mind that any differences in fundamental explosion parameters in the
different studies may be related to the particle size distribution,
all the studies were performed with some apparatuses which differed in the method of dust
dispersing, in their capacity, also in the turbulence level of the dust cloud at the moment of
the ignition. This could explain why it was quite difficult to give a precise definition of the
basic explosion parameters from the data available (Denkevits et al., 2006; Wood, 2007)
this could also explain the differences in results for the same range of particle sizes between
two experiences performed in two different apparatuses,
the minimum ignition energy value seems to lie around 2 kJ (one case was got with a 1 kJ)
but we noticed that it was not possible to draw a curve representing this energy as a function
of dust concentration because of a lack of data. Moreover, according to Japanese results, it
was noticeable that the ignition energy seems to increase with the presence of other
particles, such as iron powder and FeO powder.

4. Application to French power UNGG reactors decommissioning
The licensee assessed the risk of graphite dust explosion and classified it as weak during
graphite-moderated reactors decommissioning. However, some information is missing to confirm
this conclusion; thus some works are still in progress in France at the moment. Dust samples
were collected within Bugey reactor. The estimated dust weight, which was estimated in 1997,
was several tons. Dust is mainly constituted by carbon (more than 60 % weight), iron (less than
35 % weight) and other reduced metals (under 1 % weight). If we suppose that all the dust
disperse within the vessel, the dust concentration would reach a value which lies in the
explosible dust concentration range. However, in order to assess precisely the dust explosion
hazard, it would be necessary to have precisions about dust granulometries and composition. If
it were the case, it would be interesting to assess more precisely the inerting effect on the lower
ignition energy, caused by the presence of other particles than graphite in dusts.
Then, concerning the ignition energy which could be generated by blocks falling, some research
works should be initiated to determine the part of energy that would be really dissipated and
transmitted as ignition energy. Some other works should be necessary to evaluate the quantities
and the sizes of graphite dusts generated during such a falling and during the removal of

graphite blocks (for dry dismantling). Last but not least, without assumptions about the ignition
source and if we suppose that a graphite dust explosion would occur in a reactor vessel, a more
detailed determination of the likely overpressures in the reactor geometry needs to be achieved.
The use of a numerical tool would lead to a request for knowledge about laminar flame speeds
and temperatures consecutive to a graphite dust explosion. Nevertheless, it may be noted that
the containment vessel of a large UNGG reactor is made of several meters of pre-stressed
concrete.This risk will be evaluated by IRSN at the different stages of the reactor dismantling, on
the basis of the safety analysis performed by the licensee.

5. Conclusions
Some test studies have been performed by France, Germany, Italy, Japan and the United
Kingdom concerning graphite dust explosion risk. Basic explosion parameters, such as minimum
ignition energy, minimum explosible concentration, maximum explosible concentration,
maximum rate of pressure rise, deflagration index and maximum overpressure have been
assessed on the basis of different types of apparatuses. But, due to the differences between
these apparatuses and their operating conditions, it was quite difficult to give a precise definition
of the basic explosion parameters from the data available. So, data are needed to evaluate more
precisely this risk such as: graphite dust particle size and its chemical composition, ignition
energy during decommissioning

References
ASTM E1515-03a. Standard Test Method for Minimum Explosible Concentration of
Combustible Dusts.
ASTM E1226-05. Standard Test Method for Pressure and Rate of Pressure Rise for
Combustible Dusts.
Bonal J.P., Brsard I., Parraud S. Caractrisation du graphite des racteurs de la filire UNGG,
2002
Carson D. Etude de Risque Incendie-Explosion li au Graphite UNGG, Caractrisation du
Graphite, INERIS, France, 14th November 2000
Denkevits A., Dorofeev S. Dust explosion experiments. Measurements of explosion indices of
graphite dusts. Report on EFDA Subtask TW2-TSS-SEA5.2. FZKA 6872. 2003
Denkevits A., Dorofeev S. Explosibility of fine graphite and tungsten dusts and their
mixtures. Journal of Loss Prevention in the Process Industries 19; p.174180, 2006
Eckhoff, R.K. Dust Explosions in the Process Industries. Gulf Professional Publishing, USA,
2003
IAEA. Characterization, Treatment and Conditioning of Radioactive Graphite from
Decommissioning of Nuclear Reactors. Report -TECDOC-1521, Vienna, 2006
ISO 6184-1. Explosion Protection Systems Determination of Explosion Indices of
Combustible Dusts in Air, 1985 (also issued as European Standard EN 26184-1 [1991]
and British Standard BS 6713-1 [1986] under the title Explosion Protection
Systems: Part 1; Method for Determination of Explosion Indices of Combustible Dusts
in Air); as amended February 1992: ISBN 0 580 15204 9
Roubineau D. Premire Campagne: Approche Experimentale: Rsultats du Programme
dEssais Graphite EdF-CIDEN, CNPP Report No. PE 03 6570, 8th January 2004
Roubineau D. Explosivit des Poussires de Graphite: Deuxime Campagne dEssais
concernant le Risque Incendie et Explosion lors du Dmantlement des Centrales
UNGG. CNPP Report No. PE 03 6570-1, 14th March 2006
Wood C. Graphite Dust Deflagration: A Review of International Data with Particular
Reference to the Decommissioning of Graphite Moderated Reactors. EPRI Technical
Report CA. 1014797, 2007
1

An Arch-Based Model for Predicting Dynamic Pressure during Silo
Discharge

Xuduo Cheng
1
, Qiang Zhang
2*
1
School of Food Science and Engineering, Nanjing University of Finance and Economics,
Nanjing, China
2
Department of Biosystems Engineering, University of Manitoba, Winnipeg, MB, Canada R3T
5V6
*Corresponding author. E-mail: zhang@cc.UManitoba.ca
Abstract
A theoretical model is developed to predict dynamic pressure in storage silos during
discharge. It is hypothesized that dynamic pressure during discharge is resulted from the
formation of instantaneous arches in the silo. These unstable arches are not strong enough
to completely stop the material flow, but they may temporarily stop or slow down the moving
material in some location in the silo. When the moving material is stopped or slowed down,
the momentum is transferred to dynamic pressure. By considering the change in momentum
in the material above an arch, equations for dynamic pressure during discharge are derived.
To solve the equations, the granular material in the silo is divided into layers of equal mass
connected by elastic springs, and the acceleration of each layer is then solved numerically.
The dynamic pressures predicted by the current model are in good agreement with the
values published in the literature. The predicted dynamic pressure is affected by the
discharge rate, bulk modulus of elasticity of the granular material, internal friction, and wall
friction.
Key words: silo, discharge, dynamic pressure, arch.

1. Introduction
One of the most frequent causes of structural failures of silos is the dynamic pressure that
develops during discharge. Several hypotheses/theories have been presented by various
researchers to explain how dynamic pressure of discharge develops in silos, including:
pressure switch; and dilation (shear-induced volume increase) of granular materials.
J enike and his co-workers proposed that the active stress state changed to the passive
stress state in the silo during discharge and this pressure switch caused the pressure to
increase in the silo (J enike & J ohanson, 1968; J enike et al., 1973). However, the occurrence
of pressure switching in silos is difficult to predict, and therefore the value of dynamic
pressure is difficult to obtain.
Smith and Lohnes (1980) hypothesized that dilation of the stored material is the main cause
of the overpressure during discharge. Discharge of granular material in the silo causes the
material in different parts of the silo to move at different velocities, thus shearing occurs
within the granular material. Shearing may cause the granular material to dilate, and however
dilation in the lateral direction is restricted by the silo walls. Thus an increase in the wall
pressure is resulted. Based on the dilatancy hypothesis, Zhang and Britton (2003) developed
a two-dimensional microstructure mechanics model for predicting dynamic pressures during
discharge in silos. In their model, the granular material is considered to be an assembly of
elastic discs of equal size and dilation during discharge is calculated from the change in the
microscopic structure of granular material. A micromechanics theory is then used to transfer
the microscopic dilation to macroscopic stresses (dynamic pressures) on the silo walls.
Although the dilatancy theory is founded on the fundamental characteristic of granular
material, shearing actions within the granular material in the silo are difficult to quantify. To
2

date, no theories are universally accepted for predicting dynamic pressure of discharge in
silo design. Most design standards and codes recommend overpressure factors, which are
used to multiply the static pressure to obtain the dynamic pressure. The objective of this
study is to develop a theoretical model for predicting dynamic pressures in silos during
discharge by considering the arching effect.

2. Model Development
2.1. Hypothesis
Silo quake (vibrations or pulsations) during discharge is a well-documented phenomenon
(e.g., Benson et al., 2004). Benson et al. (2004) explained that silo quake was due to stick-
slip friction between the wall and the granular material in silos. However, pulsations have
also been observed in funnel flow bunkers, where the granular material at the walls does not
slip during discharge (Hardow, 1998). Therefore, we propose that it is also possible that
vibrations/pulsations during silo discharge are attributed to unstable arch formation and
destruction.
Arching is a well-known phenomenon in silos. When a material with relatively high
unconfined strength (e.g., powders) is discharged from a silo, there is a good possibility that
stable arches may form and material flow stops. However, for materials with relatively low
unconfined strength (e.g., cohesionless granular materials), arches may also form during
discharge, but are not stable (strong) enough to stop the material flow. It is hypothesized that
instantaneous arches form and are destructed alternately in silos during discharge, causing
vibrations/pulsations. When an instantaneous arch forms, it would temporarily slowdown or
even stop the material flow, which would exert an inertia force on the arch. This force on the
arch is then transferred to the silo wall, causing an increase in pressure on the silo wall.
2.2. Force equilibrium on instantaneous arch
Assuming the material flowing downward in a silo has a velocity v during discharge, when an
instantaneous arch (simplified as a horizontal layer of material) forms, the worst scenario is
that the material above the arch decelerates from v to 0. Taking a thin layer of material with a
thickness dy just above the arch (Fig. 1), and applying the theorem of momentum in the
vertical (y) direction, we obtain equations 1 and 2:



p
v
+dp
v



p
v

p
h

p
h



dW
D
FIGURE 1: Forces acting on a layer of material above arch
dy
3

dt gAdy Ap Cdy dp p A Adydv
v v v
] ) ( [ t + + = (1)
h
p t = (2)
where
=bulk density (kgm
-3
)
A =cross sectional area of silo (m
2
)
C =perimeter of silo cross section (m)
y =location of the arch (m)
p
v
=vertical dynamic pressure (Pa)
p
h
=horizontal dynamic pressure (Pa)
dv =increment of velocity of layer of grain (ms
-1
)
dt =time (s)
dy =thickness of material layer (m)
=coefficient of wall friction


Combing equations 1 and 2 yields:
(3)
It is assumed that the lateral to vertical pressure ratio is a constant (K):

(4)
(4)
Substituting equation 4 into 3 yields:
(5)
where
K =ratio of lateral to vertical dynamic pressures
Integrating equation 5, with the initial condition , the dynamic pressures are derived
as follows:
(6)
(7)
As discussed earlier, most design standards and codes recommend the use of overpressure
factors for estimating dynamic pressure of discharge in silo design. If the static lateral
pressure is calculated by using the J anssen equation (1895)
dy g
A
C
p
dt
dv
dp
h v
) ( + =
dy g
A
C
kp
dt
dv
dp
v v
) ( + =
0 ) 0 ( =
v
p
) 1 )(
1
1 (
y
R
k
v
e
dt
dv
g k
gR
p

+ =
) 1 )(
1
1 (
y
R
k
h
e
dt
dv
g
gR
p

+ =
h
v
p
K
p
=
4

(8)
the overpressure factor can be calculated by combing equations 7 and 8
(9)


2.3. Deceleration of material

The key variable in proposed model (equation 9) for calculating the dynamic pressure is the
material deceleration when an instantaneous arch forms. To predict the deceleration, the
granular material in the silo is divided into layers of equal mass (m) and adjacent layers are
connected by elastic springs (Fig. 2), following the dynamic model proposed by Cheng and
Zhang (2009).



FIGURE 2: Layered model to approximate granular material in silo.

If the material above the arch is divided into n layers in the cylindrical portion of the silo
(assuming that the arch forms in the cylindrical portion), the thickness and mass of each
layer are obtained respectively as follows:
) 1 (
y
R
k
hs
e
gR
p

=
)
1
1 (
dt
dv
g p
p
hs
h
+ = = o
dt
dv
5


10
11
where
D =silo diameter in the cylindrical portion (m)
H
g
=distance of the arch from the top material surface (m)
n =total number of layers above the arch in silo
d =thickness of each layer (m)
m =mass of a material layer (kg)

When the granular material is stored in the silo, the self-weight of the material causes the
material to be compressed and the compression of each layer is related to the stiffness of the
material (spring in the model, Fig. 1) as follows (Cheng & Zhang, 2009):

i =1,,n (12)
where

i
=compression of the ith layer (m)
k =(spring) stiffness for a material layer in cylindrical portion (Nm
-1
)

During discharge, the total system energy contained in the material in the silo consists of
kinetic, elastic potential, and gravitational potential energy, and each component is calculated
as follows:

(13)
(14)
(15)
(16)
where
z
i
=distance between the ith layer and arch =(H
g
- y) (Fig. 1)
T =kinetic energy (Nm)
V
e
=elastic potential energy (Nm)
n H d
g
/ =
n
H D
m
g
4
2
t
=
k i n mg
i
/ ) ( =
2
1
2
1
i
n
i
z m T

=
=
2
1
1
1
) (
2
1
i i i
n
i
i i e
d z z k V =

=
+

=
=
n
i
i g
mgz V
1
g e
V V V + =
6

V
g
=gravitational potential energy (Nm)
z
i
=coordinate of the ith layer (m)
dot =the first derivative with respect to time (ms
-1
)
g =gravitational acceleration (ms
-2
)
V =total potential energy (Nm).

The generalized forces acting on each layer are as follows:

) ( ) / (
1 1 1
nd p Cd z z Mg F
hs
+ =
(17)

) ( ) / ( id d nd p Cd z z F
hs i i i
+ = (i =2,,n) (18)
where
F
1
=generalized forces acting on the first layer (N)

F
i
=generalized forces acting the ith layer (N)
M =total mass of grain above the arch in silo (kg)

Once the energy components and the generalized forces are defined, the Lagrange equation
may be presented for the granular material above the arch in the silo as follows (Davies,
1980):

I =1,2,, (19)
Substituting equations 13-18 into equation 19, the dynamic equation for the material above
the arch in the silo is derived as follows:

= +
= + =
+
= + +

+
) ( ) / ( ) (
) 1 , , 2 ( ), ( ) / (
) 2 (
) ( ) / ( ) (
1
1 1
1 1 2 1 1
d p Cd z z kd z z k z m
n i id d nd p Cd z z
z z z k z m
nd p Cd z z kd z z k z m
hs n n n n n
hs i i
i i i i
hs

(20)
where
=velocity of ith layer (ms
-1
)
=acceleration (deceleration) of ith layer (ms
-2
).

The initial layer positions z
i
(0) are determined as follows:
(21)

i
i i i
F
z
V
z
T
z
T
dt
d
=
c
c
+
c
c

c
c
) (

i
z
i
z
)
`

= =
=
) , , 2 ( , ) 1 ( ) 0 (
0 ) 0 (
1
n i d i z
z
i

7

The initial layer velocity are determined as follows:
22

The displacement of each layer is determined as the difference between the current and the
initial positions:
i =1,2,,n (23)

Combing equations 20, 21 and 23, the dynamic equations of the system are rewritten as
follows:

= +
= + =
+
= +

+
) ( ) / ( ) (
) 1 ,..., 2 ( ), ( ) / (
) 2 (
) ( ) / ( ) (
1
1 1
1 1 2 1 1
d p Cd x x x x k x m
n i id d nd p Cd x x
x x x k x m
nd p Cd x x x x k x m
hs n n n n n
hs i i
i i i i
hs

(24)
where
=velocity of ith layer (ms
-1
)
=acceleration (deceleration) of ith layer (ms
-2
).

When the material above the arch impulse downward, it is assumed that the arch itself (layer
1) is stationary, or mathematically:
(25)
To facilitate the solving of equation 24, an intermediate variable is defined to reduce the
equation to the first order differential equation:
, i=1,2,,n (26)
Combing equations 24 , 25 and 26 yield:

+ =
= + +
+ =
+ =
= =

+ +
+
+
) ( ) / ( ) (
) 1 , , 2 ( ), ( ) / (
) 2 (
) ( ) / ( ) (
) ,..., 2 , 1 ( ,
1 2
1 1
1 1 1 2 1
d p Cd m x x x x
m
k
x
n i id d nd p Cd m x x
x x x
m
k
x
nd p Cd m x x x x
m
k
x
n i x x
hs n n n n n
hs i i
i i i i n
hs n
i n i

(27)
) 0 (
i
z
)
`

= =
=
) , , 2 ( , ) 0 (
0 ) 0 (
1
n i v z
z
i

) 0 ( ) ( ) (
i i i
z t z t x =
i
x
i
x

=
=
=
0 ) 0 (
0 ) 0 (
0 ) 0 (
1
1
1
x
x
x

i i n
x x =
+
8

Combing equations 21, 22 and 23, the initial conditions for the above equations are
determined as follows:
(28)
Solving the differential equation group 27, using the initial conditions given by equation 28,
displacement, velocity and acceleration of each layer can are obtained. The acceleration
(deceleration) of the layer just above the arch is as follows:
(29)

2.4. Determination of model parameters

The two important model parameters are the layer stiffness k and the velocity of material
movement v in the cylindrical portion. The layer stiffness is determined from the modulus of
elasticity of the granular material as follows:
(30)
where
E =modulus of elasticity of granular material (Pa)
E
v
=bulk modulus of elasticity of granular material (Pa)

The flow velocity of material in cylindrical portion is estimated from the discharge rate:
(31)
where
Q =volumetric discharge rate (m
3
s
-1
)
Q
Z
=mass discharge rate (kgs
-1
).
3. Results
A parametric study is performed for a 10-m diameter and 24-m height cylindrical silo filled
with a material with properties typical for grains (e.g., wheat) (Table 1). A range of values for
the angle of internal friction (), coefficient of wall friction (), and bulk modulus of elasticity
(Ev) are used to simulate various silo conditions for storing grains. Also three different
discharge rates of 6x10
5
, 1.0 6x10
6
, and 1.4 6x10
6
kgh
-1
are simulated.

= =
=
= =
+
+
) , , 2 ( , ) 0 (
0 ) 0 (
) , , 1 ( , 0 ) 0 (
1
n i v x
x
n i x
i n
n
i

2
x
dt
dv
=
d
A
E
d
EA
k
v
) 2 1 ( 3 v = =
A
Q
A
Q
v
Z
= =
9



TABLE 1: Material parameters used in model prediction
Value 1 Value 2 Value 3
Silo height (H), m 24.0
Silo diameter (D), m 10.0
Thickness of the layer (d), m 0.2
Bulk density (), kg/m
3
800.0
Angle of internal friction () 20.0

25.0

30.0

Coefficient of wall friction () 0.2 0.3 0.4
Bulk modulus of elasticity (E
v
), MPa 10.0 30.0 50.0
Poisson

s ratio (v) 0.4

Discharge rate (Q
Z
), 10
6
kgh
-1
(or kth
-1
) 0.6 1.0 1.4

The dynamic pressure during discharge increases as the discharge rate increases (Fig. 3),
and as the bulk modulus increases (Fig. 5). Both the dynamics pressure of discharge and the
static pressure increase as the coefficient of wall friction decreases (Fig. 7), and as the angle
of internal friction of grain decreases (Fig. 9).

The overpressure factor increases with increase in: (i) the depth of material (Figs. 4, 6, 8,
10); (ii) the discharge rate (Fig. 4); (iii) the bulk modulus (Fig. 6); and (iv) the coefficient of
wall friction (Fig. 8). The overpressure factor increases with decrease in the angle of internal
friction (Fig. 10). The predicted overpressure factors vary from 1.33 to 1.87 for the conditions
simulated in this study, and the values are within the range reported in the literature and
recommended by various silo design standards. For example, EP 433 of ASABE Standards
(ASAE, 2000) suggests an overpressure factor of 1.4 if the bin height to diameter (H/D) ratio
is greater than 2.0.


FIGURE 3: Predicted static and dynamic pressure of discharge for a 1024m (diameter
height) silo for different discharge rates (=800 kg/ m
3
, =25.0

, =0.4, E
v
=30.0MPa, v=0.4)

10


FIGURE 4: Predicted overpressure factor of discharge for a 1024m (diameter height) silo
for different discharge rates (=800 kgm
-3
, =25.0

, =0.4, E
v
=30.0MPa, v=0.4)





FIGURE 5: Predicted static and dynamic pressure of discharge for a 1024m (diameter
height) silo for different values of modulus of elasticity (=800 kgm
-3
, =25.0

, =0.4,
Q=1.0kth
-1
, v=0.4)





11


Figure 6. . Predicted overpressure factor of discharge for a 1024m (diameter height) silo
for different values of modulus of elasticity (=800 kgm
-3
, =25.0

, =0.4, Q=1.0kth
-1
, v=0.4)






Figure 7: Predicted static and dynamic pressure of discharge for a 1024m (diameter
height) silo for different values of wall friction (=kgm
-3
, =25.0

, E
v
=30.0MPa, Q=1.0kth
-1
,
v=0.4)






12


Figure 8: Predicted overpressure factor of discharge for a 1024m (diameter height) silo for
different values of wall friction (=kgm
-3
, =25.0

, E
v
=30.0MPa, Q=1.0kth
-1
, v=0.4)




FIGURE 9: Predicted static and dynamic pressure of discharge for a 1024m (diameter
height) silo for different values of internal friction angle (= kgm
-3
, =0.4, E
v
=30.0MPa,
Q=1.0kt/h, v=0.4)









13


FIGURE 10: Predicted overpressure factor of discharge for a 1024m (diameter height)
silo for different values of internal friction angle (=kgm
-3
, =0.4, E
v
=30.0MPa, Q=1.0kt/h,
v=0.4)


4. Conclusions

A predictive model based on force transfer through instantaneous arches adequately predicts
the dynamic pressure of discharge in silos for granular material. The predicted dynamic
pressure increases with the discharge rate; the bulk modulus of elasticity of the granular
material in the silo; and the coefficient of silo wall friction, but decreases with the angle of
internal friction of the granular material.

Acknowledgements
The authors thank the Natural Science and Engineering Council of Canada and the Natural
Science Foundation of J iansu Higher Education of China07KJ D130068for funding this
study.

