See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/263092752

Full-scale tests to measure stresses and vertical

displacements in an 18.34 m-diameter
agricultural steel silo roof
ARTICLE in COMPUTERS AND ELECTRONICS IN AGRICULTURE AUGUST 2014
Impact Factor: 1.76 DOI: 10.1016/j.compag.2014.05.008

READS

121

6 AUTHORS, INCLUDING:
lvaro Ramrez-Gmez

Eutiquio Gallego

Universidad Politcnica de Madrid

Universidad Politcnica de Madrid

53 PUBLICATIONS 279 CITATIONS

45 PUBLICATIONS 175 CITATIONS

SEE PROFILE

SEE PROFILE

Jos Mara Fuentes

Francisco Ayuga

Universidad Politcnica de Madrid

Universidad Politcnica de Madrid

35 PUBLICATIONS 115 CITATIONS

116 PUBLICATIONS 775 CITATIONS

SEE PROFILE

All in-text references underlined in blue are linked to publications on ResearchGate,

letting you access and read them immediately.

SEE PROFILE

Available from: lvaro Ramrez-Gmez

Retrieved on: 17 February 2016

Computers and Electronics in Agriculture 106 (2014) 5665

Contents lists available at ScienceDirect

Computers and Electronics in Agriculture

journal homepage: www.elsevier.com/locate/compag

Full-scale tests to measure stresses and vertical displacements

in an 18.34 m-diameter agricultural steel silo roof
. Ramrez-Gmez a,, E. Gallego a, J.M. Fuentes a, C. Gonzlez-Montellano a, C.J. Porras-Prieto b, F. Ayuga a
a
b

BIPREE Research Group, Universidad Politcnica de Madrid, Madrid, Spain

Departamento de Ingeniera Rural, Universidad Politcnica de Madrid, Madrid, Spain

a r t i c l e

i n f o

Article history:
Received 8 March 2014
Received in revised form 9 May 2014
Accepted 18 May 2014

Keywords:
Roof
Silo
Stress
Steel
Full-scale tests
Structural design

a b s t r a c t
The large-diameter, metallic, cylindrical silos used in agroindustry require in many cases some type of
structure in their uppermost section so that roof panels can be secured in place. These structures often
take the form of a lattice of radial and circular beams. The calculation models used in the design of these
lattices assume the existence of certain behaviours that are not usually veried experimentally. In the
present work, a 3-D beam model was used to predict the stresses and vertical displacements of a metal
silo roof structure measuring 18.34 m in diameter. To check the validity of the model, these stress and
vertical displacement values were experimentally veried at full scale. The instrumentation required
to obtain these values in such a large structure is complex and costly. In this study, slings were used
to apply load at 54 points in the roof structure, monitored by the use of dynamometers. The vertical displacements and strains suffered by the structure were recorded using eximeters and strain gauges
respectively. The experimental assays showed that the 3-D beam model did not contemplate the true
rigidity of the joints between the tension plates and the radial beams at their point of contact with the
vertical silo wall. The model therefore required adjustment in order to predict the measured results.
The results show the need to use conservative models in the design of these structures.
2014 Elsevier B.V. All rights reserved.

1. Introduction
Silos have been used as storage facilities in the agroindustrial,
mining, chemical and pharmaceutical industries since the end of
the 19th century. Cylindrical silos made of steel are probably the
most commonly used in the agroindustrial sector.
The enormous increase in computing power achieved since the
last third of the 20th century has greatly facilitated the use of
numerical methods for analysing the mechanical behaviour of
these structures (Jofriet et al., 1977; Eibl and Hassler, 1984; Ooi
and Rotter, 1990; Meng et al., 1997; Briassoulis, 2000; Guaita
et al., 2003; Vidal et al., 2006; Gallego et al., 2010; GonzlezMontellano et al., 2012). Such studies have signicantly improved
our knowledge regarding the behaviour of silos and their different
components (Ayuga, 2008).
Numerical models allow the stresses and displacements in any
structure to be predicted. All that needs to be known are the
dimensions involved, the construction materials to be used, the
cross-sections of the beams used, the design of the joints, and
Corresponding author. Address: ETSID Industrial, Ronda de Valencia, 28012
Madrid, Spain. Tel.: +34 91 3366837; fax: +34 91 3365625.
E-mail address: alvaro.ramirez@upm.es (. Ramrez-Gmez).
http://dx.doi.org/10.1016/j.compag.2014.05.008
0168-1699/ 2014 Elsevier B.V. All rights reserved.

the constrains applied to the model. However, it is difcult for

numerical models to reliably reproduce the behaviour of the nal
structure; nor can they contemplate small deviations from the
design owing to construction works. Such is often the case with
silos when the roof structure attaches to the separately-installed
wall sheeting.
The need for efcient, economically competitive silos means
engineers are faced with structural challenges that would be best
examined experimentally in situ. The literature, however, contains
few reports of full scale experiments on metallic structures with
the aim of assessing the resistance of their different elements to
deformation when under loads (Foutch et al., 1987; Nielsen,
1998; Kim et al., 2003; Teng et al., 2005; Hrtl et al., 2008; Xue
and Liu, 2009; Ramirez et al., 2010a,b,c; Pingue et al., 2011;
Couto et al., 2012; Yang and Liu, 2012; Ruiz et al., 2012). The reason
for their rarity is the cost and complexities involved in setting up
the instrumentation. Numerical models are therefore commonly
used in the design of such structures, but these are hardly ever validated by checking their predicted results with real ones.
Silo roofs are exposed to the action of wind and snow, and this
issue needs to be taken into account in their design (EN 1991-1-3,
2003; EN 1991-1-4, 2005). Their structure must also commonly
provide support to rooftop inspection gangways. They must

. Ramrez-Gmez et al. / Computers and Electronics in Agriculture 106 (2014) 5665

therefore behave in such a way as to prevent undesirable stresses

on the silo wall, which might result in the denting of the side
panels or even their buckling and failure, while allowing all gangways to be held rmly in place (Briassoulis and Pecknold, 1987;
Portela and Godoy, 2005; Ayuga, 2008).
The present case study compares the stresses and vertical displacements predicted by a 3-D beam model used to design an
18.34 m-diameter steel silo roof, with those measured in an experimental procedure performed at full scale. The instrumentation
required was complex and costly, and required many factors be
taken into account, such as the representative selection of the loading points on the roof, the monitoring of the loads transmitted, and
the selection of devices capable of measuring the desired variables.

57

bolts placed only at the web of the radial beams were used to join
this element to the tension plates. So, the anges of the radial beams
can freely move with respect to the tension plate, and then the angle
formed by both elements can vary. This is the reason why the union
of tension plates to the radial beams was supposed to be articulated
in the 3-D beam model. Detailed views of the joints between the
different elements of the roof structure are shown in Fig. 3.
3. The 3D-beam model
A 3D-beam model (PowerFrame software [BuildSoft]) was used
to predict the behaviour of the silo roof under load. The model
was based on the displacement method, and contemplated the
following (Fig. 1):

2. Description of the silo roof structure

The silo roof used in the present work had a diameter of
18.34 m. Its 24 radial beams were the main components resisting
the loads acting on the roof (Fig. 1). The same beams also provided
a surface for the xing of the covering panels.
A steel plate was used to connect the circular beams to the
radial beams only through the web of both beam types, leaving
the anges to rotate freely. This plate was L-shaped, being one side
attached to the web of the circular beam and the other side
attached to the web of the radial beam. In both cases, bolts were
used to connect the plate with the radial or circular beams. So, this
type of union was understood as articulated in the 3D-beam model.
On the other hand, the radial beams were deemed to be rigidly
joined to the ring stiffener placed at top of the structure. For this type
of union, steel plates are also used to connect the ring stiffener to the
radial beams by using bolts. In this case, the steel plate is joined both
to the web and anges of the radial beam, thus forcing the radial
beam and ring stiffener to maintain the angle formed by both elements, even if the structure is loaded. This is the reason why this
type of union was supposed to be rigid in the 3-D beam model.
Finally, at the lower part of the structure, the radial beams were
joined among them by means of tension plates (Fig. 2). In this case,