References
ASAE. (2000). ANSI/ASAE EP433 DEC01: Loads Exerted by Free-Flowing Grain on Bins.
ASABE Standards. American Society of Agricultural and Bilogical Engineers, St. J oseph, MI,
USA.
Benson, K. M., Shandon, F. Q., Sankaran, S. & Rao, K. K. (2004). Silo music and silo quake:
granular flow-induced vibration. Powder Technology 145 (3), 190-202.
Cheng, X. & Zhang, Q. (2009). A predictive model for arch destruction in hopper-bottom bins
for cohesive granular material. Transactions of ASABE 52(6), 2039-2045.
Davies, A. J . (1980). The Finite Element Method: A First Approach. Oxford University Press,
New York, N.Y.
Hardow, B., Schulze D., & Schwedes J . (1998). An experimental analysis of the silo quaking
phenomenon, Proceedings of the 3rd World Congress on Particle Technology, Brighton,
England.
J anssen, H. A. (1895). Versuche ber getreidedruck in silozellen.Zeitschriff des Vereines
Deutscher Ingenieure 39,1045-1049.
14

J enike, A. W., J ohanson J . R. & Carson J . W. (1973). Bin loads part 3: Mass flow bins.
Transaction of ASME, J ournal of Engineering for Industry, February.
J enike, A.W. & J ohanson J .R. (1968). Bin loads. J ournal of Structure Division, Proceedings
of ASCE 94(ST4), 1011-1041.
Smith, D. L. O. & Lohnes R. A. (1980). Grain bin overpressures induced by dilatancy upon
unloading. ASAE Paper No. 80-3013, American Society of Agricultural Engineers, St.
J oseph, MI, USA.
Zhang, Q. & Britton M.G. (2003). A Micromechanics model for predicting dynamic loads
during discharge in granular material storage structures. Canadian Biosystems Engineering
45:5.21-5.27.
ANALYSIS OF THE PRESSURE EFECTS IN CORN DUST
EXPLOSION USING COMPUTACIONAL FLUID DYNAMIC
Janana A. Silva
1
*, Francisco C. Gomes
1
, Jefferson L.G. Correa
2

1
Engineering Department ,Federal University of Lavras, University Campus, Lavras, PO Box
3037 CEP 37200-000
2
Food Engineering Department ,Federal University of Lavras, University Campus, Lavras,
PO Box 3037 CEP 37200-000

*Corresponding author. E-mail: nainasjdr@hotmail.com
Abstract
Dust explosions happen very easily, causing loss of valuable material and people, although
several studies related to this kind of explosion it still presents a complex containing several
variables that affect the process. The creation of a model that embraces this kind of
explosion from different angles becomes an issue relevant to the present technological
environment. A tool that helps you translate the different behaviour of an explosion and that
has been used is a computational fluid dynamics. Since this, one can evaluate different types
of explosion. In this regard, we aimed at analysing computationally dust explosions of
different materials and check the behaviour of the burst in each of these. This used the tool
ANSYS CFX 11 to model together with data explosion and thermophysical properties
taken from literature. Mostly conditions related to pressure in an explosion in a storage unit.
In this sense the present study aimed to analyse the effect of pressure in a silo, which was
submitted to an explosion of dust grain of corn. After several steps of simulations in ANSYS
CFX 11.0, this study found that higher levels of pressure occurred in the separation between
the hopper and the body of the silo.
Key words: CFD, Grain dust explosion, Pressure, Computational model, Silo.

1. Introduction
The job in storage units will be able to very dangerous, when to be occur a dust explosion.
This risk affects storage units of grain, medicine and woven. This kind of explosion caused
by dust accumulation can be having unimaginable proportions, since these explosions can
occur in the chain causing loss of valuable personnel and equipment. A dust explosion brings
great fire and the release of various gases.
The risk of explosion in storage units mainly agricultural products, occur with greater
frequency, because this product is highly explosive and the process of loading, unloading,
crushing, and others may generate sparks or flames that initiate a dust explosion (Ramirez et
al., 2009).
Corn Dust explosions are so dangerous because corn dust is very explosive and generally
the moisture content in corn is smaller than other grains, therefore its facility the ignition of
cloud dust. Researches about dust explosions are so dangerous, because there is life risk in
the most of experiment. Then a good idea for analyze is the computational fluid dynamic
(CFD) , that can simulating a model accuracy of dust explosion .
Computational fluid dynamics (CFD - computational fluid dynamics) is a set of numerical and
computational tools used to solve, visualize and interpret solutions related to the flow of
fluids. Widely used in fields like physics, chemistry, mechanical engineering, biomedical
engineering and others (Hong et al.,2010; Skjold et al. , 2008; Versteeg,1996).

This paper aims to analyze the effect of pressure in a corn dust explosion using CFD that
can efficiently modeling a dust explosion with a larger number of variables in this
phenomenon.
2. Material and Methods
The procedural processes initiated in the choice of ANSYS CFX 11 as a tool of analysis of
computational fluid dynamics. First was chosen to analyze the environment, thus determining
the geometry of a silo of approximately 1 m3 as proposed in the standard ISO 6184:1
(ABNT, 2007), which can be seen in Figure 1

Figure 1: Geometry of the proposed silo

The geometry shown in figure 1 has been transferred to the ANSYS Workbench for it to be
transformed into a 3D geometry, which software meshing ANSYS ICEM CFD 11.0 supports.
The determination of the boundary conditions occurred in the tool ANSYS CFX 11 Pre, which
were determined only the conditions of entry in order to be entered the amounts necessary
for an explosion. This being set in the turbulent regime, which turbulence model K-Epsilon
was used as a model for stable and robust and has consolidated the use of simulations in
several similar experiments (Andersen et al, 2009; COLLECUTT et al, 2009; TANIGUCHI et
al, 2011; Zhang et al, 2011).
In all simulations were used two fluids, one with the properties of a particle of a powder
material and the other air. The model of thermal radiation component P1 is the air (to
differential) in which the radiation is assumed to be isotropic (Andersen et al, 2009;
COLLECUTT et al, 2009). Since this model the data are diagnosed at each stage. The
material properties of the silo galvanized steel was chosen because it belongs to most
currently constructed silo.
The first analysis was the relationship between the rate of methane grain related breathing
and increase in temperature. The definitions of the rate of methane in the atmosphere of the
silo have been quantified as 0.05%, 0.1%, 0.5% and 1% to 5.
Considering that the methane storage units is generated by anaerobic respiration of
microorganisms or deterioration of the grains (Couto, 2004). Methane is highly explosive so
adding the factor to the reaction between methane and air in a silo used the chemical
reaction of combustion viewed in equation 1.
(1)
The tests performed were aimed at verifying the reduction of temperature initial ignition of the
explosive dust cloud according to the increase of methane. The characteristic of the corn
dust particle chosen has 520 -1250 micron, minimum concentration of ignition 60 g/m3 and
400 C is the minimum temperature of ignition.
The properties for the definition of the wall of the silo have been defined according to the
material properties of the silo, galvanized steel, which are displayed in Table 1.

Table 1: Properties of the walls in the CFD model
Adiabatic Heat Transfer Adiabatic
Thermal radiation Opaque
Emissivity 0.28
Diffuse fraction 0.72

In order that in this study fluid Ar is of extreme importance that the explosion occurs, certain
properties have been deduced to hold a turbulent flow, for example, speed. The input
properties are shown in Table 2.
Table 2: Input values for the CFD simulation model
Terms Value Unit
Pressure reference 20 Bar
Speed 10 cm/s
Air density 1.185 Kg/m
3

Molar mass of air 28.96 Kg/mol
Specific heat of air 1004.4 J.Kg/K
Volume of CO2 0.0380 %
Volume of O
2
21 %

3. Results
In a first moment the research showed more pressure in superior part in silo, after this
moment the pressure is stabilized and more pressure is showed in hopper. This situation can
looked in the figure 1. This results demonstrate that silos project has to more emphasis to
develop strong hopper and distribute the pressure trough of the structure.

Figure 1: Visual behavior of the pressure in a silo for a dust explosion
From figure 1 it is observed that there is a greater action of the pressure situated in the
separation and body of the silo and in some situations such as at the beginning of a burst
major action is exerted on the hopper of the silo.
This action may be due to the way that the silo is designed, thus having a greater number of
tensions in your hopper. As a place where there is a greater amount of stored products also
exerting pressure on the structure of the silo.
The higher concentration of pressure in the silo of the explosion occurs at the beginning as
early in the burst there is a greater amount of oxygen and product to be consumed. After
some time the violence of the explosion decreases, thus there is a decrease of pressure in
the silo structure.
With this study it was observed that the junction between the hopper and the body silo must
be extremely increased and the use of other materials for the development of silos will
reduce the damage for structures in short bursts.

4. Conclusions
After running the study found that the computational fluid can form various abort dust
explosions. Although he has not natively some products that must be entered manually
ANSYS CFX has shown to be robust and able to bring different information about the same
phenomenon. Is important to point out that the current silos should be constructed so that
they can withstand an explosion of dust, for this model as proposed can be used as an aid in
developing new projects.

5. Acknowledgements
We would like to thank the Federal University of Lavras, Engineering Department, Graduate
program in systems engineering, FAPEMIG and CNPQ for supporting this project.

6. References
ABNT. 2007. ABNT NBR ISO 6184-1Sistema de proteo contra exploso. Parte 1:
Determinao dos indices de exploso dos ps combustveis no ar, Brasil.

Andersen,J.; Rasmussen,C.L.; Giselsson,T. and Glarborg,P. 2009. Global Combustion
Mechanisms for Use in CFD Modeling under Oxy-Fuel Conditions. Energy & Fuels .
10.1021/ef8003619

Collecutt,G.; Humphrey, D. and Proud , D. 2009. CFD simulation of underground coal dust
explosions and active explosion barriers, Seventh International Conference on CFD in the
Minerals and Process Industries CSIRO, Melbourne, Australia 9-11 December 2009

Couto, J.L.V. Risco no Trabalho em silos e Armazns.2004.
http://www.ufrrj.br/institutos/it/de/acidentes/silo.htm

Eckhoff, R. K. 2003. Dust explosions in the process industries. Vol.3. Elsevier Science. 2003.

Hong, S. ; Lee ,I.; Hwang,H.; Seo, I.; Bitog ,J.; Kwon, K.; Song ,J.; Moon, O.; Kim, K. and Ko,
H. 2010. CFD modelling of livestock odour dispersion over complex terrain, part I:
Topographical modelling , Biosystems Engineering. doi:
10.1016/j.biosystemseng.2010.12.008

Ramrez, A; Torrent, J.G and Aguado, P.J. 2009. Determination of parameters used to
prevent ignition of stored materials and to protect against explosions in food industries,
Journal of Hazardous Materials.

Skjold,T.; Arntzen B. J.; Hansen O. R.; Taraldset O. J.; Storvik, I. E. and Eckhoff, R. K. 2008.
Simulating dust explosions with the first version of DESC. Process Safety and Environmental
Protection . 10.1205/psep.0423

Taniguchi,M.; Yamamoto,K.; Okazaki,T., Rehfeldt,S. and Kuhr,C. 2011. Application of lean
flammability limit study and large eddy simulation to burner development for an oxy-fuel
combustion system, International Journal of Greenhouse Gas Control.
10.1016/j.ijggc.2011.05.008

VERSTEEG, H.K and Malalasekera, W. 1996. An introduction to computational fluid
dynamics : The Finite Volume Method. Logamam

Zhang Q.; Pang L. and Zhang S.X. 2011. Effect of scale on ame speeds of methane -air
Journal of Loss Prevention in the Process Industries. 10.1016/j.jlp.2011.06.021

A note on load combinations for silos
Jrgen Nielsen
1
*, J. Michael Rotter
2
, John D. Srensen
3
1
Aalborg University, Danish Building Research Institute, Dr Neergaards Vej 15,
DK-2970 Hrsholm, Denmark
2
Institute for Infrastructure and Environment, School of Engineering, Kings Buildings,
University of Edinburgh, Edinburgh EH9 3JL, Great Britain
3
Aalborg University, Sohngaardsholmsvej 57, 9000 Aalborg, Denmark
*Corresponding author. E-mail: jni@sbi.aau.dk
Abstract
Silos are subject to several types of actions or loads. In the current Eurocode, EN 1991-4
(2006) [3] , the following are listed: loads from the stored material, wind loads, snow loads,
imposed loads and deformations, thermal loads, seismic loads, and loads from dust
explosions.
In the practical design of a silo, these actions must be combined according to the rules in
EN 1990 (2002) [2], which implies that a series of design situations must be considered and
the corresponding combination of actions selected, generally with a principal load and
accompanying loads. Finally, each accompanying load must be scaled to a value that
accounts for the reduced probability of simultaneous occurrence of the combined actions.
This scaling factor is described as the load combination factor.
When compared with similar developments of combination factors for other types of
structure, silos present some significant additional challenges. A first concern is that the key
features that must be considered in design depend on the expected mode of failure of the
structure. The actions and their combinations must be defined in such a way as to address
the critical stress states in the structure that can cause failure modes by shell bending or by
shell buckling. As a result, different load combinations are needed for different types of silos
(concrete or metal, squat or slender etc.) and for silos designed for different operating
conditions (flow pattern, method of discharge etc.). Apart from understanding the nature of
loads on silos, this aspect calls for an acute awareness of the structural consequences of
each load and a clear understanding of the safety philosophy. A second concern relates to
seismic loads, where there must be a division between on-ground and elevated structures,
since the seismic response is completely different (Trahair et al., 1983) [7], and different
degrees of partial filling can be critical for an elevated silo. Further, both the properties of the
stored solid and the operating conditions impact on many aspects of the statistical
assessment required to define load combination factors in a rational manner.
In two annexes of the current EN 1991-4 (2006) [3], load combinations and combination
factors are prescribed. However, they do not account for the differentiations described above
and these annexes were developed before several of the other silo relevant actions had even
been defined in the relevant Eurocode.
The paper discusses the current rules for load combinations in EN 1991-4 (2006) [3] and
offers some suggestions for amended rules to address these and other issues.

Key words: Silo, Loads, Load Combinations, Design, Standard

1. Introduction
Silos are subject to several types of actions or loads. In the current Eurocode, EN 1991-4
(2006) [3], the following are listed: loads from the stored material, wind loads, snow loads,
imposed loads and deformations, thermal loads, seismic loads, and loads from dust
explosions.

In the practical design of a silo, these actions must be combined according to the rules in
EN 1990 (2002) [2], which implies that a series of design situations must be considered and
the corresponding combination of actions selected for each, generally with a principal load
and accompanying loads. Finally, each accompanying load must be scaled to a value that
accounts for the reduced probability of simultaneous occurrence of the combined actions.
This scaling factor is described as the load combination factor.
It is a general principle in the European structural standards, the Eurocodes, that the three
items: safety considerations, the specification of loads, and the design of structures, may be
considered independently of each other. This means that safety considerations, as defined in
EN 1990 (2002) [2], are valid for all types of loads and all types of structures made of
different materials (concrete, steel, aluminium, timber etc.) and that the loads specified in the
action standards are valid for all types of structure.
When it comes to the practical design of silos, this philosophy is challenged for at least three
reasons. Firstly, the structural response to different load conditions is different for different
types of silo. Typically, the ultimate limit state for a concrete silo involves shell bending as
the failure mechanism, but in steel silos shell buckling is the commonest failure mode.
Secondly, silos are used for different purposes with very different operating conditions.
Typically, chemical process silos are run continuously whilst storage silos may be used to
store materials only for a short period each year. This also implies that very different sizes
and shapes are economic for different applications. Thirdly, the extreme load from a stored
material is likely to occur on a large part of the structure every time a full silo is subject to
discharge, and this extreme discharge load may be locked in if discharge is arrested before it
is complete, potentially making the extreme discharge load persistent rather than intermittent.
The general rules for the combination of loads are prescribed in EN 1990 (2002) [2], but load
combinations and the values of combination factors are temporarily prescribed in two
annexes of the current EN 1991-4 (2006) [3] and are, according to the principle mentioned
above, to be transferred to EN 1990 (2002) [2]. However, the current rules do not account for
the differentiations described above and these annexes were developed before several of
the silo relevant actions had even been defined in the relevant Eurocodes. Furthermore, the
authors of this paper have not found any publication which provides a sound technical
discussion of the considerations that should provide the basis for these rules for silos. Such
a discussion is begun here, with a focus on the current rules for load combinations in
EN 1991-4 (2006) [3] and some suggestions for amended or supplementary rules.
2. Combination rules
The general rules for verification of structures are prescribed in EN 1990 (2002) [2]:
It shall be verified that, in all relevant design situations, no relevant limit state is exceeded
when design values for actions or effects of actions and resistances are used in the design
models. It is required that the individual actions should be combined to form critical load
cases. However actions that cannot occur simultaneously, for example due to physical
reasons, should not be considered together in combination.
The rule for load combinations says that each combination, a load case, shall include:
All permanent actions
A leading variable or accidental action (typically giving a name to the load case)
Actions which have some probability of occurring simultaneously, termed
accompanying actions, each with its magnitude reduced by the load combination
factor, .
Specific rules for the verification of ultimate limit states and for serviceability limit states are
prescribed in detail in EN 1990 (2002) [2]. Different values of the combination factors are
used in these rules. The following combination factors are relevant to the discussion below:

1.
0
(Combination value) factors are chosen so that the probability that the load effects
caused by the combination will be exceeded is approximately the same as that for the
characteristic value of an individual action.
These factors are used in the ultimate limit design for persistent or transient design
situations and for the serviceability limit state.
2.
1
(Frequent value) factors are chosen so that either the total time, within the
reference period, during which it is exceeded is only a small given part (typically 1%)
of the reference period, or the frequency of it being exceeded is limited to a given
value.
These factors are used in Accidental design situations.
3.
2
(Quasi permanent value) factors are chosen so that either the total period of time
for which it will be exceeded is a large fraction of the reference period (and typically
approximately equal to the mean value of the load).
These factors are used in accidental design situations, in seismic design situations
and when evaluating the effect of differential settlements.
The combination rules and their corresponding factors are notionally based on Turkstras rule
with the Ferry-Borges-Castanheta load model [4] (ISO 2394,1998), which leads to the
equations in EN 1990 (2002) Annex C [2]. This model is based on the assumption that the
variable loads can approximated by a number of statistically independent realisations of
equal time duration and that the annual maximum load can be modelled by a Gumbel
distribution. The uncertainty of silo loads is dominated by model uncertainties and almost no
statistics are available to determine either which statistical distribution is relevant or the
magnitude of the variability. Thus current knowledge of silo loads cannot satisfy the
assumptions that underlie the EN 1990 approach to load combination factors. Therefore
currently the load combination factors can only be chosen by using engineering judgement
based on knowledge of the structural behaviour of silos and the mode of their operation.
With a focus on ultimate limit states, this paper discusses which actions should be combined,
and indicates appropriate magnitudes for the combination factors for different types and uses
of silos.
3. Some characteristics of silos and their operation
As further background to the discussion, some characteristics of silos and their usage are
described here. Further extensive information may be found in Brown and Nielsen (1998) [1].
3.1. The size of the silo
Silos are built in very different sizes. Some are small structures with a capacity below one
tonne of stored particulate solid, while others are huge structures with capacities exceeding
10 000 tonnes. In the discussions undertaken during the development of EN 1991-4 (2006)
[2] it was realized that a higher degree of rigour is necessary in the action assessment of
larger silos. The following classification was therefore introduced:
1. Action Assessment Class 1, consisting of silos with a capacity in excess of 10 000
tonnes, or if there is a large eccentricity in the outlet, or the inlet in a squat silo, and
the capacity is in excess of 1 000 tonnes.
2. Action Assessment Class 2, consisting of all silos not placed in another class.
3. Action Assessment Class 3, consisting of silos with capacity below 100 tonnes.
3.2. Silo structural behaviour
Some silos are thin walled shell structures carrying their loads predominantly by membrane
stresses, with buckling in compression or through-thickness rupture of metal plate as the
dominant modes of failure [4] (Rotter, 2001). These structures are typically made of steel,

aluminium or stainless steel. Other silos are relatively thick-walled shells in which slight
asymmetries of loading lead to bending moments as a critical feature of the structural
behaviour. These silos are typically made of reinforced concrete, but may also be made of
other materials. A fuller discussion of this distinction is presented in Rotter et al. (1986) [5].
Because the response of the structure, and the ultimate limit state induced, are so different
for these two types of silo, in drafting EN 1991-4 (2006) [3] it was decided that it is necessary
to separate silos into two types: thick-walled and thin-walled. This decision permitted load
models to be used that relate well to the critical loading conditions for the failure mechanism
that is likely to be controlling in these circular shell structures (patch loads).
3.3. Geometries of silos
The relationship between the height and diameter (or horizontal characteristic dimension for
non-circular silos) plays a major role in determining the loading on the silo walls, and this
aspect is brought out in EN 1991-4 (2006) [3] through a rather complex separation of
categories. The aspect ratio is termed the slenderness of the silo, with categories from very
slender to squat and retaining geometries. Again, the ultimate limit states are very different
in these different geometries, so that a load combination that is critical for one shape may be
irrelevant or unlikely in another. Furthermore, a squat silo carries a considerable part of the
weight of stored solids directly to the bottom, so that the scatter in the wall load is much
smaller, leading to a small coefficient of variation in the definition of the load.
Furthermore, many silos are elevated. For these structures, much of the well-defined weight
of the stored solid may be carried by a hopper, which is susceptible to quite different ultimate
limit states. This situation suggests that a lower partial coefficient could be used for well-
defined loads, but such a discussion is beyond the scope of this paper.
Silos constructed in either reinforced or prestressed concrete may be either single cylindrical
vessels, or may be combined into interconnected batteries of silo cells to form a block that
has very specific characteristics. In particular, load cases involving some filled cells and
others empty, and conditions involving imposed deformations, need careful treatment in the
consideration of load combinations. It is not clear how differentially filled cells can be
incorporated into the load combination model without making special provision for batteries.
3.4. Operating conditions and usage
The mode of usage of the silo has a major impact on the loads and their probability of
occurrence. The silo may be at an industrial plant, where it is being filled and discharged
almost continuously and is rarely full and rarely empty, but no data will be available to the
designer on this issue. Or it may be a grain silo that is filled once a year after the harvest
and then emptied once a year. Or it may be at a harbour, where it is filled to capacity
awaiting the arrival of a ship, at which point it is fully discharged and then filled progressively
by trains arriving to refill it. The latter could have 10 fillings per year or 3 or 50. The designer
is unlikely to know in advance. So silos are much more varied in usage than buildings or
bridges, and the only way that this could be dealt with in choosing the load combination
factor would be to have categories of usage relating to the number of expected filling and
discharge cycles to be expected annually.
Further, the stored solid and the manner of operation may imply thermal actions from hot
solids on filling, or additional air pressure which may have multiple causes: used as a
discharge aid, or for homogenisation, or for control of moisture or temperature, or due to
filling with fine grained solids that entrain interstitial air. These matters are crudely dealt with
in EN 1991-4, but it may be more appropriate for them to be treated as load combinations.
4. Actions for load combinations
In the current Eurocode, EN 1991-4 (2006) [3], the following non-permanent actions are
listed: filling loads from the stored material, discharge loads, imposed loads, wind loads
(when the silo is full or empty), snow loads, imposed deformations: foundation settlements,

thermal action, seismic loads, and loads from dust explosions. Three accidental actions
should be added to this list: impact and fire and actions during execution.
Some of these were discussed above. Here, three parts of the silo structure deserve
separate recognition: the top of the silo, including the roof and any additional structure
carrying conveyors etc., the silo vessel, involving vertical walls and hopper or bottom which
are directly exposed to loads from the stored material, and the substructure, including the
foundation which in many cases carries the total weight.
4.1. Filling and discharge loads
Filling and discharge loads have to be treated independently as two different loads because
they may each cause the greatest load effects on different parts of the silo.
These loads are likely to occur with the characteristic value corresponding to the full
condition as defined in EN 1991-4 (2006) [3] every time the silo is full or is being discharged
from full. Because the discharge pressure state often persists after a partial discharge, it is
necessary to include the discharge load as quasi-permanent, unless specific knowledge of
the manner in which the silo will be operated dictates otherwise.
Thus, if the silo is used for storage most of the time, the duration of these extreme loads is
much longer than that of other variable loads (e.g. wind loads), and they are therefore likely
to occur simultaneously with other extreme loads. As a result, the combination value as well
as the frequent value of the combination factor should be high. Given that the coefficient of
variation of extreme loads in a silo is fairly small, a value of 0.8 or 0.9 could be chosen.
However, these values are probably only relevant to silos that are used for long term storage,
and not those in industrial applications where mean loads are lower. A further discussion
should be held on the magnitude of the partial coefficient, which should perhaps be lower
than 1.5, but such a discussion is again beyond the scope of this paper.
The stochastic variability associated with filling and discharge loads is mainly a problem for
the design of the vertical walls of the silo because of the potential for redistribution of load
between different parts of the wall and between the wall and the bottom. Under static
conditions, the substructure carries a load that is much better defined.
The value of the quasi permanent combination factor depends on the manner in which the
silo will be used and on the load case to be considered. Whilst the designer probably has
much information on this when the design is undertaken, it is difficult to codify into EN 1990.
If the design situation involves differential settlements and the silo is used intensively, the
value might be around 0.8, but it might be only 0.2 for a silo that is used for shorter periods.
If the design situation is seismic, it seems inappropriate to use the quasi permanent
combination factor as prescribed. The total mass of stored material at the time of the seismic
event is relevant. If the silo is a grain silo, it is most likely that the silo is either full or empty,
so that the partially full concept is not relevant, but if it is an industrial silo, it will usually be
partially full, and its seismic response will be greatly affected by the level of filling. Thus it is
difficult to define a proper value for this combination factor without having more specific
information about the silo and its usage. It may be appropriate to take the combination value
as the frequent value, but only for some applications. This discussion is relevant only to
global horizontal excitations. Further challenges that need to be considered are a seismic
action that affects the friction coefficient between the stored mass and the wall (potentially
leading to significantly increased normal pressures), and vertical accelerations which can be
overwhelmingly important for the substructure of elevated silos.
4.2. Wind loads
Wind is an important load for some silos. For the roof structure it plays the same role as in
buildings. For the vertical walls, the critical situation in thin-walled structures is when the silo
is empty, but the critical case for the substructure is always when the silo is full.

4.3. Imposed deformations
Imposed deformations may arise from foundation differential settlements, thermal action, or
differential filling of connected silo cells in batteries. Because shell structures are highly
indeterminate and lack ductility in some failure modes, imposed deformations usually induce
stresses that must be considered, notably for the walls and substructure.
4.4. Thermal actions
Thermal actions may be climatic, as for other structures, but may also be caused by hot
solids from industrial processes being placed in the silo. For silos, climatic thermal actions
are rarely extreme, but thermal actions due to special operational conditions may occur at
their full value every time the silo is filled, so the combination value should be 0.8 or 0.9. In
some cases thermal actions may be relevant for the serviceability limit state only, but this
discussion is again beyond the scope of this paper.
4.5. Accidental actions
Four accidental actions are relevant for silos:
Seismic action on silos has been noted above.
Design for dust explosions is relevant to silos that store solids susceptible to explosions.
Such cases are best controlled by the design of venting openings [3] (EN 1991-4, 2006).
Vehicle impact is especially relevant to elevated silos that are supported on columns, where
discharge is achieved by loading trucks that pass beneath the silo vessel.
Design for fire is generally not relevant for silos, though this situation is important for tanks
that store highly combustible or explosive fluids.
5. Load cases
As noted above, each design situation is formulated as a load case involving all permanent
actions, a leading variable action and relevant accompanying loads. The accompanying
loads are scaled by the load combination factor.
For normal ultimate limit states (persistent or transient) the following combination factors are
suggested here: filling loads (0.8 or 0.9), discharge loads (0.8 or 0.9), imposed loads (0.7),
wind loads when the silo is full or empty (0.6), snow loads (0.5 or 0.7 depending on the
location and altitude of the site: EN 1990 (2002) [2]), imposed deformations: foundation
differential settlements (1.0), thermal action - climatic (0.6) and thermal action from hot solids
(0.8 or 0.9). Additional air pressure should be considered as specified in the design, with a
higher value of air pressure being considered as accidental. For silo batteries, defined load
cases for patterns of full and empty cells and imposed deformations must be devised and
included in a revision of EN 1991-4. Filling loads and discharge loads cannot occur
simultaneously, while the others may all occur at the same time. In load cases where the
leading variable is not filling or discharge, filling loads should be disregarded and only
discharge loads considered.
In applying the above principles, the following cases should be considered. To avoid
repetition, the full list of accompanying loads is only given for the first case:
A. Filling load in combination with imposed load (0.7) , wind loads (0.6), snow loads (0.5 or
0.7), imposed deformations: foundation settlements (1.0), thermal action - climatic (0.6)
and thermal action from hot solids (0.8 or 0.9).
B. Discharge load - in combination with ..
C. Imposed load in combination with discharge loads (0.8 or 0.9), ..
D. Wind load in combination with discharge load (0.8 or 0.9),

E. Wind load (empty silo) in combination with imposed load (0.7), . (If the presence of
the stored solid is used in some structural design treatments to enhance the buckling
strength or to restrain bending moments in silo walls, it is important that the ABSENCE of
solid within the silo is specifically identified for this load combination).
F. Snow load in combination with discharge loads (0.8 or 0.9),
G. Climatic thermal action in combination with
6. Conclusions
This paper has presented a wide-ranging discussion of the many factors that affect the
definition of load combinations for silo structures, with special reference to ultimate limit
states. These structures are clearly identifiable as requiring considerably more information
concerning the planned usage and operation of the silo to be available to the designer if load
combination factors are to be based on rational decisions.
It is recommended that the load combinations to be considered for silos incorporate the
following aspects:
1. Load combinations and combination factors should be set by dividing silos into different
categories, at least according to the following:
a) the relative thickness of the silo walls;
b) the aspect ratio of the silo;
c) whether the silo is on-ground or elevated;
d) operating conditions.
2. The specific values of factors can only be determined on the basis of engineering
judgement at the present time.
Reference list
[1] Brown, C.J. and Nielsen, J. (Eds) (1998) Silos: Fundamentals of Theory, Behaviour and
Design, E & FN Spon, London.
[2] EN 1990 (2002). Eurocode Basis of structural design. European Standard. European
Committee for Standardization (CEN).
[3] EN 1991-4 (2006). Eurocode 1- Actions on structures Part 4: Silos and Tanks.
European Standard. European Committee for Standardization (CEN).
[4] ISO 2394 (1998). General principles on reliability for structures. (Annex F).
[5] Rotter, J.M. (2001) Guide for the Economic Design of Circular Metal Silos, London:
Spon.
[6] Rotter, J. M., Pham, L. & Nielsen, J. (1986). On the specification of loads for the
structural design of bins and silos. 2nd Int. Conf. On Bulk Materials, Storage, Handling
and Transportation (pp. 241-247) The Institution of Engineers, 7-9 July, Wollongong,
NSW, Australia.
[7] Trahair, N.S., Abel, A., Ansourian, P., Irvine, H.M. and Rotter, J.M. (1983) Structural
Design of Steel Bins for Bulk Solids, Australian Institute of Steel Construction, Sydney.


Determination of the mechanical properties of biomass briquettes
for use in DEM simulations
Carlos Gonzlez-Montellano, lvaro Ramrez Gmez, J os M. Fuentes, Eutiquio
Gallego, Francisco Ayuga*
BIPREE Research Group, Universidad Politcnica de Madrid, ETSI Agrnomos, Avda
Complutense s/n, 28040 Madrid, Spain
*Corresponding author. e-mail: francisco.ayuga@upm.es
Abstract
Biomass pellets and briquettes are granular materials that are gaining importance nowadays
in the industry but many design issues regarding the storage and handling equipment still
pose a challenge for engineers. In this sense, the Discrete Element Method (DEM) is a good
numerical technique to assist designers but requires certain microscopic properties of the
material under study be known. Literature generally contains little information in this respect
and there is a lack of standardized procedures due to the difficulties for obtaining these
properties at particle level. In this work a large proportion of the microscopic properties of
biomass briquettes required in DEM simulations have been experimentally determined. In
particular, most of the microscopic properties required in the Hertz-Mindlin contact model are
considered: particle density, modulus of elasticity, particle-particle coefficient of restitution,
and the particle-wall coefficient of friction. The experimental protocols and the test
apparatuses used in this work were mainly based on procedures previously proposed by
other researches but adapted to the specific characteristic of biomass briquettes.
Key words: Mechanical properties, Discrete Element Method, Biomass.

1. Introduction
The use as fuel of pellets and briquettes made from cereal straw, sawdust and wood
shavings is rapidly increasing throughout Europe. In 2010, biomass contributed
around 5%, some 85 million tonnes oil equivalent (Mtoe), of the energy needs of the
EU. By 2020, it is anticipated that renewable sources should contribute 20% of the
EU primary energy demand, with more than 20 million hectares used for energy
crops (European Biomass Association, 2011). The big amounts of biomass that are
expected to be produced in the upcoming future will require appropriate machinery,
transport systems and storage facilities to handle it. Different biomass materials (or
even different grades of the same material) can have very different properties. Thus,
different handling, transport and/or storage solutions would be required. Advanced
numerical techniques such as the finite element method (FEM) or the discrete
element method (DEM) may be used to analyse the behaviour of biomass for storage
and transport in order to avoid problems in the industries such as blockages or
damages in the machinery and equipments.
Discrete Element Modelling (DEM) is a numerical technique specifically designed to
simulate the behaviour of granular materials (i.e. biomass pellets or briquettes). It is
based on an explicit numerical procedure in which the interactions between the
particles (or between the particles and the limiting walls) are monitored contact by
contact. The movement is modelled particle by particle using Newtons second law of
motion to describe translational movement, and the general rotational dynamics
equation to describe rotational movement (Cundal & Strack, 1979). The trajectory of
each particle is predicted via the cyclical repetition of an algorithm executed at the
different time intervals into which the total studied time is divided.
The development of DEM models requires the use of contact models that reliably
represent the characteristics of the simulated material and values that adequately
describe the mechanical properties of the material under study. However, these
properties must be obtained at a particle (microscopic) level and it sometimes makes
the direct determination difficult. Until now very little accurate information about the
values of the material properties is available.
The present paper focuses on the direct determination of the properties of different
biomass briquettes (vine shoots; colza straw; rice husk; corn stalk; corn stalk mixed
with pine wood chips; sawdust and cereal straw; cereal straw) required in DEM
simulation (Fig. 1). The microscopic properties analysed are: particle density,
modulus of elasticity, particle-particle coefficient of restitution and particle-wall
coefficient of friction. Results obtained by laboratory essays are discussed, and some
practical recommendations about the use and improvement of these experimental
methodologies for the case of irregular particles are presented.

FIGURE 1: Samples of biomass briquettes tested, that were provided by FERRO S.A.

2. Materials and Methods
The microscopic properties of different biomass briquettes required in DEM
simulations have been experimentally tested. The experimental protocols and the
test apparatuses used in this work were mainly based on procedures previously
proposed by other researchers (Chung et al, 2004; Wong et al, 2009), properly
adapted to the specific characteristic of biomass briquettes. Essays of other
agricultural materials have been previously developed by the authors (Gonzalez-
Montellano et al., 2012)
The particle density (
p
) was estimated by the mercury immersion technique. In
particular, the total volume of each briquette was obtained by measuring the mass of
mercury displaced when immersing it in a recipient levelled up with mercury, whereas
the sample mass of the briquette was obtained by using a precision balance.
Between 10 to 15 samples of each material, randomly selected, were essayed.
The youngs modulus (E
p
) is defined as the relationship between the tension applied
to an object in a certain direction and the unit deformation suffered by the object in
that direction. The determination of its value has been carried out as described in
standard ASAE 368.4 (2006), based on the Hertz contact model using data obtained
in a compression assay on an individual particle. The compression tests were carried
out using a Texture Analyser TX2 machine. The spherical indenter used consisted of
a steel ball, with a diameter of 19.5 mm. A total number of 25-30 briquettes of each
material were analyzed in the axial and radial directions, the compression force being
about 40 N applied at a speed of 0.3 mm/s.
The value of the particle-particle coefficient of restitution (e
p
) was obtained by using a
pendulum test similar to the one described in Wong et al (2009). The experimental
apparatus consisted of a double pendulum formed with 0.22 mm diameter nylon
string that allowed a collision between two particles. The test particles were glued at
the middle of each string and both pendulums were fixed to a horizontal bar with the
suspended particles perfectly aligned with one another. One of the particles (Particle
1) was then displaced laterally to a height H
0
with respect to the position of Particle 2,
and maintained in this position. After releasing Particle 1, it impacted against Particle
2, and both particles reached heights of H
1
and H
2
, respectively. H
0
, H
1
and H
2

heights were determined from images taken with a Genie H1400-Monochrome high
speed camera (50 frames per second). Three samples of each material were tested,
involving three different initial heights H
0
. All tests were replicated 15 times.
The determination of the particle-wall friction coefficient (
w
) was based on a sliding
test similar to the one described in Chung (2006). The assay apparatus consisted of
an inclinable platform on which a sample tray was placed. Three test particles were
fixed on this tray in a triangular arrangement. With the inclinable platform in a
horizontal position, a piece of wall material (steel or methacrylate in the present work)
was laid over the three particles. The platform was then raised progressively at one
end until the wall material lying over the particles began to slide. At this moment the
assay was stopped. The angle of inclination of the platform () at this point is related
to the value
w
in the following manner:
w
=tg (). Three different samples were
tested in each particle/wall material combination, with 14 replicates of each assay.

3. Results
3.1. Particle density (
p
)
Table 1 shows the results obtained for the particle density determination
accomplished with all the materials considered. The values obtained in table 1 were
deduced from 10-15 samples of each material. It can be seen that in all materials the
average particle density lies in the range 9001030 kg/m
3
, except for cereal straw
briquettes, whose density is 680.9 kg/m
3
. It is also important to note the coefficient of
variation is always lower than 10%, which is indicative of the homogeneity of the
samples tested.

TABLE 1: Particle density obtained for all briquettes materials considered
Parameter Corn stalk
mixed with pine
wood chips
Rice
husk
Vine
shoots
Colza
straw
Corn
stalk
Sawdust
and cereal
straw
Cereal
straw
Mean particle
density,
p (kg/m
3
)
1026.7 914.2 1010.8 1003.0 1017.2 965.4 680.9
Coefficient of
variation,
CV (%)
3.0% 4.8% 8.3% 2.8% 4.0% 4.3% 6.3%
3.2 Modulus of elasticity: axial (E
a
) and longitudinal (E
l
)
Fig. 2 shows the results obtained for the axial and longitudinal modulus of elasticity in
the materials tested. The results were obtained from more than 25-30 samples for
each material, and by using briquettes with an average radius of 34.56 mm. A single
Poissons coefficient of 0.33 and average compression diameter of 19.50 mm were
used in those materials that required these parameters to calculate the modulus of
elasticity. It can be seen that four materials (vine shoots, corn stalk, corn stalk
with pine wood chips and sawdust and cereal straw) have an axial modulus of
elasticity close to 35 MPa. On the other hand, corn stalk is the material with the
lower Axial Modulus of Elasticity, 6.39 MPa.
0
5
10
15
20
25
30
35
40
VineShoots ColzaStraw RiceHusk Cornstalk
briquettes
Cornstalk andpine
Woodchips
Sawdust andcereal
straw
Cornstalk
M
o
d
u
l
u
s

o
f

E
l
a
s
t
i
c
i
t
y

(
M
P
a
)
AxialModulus ofElasticity (MPa)
LongitudinalModulus ofElasticity (MPa)

FIGURE 2: Axial (E
a
) and Longitudinal (E
l
) modulus of elasticity for the materials tested
The longitudinal modulus of elasticity shows lower values than those corresponding
to the axial modulus of elasticity in any material. Thus, it can be seen that corn stalk
with pine wood chips has the greater E
l
, 22.19 MPa, which is about a 37.5% of E
a
. It
is also important to remark that all materials show a great variability in the values for
the modulus of elasticity, with coefficients of variation that ranges in all cases
between 30% and 50%.

3.3 Particle-particle coefficient of restitution (e
p
)
Fig. 3 shows the mean values of e
p
depending on the impact velocity for the different
materials tested. Three samples (MA, MB and MC) were used for each material, and
three different impact velocities were also considered for each sample. In addition,
the value of e
p
for each sample was obtained by using the results obtained in 15
repetitions. If the impact velocity considered is higher, then it can be observed less
variability among the e
p
obtained with the three samples. In addition, most of the
materials tested show that e
p
tends to grow with increasing impact velocities. This
trend is not detected for corn stalk and vine shoots briquettes, where e
p
decreases
if the impact velocity is higher.
It is also interesting to note that mean e
p
lies in the interval 0.350.50 for all materials
with independence of the impact velocity, except for rice husk briquettes. In this
case, there appears a great difference in mean e
p
from low impact velocity (0.16 for
0.72 m/s) to high impact velocity (0.37 for 1.60 m/s). Finally, it has been obtained a
high coefficient of variation (over 30%) for the repetitions with the same particle in
any material. This is a symptom of the heterogeneity of the samples and the
difficulties of this essay to determine e
p
in briquettes.
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.50 1.00 1.50 2.00
R
e
s
t
i
t
u
t
i
o
n

C
o
e
f
f
i
c
i
e
n
t

(
e
p
)
Impactvelocity(m/s)
MA MB MC MEAN

0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.50 1.00 1.50 2.00
R
e
s
t
i
t
u
t
i
o
n

C
o
e
f
f
i
c
i
e
n
t

(
e
p
)
Impactvelocity(m/s)
MA MB MC MEAN

a B
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.50 1.00 1.50 2.00
R
e
s
t
i
t
u
t
i
o
n

C
o
e
f
f
i
c
i
e
n
t

(
e
p
)
Impactvelocity(m/s)
MA MB MC MEAN
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.50 1.00 1.50 2.00
R
e
s
t
i
t
u
t
i
o
n

C
o
e
f
f
i
c
i
e
n
t

(
e
p
)
Impactvelocity(m/s)
MA MB MC MEAN

c D
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.50 1.00 1.50 2.00
R
e
s
t
i
t
u
t
i
o
n

C
o
e
f
f
i
c
i
e
n
t

(
e
p
)
Impactvelocity(m/s)
MA MB MC MEAN

E
FIGURE 3: Change in e
p
with impact velocity for a) Corn stalk mixed with pine wood chips; b)
Colza straw; c) Corn stalk; d) Rice husk; and e) Vine shoots briquettes

3.4 Particle-wall coefficient of friction (
w
)
Table 2 shows the results for
w
for the different biomass materials in combination
with either steel or methacrylate. The mean CV of the repetitions with each sample
was 12%, with all individual values involved in the calculation of this mean below
17%. Independent of the particle type, the (
w
) is larger when the wall material is
methacrylate. The mean value of
w
for all materials lies in the interval 0.230.26 for
a steel wall, and 0.310.37 for a methacrylate wall.
Some procedural factors need to be carefully controlled. For example, the system for
raising the inclinable platform must do so with no sudden changes or jumping, and a
reliable system for measuring (directly or indirectly) the angle at the moment of
sliding must be available. This moment itself must be well identified. Finally, the
possibility of the platform wearing with use must be taken into account.
TABLE 2: Particle wall coefficients of friction (
w
) for different biomass briquettes with steel
and methacrylate surfaces
Parameter Rice husk Colza straw Vine shoots Corn stalk
Pine wood chips
and corn stalk
Steel Methacrylate Steel Methacrylate Steel Methacrylate Steel Methacrylate Steel Methacrylate
n 14 14 14 14 14 14 14 14 14 14
m 0.25 0.37 0.26 0.33 0.24 0.31 0.25 0.34 0.23 0.32
0.03 0.06 0.02 0.04 0.04 0.05 0.02 0.04 0.02 0.03
CV (%) 12 17 9 11 15 15 9 11 10 11
n: number of samples m =mean value of
w
=standard deviation CV =Coefficient of variation (%)

4. Conclusions
The essays for the determination of particle density and the particle-wall
coefficient of friction show the lowest coefficient of variation, which reflects the
reliability of the values obtained.
Most of the biomass briquettes tested show values for the particle density,
particle-particle coefficient of restitution and particle wall coefficient of friction that
lay in a reduced range.
Some procedural factors need to be carefully controlled in the determination of
particle-particle coefficient of restitution and the particle-wall coefficient of friction.

Acknowledgements
The authors thank the Spanish Plan for Research, Development and Innovation for financing
this research via the project Study on Buckling and Dust Explosions in Agricultural Steel
Silos. Simulation with Numerical methods and Experimental Validation (AGL2009- 13181-
C02-01). The authors also thank FERRO S.A. for providing the biomass briquettes to be
tested and the research group LPF-TAGRALIA, from the Technical University of Madrid,
for having offered the XT2 Texture Analyzer used in this work.

References
ASAE 368.4 (2006). Compression Test of Food Materials of Convex Shape. St. J oseph, MI
(USA): American Society of Agricultural and Biological Engineers (ASABE).
Chung, Y.C. (2006). Discrete element modelling and experimental validation of a granular
solid subject to different loading conditions. PhD. Thesis. University of Edinburgh.
Cundal, P. & Strack, O. (1979). A discrete numerical model for granular assemblies.
Geotechnique, 29 (2), 4765.
European Biomass Association (2011). AEBIOM Annual report 2011. Brussels: European
Biomass Association. Available online at: http://www.aebiom.org/?cat=29 (Last accessed:
09-05-2012)
Gonzlez-Montellano, C., Fuentes, J .M., Ayuga-Tllez, E. & Ayuga, F. (2012). Determination
of the mechanical properties of maize grains and olives required for use in DEM simulations.
J ournal of Food Engineering, 111 (4), 553562.
Wong, C., Daniel, M. & Rongong, J . (2009). Energy dissipation prediction of particle
dampers. J ournal of Sound and Vibration, 319 (12), 91118.
Test silo for dust explosions
lvaro Ramrez-Gmez
1
*, Alberto Tascn
2
, Pedro J os Aguado
2
, Francisco Ayuga
1


1
BIPREE Research Group, Universidad Politcnica de Madrid, Ciudad Universitaria S/N,
28040 Madrid, Spain
2
INRUMA Research Group. Dpto. Agricultura y Alimentacin. Universidad de La Rioja, Avda.
de la Paz, 93. 26006 Logroo. Spain.
3
INRUMA Research Group. Dpto. Ingeniera y Ciencias Agrarias. ESTI Agraria. Universidad
de Len. Avda. de Portugal 41. 24071 Len. Spain.

*Corresponding author. E-mail: alvaro.ramirez@upm.es
Abstrat
It is common to find storage installations in the agri-food industries. The processes of filling
and discharge carried out in these installations usually lead to dust clouds, and therefore
potential explosive atmospheres. It is known that prevention and protection are the main
measures to be followed in these situations. Nowadays, there are many technical and
economical difficulties that avoid to achieve a viable protection in the agricultural silos. Some
studies have shown differences between standards in this regard, this is the case of the
European code (EN 14491, 2006) and the North American code (NFPA 68, 2007). This
implies that research is needed in this direction. Further knowledge about the explosion
process and venting is needed in order to arrive to optimal designs. Experimental work is
essential, because of that three research groups, two from the Universidad Politcnica de
Madrid and one from the Universidad de Len, have joined efforts to develop an installation
that will allow the study of the explosion phenomenon and will allow to optimize the design of
protection devices.

Key words: Silo, dust, explosion, design, venting.

1. Introduction
When the organic material is finely divided and on the air forming a cloud of dust, a potential
explosive atmosphere can be formed if the dust concentration is enough (Eckhoff, 2003).
Operations as transport, load and discharge may generate dust clouds able to produce
explosions in agricultural industries involving loss of goods and materials.
When a dust cloud is ignited in a confined environment exothermic reactions are produced,
being developed a rapid increase of pressure until values of 7-10 bar (700-1000 kPa).
The EU legislation establishes the conditions that must fulfil the equipment to be mounted in
potential explosive atmospheres and the working conditions in two directives (ATEX 94/9/CE,
1994 and ATEX 1999/92/CE, 1999). These Directives receives the colloquial acronym ATEX
(ATmosphre EXplosible).
The basic system of protection against explosions in silos are venting devices, which allow to
release pressure generated when a dust cloud ignite in a controlled manner by means of its
rupture.
During the 80s and 90s several researchers have contributed to the knowledge of the dust
explosions in silos by means of the development of explosion tests. Remarkable tests are
those developed by Eckhoff in a silo of 500 m3 (Eckhoff, 1984); and in one of 236 m3
(Eckhoff, 1988); results by Bartknecht in 20 m3 (1985); the test of Siwek in 10, 25 and 250
m3 (Siwek, 1989); the tests of Tamanini in 63,7 m3 (Tamanini, 1996); Deng in 95 m3 (Deng,
1993); Hattwig and Hesel in silos of square section of 45 and 85 m3 (Hattwig, 1993); and
those developed by Eibl, Nars, Leuckel and Hochst in 50 m3 (Hchst, 1998).
However, nowadays there are still technical and economical difficulties to efficiently protect
agricultural silos (Tascn, 2009a). On the other hand, recent works (Tascn, 2009b) have
remarked significant differences between codes (EN 14491, 2006 and NFPA 68, 2007) when
venting surfaces must be determined to protect a silo. Because of that, it is needed to
continue doing research to know much better the explosion process and the venting
protection to find rules for the optimal design of silos and their protection. The development
of new design methods would allow contributing to the current normative, which is in a
constant review process.
Dust technology plays an important role to understand the processes implied: dust dispersion
and cloud formation, ignition, explosion and venting (Eckhoff, 2008). The complexity of these
processes, together with the numerous situations that can be presented inside silos, makes
necessary to research in this direction.
The main objective of this research work is the design of an experimental silo to develop dust
explosions for venting studies. This facility will contribute to the knowledge of those variables
that take part in the explosion phenomenon and venting; and also will provide information
useful for the design of more efficient protection equipment.
2. Description of the experimental silo
The design of the silo has been carried out according to the recommendations of the UNE
normative for the construction materials (UNE-EN 13445-2, 2010), the design of collars,
heads and bolted joints (UNE-EN 13445-3, 2010), the design of resistant equipment to
explosions (UNE-EN-14460, 2006) and also the Spanish Technical Building Code, (CTE DB
SE, 2006).

2.1 Silo
The proposed design is a silo with a circular section composed of 4 modules, 1.6 m long and
2 m diameter; the volume of each module is about 5 m3 (Fig. 1). The shell will be made of
steel P265GH and will have a nominal thickness of 12 mm that will allow to achieve a
nominal pressure of about 10 bar (1 MPa) without deformation.
The collars will be joined by using normalized flanges of circular section with fire-resistant
joints. The material for the flange assembly will be of steel S275J R.
Concerning the heads, they will be of two types: One of the heads will be a torispherical head
(Klpper type), made also with steel P265GH of 12 mm thick and the other is built with the
same type of steel but with a square reinforced opening 1x1 m2:
To screw the modular frame where the venting panels will be mounted to be tested.
To allow cleaning and maintenance tasks inside the silo.



Figure 1. Sketch of the experimental silo. Front, lateral and back view.

2.2 Support structure
The silo will be placed in horizontal position, being each collar supported on two legs. The
support structure consists of legs, tie bars and gusset plates made of S275R steel.
The legs will be made with HEB-160 profiles and they will be welded in the upper part to
plates of the following dimensions 360x340x10 mm.
To gain stiffness the structure will be tighten in the longitudinal direction of the silo with a L
45.5 profile bolted to a plate of 10 mm thick that will be welded in the lower part of each leg.
The cross bracing system consists of steel bars of 180 mm2 net area. In the transverse
direction the legs will be tightened by a L60.5 steel profile bolted to identical plates that are
used to hold the steel bars in the longitudinal plane.
To make easier dismounting and handling the collars, lifting lugs will be welded on a plate
270x150x10 mm on the shell and they will be 10 mm thick with a shape adapted to the hook
crane that will be used.
The silo will be protected against the weather conditions inside a warehouse. Wheels 250
mm diameter and made of steel and polyurethane have been considered in the legs of the
silo to pull the silo out, where the explosion tests will be carried out.

2.3 Dust injection
The dust injection aims at generate the dust cloud inside the silo (Fig. 2). It will be similar to
the used in the German experimental silo of 50 m3 (Hchst, S., 1998). The system consists
of the injection of dust by means of pressurized bottles at 10 and 20 bar and nozzles in the
interior of the silo.


Figure 2. Sketch of the experimental silo. Dust injection system and instrumentation

To achieve the injection is needed the installation of a compressor that will be able to
pressurize all the bottles. Once is achieved the pressure needed, by means of the opening of
the quick opening valves, the dust will be introduced in the silo, being disperse by means of
nozzles similar to the used in the normalized explosion tests in the 1 m3 vessel (UNE-EN
14034-1, 2005) or other designs.
A pressurized bottle will be mounted in each collar. The content of dust will be modified to
get different concentrations of dust inside the silo. For tests with maize starch the
concentration that produces the highest values of overpressure is about 500 g/m3 (Eckhoff,
2003).

2.4 Instrumentation
The main parameter to be measured in the explosion tests will be the overpressure reached.
Studies modifying variables such as the aspect ratio, venting area, dust concentration,
ignition delay and so on will be carried out to observe the variation of the overpressure
values.
To be able to measure the generated overpressures in an explosion, that is a quick
phenomenon, it will be needed high speed data-acquisition systems and with enough
resolution- 16 bits-.
The instrumentation will be composed of 4 pressure cells, distributed along the silo (Fig. 2),
connected to a data-logger storing the measurements in a PC. The pressure cells selected
are piezo-resistive, model 4043 from Kistler and able to work in a wide range of pressure
from 0 to 10 bar. They can bear high temperatures and they will be screw in the silo, type of
thread M 14x125.
The data-logger is a Traveller CF from ESAM. It has a sampling rate of 250 kS/s, bandwidth
of 10 kHz with possibility of Wheatstone bridges connection and analogical trigger.
The ignition of the dust cloud will be achieved with chemical igniters. The normalized values
(UNE-EN 14034-1, 2005) are of 10 kJ , but also can be studied the effect of the intensity of
ignition energy over the pressure generated using different ignition sources.

3. Research possibilities that offer the silo
The modular character of the silo offers a wide range of possibilities to carry out tests in
different volumes and different aspect ratios.
The tests that can be carried out in the silo are the following:
Explosion tests in normalized conditions (UNE-EN 14034-1, 2005) defining as a
variable the silo volume and the aspect ratio L/D to research on the influence of the pressure
devloped.
Explosion tests in normalized conditions for a specific volume and aspect ratio L/D,
defining as a variable the venting area to study the pressure developed and compare it with
the normative used for venting design. 4 different geometric configurations can be studied (1-
4 modules) in which different venting areas will be tested.
Explosion tests in normalized conditions modifying the moisture content, size
distribution or uniformity in the dispersion of the dust.
Explosion tests in normalized conditions, varying the concentration and turbulence of
the dust cloud.
Dispersion tests of dust within the equipment, without ignition and explosion, to study
different designs or dispersion conditions and get a better understanding about the dust
clouds generated.
If in the next future are mounted in the silo measuring devices for measuring dust
concentration and air velocity (Hauert, 1996) studies about the dust formation could be
carried out in an accurate manner. This will allow to carry out all the tests aforementioned for
different conditions different to the normalized ones.
Finally, it is important to remark that it would be possible to add more modules, extending the
silo, and achieving in this way higher volumes and more aspect ratios.

4. Conclusions
This work presents a design of an experimental steel silo to carry out explosion tests. The
facility will be modular, which will allow to modify in a easy way the volume and the aspect
ratio.
The equipment has been designed in a manner that will allow to study the effect of all the
parameters that have influence in the generated pressure in an explosion: volume, aspect
ration, venting area, venting type, ignition energy and characteristics of the dust cloud.
This silo will allow to get a better understanding of the explosion process and the venting
systems in a considerable volume, initially from 5 m3 until 20 m3, with the objective to
contribute to solve the difficulties that are still not solve in the protection of agricultural silos
against dust explosions.

5. Acknowledgements
The authors thank to the Ministry of Science and Innovation for funding this research Project
through the national Programme of R&D: Project AGL2009-13181-C02.

6. References
ATEX 94/9/EC (1994) Directive 94/9/EC on the approximation of the laws of the member
states concerning equipment and protective systems intended for use in potentially explosive
atmospheres.
ATEX 1999/92/EC (1999) Directive 1999/92/EC on minimum requirements for improving the
safety and health protection of workers potentially at risk from explosive atmospheres.
Bartknecht W. (1985) Effectiveness of explosion venting as a protective measure for silos.
Plant/Operations Progress, Vol. 4, 4-13.
CTE DB SE (2006). Documento Bsico Seguridad Estructural. Cdigo Tcnico de la
Edificacin. Ministerio de la Vivienda, Espaa.
Deng X., Zang J , Xie L. (1993) Maize dust explosions in the 94,4 m3 experimental silo for
venting or deflagrations in low strength silos. Proc. 5th Int. Coll. Dust Explosions. Abril 1993,
Pultusk, Polania.
Eckhoff R. K., Fuhre K. (1984) Dust explosion experiments in a vented 500 m3 silo cell. J .
Occup. Accid, Vol. 6, 229-240.
Eckhoff R. K., Alfert F., Fuhre K., Pedersen G. H. (1988) Maize starch explosions in a 236
m3 experimental silo with vents in the silo wall. J . Loss Prev. Process Ind., Vol. 1, 16-24.
Eckhoff R.K. (2003) Dust explosions in the process industries, Gulf Professional
Publishing/Elsevier, Amsterdam.
Eckhoff R.K. (2008). Understanding dust explosions. The role of powder science and
technology. J . Loss Prev. Process Ind., Vol. 22, pp. 105-116.
EN 14491 (2006). Dust explosion venting protective systems. CEN.
Hattwig M., Hensel W. (1993) Applicability of the new VDI-guideline 3673 to silos of
rectangular geometry. 5th Int. Coll. Dust Explosions. Abril 1993, Pultusk, Polonia.
Hauert F., Vogl A., Radant, S. (1996) Dust cloud characterization and its influence on the
pressure-time-history in silos. Process Saf. Prog., Vol.15 (3), 178-184.
Hchst S., Leuckel W. (1998). On the effect of venting large vessels with mass inert panels.
J . Loss Prev. Process Ind., Vol. 11 (2), 89-97.
NFPA 68 (2007). Standard on Explosion Protection by Deflagration Venting. National Fire
Protection Association, Quincy, MA, USA.
Siwek R. (1989) Drukentlastung von Staubexplosionen beim pneumatischen Befllen von
Behltern. VDI-Berichte 701, Verein Deutscher Ingenieure, Dusseldorf.
Tamanini F., Valiulis J . (1996). Improved guidelines for the sizing of vents in dust explosions.
J . Loss Prev. Process Ind., Vol. 9 (1), 105-118.
Tascn, A. (2009a) Aplicacin de la Dinmica de Fluidos Computacional al estudio de las
explosiones de polvo en silos y su proteccin mediante venteos. Tesis doctoral. Dpto.
Ingeniera y Ciencias Agrarias. Universidad de Len.
Tascn, A., Aguado, P.J , Ramrez, A. (2009b) Dust explosion venting in silos: A comparison
of standards NFPA 68 and EN 14491. J . Loss Prev. Process Ind., Vol. 22, 204-209.
UNE-EN 14034-1 (2005) Determinacin de las caractersticas de explosin de nubes de
polvo. Parte 1: Determinacin de la presin mxima de explosin pmx de nubes de polvo.
AENOR.
UNE-EN 14460 (2006) Equipos resistentes a las explosiones. AENOR.
UNE-EN 13445-2 (2010) Recipientes a presin no sometidos a llama. Parte 2: Materiales.
AENOR.
UNE-EN 13445-3 (2010) Recipientes a presin no sometidos a llama. Parte 3: Diseo.
AENOR.




Simulations of dust explosions in a 16 m
3
silo

Alberto Tascn
1
*, ngel Ruiz
2
, ngel Couto
2
, Pedro Aguado
2


1
University of La Rioja, Departamento de Agricultura y Alimentacin, C. Madre de Dios 51 -
Logroo 26006, Spain
2
University of Len, Departamento de Ingeniera y Ciencias Agrarias, Avenida de Portugal
41 - Len 24071, Spain
* Corresponding autor. E-mail: alberto.tascon@unirioja.es

Abstract
Venting devices are the most common protective systems employed in silos against dust
explosions. However, the installation of vents is technically difficult and costly in many cases.
In the present work, dust explosions were simulated in a 16 m
3
silo using a commercial CFD
program, the DESC code. The aim was to determine the overpressures developed in vented
explosions for different vent areas. The pressures and associated vent areas in these
simulations were compared to those contemplated in two venting standards. The simulated
pressures showed the expected trends and agreed reasonably well with the standards.
However, the results suggested that the standards could oversize the vent areas for silos.
Key words: Silo, dust, explosion, venting, CFD.

1. Introduction
Dust explosions occur when combustible dust is dispersed in the air as a cloud and reacts
with oxygen in the presence of an ignition source, generating a violent combustion reaction in
the form of flame propagation (Eckhoff, 2003). Many materials stored in silos can cause dust
explosions.
Vent devices are the most commonly implemented protective measure in silos. Venting
consists of the release of the explosion through an opening to prevent the pressure from
reaching unacceptably high levels. Standards EN 14491 (2006) and NFPA 68 (2007)
propose empirical formulae to size the vent areas. However, the above standards are
somewhat contradictory when are applied to silos in certain situations (Tascn et al., 2009).
In addition, the cost of vents could represent an important burden that must be taken into
account.
Computational Fluid Dynamics (CFD) codes have become a useful tool to study complex
fluid problems involving heat and mass transfer and chemical reactions. A number of
researchers have already published interesting reports on the application of CFD simulations
to dust explosions (Van Wingerden et al., 1995; Zhong & Deng, 2000; Kosinski & Hoffmann,
2006).
The object of this work was to simulate dust explosions using a CFD commercial code, the
Dust Explosion Simulation Code (DESC) by GexCon (Skjold, 2007), to study the venting of
dust explosions in silos. The ultimate aim is to reduce the technical and economical
difficulties that frequently arise in the protection of silos. This research completes some
previous works (Tascn et al., 2011).
2. Method
Dust explosions in a 16 m
3
cylindrical steel silo were simulated using the first version of
DESC. Such a silo is part of an experimental installation located at the Polytechnic University
of Madrid (Ramrez et al., 2010) and it consists of a conical concentric hopper, a cylindrical
part, and a conical roof with a ventilation chimney. The geometry of the vessel that has been
simulated was slightly modified with respect to the real silo in order to adapt it to the
orthogonal 3-D grid used in DESC.
The silo simulated had a cylindrical part 5.20 m in height and 1.90 m in diameter. The height
of the hopper was 1.50 m and the diameter of its circular outlet, supposed closed during the
simulations, was 0.30 m. The roof was flat and included a 0.01 m
2
ventilation opening. The
total volume of the vessel was 16.4 m
3
.
Four inertialess pressure relief panels were contemplated in the flat roof, each one of 0.40 x
0.40 m
2
. The vent area could be set to 0.16, 0.32, 0.48 or 0.64 m
2
. Simulations were carried
out for activation pressures of the venting devices of 0.10 bar and 0.05 bar, both of which
represent common static activation overpressures of commercial vent panels.
The ignition was located at 0.95 m above the outlet of the conical hopper of the silo, i.e., at
approximately half height of the hopper. The ignition energy was set to 10 kJ , which is the
standard energy used in explosion severity tests (EN 14034-1, 2004).
The initial characteristics of the dust cloud were taken from the studies carried out by Hauert
& Vogl (1995) in a 12 m
3
silo for mechanical feeding. An average dust concentration, air
velocity and turbulence intensity were considered. Thus, the silo was considered completely
filled by a homogeneous dust cloud. Maize starch (P
max
=8.63 bar and K
st
=149 bar.m/s)
was contemplated in all simulations. The fuel file with the explosion characteristics of the
dust was input in DESC as supplied by GexCon.
To obtain the values of essential parameters during the explosion, several monitoring points
were selected along the vertical axis of the silo and on a plane near the roof.
The pressures and associated vent areas in these simulations were compared to those
contemplated in the standards. Further simulations involved different initial conditions prior to
the ignition.
3. Results
The DESC simulations showed the expected trends. Overpressures were recorded as text
files and graphs. Fig. 1 shows the comparison of the overpressure time histories for four
different vent areas at a monitoring point situated in the middle of the silo (at 3.45 m from the
bottom) over the z-axis. Note that the pressure denotes gauge pressure, i.e., pressure above
atmospheric pressure.
Negative pressure phases were seen over the pressure-time curves; the maximum values
reached were about 0.05 bar (4.56 kPa for a 0.64 m
2
vent area and an activation pressure
of 0.05 bar). In the present simulations, such underpressures increased with the vent area.
Some silos could be vulnerable to the stressed developed by partial vacuums.
Fig. 2 shows the geometry of the silo fitted to the 3-D grid and the propagation of the flame
along the time for a total vent area of 0.16 m
2
, i.e., one panel venting on the right hand of the
silo roof.
A comparison of the DESC simulations with the standards EN 14491 (2006) and NFPA 68
(2007) is shown in Fig. 3. The overpressures plotted for the simulations are the maximum
values recorded by any monitoring point during the explosion process. The trend of the curve
for the DESC simulations is similar to those arising from the standards. However, the
overpressures obtained by DESC were lower than those predicted by either.

FIGURE 1: Overpressure time histories. Maize starch. P
stat
=0.1 bar. Vent area =0.16 m
2
(black), 0.32 m
2
(red), 0.48 m
2
(blue) and 0.64 m
2
(green).

FIGURE 2: Flame development (represented as combustion products); three time steps are
shown: 0.000 s, 0.477 s and 0.593 s. Maize starch. Vent area =0.16 m
2
The present computer simulations seemed to confirm the widespread idea that vent areas
calculated according to standards are conservative compared with what might really be
necessary for most silos (Eckhoff, 1993).
It should be note that the results of the present DESC simulations were very dependent on
the initial conditions of the dust cloud: dust concentration, air velocity and turbulence.
Therefore, the present results should be extrapolated with caution and experimental
validation is still necessary.
4. Conclusions
CFD simulations of maize starch explosions in a 16 m
3
silo protected with vent panels have
been carried out. The DESC program has proven to be a very useful tool to study the
propagation of dust explosions and the protection of vessels.
Negative pressures could be developed during the venting process. It may therefore not be
enough to simply consider the maximum overpressure to correctly design silos protected with
venting panels.
The results of the present simulations suggested that the venting standards could oversize
vent areas for silos. However, the results were very dependent on the initial conditions of the
dust cloud. Improving the design of vented silos will require further research into the venting
process and the characteristics of the dust clouds in the actual equipment currently used in
industry.

FIGURE 3: Comparison of DESC simulations with standards EN 14491 and NFPA 68. Maize
starch. Pstat=0.1 bar.
Acknowledgements
This work has been supported by the Community of Castilla y Len (Spain) and the
European Social Fund via project LE013A07 and a research grant.
References
Eckhoff, R. K. (1993). Influence of initial and explosion-induced turbulence on dust
explosions in large vented silo cells. Safety Science, 16, 511525.
Eckhoff, R. K. (2003). Dust Explosions in the Process Industries. (3rd ed.). Boston: Gulf
Professional Publishing/Elsevier.
EN 14491. (2006). Dust explosion venting protective systems. Brussels: European
Committee for Standardization.
EN 14034-1 (2004). Determination of explosion characteristics of dust clouds, Part 1,
Determination of the maximum explosion pressure p
max
of dust clouds. Brussels: European
Committee for Standardization.
Hauert, F., & Vogl, A. (1995). CREDIT-Project: Measurement of Dust Cloud Characteristics
in Industrial Plants, Final Technical Report. Mannheim, Germany: FSA.
Kosinski, P., & Hoffmann, A. C. (2006). An investigation of the consequences of primary dust
explosions in interconnected vessels. J ournal of Hazardous Materials, 137, 752761.
NFPA 68. (2007). Standard on explosion protection by deflagration venting. Quincy, MA,
USA: National Fire Protection Association.
Ramrez, A., Nielsen, J ., & Ayuga, F. (2010). Pressure measurements in steel silos with
eccentric hoppers. Powder Technology, 201, 720.
Skjold, T. (2007). Review of the DESC project. J ournal of Loss Prevention in the Process
Industries, 20, 291-302.
Tascn, A., Aguado, P. J ., & Ramrez, A. (2009). Dust explosion venting in silos: A
comparison of standards NFPA 68 and EN 14491. J ournal of Loss Prevention in the Process
Industries, 22, 204-209.
Tascn, A., Ruiz, A., & Aguado, P. J . (2011). Dust explosions in vented silos: Simulations
and comparisons with current standards. Powder Technology, 208, 717724.
Van Wingerden, K., Pedersen, G. H., & Eckhoff, R. K. (1995). Violence of dust explosions in
integrated systems. Process Safety Progress, 14 (2), 131138.
Zhong, S., & Deng, X. (2000). Modeling of maize starch explosions in a 12 m
3
silo. J ournal of
Loss Prevention in the Process Industries, 13, 299309.


A study on the discharge rate of biomass briquettes stored in hoppers: Discrete
element analysis

Carlos Gonzlez-Montellano*, lvaro Ramrez Gmez, J os M. Fuentes, Eutiquio Gallego,
Francisco Ayuga
BIPREE Research Group, Universidad Politcnica de Madrid, ETSI Agrnomos, Avda
Complutense s/n, 28040 Madrid, Spain
*Corresponding author. e-mail: carlos.gonzalez.montellano@upm.es
Abstract
Silos and hoppers are structures designed for storing and handling very different granular
materials in the industry. The discharge rate produced during the discharge processes must
be perfectly known in many cases. There exist some classical analytical procedures for
estimating the discharge rate and their validity has been checked for many granular materials
usually handled in the industry. However, this is not the case of new materials, such us
biomass pellets and briquettes. In this work, different discrete element models have been
used to predict the discharge rate of biomass pellets from a hopper under different conditions
The preliminary results of these models were compared with those obtained using analytical
procedures and a good level of agreement has been found.
Key words: biomass pellets, discrete element method, hopper, discharge rate

1. Introduction
Silos and hoppers are structures designed for the storage and handling of very different
granular materials in the industry. Their correct design is essential to ensure an adequate
functioning of the industrial process where they are included. One of the design factors to be
accounted for is the achievement of a reliable and continuous discharge rate during
emptying. For many typical materials, discharge rate issues are well known and analytical
procedures exists to describe this process. However, this is still an unknown issue for new
materials such us biomass which are nowadays being stored and handled in silos and
hoppers.
The Discrete Element Method (DEM) (Cundal and Strack, 1979) is a numerical technique
specifically designed for the simulation of the mechanical behaviour of granular materials and
nowadays is becoming increasingly popular. It has proven to be a good prediction tool in
many different scenarios involving granular materials (Gonzalez-Montellano et al, 2011). In
this work, DEM has been used for simulating the discharge process of biomass briquettes
stored in a hopper under different scenarios.
The aim of this work is to compare the numerical predictions given by the DEM models with
those obtained using analytical procedures. This comparison will allow assessing whether
the analytical procedures are valid for biomass materials under different working conditions.

2. Materials and Methods

2.1. Numerical models
The DEM models developed in this work aim at describing the filling and discharge of
biomass briquettes (made up of crushed vine shots) from a 3D pyramidal hopper with a
square outlet. The generic geometry of this hopper can be seen in Fig. 1a. Twelve different
DEM models were defined in this paper by particularising the generic dimensions shown in
Fig. 1 in order to produce a wide range of different hopper configurations. Each configuration
is characterised by particular values of the hopper angle of inclination () and the side of the
hopper outlet (D). These particular values are shown in Table1 In all cases, a small vertical
section of height H =2m and side of L =7.3 m was included on top of the hopper.
Model () D (m) T (m)
M1 60 0.5 1.95
M2 60 1 1.81
M3 60 1.5 1.67
M4 60 2 1.52
M5 45 0.5 3.39
M6 45 1 3.14
M7 45 1.5 2.89
M8 45 2 2.64
M9 30 0.5 5.86
M10 30 1 5.43
M11 30 1.5 5.00
M12 30 2 4.56
FIGURE 1: Hopper Geometry TABLE 1.- Dimensions of the models
All simulations were performed using EDEM Academic 2.3 (2010) commercial software. To
reduce the simulation time, the hopper was filled en masse, i.e., the whole set of particles is
generated randomly within the hopper volume at the beginning of the simulation and then
allowed to fall under gravity until the particles reach a static situation. Once the filling process is
complete, the hopper outlet is opened such that the particles began to discharge. The
simulation of the discharge process continued until all the particles had exited the silo (Fig. 2a).
(a) (b)
FIGURE 2: Filling and discharge processes (a); Particle geometry (b)

The contact model taken into account for the simulation of both particle-particle and particle-
wall contacts was based on the Hertz-Mindlin contact model (Tsuji et al., 1992), assuming
viscous damping in the perpendicular and tangential directions, and frictional damping in the
tangential direction. A constant time step value of t = 1.7710
-4
s was used in all
simulations, which was obtained as 20% of the critical time step (t
c
) calculated from the
Rayleight time (t
R
) (Li et al., 2005), a function of the properties of the material.
The values for the mechanical properties of the simulated briquettes were obtained largely by
direct measurement. When no standardised or reliable procedures were available, the values
of the material properties were estimated based on the experience of the authors. Table 1
summarises the values of the material properties considered in this work as well as the
references to the measurement methodologies when appropriate. The real shape of the
briquettes was approximated using the multisphere method (Favier et al., 1999); this led to
particles composed of five spheres (Fig. 2b). The size distribution of these particles was
represented by a normal distribution defined in terms of volume, adopting a mean volume of
V =644 cm
3
and a standard deviation of =32.2 cm
3
.

PROPERTY VALUE DETERMINATION METHODOGOY
Material properties
Density (kg/m
3
)
p
1011 Pycnometer Test (ASTM D854-10 (2010))
Stiffness (Pa) E
p
1.310
7
Parallel Plate Compression Test (ASAE S368.4 (2006))
Poissons ratio
p
0.33 Estimated
Interaction properties
Particle-Wall Restitution coeff. e
w
0.4 Drop Test (Gonzalez-Montellano et al, 2012)
Particle-Particle Restitution coeff. e
p
0.15 Pendulum Test (Gonzalez-Montellano et al, 2012)
Particle-Wall Friction coeff.
w
0.25 Sliding Test (Gonzalez-Montellano et al, 2012)
Particle-Particle Friction coeff.
p
0.25 Estimated
TABLE 1. Particle values of the properties used in the DEM models

2.2. Determination of the discharge rate from the numerical results.
The hopper discharge rate is defined as the mass (kg) of material evacuated through the
outlet per unit time (s). In this work, two possible values of the discharge rate were
calculated: the mean discharge rate and the instantaneous discharge rate. The mean
discharge rate (W) can be estimated as the slope of the regression line calculated for a
function M
o,t
=f(t), where M
o,t
is the total mass of particles that has already exited the hopper
at a particular instant t after the beginning of the discharge process. However, in order to
appreciate possible fluctuations of the mean discharge rate, the instantaneous discharge rate
W was also calculated. The instantaneous discharge rate at a time t is given by Eq. 1, where
t is the averaging period (a value of t =0.1 s has been adopted in this work).
t
M M
t W
t t t t

=
+ + 2 / , 0 2 / , 0
) ( '

(1)

2.3. Analytical expressions for estimating the discharge rate in pyramidal hoppers.
There exist several analytical expressions that allow the mean discharge rate (W) to be
estimated based on the silo or hopper geometry and the material parameters. Beverloo et al.
(1961) provided a correlation developed for flat-bottomed cylindrical silos with a circular opening:
( )
2 / 5 2 / 1
kd D g C W = (2)
Where k is the Beverloo constant, which depends on the shape of the particles, C is a
constant found empirically to be in the range of 0.55 <C<0.65, D is the diameter of the outlet,
g is the gravity acceleration, d is the mean particle diameter and the mean bulk density in
the hopper after the filling process. The Beverloo correlation has been improved and
modified several times by different researchers. Myers and Sellers (1971) extended the
correlation to a flat-botomed rectangular silo with a squared outlet:
( )
2 / 5 2 / 1
03 . 1 kd D g W
f
=

(3)
Were
f
is the flowing bulk density at the hopper outlet.
To take into account the effect of an inclined hopper, a correction factor F is usually used to
modify the value of W obtained for a flat-bottomed silo. The factor F, which is multiplicative of
the value of W for a flat-bottomed silo, can be calculated in different ways. For example,
Brown and Richards (1965) proposed the Eq. 4 to estimate the value of F.
( )
2 / 5 2 / 3
sin / cos 1 = F

(4)
A different option was proposed by Rose and Tanaka (1959) to calculate de factor F:
( )

<
>
=

1 tan tan tan tan
1 tan tan 1
35 . 0


F
(5)
Where is the angle from the horizontal of the stagnat-flowing boundary, which is considered
to be constant for a given particulate material. The value of is different from the angle of
repose of the material and is usually estimated by in situ observation. However, when no
other information is available, a value of =45 is typically assumed.
In the present work, the analytical prediction of W for each of the models considered has
been carried out using Eq. 3 combined either with Eq. 4 or Eq. 5. The following
considerations have been adopted in the analytical calculations carried out for all models:
A value of k =2 has been used in all cases, as it is the value recommended for
elongated particles.
The value of D used for each model was taken from Table 1
The value of the
f
was calculated from the numerical results as the mean bulk
density of the material close to hopper outlet. The value used in the analytical
approach was taken as the mean value of
f
calculated for each individual model.
The value of d (mean particle diameter) was fixed as an effective value in the range
7-20 cm (the minimum and the maximum dimensions of the particles) that allows the
differences between the numerical and the analytical results to be minimised.

3. Results
The work presented in this paper is part of an on-going research project which is currently
being developed. Because of that, results for all the models included in Table 1 are not
shown in this section. In particular, only the results for models M6, M7 and M11 are included.
These preliminary results will allow assessing the quality of the results before developing all
numerical models described in Table 1.

3.1. Discharge rate obtained from the numerical results.
Fig. 3 shows the values of discharge rate obtained for models M6, M7 and M11. The red line
shows the change in W over the simulation period whereas the black horizontal line is
indicative of the mean discharge rate (W). The value of W is indicated on each graph, together
with the value of the coefficient of variation of W (CV), indicative of its oscillation level.
0
500
1000
1500
2000
2500
3000
3500
4000
0 5 10 15 20
1.21W
0.81W
W=815 kg/s
CV =8.3%

0
500
1000
1500
2000
2500
3000
3500
4000
0 5 10 15 20
1.21W
0.78W
W=2688 kg/s
CV =5.2%
0
500
1000
1500
2000
2500
3000
3500
4000
0 5 10 15 20
1.11W
0.87W
W=3084 kg/s
CV =3.5%
M6 M7 M11
FIGURE 3: Instantaneous and mean discharge rate for models M6, M7 and M11
The mean discharge rate W is different depending on the model under consideration. In
general bigger values of W are obtained for bigger hopper outlets and for steeper hoppers. It
is the expected behaviour and, in addition, this is what analytical procedures predict via the
variables D and F described in section 2.2
All the models show W to oscillate around the mean W, their magnitude remaining relatively
constant over the process. In absolute terms, these oscillations tend to be lower for lower
discharge rates, something that can be seen by comparing model M6 with models M7 and
M11. However, in relative terms, the oscillation levels (which can be quantified through the
value of the CV shown in Fig. 3) behave just in the opposite manner. In this case, bigger
values of the CV (bigger oscillation level) are found for smaller hopper outlets. This is a
consequence of the higher ratio particle size/hopper outlet, which provokes a more stop/start
flow. This effect can be seen clearly when comparing models M6 (CV =8.3 %) and M7 (CV =
5.2 %), both of them with the same but with a bigger outlet in M7. In addition, a more
oscillation level is found when considering a shallower hopper, as it is the case of model M7
when compared with model M11. Both models have the same outlet size but model M11
represent a steeper hopper. When a steeper hopper is considered, particles tend to orientate
in a more vertical way as they approach the hopper outlet. Then, particles can come out the
hopper more easily and, therefore, a more continuous flow is presented.

3.2 Comparison between numerical y analytical results.
The value of W was analytically estimated using Eq. 3 in combination with either Eq. 4 (factor
F by Brown and Richards (1965)) or Eq. 5 (factor F by Rose and Tanaka (1959)). Table 2
shows the values of the key variables used in the analytical estimation of the mean discharge
rate (W) as well as the corrections factors F calculated with Eq. 4 or Eq. 5. For both
correction factors, the value of the mean discharge rate W is given together with its relative
error (%) with respect to the reference value obtained via numerical calculation (W (DEM)).
Variable M6 M7 M11
0
1000
2000
3000
4000
5000
6000
10 20 30 40 50 60 70 80
W

(
k
g
/
s
)
()
BrownandRichads(1965)
RoseandTanaka(1959)
DEM
D=1.5 m
D=1 m
M7
M11
M6
Analytical
calculation

f
(kg/m
3
) 560 560 560
D (m) 1 1.5 1.5
k 2 2 2
d (mm) 150.5 150.5 150.5
() 45 45 30
45 45 45
Numerical
calculation
W (DEM) 815 2688 3084
Brown
and
Richards
(1965)
F 0.96 0.96 1.10
W 709.1 2775.3 3084.0
Diff. (%) 13.0 3.2 0.0
Rose and
Tanaka
(1959)
F 1 1 1.21
W 735.4 2878.4 3404.6
Diff. (%) 9.77 7.08 10.40
TABLE 2.- Parameters used in the
analytical calculations. Comparison with
numerical results.
FIGURE 4.- Comparison between numerical
results and analytical results.

In addition, Fig. 4 shows a graphical comparison between the analytical predictions and the
numerical results for models M6, M7 and M11. The curves shown in Fig. 4 represent the
analytical functions W=f() for two possible values of the size of the hopper outlet (D =1 m
and D =1.5 m) and for the two correction factors described in 2.2 .The dots included in Fig. 4
represent the mean discharge values (W) obtained with the numerical models.
The numerical predictions are in good agreement with the values analytically predicted for all
models and correction factors. However, for low discharge rate models (M6), the difference
between numerical and analytical predictions is a bit higher than for high discharge rate
models (M7 and M8). Both correction factors allow the prediction of acceptable values of W.
However, the correction factor of Brown and Richards (1965) seems to be more accurate.


4. Conclusions
Different DEM models were developed to predict the discharge rate in a hopper filled
with biomass briquettes.
The numerical results were compared with classical analytical predictions. In general, a
good agreement between both methods was found, with differences lower than 10 % in
all cases.
The results shown in this paper are the preliminary results of a wider research project
which is currently being developed. These preliminary results are considered to be good
enough to continue with the rest of de models included in the research project.

References
ASAE S368.4 (2006). Compression Test of Food Materials of Convex Shape. American
Society of Agricultural and Biological Engineers (ASABE)
ASTM D854-10 (2010). Standard Test Methods for Specific Gravity of Soil Solids by Water
Pycnometer. American Society for Testing and Materials (ASTM).
Beverloo, W.A., Leniger, H.A., Van de Velde, J . (1961).The flow of granular solids through
orifices Chemical Engineering Science 15, 260269
Brown, R.L., Richards, J .C. Profile of flow of granules through apertures. (1960).
Transactions of the Institution of Chemical Engineers, 38, 243250
Cundal, P. & Strack, O. (1979). A discrete numerical model for granular assemblies.
Geotechnique 29(2), 47-65.
EDEM, 2010. EDEM 2.3 User Guide. DEM Solutions, Ltd., Edinburgh, Scotland, UK.
Favier, J . F.; Abbaspour-Fard, M. H., K. M. & Raji, A. (1999). Shape representation of axi-
symmetrical, non-spherical particles in discrete element simulation using multi-element
model particles. Engineering Computations 16(4), 467 - 480.
Gonzlez-Montellano, C.; Fuentes, J .; Ayuga-Tllez, E. & Ayuga, F. (2012). Determination of
the mechanical properties of maize grains and olives required for use in DEM simulations.
J ournal of Food Engineering 111(4), 553-562.
Gonzlez-Montellano, C.; Ramrez, ..; Gallego, E. & Ayuga, F. (2011). Validation and
experimental calibration of 3D discrete element models for the simulation of the discharge
flow in silos. Chemical Engineering Science 66(21), 5116-5126.
Li, Y.; Xu, Y. & Thornton, C. (2005). A comparison of discrete element simulations and
experiments for sandpiles' composed of spherical particles. Powder Technology 160(3), 219-
228.
Myers, M.E., Sellers, M., 1971. Chemical engineering, tripos part 2. Research Project
Report, University of Cambridge.
Rose, H.E., Tanaka, T. (1959). Rate of discharge of granular materials from bins and
hoppers. Engineer, 208, 465469
Tsuji, Y.; Tanaka, T. & Ishida, T. (1992). Lagrangian numerical simulation of plug flow of
cohesionless particles in a horizontal pipe. Powder Technology 71(3), 239-250.


Experimental study of the pressures exerted by the wheat stored in
slender cylindrical silos. Comparison with Eurocode 1 part 4.
Angel Couto
1*
, Pedro J . Aguado
1
,

Angel Ruiz
1
, Alberto Tascn
2
1
Department of Agricultural Engineering and Sciences, ESTI Agricultural, University of Leon,
Av. Portugal 41, 24071 Len, Spain.
2
University of La Rioja Madre de Dios 51, 26006 Logroo (Spain)
*Corresponding author. E-mail: acouy@unileon.es

Abstract
Very few experimental installations in the world have full-scale silos, and very few assays
have been conducted on them. Consequently, many uncertainties remain which require
further research in order to reliably predict the behaviour of material stored in this kind of
structure. In this paper the results of the tests carried out using a real scale experimental silo
are presented by the authors. The different obtained pressures are analysed and compared
with those calculated using the Eurocode 1, part 4.

Key words: Silo, experimental test, wheat, pressures

1. Introduction
Very few experimental installations in the world have full-scale silos (Brown et al., 2000; Hrtl
et al., 2008; Ramirez et al., 2010; Schurich et al., 2001; Teng and Lin, 2005; Teng et al.,
2001; Zhong et al., 2001) and very few assays have been conducted on them (Askegaard
and Munch-Andersen, 1985; Chen et al., 2005, 2007; Hrtl et al., 2008; Ooi et al., 1990;
Rotter et al., 2002; Teng et al., 2005; Wu et al., 2009; Yang and Hsiau, 2001). Such
structures and tests demand a considerable investment, and although they have yielded
extremely important results, many uncertainties still remain which require further research in
order to reliably predict the behaviour of material stored in this kind of structure.
For determination of the actions due to the stored material it has been employed a real size
testing station. The installation basically consists of a full-scale, cylindrical silo equipped with
load cells to measure pressure and variable-frequency drives connected to each of the
electric motors driving the filling and discharge screw conveyors, in order to study the effect
on pressure of the speed at which the silo is filled or discharged. As a result of this novel
design, it is possible to obtain most of the parameters which control the behaviour of stored
material and to compare and validate the different theoretical models used for performing
calculations and establishing current standards. (Couto et al., In Press-a, b).
This paper presents the results of pressures on the walls of cylindrical steel silos obtained
when the stored material is wheat. The results are compared with those values obtained
using the European Code dealing with actions in silos (CEN, 2006).

2. Material and methods
2.1. General description of the installation
The principle components of the test station consist of a silo for storing the material to be
tested and another, experimental silo containing the instrumentation necessary for
measuring the actions of the stored material. The stored material is transferred between
both silos using two screw conveyors powered by electric motors (Figure 1).



2.2. Geometry of the experimental silo
The geometry and dimensions of the test silo are shown in Figure 2. The silo is cylindrical,
with a central hopper and a roof at the top. The silo body (a vertical cylinder), the hopper
and the roof are independent and can be dismantled.
The silo walls were constructed from sheets of stainless steel 3 mm thick, whilst the cylinder,
hopper and roof reinforcement rings were 50 mm wide and 10 mm thick. Given the
dimensions and the type of material used for the walls, the silo can thus be considered a
rigid, smooth-walled steel silo.
According to the Eurocode classification, it corresponds to a slender silo, since the cylinder
height/diameter ratio is equal to 2 (h
c
/d
c
=2).

x
p
h
p
h
p
v
p
w
z
z
t
h =2000
c
d =1000
c
p
n h =480
t

d =350
h
h =738
h
e =0
0
p
n
250
250
250
250
250
250
150
150

FIGURE 1: General View of the Test
Station
FIGURE 2: Symbology and dimensions of
the test silo

2.3. Measuring horizontal pressures
To measure normal wall pressures, a vertical generatrix was located on the cylinder wall,
along which 7 readings were taken at the different heights indicated in (Figure 2 and 3),
whilst hopper pressures were measured immediately below the silo-hopper transition, at the
prolongation of the generatrix mentioned above for measuring cylinder pressures.
Double bending beam load cells were used to measure normal wall pressures. For this, 150
x 150 mm openings were made in the wall, into which panels of the same curvature were
inserted, leaving a small gap around the edges. The panels were attached to the load cells
as indicated in Figure 3. Hopper wall pressure readings were taken in the same way.


FIGURE 3: Measurement of normal
cylinder wall pressures
FIGURE 4: Measurement of vertical
forces.

2.4. Measuring vertical forces
In order to measure vertical forces caused by friction of the stored material against the wall,
the three separate and independent parts of the silo (cylinder, hopper and roof) were
connected by mounting the external reinforcing rings of each on the tension/compression
load cells. The rings were very close but did not come into contact, and thus all vertical
pressures were absorbed by the load cells (Figure 4).

2.5. Description of the tests
Tests were conducted using common wheat (Triticum aestivum), Galera R2 variety, with a
minimum purity of 98%. The properties of this material were previously determined
experimentally in the School of Agricultural Engineering (E.S.T.I.A.) laboratories in
accordance with the method proposed in Eurocode 1, part 4 (CEN, 2006). The values
obtained were: specific weight, 8.38 kN / m
3
; angle of repose, 34.22; steel wall friction
coefficient, 0.20; and humidity, 10.3%. The values obtained in these tests, both as regards
the specific weight and grain-wall friction, were very close to those obtained in the
experimental silo.
Using this granular material, 12 tests were conducted which consisted of collecting data
during three different phases: during silo filling, under static conditions and during discharge.
Silo filling was performed centrally at a constant speed. Once this phase was completed, the
material was left inside the silo under static conditions, and lastly, the material was
discharged.
The behaviour and values obtained for the different parameters analysed were similar in all
12 tests: consequently, in order to avoid presenting an elevated number of figures or
overloaded graphics, only the results obtained for one of the tests conducted will be given in
the following section
3. Results and discussion
Figures 5 shows the normal silo wall pressures and the vertical pressures of the stored
material at the silo-hopper transition during the three phases. As expected, normal wall
pressures reached their highest values during discharge, at the hopper wall at the level of the
silo-hopper transition, in accordance with European standards (CEN, 2006) and in
agreement with international tests on real silos (Askegaard and Munch-Andersen, 1985;
Hrtl et al., 2008; Ramirez et al., 2010; Zhong et al., 2001)
The same figure also shows that in the interval between completion of filling and start of
discharge, in other words, during the static phase, pressures inside the silo did not remain
constant. This is because the material undergoes a process of settling inside the silo, a
process which occurs in surges separated by increasingly longer intervals.

FIGURE 5: Normal silo wall pressures (phGA,t) and vertical stress in the stored material at
the transition (pvt,t), at each time t
In the same Figure it can be seen that silo filling occurred at a constant speed, since the GA
cylinder cells placed at regular intervals around the wall began to detect stresses at regular
time intervals. Once filling was completed and the screw conveyor motors disconnected, the
phenomenon of grain settling began. As can be observed, the time interval between each
surge became longer, a phenomenon which, as we mentioned earlier, continued to occur
throughout the static phase of the test.
It can also be seen in Figure 5 that the highest pressures inside the silo occurred at the
beginning of discharge, a behaviour which is in agreement with Eurocode 1, part 4 (CEN,
2006), and with international publications on silos.
The extremely short time interval that we established for reading data during the test (0.5 s)
enabled us to observe that the overpressure produced during discharge did not occur
instantaneously, to then decrease gradually, but rather, it reached a maximum value which
remained almost constant for some seconds.
Figure 6 provides a comparison of the maximum pressures obtained during tests under static
conditions and those obtained during discharge using Eurocode 1, part 4 (CEN, 2006). As
could be seen in the previous figures, the values obtained in real-life tests were lower than
those obtained using the calculation methods proposed in the European standard. Thus, the
biggest differences were observed during discharge at normal pressures on the cylinder wall
at the silo hopper junction, where variation with respect to the standard reached 75.7%.


FIGURE 6: Comparison between normal wall pressures obtained with the tests and using the
Eurocode 1, part 4

4. Conclusions
Before discharge, in other words, during filling and when the material stored inside the silo is
static, the highest normal cylinder wall pressures occurred just at the end of filling. In
contrast, the opposite was true for the hopper wall, were normal wall pressures increased
over storage time.
The highest pressures inside the silo occurred during the first seconds of discharge, in
agreement with real-life tests conducted by several other authors (Askegaard and Munch-
Andersen, 1985; Hrtl et al., 2008; Ramirez et al., 2010; Wu and Schmidt, 1992; Zhong et
al., 2001) and with the European standard on silos (CEN, 2006). The highest normal wall
pressures occurred at the hopper wall, at the silo-hopper transition. This was also where the
highest normal wall pressures were produced in the cylinder, although these pressures were
considerably lower than those reached in the hopper. In our tests, normal hopper wall
pressures were of the order of 180% higher than those obtained for the cylinder wall.
Under static conditions, the stored material undergoes a process of settling. During this
process, the cylinder walls released pressure, increasing the normal wall pressure of the
hopper and the vertical pressure of the stored material at the silo-hopper transition. This
phenomenon occurred in surges over increasingly longer time intervals. This aspect requires
further in depth study before drawing definitive conclusions as it may be due to, or be
amplified by, the specific design of the silo.
A study of pressure patterns under static conditions inside the silo due to grain settling, and
of overpressure patterns along the vertical generatrix in the cylinder during discharge,
suggests that the increase in pressures observed in certain areas was not a consequence of
chaotic behaviour but rather obeyed certain physical laws which have not yet been studied
sufficiently. This point is considered in the Eurocode (CEN, 2006) in the section on Patch
load, where these laws are taken into account both in static and dynamic states. They
comprise a number of phenomena that produce asymmetries.

Acknowledgements
The authors thank the Spanish Research and Technology Commission (CICYT) (Research
Project AGL2005-07430-C02-01/AGR) and the Regional Executive of Castilla y Len
(Research Project LE020A10-2) for financing this research.

Reference list
Askegaard, V., Munch-Andersen, J ., 1985. Results from tests with normal and shear stress
cells in a medium-scale model silo. Powder Technology 44, 151-157.
Brown, C.J ., Lahlouh, E.H., Rotter, J .M., 2000. Experiments on a square planform steel silo.
Chemical Engineering Science 55, 4399-4413.
CEN, 2006. EN 1991-4:2006. Eurocode 1: Actions on structures. Part 4: Silos and Tanks.
European Committee for Standardization. Brussels.
Couto, A., Ruiz, A., Aguado, P., In Press-a. Design and Instrumentation of a Full-Scale Test
Station for Measuring Static and Dynamic Pressures in Silos under Different Conditions.
Part I. Computers and Electronics in Agriculture.
Couto, A., Ruiz, A., Aguado, P., In Press-b. Design and Instrumentation of a Full-Size Test
Station for Measuring Static and Dynamic Pressures in Silos under Different Conditions.
Part II: Construction and Validation. Computers and Electronics in Agriculture.
Hrtl, J ., Ooi, J .Y., Rotter, J .M., Wojcik, M., Ding, S., Enstad, G.G., 2008. The influence of a
cone-in-cone insert on flow pattern and wall pressure in a full-scale silo. Chemical
Engineering Research and Design 86, 370-378.
Ramirez, A., Nielsen, J ., Ayuga, F., 2010. Pressure measurements in steel silos with
eccentric hoppers. Powder Technology 201, 7-20.
Schurich, T., Frll, C., Enstad, G.G., 2001. Full scale silo tests and numerical simulations of
the "cone in cone" concept for mass flow, In: Kalman, A.L.a.H. (Ed.), Handbook of Powder
Technology. Elsevier Science B.V., pp. 175-180.
Teng, J .G., Lin, X., 2005. Fabrication of small models of large cylinders with extensive
welding for buckling experiments. Thin-Walled Structures 43, 1091-1114.
Teng, J .G., Zhao, Y., Lam, L., 2001. Techniques for buckling experiments on steel silo
transition junctions. Thin-Walled Structures 39, 685-707.
Wu, Y.H., Schmidt, L.C., 1992. A BOUNDARY ELEMENT METHOD FOR PREDICTION OF
SILO PRESSURES. Computers & Structures 45, 315-323.
Zhong, Z., Ooi, J .Y., Rotter, J .M., 2001. The sensitivity of silo flow and wall stresses to filling
method. Engineering Structures 23, 756-767.




Discrete Element Modeling of Cohesive Grain Materials

Itzhak Shmulevich*
1
, Zvika Asaf
1
, Ramy Catav
1

1
Technion, Faculty of Civil Engineering, Research Center of Agricultural
Engineering, Hifa 32000, Israel,
*Corresponding author. E-mail: agshmilo@technion.ac.il.
Abstract.
The Discrete Element Method (DEM) has been found suitable for modeling and investigating
granular materials. The model parameters for evaluation of non-cohesion materials has been
presented in previous research, but the correct model for cohesive material and the calculation
of its parameters has not yet been achieved. The highly cohesive response of partially saturated
granular material depends upon the interaction of the particles, as well as interactions due to the
capillary pressure and surface tension that produce attraction forces between the particles. To
evaluate the parameters of the attraction force resulting from hydrostatic pressure and surface
tension for DEM models at the micro-scale level, a laboratory measurement system was
designed built and tested.
The measurement system is based on Atomic Force Microscope (AFM) principles, in which
force is measured by means of cantilever bending which is measured by an optical microscope
with digital image processing. The system enables accurate evaluation of the forceseparation
distance curve for a wide range of liquid volumes, and grains of different diameter sizes and
textures. Tests of the liquid bridge between two glass spheres of 2 mm diameter obtained good
agreement with the theoretical force-displacement curve. The attraction force that was
measured in the sphere glass tests was in the order of 0.2 mN, and the separation distance was
up to 850 m. A full description of the measurement system will be presented, including test
results of agricultural material grains like rice. In addition, the test results will be compared with
a simple case of applying the parameters to wet granular material in the DEM model using a
softening contact model.
Keywords.
Discrete Element Method, granular materials, cohesive materials, liquid bridge, force separation
curves.
1. Theoretical Background
1.1 Liquid bridge definition
The systems investigated in the present research contain two rigid spherical elements, with
different amounts of liquid between them (Fig. 1). The presence of liquid causes attraction
forces between the two elements. The equilibrium of the cohesion forces in the liquid air surface
interface is achieved by the geometric curvature of the liquid surface and by the difference
between the air pressure and the liquid pressure (Lambert et al., 2003).








FIGURE 1: Two solid spheres with liquid between them.

The equilibrium equation is known as the Young-Laplace equation (Eq. 1).
[1]
where: P is the differential pressure between the air and the liquid; R
X
is the surface
curvature in the x direction; R
y
is the surface curvature in the y direction; and
lg
is defined
as the liquidgas surface tension coefficient, which is specific for every kind of liquid and gas.
The total force, F, acting between two spherical bodies of radius a (Fig. 2), is obtained by the
sum of two components: (a) the axial component of the surface tension acting on the three-
phase (airliquidsolid) contact line, and (b) the hydrostatic pressure acting on the axial
projected area of the liquid contact on either sphere. This leads to the expression:

t u t
2 2
sin ) sin( sin 2 p a a F A + + =
[2]
where: - is the liquid surface tension, - is the half- filling angle; and - is the contact angle.
The pressure deficiency across the airliquid interface is given by the Young-Laplace equation,
expressed in the dimensionless form:
2 / 1 2 2 / 3 2
) 1 (
1
) 1 (
* 2
Y Y Y
Y
H


+

+
=
[3]
where the dimensionless mean curvature of p- H*, is given by 2 / * pa H A = . X and Y are
dimensionless coordinates with respect to the sphere radius, a, and the dot notation refers to
the differentiation with respect to the X coordinate, which is the axisymmetric horizontal center
line. The solution of Eq. [3], Y
(X)
, represents the profile of the liquid bridge (Fig. 2).







FIGURE 2: Illustration of the liquid bridge and the geometric parameters.

Rigid element #1
Liquid
a a
Y(x)
Y0
s s


x
x
c

Rigid element #2

Fisher (1926) proposed a simpler approach to estimate liquid bridge forces. It was based on a
toroidal approximation, which involved treating the meridional profile of the liquidair interface
as an arc of a circle .In the case of toroidal approximation, the curvature profile Y
f
is
represented by
2 2
1 2 1
) * ( * * X Y
F
+ =
[4]
where *
1
and *
2
are the dimensionless radii of the toroid with respect to the radius of the
spheres, a, *
1
corresponds to the radius of the meridian profile, and *
2
is the radius of the
neck.The formulation for the total force that obtained from Eq. [2] is
} { | u | | t 2 sin ) * ( ) sin( sin 2 *
1 F F
H F + + =
[5]
where H*
F
is the maximum value at the three-phase contact line.
1.2 DEM simulation for cohesive materials
The DEM simulation examines the behavior of granular material according to the properties of
the contact between the particle themselves and between the particles and the boundary
surface. The behavior of the system is determined by the solution of equations of motion for
each particle. The features of this model are described in Fig. 3 .










FIGURE 3: Softening contact model between two particles.
The model consists of a linear spring with a constant (Kn) in the normal direction and a linear
spring with a constant (Ks) in the tangential direction in a series with a friction coefficient ().
And g). The cohesion and adhesion forces represented with a softening
model (F
C
).

2. Experimental Instrumentation
2.1 The measurement method
The measuring is based on cantilever bending. A schematic diagram of the system is shown in
Fig 4. The cantilever bending is calculated by Eq. 6 as follows:
S X A A = o
[6]

where is the bending value at the free edge of the cantilever, X is the displacement of the
linear stage, which is the displacement of the fixed edge of the cantilever, and s is the
separation displacement between the solid spheres.







FIGURE 4: Description of the measurement system. The cantilever is in loaded position due to
the liquid bridge.
The system is illustrated in Fig 5.












FIGURE 5: General view of the measurement system.

3. Results and discussion
For comparing the test results with the theoretical equations, the geometric parameters of the
liquid bridge was measured (contact angle , half-wetting angle , separations distance S, and
surface tension ). These geometric parameters can be evaluated from the snapshot of the
liquid bridge with an appropriate picture analysis program (Fig. 6).



X -
LINEAR
STAGE


S


MOVEMENT DIRECTION OF
THE LINEAR STAGE

X



Microscope
Cantilever
Humidity
controlled
chamber
Linear Stage
Light source
Moisture
Generator

-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
-100 0 100 200 300 400 500 600 700 800 900 1000
f
o
r
c
e

(
m
N
t
)
separation distance -(microns)
calculated
data
test
calculatedcurve
measured curve





FIGURE 6: Microscope picture of the solid spheres and the liquid bridge.
A comparison between the measurements and calculations (based on geometric
measurement) is shown in Fig. 7.









FIGURE 7: Comparison of force separation curve between test results (solid
line) and calculated curve (dashed line).
Translating the force separation curve to the DEM contact model
A simple test was performed to provide the DEM with the parameters for cohesive material. The
test configuration is described in Fig. 8. Four hundred glass balls were placed in a rectangular
box. The balls were pushed to one corner by a moving plate and then immediately released,
once with completely dry balls, and again with balls that were completely wet.



(a) (b) (c)
FIGURE 8: (a) 400 balls spread in a rectangular box; (b) a moving plate compressed the balls to
the left corner; (c) the moving plate was removed and the particles spread again The DEM
simulation was performed using the Itasca 3D program. The test results are presented in Figs.
9 (a) and 9 (b).



(a) (b)
FIGURE 9: (a) Spread of the balls in the dry test; (b) spread of the wetted glass balls test. The
black curve in pictures describes the contour of the spread of the balls

The simulation results are presented in Figs. 10 (a) and 10 (b).





(a) (b)
FIGURE 10: DEM simulation results for test case with dry balls (a) and wet balls (b)
It can be seen that there is good correlation between the configurations of the balls in the test
and the simulation results.
4. Conclusion
The measurement system enabled the production of a force-displacement curve for the normal
attracting forces between two particles with a liquid bridge between them.
The measurement system can be adapted for use with different shapes of particles. By
changing the cantilever size enables measurement of particles that are at least one order
smaller, that is, a diameter of less than 200 m and a force range of less than 0.001mN.
The softening model provides a good description of the force separation curve, at least for two-
sphere case. The true force separation curve that was measured is close to the simplified
triangular description of the softening model.
The DEM simulations with the cohesive parameters showed good agreement with the
characteristic results in the tests. These capabilities contribute another step towards enabling
more reliable simulation of particles of different geometry and materials and other types of
liquids

References
Clark, W.C. Haynes, J.M., and Mason, G.,1965, Liquid bridge between spheres, Chem.Eng.Sci
20, 859.
Cundall P. A. and O. D. L. Strack. 1979. A discrete numerical model for granular assemblies. J.
Geotechnique 29(1):47-65.
De Bisschop, F.R.E., and Rigole, W.J.L.,1982, A physical model for liquid capillary bridges
between adsorptive solid spheres, J. Colloid interface Sci. 88, 117.
Fisher, R, A, , 1926, On the capillary forces in an ideal soil, J.Agric.Sci 16, 492.
Guoping Lian, Colin Thornton, Michael J. ADAMS, 1993, A theoretical study of the liquid bridge
forces between two rigid spherical bodies", Journal of colloids and interface science:
161, pages 138-147.
Itasca, 2002. PFC2D Users Manual, Version 3.0., Itasca Consulting Group Inc.,Minneapolis,
MN, 55415 USA.
Lambert, P.; Letier, P.; Delchambre, A. ,2003, Capillary and surface tension forces in the
manipulation of small parts, Proceedings of the 2003 IEEE International Symposium on
Assembly and Task Planning, p 54-9.

The Household Metal Silo: A Helpful Technology For Food Security


Danilo J. Meja Loro
1*
, Divine N. Njie
1
1
Food and Agriculture Organization of the United Nations, FAO-Rome, Italy.
* Corresponding author. E-mail: Danilo.Mejia@fao.org

Abstract

Postharvest losses of grains and cereals are considerable among rural farmers of developing
countries. Most of these losses occur during storage due to the use of inefficient structures that
allow pests and moisture to enter the product, thus causing quantitative and qualitative losses. As
a way to reduce these losses and enhance food security, the Food and Agriculture Organization of
the United Nations (FAO) has validated and promoted the household metal silo, a hermetic
cylindrical structure which is made with galvanised metal sheets and can be used to efficiently
preserve grains or cereals. FAO has introduced the silo in twenty four countries in Africa, Asia and
Latin America during the last fifteen years. In this paper, the technology of the household metallic
silo and the approach used for its transfer are described. Experiences and lessons learnt in
implementing the technology are also described with emphasis on technical and socio-economic
aspects.

Keywords: Postharvest, quantitative and qualitative losses, household metal silo, food security.


1. Introduction

The agriculture sector needs to have efficient food production and post-production systems
which allow healthy crops with adequate yields to be obtained and permit proper handling of the
harvest so as to reduce losses and achieve food security. Indeed this should be a priority
especially in developing countries where there is still insufficient access to technical knowledge to
be applied along the food chain from the field production to consumption.

In general, postharvest losses for agricultural crops are considerably higher in developing
countries than in developed countries, with the main reason being the lack of technical knowledge.
For instance, in the case of cereals such as wheat, sorghum and maize postharvest losses in east
and southern Africa may reach 10-20 percent (The World Bank, 2011). These losses may occur
during the different operations which the grain or cereal has to undergo before it reaches the final
consumer; therefore the magnitude of the losses varies depending on how efficiently these
operations are conducted.

Grains and cereals are exposed to qualitative and quantitative losses during operations
such as harvesting, drying, threshing, transport, storage, shelling/dehusking, cleaning, milling,
bagging, distribution and processing. The losses are more significant in some of these operations
than others. Storage is a serious problem particularly in developing countries where the traditional
storage structures used are inefficient because they do not keep the product in hermetic and safe
conditions that impede the entrance of moisture and pests. Grain preservation is evidently difficult
for small and medium farmers everywhere, but it becomes critical in tropical countries because of
the high temperatures, frequent rains and high relative humidity in the environment which is
favourable to pests such as rodents, insects and fungi (Meja et al. 2012).

There are diverse methods of storage of grains and cereals available for small and medium
farmers. These include wood cribs, metal drums, plastic containers, household metal silos, bags
of plastic or jute and sheds among others. The selection by the user of any of these methods will
depend upon factors such as availability, convenience of use, the particular situation, the efficiency
of the method and suitable benefit-cost ratio (Meja et al. 2012). In the 1980s, the Swiss
Cooperation had a positive experience in Central America in using household metallic silos for the
prevention of postharvest losses of staple grains and cereals (COSUDE, 1998; FAO, 2008).
Drawing from this experience FAO validated and promoted the household metal silo technology for
grain storage at the end of the 1990s. Since then, the Organization has promoted the household
metal silo through projects implemented in a diverse range of developing countries in Africa, Asia
and Latin America.

2. Technology description

The household metal silo is a simple cylindrical structure which can be constructed using
simple tools and easily available materials, galvanised metal sheets being the most expensive
material required. In order to build durable silos at relatively low cost, the galvanised sheets used
should have particular technical specifications. Silo construction normally requires galvanized
sheeting of 100x200 cm and 0.5 mm of thickness (known as standard No.26). With good
maintenance, the average useful life of a silo is around 15 to 20 years. Silos of different sizes can
be made, ranging in capacity from 100 to 3000 kg. Examples of intermediate sizes are 250, 500,
750, 1000 and 1500 kg. Although silos of 4000 kg capacity can be constructed, they present some
difficulties for handling and are therefore not in high demand. Silos of small capacities such as 100
kg or 120 kg are in general recommended for storing seeds for planting, while silos of larger
capacities are recommended for storing grains for human consumption. A 1000 kg silo can hold
enough grain to feed a family of five for one year.

A small or medium farmer owning more than one silo can store surpluses for off-season
sale when prices are more attractive, thus increasing household income (FAO, 2008). Apart from
being used to store grains for families in rural areas, the metallic silos have been employed in
countries like Kenya for storing grains used for preparing meals in institutions (e.g. secondary
schools). Small poultry farmers have successfully used the silo for storing feed (Meja, 1998). In
order to obtain excellent results in the preservation, the cereals and grains must be properly dried
before placing them inside the silo. The maximum moisture content should be 14 percent for stored
cereals such as maize and rice and 10 percent for pulses and oil seeds. Inadequate drying of the
grain or cereal may result in increased losses. It is also important to place the silo on a wooden
pallet so that its underside is protected, and to place the silo in an area protected from sunlight and
rain so as to avoid corrosion (FAO, 2008, COSUDE 1998).

In comparison to other storage methods used by small and medium farmers in the
developing countries, the silo has the following advantages: the quality of the product stored in it is
maintained; its air tightness permits effective fumigation and avoids the use of insecticides which
could leave chemical residues on the product; it requires little space and can be installed close to
the home; quantitative losses of the product stored in it are close to nil; it allows farmers to take
advantage of fluctuating grain prices; prevents entrance of rodents and other pests that could harm
consumer health; easy to use, relatively low cost; facilitate womens work; can be built in situ at the
location of final use with local labour and easily available materials; it is a decentralised storage
system; and the technology has been tried and tested in several countries (COSUDE, 1997; FAO,
2008).

3. Technology transfer method

Since 1996, the Rural Infrastructure and Agro-Industries Division of FAO has introduced
household metal silos through different projects in a wide range of countries in the developing
world. Some of these projects have been financed by FAOs regular programme of work and
external donors (international agencies, governments, etc), while others have been implemented at
the request of the beneficiary country. In the last few years FAOs emergency programme has
implemented projects involving the dissemination of household metal silos in several countries
affected by natural disasters or conflicts.

In these projects the method that has been employed to disseminate the metallic silo
technology is Training of Trainers (ToT). This involves training a national master group of about 25
people on silo construction, use and maintenance. The training is provided by an international
expert who is provided generally through a South-South Cooperation (SSC) agreement between
an emerging developing country and the beneficiary country.
The national master group is made up of agricultural technicians, blacksmiths/tinsmiths,
skilled interested farmers and staff of NGOs and related institutions working in the agriculture
sector. The group is trained so that they can subsequently deliver the training to other national
beneficiaries. Usually the initial master group receives a free toolbox and kit for silo construction,
as well as a technical manual on construction, use and maintenance of the metal silo. As the
material required in the training are very common, it is not difficult to implement the training
workshops which last about one to two weeks.
In general, the beneficiary country is requested to establish a demonstration centre for use
and maintenance of the silo in order to provide support to farmers needing silos for grain storage.
The local tinsmiths or artisan blacksmiths trained in silo construction are also motivated to become
entrepreneurs and promoters of the silo through providing this service to their communities. In such
a way they can establish business agreements with farmers, grain traders and users in general
wishing to acquire silos. Additionally, during the implementation of projects in which silos form an
important component, the silos are promoted through demonstrations, radio spots, brochures,
posters, lectures, etc.

4. Technology adoption and lessons learned

FAO has introduced and disseminated the household metal silos in 24 countries of Africa,
Asia and Latin America with the objective of reducing post harvest losses and enhancing food
security (Table 1). In these projects, most national authorities, farmers, silo artisans and other
direct or indirect beneficiaries and stakeholders have been very satisfied with the technology. In
some countries and rural communities farmers have adopted and retained well the silo technology.
For instance, in the last five years more than 65 000 household metal silos have been constructed
and delivered through projects conducted by FAO in Afghanistan and the feedback from
beneficiaries has indicated satisfaction with the adoption of the technology.
Table 2 shows the cost of production of silos in a selection of countries that have benefited
from FAO projects. These costs include materials, labour and depreciation of tools, but do not
include utility, transportation of silos to destination and other costs. Most beneficiaries who have
adopted the technology during the last 10 years have considered that silos are affordable and
present a fast return of investment in the first harvesting year under normal crop production
conditions (Meja & Francescutti, 2011).
One of the challenges commonly faced is that silo artisans lack capital and access to credit
to finance the creation of their own silo construction business. Although the cost of production of
the silos varies according to the socio-economic and political circumstances in each country, it is
clear that the most expensive material used to build silos is the imported galvanised metal sheet. In
some cases the high cost of galvanised sheets makes the silos unaffordable for farmers.
The experience from the projects implemented suggests that the initial enthusiasm about
the technology gradually disappears if the national government is not strongly committed to provide
follow up support. Although the silo adoption process is different in each country, complementary
measures must be taken by the government in order to make adoption successful and sustainable.
These include: providing incentives to farmers, artisans and other stakeholders linked to grain and
cereal production and commercialization; organizing farmers in groups that can easily acquire and
more profitably operate silos; taking appropriate regulatory measures such as reducing import
taxes on galvanised sheets; and setting up revolving funds and loans to farmers and artisans.

5. Conclusions

(1) The household metal silo offers several advantages in providing a solution to the high
postharvest losses of grains and cereals at the level of small and medium farmers in developing
countries.
(2) Governments of countries where projects to introduce silos are implemented must put in place
parallel strategies to provide an incentive for the sustainable and large-scale adoption of the
technology.
(3) The household metal silo represents an important element in the grain distribution system.
Within a well organized system, it can create a positive socio-economic impact among players who
are directly or indirectly involved in the grain production and distribution chain such as farmers,
artisans, hardware owners, middlemen, consumers and credit institutions, among others.


References
Arancibia, A. 2012. Personal communication.
COSUDE, 1997. Everybody gets a grain silo. Postcosecha in Central America: The story of a
successful project in post-harvest protection. Text by Kurt Scheneider and edited by S. Brunold &
P. Schmidt, LBL
COSUDE, 1998. Silos Metlicos: Manual para la fabricacin de silos metlicos para almacenar
granos. Programa Regional de Poscosecha: Guatemala, Honduras, El Salvador y Nicaragua, CA.
Agencia Suiza para el Desarrollo y la Cooperacin. Cuarta Edicin. Managua, Nicaragua.
FAO, 2008. Silos Metlicos Familiares: Aliados fundamentales de la FAO en la lucha contra el
hambre. Servicio de Tecnologa de Ingeniera Agrcola y Alimentaria, Organizacin de las
Naciones Unidas para la Agricultura y la Alimentacin, AGST-FAO. E/I/F-AD/I/I0022S/1/3.08/500.
Meja, D. 1998. Estudio de Impacto en Aspectos Tcnicos y Socioeconmicos del Programa
Poscosecha en Bolivia. FAO-AGSI Reporte Final en base al proyecto GCP/Bol/020-029-032/Net.
Meja, D. & Francescutti, D. 2011. Evaluacin econmica de los silos metlicos familiares para
almacenamiento de granos y cereales. Revista Granos: postcosecha latinoamericana de la semilla
al consumo. Ao 18 No. 84 pg. 30-32. www.consulgran.com
Meja, D., Arancibia, A., & Howell, M. 2012. Manual Tcnico para la Construccin y Uso de los
Silos Metlicos Familiares para Almacenar Cereales y Leguminosas de Grano. FAO Publication in
edition for printing, Rome, Italy, May 2012.
The World Bank, (WB); Natural Resource Institute, (NRI); & The Food and Agriculture Organization
of United Nations, (FAO), 2011. Missing Food: The case of postharvest grain losses in Sub-
Saharan Africa. Pgs. 18-19. Report No.60371-AFR.

Table 1. Countries where the household metal silo has been introduced and disseminated
*
by
FAO (1996-2012)
Africa Asia Latin America
Burkina Faso, Chad, Gambia,
Guinea, Madagascar, Mali,
Malawi, Mozambique,
Namibia, Senegal, Somalia,
Uganda

Afghanistan, Cambodia, Iraq,
East Timor
Bolivia, Ecuador, Panama,
Nicaragua
*
, Honduras
*
,
Guatemala
*
, Peru, Haiti.
* In these countries silos have been only disseminated.

Table 2. Cost of production for household metal silos of different capacities introduced by
FAO projects in some countries (US$)
Country 120 kg 250kg 500kg 900kg 1800kg
Afghanistan
1
- 28 70 - 92
Bolivia
2
22 40 60 78 112
Burkina Faso
3
26 29 42 56 70
Chad
4
- 66 97 128 187
Adapted from (FAO, 2008)
1= year 2006; 2 = year 2012 (A. Arancibia personal communication); 3 = year 2001; 4= year 2007


EXPLOSION PROTECTION IN GRAIN HANDLING FACILITIES:
FROM COUNT MOROZZO TO COMPUTATIONAL FLUID DYNAMICS
Trygve Skjold
1,2
* & Rolf K. Eckhoff
2
1
GexCon, Fantoftvegen 38, 5892 Bergen, Norway
2
University of Bergen, Allgaten 55, 5007 Bergen, Norway
*Corresponding author. E-mail: trygve@gexcon.com
Abstract
Dust explosions represent a serious hazard in the food and feed industry. This paper reviews
some milestones in the history of explosion protection in grain handling facilities, including an
incident that took place in a bakery in Turin in 1785, early research motivated by a series of
devastating flour mill explosions during the second half of the nineteenth century, selected
research campaigns in the twentieth century, and finally prospects for safe design of grain
handling facilities based on computational fluid dynamics (CFD) in the twenty-first century.
To illustrate some of the challenges associated with realistic modelling of industrial dust
explosions, some large-scale dust explosion experiments, performed in a 250 m
3
spherical
vessel, have been simulated with the CFD code DESC.
Key words: dust explosions, explosion protection, computational fluid dynamics
1. Introduction
Dust explosions pose a hazard whenever combustible material is present in the form of fine
powder, there is a possibility of dispersing a sufficient mass of dust in air to form an explosive
cloud within a relatively confined/congested volume, and there is an ignition source present.

FIGURE 1: Explosion pentagon, after Kauffman (1981).
Dust explosions are inherently complex phenomena. The flow in a mechanical suspension is
inherently turbulent, the overall process is inherently transient, and the course of events tend
to escalate due to flame acceleration, dust lifting ahead of the flame front, and possibly
pressure-piling in complex geometries. Flame propagation in dust clouds entails premixed
combustion with non-premixed substructures (Williams, 1986), and no fundamental theory
for predicting basic combustion parameters for such systems exists today.
1. Count Morozzo di Bianz
The incident that took place in the bakery of Mr. Giacomelli in Turin on the evening of 14
December 1785 was not the first accidental dust explosion in the history of mankind, but it
may well be the first to be scientifically investigated and documented. A boy was employed in
collecting flour from an upper to a lower chamber in a warehouse, when a large amount of
flour suddenly fell; creating a dust cloud that was ignited by a lamp. The flame propagated
into neighbouring rooms, creating an overpressure that blew out the windows in the bakery.
The boy suffered modest burn injuries, and another boy broke his leg after jumping from a
scaffold. From a safety point of view, it is interesting to note that Mr. Giacomelli attributed the
accident to the extraordinary dryness of the corn that year. After investigating the incident,
Count Morozzo offered the following explanation (Morozzo di Bianz, 1795):
as the flour fell down, a great quantity of inflammable air, which had been confined in its
interstices, was set free; this, rising up, was inflamed by contact with the light; and, mixing
immediately with a sufficient quantity of atmospheric air, the explosion took place on the
side where there was the least resistance. As to the burning of the hair, and the skin, of
the boy who was in the warehouse, the cause of it must be attributed to the fine particles
of the flour, which, floating in the atmosphere, were kindled by the inflammable air, in the
same manner as powder from stamina of certain vegetables, (particularly of pine, and of
some mosses,) when thrown into the air, takes fire if any light is applied to it.
This statement identifies the oxidant, the mixing, the ignition source, the confinement, and
partly the true nature of the fuel (Fig. 1). The reference to inflammable air is consistent with
the prevailing theories in chemistry at the time: Cavendish (1766) had produced inflammable
air by combining metals with strong acids, and Lavoisier named this particular type of
inflammable air hydrogen in 1783. Hence, given the terminology of the time, it can be
argued that Count Morozzo accounted for the release of combustible volatiles from the flour:
We must recollect that flour also furnishes alkaline inflammable air, which is produced
from the glutinous vegeto-animal part of the corn: and we know that this kind of
inflammable air is of a very active nature.
The mechanisms involved in rapid flame propagation in mechanical suspensions of flour or
other finely divided combustible materials remain a subject for scientific investigations, more
than two centuries after Count Morozzo di Bianz investigated the accident in Turin.
2. Disasters and awareness
Although accidental dust explosions no doubt occurred in pre-industrial societies, the hazard
increased dramatically as a result of the technological changes in agriculture, manufacturing,
mining, transportation and storage that characterized the industrial revolution (1750-1850).
Dust explosion research in the Eighteenth century focused primarily on the role of coal dust
in colliery explosions, summarized by Rice (1911), and on flour dust explosions. Numerous
disasters in mills triggered widespread research on both sides of the Atlantic: Tradeston mill
in Glasgow, 9 J uly 1872: 18 people killed; Werdermhle in Hameln, 7 November 1887: 11
people killed; Washburn mills in Minneapolis, 2 May 1878: 18 people killed; etc. It is
interesting to note how the awareness of the dust explosion hazard evolved during this
period: Behrns (1873), Abel (1875), Anon. (1878), Peck, (1879), MacAdam (1880), Cassels
(1874), Cornell (1879), Cordes (1881), J ohnson (1881), Tobin (1882ab), etc. Cassels
summarized the explanation put forwards by Dr. Weibe from Germany:
It is based on a demonstrable fact that flour, in the form of impalpable powder,
undergoes rapid combustion when ignited, and during this combustion a highly explosive
hydrocarbon gas is generated, similar to that arising from heated petroleum or goal gas,
which, like these gases, when mixed with atmospheric air, and more or less confined, is
powerfully explosive.
This explanation was in agreement with the views held of other researchers at the time,
including Watson Smith from Manchester and Stevenson McAdam from Edinburg. Cassels
pointed out that larger flour mills was more liable to accidental explosions than small ones.
Peck (1879) demonstrated by a few simple experiments, the fact that all combustible martial
when finely divided, forming a dust or powder, will, under proper conditions, burn with
explosive rapidity. Fig. 2 illustrates one of his experiments: an assistant placed a dust
sample in the corner of a box, lighted a lamp inside the box, placed the cover, and stepped
upon it: Take notice that upon blowing through the hole, and filling the box with a cloud of
flour, the cover comes up suddenly, man and all, until hot gas gets a vent, and a stream of
fire shoots out in all directions.

FIGURE 2: Dust explosion experiment by Peck (1879) and the flour gun of Tobin (1882a).
3. Explosion protection systems
Much of the early work on explosion protection focused on preventive measures, in particular
prevention of ignition sources. One notable exception is the contribution of Hexamer
(1883ab). In his first paper, Hexamer emphasized the importance of preventing the
propagation of explosions between connected process units, preferably by passive means:
The great danger in malt mills is that, in case of an explosion, the explosive force and fire
is not confined to the mill only, but runs from the mill into the elevator, and from there is
distributed over the entire building. To check the force and fire of an explosion we must
put a barrier between the mill-box and the elevator. This can best be accomplished by
placing a receiving hopper below the rolls, kept well filled while the mill is in operation,
and from this hopper feed into the elevator cups. We thereby have a barrier of meal
between the elevator and the mill, a barrier almost as effective as one of sand.
He went on to propose ways of mitigating the explosion and extinguishing fires:
That the force of explosions may be spent without harming the building, large vent pipes
leading into the open air should be introduced into the mill-box. Iron pipes closed on
the outside by caps, similar to our common stove pipe caps which in the case of an
explosion would be readily blown off would be most effectual. [] Steam jets should be
introduced into the mill-box and elevator. One of our breweries has extinguished two
explosive malt mill fires by this means. [] I will not take up your valuable time in
describing other minor devices, which I have invented, such as an automatic contrivance
by which I close all communicating openings between the mill-house and brewery, and
turn on the steam by the pressure of the explosions.
In his second paper, Hexamer described his automatic suppression and isolation system in
more detail. The overall system included several levels of protection, dealt with the human
factor, and both carbon dioxide and water were proposed as suppression agents.

FIGURE 3: Vertical section of mill room with explosion protection system (Hexamer, 1883b).
The development of modern systems for active explosion protection was primarily motivated
by the significant losses that occurred during World War II from exploding fuel tanks in
combat airplanes resulting from incendiary strikes (Maisey, 1980). The concept was patented
and experimentally proven by two scientists, Glendinning and MacLennan, at Royal Aircraft
Establishment Farnborough. The key to the invention was detection of the explosion during
the early stages, when the rate of pressure growth is relatively slow, and to extinguish it by
using a small explosive charge to disperse extinguishing material into the space where the
flame propagates. Graviner pioneered the development of systems applicable to industrial
type of hazards (Grabowski, 1959; Maisey, 1980; Moore, 1979). Active explosion protection
is currently used for both suppression and isolation, for instance in connection with grain
elevators (Hauert & Radant, 2007).
The most frequently used method used for explosion protection is venting, i.e. the relief of
damaging overpressures by allowing combustion products and unburned dust to escape
through vent openings. The early guidelines for vent sizing were primarily of qualitative
nature. Following a series of disastrous explosions on the British Isles in 1911, one of the
recommendations given by Her Majesties Inspector of Factories was simply: The roof should
be such as to offer little resistance in the event of an explosion (Price & Brown, 1922). Since
then, various experimental campaigns have resulted in increasingly more quantitative
guidelines for design of vent openings (Skjold et al., 2008). Fig. 3 illustrates large-scale
experiments in vented enclosures: the 500 m
3
silo at Boge and 236 m
3
silo at Stordalen, both
Norway (Eckhoff, 2003), and the 250 m
3
spherical vessel used by Bartknecht (1985).

FIGURE 4: Large-scale vented dust explosion experiments.
3. Simulations
Recent development of advanced methods for predicting the consequences of industrial dust
explosions include phenomenological tools and computational fluid dynamics (CFD). With
improved modelling of the physical and chemical phenomena that govern the course of dust
explosions, the predictive capabilities of advanced models can be extended significantly
beyond the limited range of scenarios that is likely to be covered by experiments, and hence
beyond the scenarios that can be represented by simple empirical correlations.
It is not straightforward to model large-scale dust explosion experiments, not only because of
the inherent complexity of particle-laden flows and turbulent combustion, but also due to the
transient nature of the dust dispersion process typically adopted for this type of tests. To
illustrate some of these challenges, Figs. 5 and 6 summarize results where the CFD code
DESC has been used to simulate dust explosion experiments performed in a 250 m
3

spherical vessel (Bartknecht, 1985), shown on the right in Fig. 4. The numbers 100 and
125 in the legends indicate the relative values used for the laminar burning velocity, S
L
,
taken either directly from the procedure described by Skjold (2007), or the values increased
by a factor 1.25 (Skjold, 2010). The four-digit numbers in the legend indicate the ignition
delay time t
v
: 0600 =1.2 s, 0850 =0.85 s, etc. Finally, the experimental results for the
vented cases in Fig. 6 include data for two vent areas: 3.14 m
2
and 4.62 m
2
.

FIGURE 5: Simulated explosion pressure and rate of pressure rise: Constant volume.

FIGURE 6: Simulated explosion pressure and rate of pressure rise: Vented scenarios.
Figs. 5 and 6 illustrate that the results from the simulations are quite sensitive to the reactivity
of the dust-air mixtures. Although the results are in reasonable agreement for the lowest dust
concentration, the simulated maximum rates of pressure rise, (dp/dt)
max
in Fig. 5, are
significantly higher than the experimental values for higher concentrations. The over-
prediction is again reflected in the simulated reduced explosion pressures P
red
and rates of
pressure rise (dp/dt)
red
in Fig. 6. There can be several explanations for the discrepancies
between experimental and simulated results. For the experiments in the 250 m
3
vessel,
Bartknecht (1985) used maize starch with median particle size 10 m. For the combustion
model in DESC, Skjold et al. (2006) used data for dried maize starch with median particle
size 13 m, obtained from experiments in a 20-litre vessel. Particle size distribution, moisture
content and dispersion system influence the experimental results significantly, and the same
nominal dust concentrations may not correspond to the same actual dust concentrations in
the two vessels. The turbulent flow conditions will not be the same in the two vessels, and
the effect of radiation on flame propagation in dust clouds is likely to depend on scale. There
are also inherent limitations in the models implemented in DESC.
4. Conclusions
Although technology for protecting personnel and property from dust explosions have
developed significantly since Count Morozzo di Bianz investigated the flour dust explosion
that occurred in a bakery in Turin in 1785, severe losses still occur during handling of grain,
food and feed products. Current standards and guidelines for explosion protection are
reasonably well developed, but there is still a need for improved methodologies, particularly
for design of relatively complex geometries: grain elevators, pneumatic conveying systems,
large silo complexes, etc., and it is not obvious that this challenge can be met simply by
introducing additional or improved empirical correlations in codes or standards.
Ignorance [of phenomena] and a culpable negligence of those precautions which ought
to be taken, have often caused more misfortune and loss than the most contriving malice:
it is therefore of great importance that these facts should be universally known, that public
utility may reap from them every possible advantage.
Count Carlo Ludovico Morozzo di Bianz (1743-1804)
References
Abel, F.A. (1875). Accidental explosions. Nature, 11: 436-439, 477-478, 498-499.
Anon. (1878). Flour mill explosions. American miller, VI: 51, 77, 98, 122, 159, 184, 206, 236.
Bartknecht, W. (1985). Druckentlastung von Staubexplosionen in Grossbehltern. CIBA-
GEIGY, Project number 01 VD 011 1, Basel.
Behrns, G. (1873). Mill explosions and fires their cause and prevention. American miller,
VI: 208, 237, 261.
Cassels, J .L. (1874). Explosions in flour mills. American miller, II: 118-119.
Cavendish, H. (1766). Three papers containing experiments on factitious air. Philosophical
Transactions, 56: 141184.
Cordes, W. (1881). Explosion of flour dust. American Miller, IX: 181-182.
Cornell, W.B. (1879). Hazards of flour mills. American miller, VII: 361-363, 399-401.
Eckhoff, R.K. (2003). Dust explosions in the process industries (3
rd
ed.). Gulf Professional
Publishing, Amsterdam.
Grabowski, G.J . (1959). Industrial explosion protection. J ournal of the American Oil
Chemists' Society, 36: 57-59.
Hauert, F. & Radant, S. (2007). Brand- und Explosionsschutz in Mhle, Mischfutterwerk und
Landhandel. AgriMedia, Clenze.
Hexamer, C.J . (1882). Dust explosions in breweries. J . Franklin Institute, 115: 121-126.
Hexamer, C.J . (1883). Prevention of dust explosions and fires in malt mills. J . Franklin
Institute, 115: 200-206.
J ohnson, E.C. (1881). The fire hazard of flour mills. American Miller, IX: 418, 459, 506.
Kauffman, C.W. (1981). Agricultural dust explosions in grain handling facilities. Proceedings
International Conference on Fuel-Air Explosions, McGill University, Montreal: 305-347.
MacAdam, S. (1880). On flour mill explosions. American Miller, VIII: 218-219.
Maisey, H.R. (1980). Explosion suppression & automatic explosion control. IChemE
Symposium Series, 58: 171-191.
Moore, P. (1979). Explosion suppression in industry. Physics in Technology, 10: 202-207.
Morozzo di Bianz, C.L. (1795). Account of a violent explosion which happened in a flour-
warehouse, at Turin, December the 14
th
, 1785, to which are added some observations on
spontaneous inflammations. The Repertory of Arts and Manufactures, 2: 416432.
Peck, L.W. (1879). Explosions from combustible dust. American Miller, VII: 16-17.
Price, D.J . & Brown, H.H. (1922). Dust explosions theory, and nature of, phenomena,
causes and methods of prevention. NFPA, Boston.
Rice, G.S. (1911). The explosibility of coal dust. Dept. Interior, Bureau of Mines, Bulletin 20.
Skjold, T. (2007). Review of the DESC project. J . Loss Prev. Process Ind., 20: 291302.
Skjold, T. (2010). Flame propagation in dust clouds: challenges for model validation. Eighth
International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions
(ISHPMIE), 5-10 September 2010, Yokohama, J apan: 11 pp.
Skjold, T., Arntzen, B. J ., Hansen, O. J . Storvik, I. E. & Eckhoff, R. K. (2006). Simulation of
dust explosions in complex geometries with experimental input from standardized tests.
J ournal of Loss Prevention in the Process Industries, 19: 210217.
Skjold, T., van Wingerden, K., Hansen, O.R. & Eckhoff, R.K. (2008). Modelling of vented dust
explosions - empirical foundation and prospects for future validation of CFD codes.
IChemE Symposium Series 154: 838-850.
Tobin, T.W. (1882a). Flour and flour dust explosions. American Miller, X: 259-260, 262-265.
Tobin, T.W. (1882b). Explosive and dangerous dust. J . Franklin Institute, 114: 412-425.
Williams, F.A. (1986). Lectures on applied mathematics in combustion: Past contributions
and future problems in laminar and turbulent combustion. Physica D, 20D, 21-34.

JTI Institutet fr jordbruks- och miljteknik
Safe storage time of cereal grain summarized in a mathematical
model

Nils Jonsson
1
*, Pernilla Johnsson
2
, Aldo Ritzzo
3
Monica Olsen
2
and Leif Gustafsson
4


1
JTI Swedish Institute of Agricultural and Environmental Engineering, P.O. Box 7033, SE-
750 07 Uppsala, Sweden,
2
National Food Administration, P.O. Box 622, SE-751 26 Uppsala, Sweden,
3
Department of Chemistry, P.O. Box 368, National Veterinary and Food Research Institute,
FIN-00231 Helsinki, Finland
4
Department of Technology, ngstrmslaboratoriet, Uppsala Universit, Box 534, SE-751 21
Uppsala, Sweden
*Corresponding author.E-mail: nils.jonsson@jti.slu.se

Abstract
In temperate climate regions, grain preservation and storage are primarily questions of
avoiding damage by fungi. Ochratoxin A (OTA) is probably one of the most common
mycotoxin produced in stored grain in these regions. In a study repeated during three harvest
seasons the influence of moisture content (m.c.) (16-25 %), temperature (10-25 C) and time
on growth of moulds and production OTA in freshly harvested winter wheat has been
evaluated. Preliminary results showed that the storage time without microbial growth (lag
phase) was halved when the grain was stored with 1-3% higher m.c. ( 0.05 a
w
) or 5C higher
temperature. Respiration rate during lag phase was doubled when the grain was stored at
5C higher temperature or at 0,3-0,9% higher m.c. Furthermore, OTA and a decrease in
viability could generally be detected quite soon after microbial growth had started. The lowest
m.c. permitting production of OTA was 17-18%. A mathematical model of safe storage time
before mould growth and ochratoxin A production has been developed.
Keywords: cereal grain, mathematical model, ochratoxin A, safe storage time.
1. Introduction
The annual world production of cereal grains, which are mans principal source of food,
exceeds 1800 million tonnes. Probably between 10 and 30% of harvested grain is lost due to
mould-induced deterioration (Chelkowski, 1991). In temperate climates, ochratoxin A (OTA)
is one of the most common mycotoxins produced in grain after harvest (Frisvad, 1995).
Occurrence of ochratoxin A in human blood has been investigated in different parts of
Europe and Canada (Frolich et al., 1991; Fukal et al., 1990; Hald et al., 1991; Bauer et al.,
1987; Breitholtz-Emanuelsson et al., 1993 and 1994; Breitholtz et al., 1991; Golinski et al.,
1991; BAG, 1994). In studies with low detection limit, it was possible to find ochratoxin A in
all blood samples.
OTA is a mycotoxin of considerable concern for human health and is classified as a possible
human carcinogen (IARC, 1993). Cereals are considered to be the main contributor (50 to
80%) to the OTA intake among European consumers. The European Commission has stated
a maximum legislative limits for OTA in unprocessed cereals to 5 g kg-1(Commission
Regulation 472/2002 amending Regulation 466/2001). Similar maximum levels for cereals
are also being discussed within the Codex Alimentarius.

JTI Institutet fr jordbruks- och miljteknik
2
The most important factors limiting the colonisation of fungi are moisture content and
temperature of the grain. Penicillium spp. is probably the most important mycotoxin
producing storage fungi in the temperate climate zone. In Sweden, naturally occurring OTA
in cereals has been linked to Penicillium verrucosum (Holmberg et al., 1991; Jonsson &
Pettersson, 1999) but the mycotoxin is also produced by fungi belonging to Aspergillus
ochraceus group. For Penicillium verrucosum, the minimum moisture content for growth is
about 16-17% (Northolt et al., 1979). For production of ochratoxin A, the moisture content
probably needs to be about 1% higher. A Swedish study of the grain quality on farms
indicates that the occurrence of OTA in grain is attributed to insufficient drying or to
excessively long pre-drying storage (Jonsson & Pettersson, 1999).
To be able to develop safe and cost-efficient methods for handling and preservation of grain
studies have been conducted designed to determine rates of mould growth and OTA
production under various conditions.
The results from a pre-study will be used
in recommendations for storage of wet cereal grain before drying
when evaluating different analytical methods of mould growth concerning modelling
The results from the main-study will also be used
for developing models of safe storage time before mould growth and
ochratoxin A production
when evaluating maximum limit values for growth of moulds in grain
This paper is only dealing with results from the pre-study, while results from the main-study
will be presented at the conference.
2. Material and methods
Freshly harvested winter wheat was stored at about 16, 17, 18, 20 and 24% moisture content
(m.c.) (corresponding to aw 0.75, 0.80, 0.85, 0.90 and 0.95) and four different
temperatures (10, 15, 20 and 25C) in an experimental plant, constructed by JTI, in which the
water activity and temperature of the grain were controlled by saturated salt solutions (NaCl,
(NH4)SO4, KCl, BaCl2 and KNO3) and thermostatically regulated bath. An infrared monitor
connected to the plant measured the respiration of the grain (production of CO2). Prior to the
experiment the grain was inoculated with OTA-producing strains of P. verrucosum.
The microbial activity in the grain was followed in all experiments by measuring the
respiration rate (CO
2
), determination of colony forming units (CFU) and ochratoxin A. In
some experiments also internal infection of fungi, colony forming units of P. verrucosum,
content of ergosterol and the odours were followed. The viability, m.c. and water activity of
the grain were also determined.
3. Results and Conclusions
The results suggested that respiration rate was a sensitive measure of early fungal growth
and permitted only small internal infection by storage moulds before an increase. Generally it
showed a sigmoidal development with an initial lag phase, an exponential phase when
storage moulds started to grow, and a decline phase probably due to lack of nutrients.
Production of OTA and decrease in grain viability could, generally, be detected quite soon
after microbial growth had begun.
The mould growth and production of OTA started earlier and with a higher rate when the
grain had a higher m.c. and temperature.

JTI Institutet fr jordbruks- och miljteknik
3
Preliminary results showed that the storage time without microbial growth (lag phase) was
halved when the grain was stored with 1-3% higher m.c. ( 0.05 aw) or 5C higher
temperature. Respiration rate during lag phase was doubled when the grain was stored at
5C higher temperature or at 0,3-0,9% higher m.c. Furthermore, OTA and a decrease in
viability could generally be detected quite soon after microbial growth had started.The lowest
moisture content permitting production of OTA was 17-18%.
A mathematical model of maximum safe storage time before mould growth and ochratoxin A
production (f(aw,T)) has been developed based on data of the lag time according to data
about respiration rate from the main-study.
Acknowledgements
This work was financially supported by the European Commission, Quality of Life and
Management of Living Resources Programme (contract no. QLK1-CT-1999-00433), the
Cerealia Foundation R&D and the Swedish Farmers Foundation for Agricultural Research..
References
BAG. 1994. Ttigkeitsbericht der Abteilung Vollzug Lebensmittelrecht unde Lebens-
mittelwissenschaft des Bundesamtes fr Gesundheitswesen 1993, Abteilung
Lebensmittelwissenschaft, Sektion Lebensmittelchemie, c) Ochratoxin in Humanblutserum.
Mitteilungen aus dem Gebieter der Lebensmitteluntersuchung und Hygiene 85, 428-429.
Bauer, J. & Gareis M., 1978. Ochratoxin A in der Nahrungsmittelkette. Journal of Veterinary
Medicine 34, 613-627.
Breitholtz, A., Olsen M., Dahlbck . & Hult K. 1991. Plasma ochratoxin A levels in three
Swedish populations surveyed using an ion-pair HPLC technique. Food Addit. Contam. 8,
183-192.
Breitholz-Emanuelsson, A., Minervini F., Hult K. & Visconti A., 1994. Ochratoxin A in human
serum samples collected in southern Italy from healthy individuals suffering from different
kidney disorders. Natural Toxins 2, 366-370.
Breitholz-Emanuelsson, A., Olsen M., Oskarsson A., Palminger I. & Hult K., 1993. Ochratoxin
A in cows milk and in human milk with corresponding human blood samples. Journal AOAC
International 76, 842-846.
Chelkowski, J. 1991. Preface and Introduction. In: Cereal grains, mycotoxins, fungi and
quality in drying and storage ed. J. Chelkowski. Elsevier. s. I-XXii.
Frisvad, J.C. & Thrane U., 1995. Mycotoxins and mycotoxigenic fungi in storage, pp.
251-288. In: Stored Grain Ecosystems, eds. D.S. Jayas, N.D.G. White & W.E. Muir, Marcel
Dekker (New York).
Frohlich, A.A., Marquardt R.R. & Ominski K.H., 1991. Ochratoxin A as a contaminant in the
human food chain: a Canadian perspective. In: Mycotoxins, endemic nephropathy and
urinary tract tumours. IARC Scientific Publications. 115, Edited by M. Castegnaro, R.
Plestina, G. Dirheimer, I.N. Chernozemsky & H. Bartsch, pp. 139-143. International Agency
for Research on Cancer, Lyon.
Fukal, L. & Reisnerova H., 1990. Monitoring of aflatoxins and ochratoxin A in Czechoslovak
human sera by immunoassay. Bulletin of Environmental Contamination and Toxicology 44,
345-349.
Golinski, P., Grabarkiewicz-Szczesna J., Chelkowski J., Hult K. & Kostecki M., 1991.
Possible sources of ochratoxin A in human blood in Poland. In: Mycotoxins, endemic
nephropathy and urinary tract tumours. IAARC Scientific Publications. 115, Edited by

JTI Institutet fr jordbruks- och miljteknik
4
M. Castergnaro, R. Plestina, G. Dirheimer, I.N. Chernozemsky and H. Bartsch, pp. 153-158.
International Agency for Research on Cancer, Lyon.
Hald, B., 1991. Ochratoxin A in human blood in European countries. In: Mycotoxins, endemic
nephropathy and urinary tract tumours. IAARC Scientific Publications. 115, Edited by
M. Castergnaro, R. Plestina, G. Dirheimer, I.N. Chernozemsky and H. Bartsch, pp. 159-164.
International Agency for Research on Cancer, Lyon.
Holmberg, T., Breitholtz-Emanuelsson A., Hggblom P., Schwan O. & Hult K., 1991.
Penicillium verrocosum in feed of ochratoxin A positive swine herds. Mycopathologia 116,
169-176.
IARC. 1993. Ochratoxin A. In: Some naturally occurring substances: Food items and
constituents, heterocyclic aromatic amines and mycotoxins. IARC Monographs on the
evaluation of carcinogenic risks to humans. 56, Edited by IARC and WHO, pp. 489-521.
IARC, WHO, Lyon.
Jonsson, N. & Pettersson H., 1999. Utvrdering av olika konserveringsmetoder fr spannml
baserad p analyser av hygienisk kvalitet. JTI-rapport Lantbruk & Industri nr 263, JTI
Institutet fr jordbruks- och miljteknik, Uppsala, Sweden.
Lacey, J. & Magan N., 1991. Fungi in cereal grains: Their occurrence and water and
temperature relationships. In: Cereal grains, mycotoxins, fungi and quality in drying and
storage ed. J. Chelkowski. Elsevier. s. 77-118.
Northolt, M.D., Van Egmond H.P. & Paulsch W.E., 1979. Ochratoxin A production by some
fungal species in relation to water activity and temperature. Journal of Food Protection 42,
485-490.

Vibration Induced Granular Flow Through An Elbow
Ivn Snchez
1
*, J.R. Darias
2
, O. P. Michelangelli
3
1
Centro de Fsica, Instituto Venezolano de Investigaciones Cientficas, Apartado Postal
21827, Caracas 1020, Venezuela
2
Departamento de Fsica, Universidad Simn Bolvar, Apartado Postal 89000, Caracas 1080,
Venezuela
2
Institute for Polymers and Composites/I3N, University of Minho, 4800-058 Guimaraes,
Portugal
*Corresponding author. E-mail: ijsanche@ivic.gob.ve
Abstrat
Vertical vibration is often used to excite granular materials in order to enhance their ability to
flow. The way the granular material is affected by vibration is a complex problem. To gain
insight on the effect of vibration on the flowability of the granular material in realistic
geometries, we study the motion of a vertically vibrated granular assembly inside simple
containers with corners. We focused on studying two basic geometries in 2 and 3
dimensions. L-shaped containers and U-shaped containers. We study the problem both
experimentally and computationally. We have experiments using glass spheres of different
sizes in three dimensional U-shaped containers, vertically vibrated using electromagnetic
shakers. Also we have discrete element simulations of two dimensional U-shaped and three
dimensional L-shaped containers in the absence of an interstitial fluid. The transport of grains
through L-shaped and U-shaped tubes is sensible to the vibration amplitude and frequency,
and to the dissipative characteristics of the granular material. A minimum acceleration is
needed to induce transport. In U-shaped containers, the effect of interstitial air is crucial to
induce a symmetry breaking, where the granular material accumulates in one side of the
tube. However, in the absence of an interstitial fluid, there is still some accumulation, that is
sensible to the grain/tube diameter ratio. Studying U-shaped and L-shaped containers, if one
looks into a single cycle of oscillation, it is possible to identify the portion of the cycle (phase
interval) where the transport through an elbow takes place. We discuss on the importance of
this effective transport phase interval, on the overall transport observed after many cycles of
oscillation. .
Key words: Granular, vibration, elbow.

1. Introduction
Bulk solids management is one of the main tasks of modern agricultural, food,
pharmaceutical, and construction industries. Several granulate handling processes benefit
from the induced flowability that can be achieved with the help of vibrations (Gutman 1998,
Masuda et al. 2006). Common examples are vibratory conveyors and vibrating chutes.
Despite the variety of container geometries used in industrial applications, basic models of a
vibrating bed deal with rectangular containers (Kroll 1954, Gutman 1976, Akiyama & Naito
1987). In basic studies, the simple addition of a partition in a rectangular container can
induce a complex behavior: as a consequence of the dissipative nature of the granular
materials, accumulations (Ohtsuki et al. 1998, Akiyama et al. 2001, Chen) and ratchet effects
(a biased response out of a random excitation) can take place (van der Meer et al. 2004).
In an effort to explore the response of a granular material to vibration in more realistic
geometries we study the flow of grains in containers with corners. More specifically in U-
shaped and L-shaped containers, using experiments and molecular dynamics computer
simulations. We discuss several behaviors observed depending on the excitation frequency
and amplitude. We also relate the overall transport (taking place after several cycles of
oscillation) to the response of the material in a single cycle of oscillation.

2. Materials and Methods
2.1. Experimental setup:
To generate vertical vibrations we use a standard setup described earlier (Darias et al. 2010,
Prez & Snchez 2011). We have performed experiments in 3-dimensional U-shaped
containers like the one depicted in Fig. 1., made of acrylic. We also have experiments on
quasi-2-dimensional L-shaped containers like the one depicted in Fig. 1, able to hold a single
layer of grains. The L-shaped container has an exit at the bottom right to allow the grains to
escape. We used both continuous sinusoidal vibration and single taps. For continuous
vibration the container's vertical displacement is given by ) sin( t A y = with A the amplitude,
the angular frequency and t the time. The intensity of oscillation is quantified through the
maximum dimensionless acceleration g A /
2
= , with g the acceleration of gravity. Single
taps are produced inducing a single sine cycle to the electromagnetic vibrator. The resulting
displacement is a pulse whose intensity is quantified through its maximum acceleration,
measured with a piezoelectric accelerometer. Video tracking, low resolution high speed
photography and high resolution low speed photography are used to track the motion of the
granular material.
2.1. Computational model:
Our computational model is a discrete element simulation, consisting of particles whose
resistance to overlap is expressed in terms of a continuous potential function (GDR-MIDI
2004, Michelangelli et al. 2011). Given a pair of spherical particles i and j, with radius R
i
and
R
j
, that are in contact, the value of the overlap can be defined as = R
i
+ R
j
- |

|, where

= |

| is the vector connecting the centres of the i

th
and j
th
particles and is the unit
vector in that direction. The model calculates the normal (

) and tangential force (

)
between i and j, assuming the existence a spring constant:

(1)
Where

= (

(2)

= (

(3)
The normal force

= (

has two terms, a contact force given by a spring
model using the k
n
constant and a damping force defined by the normal coefficient and the
relative velocity between the particles. The tangential force

= (

also has 2
terms: a shear force given by the tangential overlap and a damping term defined by the
tangential relative velocity, in order to obey the sliding friction the tangential overlap is
truncated to fulfil the condition

. The values of the spring constant and the damping

coefficient are chosen according to the granular material. Similar considerations apply to the
interactions between particles and walls. The equations of motion are integrated numerically
using the Verlet method and the numerical method was implemented using the open source
solution (Klos & Goniva 2010).
A series of flow measurements have been carried out for a L-shape elbow, the dimension of
the horizontal square cross section is given 0.003x0.025 mm
2
. The vertical arms are 0.385 m
high and the horizontal is 0.1 m wide. The particles have a diameter of 0.00298 m, thus, the
system was conceived in order to study single-layer (quasi-2-dimensional) flow through the
elbow. 820 grains fill the vertical portion of the container. The elbow is initially filled setting
random positions along the vertical arm and waiting for the equilibrium state.

3. Results from U-shaped containers: Differe
3.1 Air mediated regime:
3.1.1 The importance of a pressure gradient
height difference h between the arms of the U, shows a tendency to increase when the
frequency of vibration is small
m) are used as granular material (Darias et al. 2011)
figure 1, where an initial height difference quickly grows due to agitation. This phenomenon
has been called the U-tube instability. The interstitial fluid plays a crucial role in the
development of this instability. S
at ambient pressures, disappears
diameter 2x10
-4
m. Clement et al. 2009 using water as the interstitial fluid, observe the
instability with grains 3-4 times larger. An important detail is that
interstitial fluid is not enough for the instability to occur. It is the development of a vertical
pressure gradient, the key factor for the instability.
while maintaining the interstitial ai
the recipient from bellow during vibration (but with a fine mesh preventing grains from falling).
The evolution of h with time is shown in figure 1 for a regular solid bottom container and a
porous bottomed container. The height difference levels in the porous bottomed container,
exactly the same result as in the absence of air, or if large grains are used (diameter >
whose resistance to the airflow induced is negligible.
3.1.2 Time resolved mobility:
of the U determines if the overall transport is negligible (at very low vibration intensities),
it corresponds to an increasing
a fixed frequency varying the amplitude of oscillation. Using low resolution high speed video,
we recorded the motion of a small section of the horizontal portion of the U
cases (see figure 2). Tracking a fixed point in the container, it is possible to crop each image
so as to "eliminate" the vibration. C
between consecutive images, it is possible to obtain a measure of the granular m
mobility. The plot at the right side of figure 2 shows this mobility as a function of time in a
single oscillation cycle. In the case of no fl
FIGURE 1: Snapshot sequence of two U
180-210 m. The granular material appears black. Vibration is started at time 0 s, at 10 Hz
and =2,5. Upper row: solid bottomed container. Bottom row: porous bottomed container.
shaped containers: Different regimes
The importance of a pressure gradient: In vertically vibrated U-shaped contai
h between the arms of the U, shows a tendency to increase when the
frequency of vibration is small (frequency 10 Hz), and if fine grains (grain diameter
m) are used as granular material (Darias et al. 2011). See for example the upper side of
figure 1, where an initial height difference quickly grows due to agitation. This phenomenon
tube instability. The interstitial fluid plays a crucial role in the
development of this instability. Snchez et al. showed that the exponential growth of
disappears at pressures of 13.33 Pa, using glass spheres of grain
Clement et al. 2009 using water as the interstitial fluid, observe the
4 times larger. An important detail is that the sole presence of an
interstitial fluid is not enough for the instability to occur. It is the development of a vertical
pressure gradient, the key factor for the instability. One way to eliminate the pressure drop,
while maintaining the interstitial air, is using a porous bottomed recipient, allowing air to enter
the recipient from bellow during vibration (but with a fine mesh preventing grains from falling).
h with time is shown in figure 1 for a regular solid bottom container and a
The height difference levels in the porous bottomed container,
result as in the absence of air, or if large grains are used (diameter >
whose resistance to the airflow induced is negligible.
the direction of the granular flow through the horizontal portion
of the U determines if the overall transport is negligible (at very low vibration intensities),
corresponds to an increasing h or to a decreasing h. The three cases can be studied at
a fixed frequency varying the amplitude of oscillation. Using low resolution high speed video,
we recorded the motion of a small section of the horizontal portion of the U
e figure 2). Tracking a fixed point in the container, it is possible to crop each image
to "eliminate" the vibration. Computing the mean pixel-to-pixel cross correlation
between consecutive images, it is possible to obtain a measure of the granular m
mobility. The plot at the right side of figure 2 shows this mobility as a function of time in a
single oscillation cycle. In the case of no flow, the mobility is almost flat.
Snapshot sequence of two U-shaped tubes filled with glass spheres of diameter
m. The granular material appears black. Vibration is started at time 0 s, at 10 Hz
=2,5. Upper row: solid bottomed container. Bottom row: porous bottomed container.
shaped containers, the
h between the arms of the U, shows a tendency to increase when the
z), and if fine grains (grain diameter 2x10
-4

ample the upper side of
figure 1, where an initial height difference quickly grows due to agitation. This phenomenon
tube instability. The interstitial fluid plays a crucial role in the
showed that the exponential growth of h seen
at pressures of 13.33 Pa, using glass spheres of grain
Clement et al. 2009 using water as the interstitial fluid, observe the
the sole presence of an
interstitial fluid is not enough for the instability to occur. It is the development of a vertical
One way to eliminate the pressure drop,
r, is using a porous bottomed recipient, allowing air to enter
the recipient from bellow during vibration (but with a fine mesh preventing grains from falling).
h with time is shown in figure 1 for a regular solid bottom container and a
The height difference levels in the porous bottomed container,
result as in the absence of air, or if large grains are used (diameter >3 mm)
the direction of the granular flow through the horizontal portion
of the U determines if the overall transport is negligible (at very low vibration intensities), or if
The three cases can be studied at
a fixed frequency varying the amplitude of oscillation. Using low resolution high speed video,
we recorded the motion of a small section of the horizontal portion of the U-tube for three
e figure 2). Tracking a fixed point in the container, it is possible to crop each image
pixel cross correlation
between consecutive images, it is possible to obtain a measure of the granular material
mobility. The plot at the right side of figure 2 shows this mobility as a function of time in a
ow, the mobility is almost flat. The increasing h

shaped tubes filled with glass spheres of diameter
m. The granular material appears black. Vibration is started at time 0 s, at 10 Hz
=2,5. Upper row: solid bottomed container. Bottom row: porous bottomed container.
case, shows a small peak at 0,2T followed by an almost flat behavior
case shows two peaks in the mobility, as well as a greater average value. This fine temporal
differences may be associated with the G dependent fluidization time dis
et al 2009, as a limitation to their model for the U

3.2 Friction-dissipation mediated regime:
Some degree of accumulation can also be achieved in the absence of an interstitial fluid. Our
simulations in 2D U-shaped containers show that this accumulation is sensitive to the grain
diameter over tube diameter ratio
than the air mediated one, since the latter shows a tendency to move all the granular
material in the tube to one side, until the other side is emptied. In our 2
simulations, the accumulation takes place ge
between both sides of the recipient. We have evidence of the sensitivity of this height
difference on the density of the grains and on the dissipation characteristics of the gr
interaction, however they will be given elsewhere.

FIGURE 2: Left: Snapshot sequence of a
shaped tube at 12,5 Hz, obtained with low resolution high speed video
cycle is shown for three cases: A) no flow
h (=6.3). Right: Granular mobility estimated from cross correlation of consecutive images,
plotted as a function of time in a single oscillation cycle (plots are obtained after averaging
FIGURE 3: Left: Snapshot sequence of
after a number of cycles. The height difference reaches a
Dependence of y
max
with the diameter of the tube (
case, shows a small peak at 0,2T followed by an almost flat behavior. The decreasing
shows two peaks in the mobility, as well as a greater average value. This fine temporal
differences may be associated with the G dependent fluidization time discussed by Snchez
et al 2009, as a limitation to their model for the U-tube instability.
dissipation mediated regime:
Some degree of accumulation can also be achieved in the absence of an interstitial fluid. Our
shaped containers show that this accumulation is sensitive to the grain
ratio (see figure 3). The nature of this accumulation is different
than the air mediated one, since the latter shows a tendency to move all the granular
material in the tube to one side, until the other side is emptied. In our 2
he accumulation takes place generating a height difference of merely ten grains
between both sides of the recipient. We have evidence of the sensitivity of this height
difference on the density of the grains and on the dissipation characteristics of the gr
interaction, however they will be given elsewhere.
Snapshot sequence of a portion of the horizontal section of a vibrated U
, obtained with low resolution high speed video. A
for three cases: A) no flow (=1.5), B) increasing h (=3.5)
Granular mobility estimated from cross correlation of consecutive images,
ted as a function of time in a single oscillation cycle (plots are obtained after averaging
over 15 cycles).
Snapshot sequence of a 2D U-tube simulation under continuous vibration,
. The height difference reaches a saturation value
with the diameter of the tube (in units of a single grain
The decreasing h
shows two peaks in the mobility, as well as a greater average value. This fine temporal
cussed by Snchez
Some degree of accumulation can also be achieved in the absence of an interstitial fluid. Our
shaped containers show that this accumulation is sensitive to the grain
The nature of this accumulation is different
than the air mediated one, since the latter shows a tendency to move all the granular
material in the tube to one side, until the other side is emptied. In our 2-dimensional
nerating a height difference of merely ten grains
between both sides of the recipient. We have evidence of the sensitivity of this height
difference on the density of the grains and on the dissipation characteristics of the grain-wall

portion of the horizontal section of a vibrated U-
A single oscillation
=3.5), C) decreasing
Granular mobility estimated from cross correlation of consecutive images,
ted as a function of time in a single oscillation cycle (plots are obtained after averaging

simulation under continuous vibration,
value y
max
. Right:
of a single grain diameter).
4. Results from L-shaped containers:
An important question that arises from the fact the motion in a singe cycle determines the
overall transport in U-tubes is how a single perturbation affects the motion of grains through
a corner of the tube. A simple configuration where we can track the position of each grain, is
a 2-dimensional experiment with a single corner, like the L-shaped tube shown in figure 4. As
a first measure of the effect of a single perturbation we track the height of the vertical portion
as a function of the number of single perturbations (taps), and study the dependence with the
tap intensity. Figure 5 shows that greater tap intensities, empty the column in a shorter
number of taps. All curves show a deflection when the column height reaches the top of the
horizontal channel (indicated by a dashed grey line). After this point, we no longer have an
elbow but rather an horizontal channel filled with grains. Before this point, we noticed that,
except for the minimum tap intensity, the variation of column height with each tap is almost
constant, indicating that the granular flow is practically independent of the amount of grains
in the vertical portion. The almost linear behaviour of the column height with the number of
taps is lost for the lowest tap intensity shown, where a dependence of the column height
variation per tap with respect to he column height is suggested: a larger flow is achieved
when the column in larger.
Further decreasing the tap intensity leads to a intermittent flow regime, where there can be
many taps with no height change, followed by some taps with a small variation. For tap
intensities of =1,2, the number of taps needed to empty the vertical portion of the L-shaped
container can reach 1000 taps. When the tap intensity is below =1 there is no flow at all.
5. Closing remarks:
We studied the transport of grains in recipients with corners. In U-shaped containers we
studied the development of an accumulation of grains towards one of the arms of the

FIGURE 4: Snapshot sequence of a quasi-2D L-shaped tube filled with Delrin spheres of
diameter 3 mm. Each picture is taken after the indicated number of taps, with = 2.3.

FIGURE 3: Evolution of the vertical column height of a quasi-2D L-shaped container with
Delrin spheres of diameter 3 mm. Each curve corresponds to a different tap intensity.
recipient. We show that there are two regimes. One regime is fluid mediated, and the key
ingredient for the development of an accumulation is not merely the presence of an interstitial
fluid, but the development of a pressure gradient in the vibration direction. This pressure
difference can be quenched using a porous bottomed container, and this eliminates the
accumulation effect. Another regime is found in the absence of an interstitial fluid. Our
computer simulations show that there is an optimal grain diameter to tube diameter ratio that
maximizes the accumulation observed after hundreds of oscillation cycles. In L-shaped
containers we studied the effect of varying the intensity of individual taps on the vibration
induced flow. The flow around the corner is affected by the intensity, showing greater
transport at higher intensities. A minimum intensity of oscillation is needed to induce flow.
Intensities slightly larger than this minimum will induce an intermittent flow. Larger intensities
will induce a flow per tap that is practically independent of the amount of grains in the vertical
portion of the L-shaped container.

Reference list
Akiyama, T. & Naito, T. (1987). Vibrated bed of powders: a new mathematical formulation.
Chemical Engineering Science, 42, 1305-1311.
Akiyama, T., Shinmura, K., Murakawa, S. & Aoki, K.M. (2001). A surface instability of
granules under vibration in partitioned containers. Granular Matter 3, 177183.
Clement, C.P., Pacheco-Martinez, H.A, Swift, M.R. & King, P.J. (2009). Partition instability in
water-inmersed granular systems. Physical Review E, 80, 011311-7.
Darias, J.R., Snchez, I. & Gutirrez, G. (2011). Experimental study on the vertical motion of
grains in a vibrated U-tube. Granular Matter 13, 13-17.
MiDi, GDR. (2004). On dense granular flow. European Physical Journal E, 14, 341-360.
Gutman, I. (1968). Industrial uses of mechanical vibrations. London: Business Books Ltd.
Gutman, R.G. (1976). Vibrated bed of particles: A simple model for the vibrated bed.
Transactions of the Institution of Chemical Engineers, 54, 174-183.
King, P., Lopez-Alcaraz, P., Pacheco-Martinez, H., Clement, C., Smith, A. & Swift, M. (2007).
Instabilities in vertically vibrated uid-grain systems. European PhysicJournal E 22, 219226.
Kloss, C., Goniva, C. (2010). Liggghts a new open source discrete Element simulation
software. In Munjiza, A. (Ed.) Proc. of The 5th International Conference on Discrete Element
Methods. (pp. 25-26). London: Queen Mary, University of London.
Masuda, H., Higashitani, K., & Yoshida, H. (2006). Powder Technology Handbook. Boca
Ratn: CRC Press Taylor & Francis.
Michelangelli, OP, Yamanoi, M, Gaspar-Cunha A. & Covas, JA. (2011). Modelling pellet flow
in single extrusion with DEM. Journal of Process Mechanical Engineering, 255-268
Ohtsuki, T., Kinoshita, D., Nakada, Y. & Hayashi, A. (1998). Surface level migration in
vibrating beds of cohesionless granular materials. Physical Review E, 58, 76507656.
Prez, B. & Snchez, I. (2011). Effect of friction on the granular U-tube instability. Mechanics
Research Communications 38, 244-248.
Snchez I, Darias, J.R., Paredes, R., Lobb, C.J., Gutirrez, G. (2009) Vertical Granular
Transport in a Vibrated U-Tube. In Appert-Rolland, C., Chevoir, F., Gondret, P., Lassarre, S.
(Eds.) Traffic & Granular Flow'07 Proceedings. (pp. 545). Berlin: Springer-Verlag.
van der Meer, D., Reimann, P., van der Weele, K. & Lohse, D. (2004). Spontaneous Ratchet
Effect in a Granular Gas. Physical Review Letters, 92, 184301-184304.