Radial beams and circular beams with a Sigma prole (height

250 mm, thickness 3 mm).
A top ring stiffener with an European Standard Channel
UPN-300 prole.
18 cm-high, 3 mm-thick steel tension plates joining the radial
beams at their lower part, close to the vertical wall sheet.
Constrains contemplated only at the exterior perimeter nodes.
Material properties, dimensions and cross sections of the beams
of the real structure were introduced in detail in the 3-D beam
model. Only an assumption was made concerning the lower support of the radial beams. These beams rested directly on vertical
stiffeners of the side wall panels, therefore only vertical displacement at these nodes were impeded in this model, by using
constrains applied at the exterior perimeter nodes.
According to the mounting of the structure described in
Section 2.1, the joints between the radial and circular beams were
deemed to be articulated, as were those between the tension plates
and the radial beams, while the joints between the radial beams
and the ring stiffener were considered to be rigid.
Concerning the stiffness of the tension plate, this was calibrated
by a numerical trial and error approach in order to account for the
real behaviour of this element as it was mounted.
4. Test procedure
4.1. Load application system
Loads were applied to the roof structure via the use of slings set at
54 points (Fig. 4). The slings were xed at one end to a concrete oor
slab using bolts (Fig. 5); the other end was wrapped over and
anchored to the joints between the radial and circular beams. Equal
loads were applied simultaneously to the loading points in increments of 1 kN, up to a maximum of 7.2 kN (approximately 80% of
the load the roof could resist) using the ratcheting system on each
of the slings. This was performed by staff members of SYMAGA, there
was one person operating one sling system. The assay was repeated
seven times and means were calculated for the variables measured.
The loads applied were monitored using dynamometers, the
strains were measured with strain gauges, and the vertical displacement experienced by the roof was recorded using cableactuated position sensors (eximeters). The strains measured can
be converted into stresses multiplying their values by the steel
modulus of elasticity. All these pieces of equipment and the procedures followed are described in Section 4.2.
4.2. Instrumentation

Fig. 1. Structure of the silo roof used in the present work.

4.2.1. Dynamometers
Eight dynamometers were installed on slings attached to
the radial beams to measure the loads applied (Fig. 6). These

58

. Ramrez-Gmez et al. / Computers and Electronics in Agriculture 106 (2014) 5665

Fig. 2. Detailed drawing of the joint between the tension plates and a radial beam at the lower part of the structure.

Fig. 3. Detailed views of the different joints: Radial beams to the ring stiffener (left), circular beams to the radial beams (center) and tension plates to the radial beams (right).

dynamometers were divided into three groups: one (D1) was a

load cell with an accuracy of 0.05% and a measuring limit of
20 kN, ve (D2-D6) were hook balances with a measuring capacity
of up to 10 kN (resolution 2 N), and two were hook balances (D7
and D8) with a measuring capacity of 10 kN (resolution 5 N).
4.2.2. Fleximeters
Three eximeters for measuring vertical displacement (range
0100 mm, resolution 0.025 mm) were installed at the points
shown in Fig. 7, i.e., one hanging from the ring stiffener, and two
from the second vein of radial beams 1 and 13. This type of sensor
transforms the linear movement of a cable (diameter 0.4 mm) into
angular movement via the use of a drum. This angular movement
is then converted into an electrical signal using a precision potentiometer, and the result captured using a datalogger. Displacement
values can then be derived from the electrical signals recorded.
The horizontal displacement at the bottom of the radial beams
was measured using four displacement gauges (resolution

0.01 mm) located on the vertical stiffeners attached to the radial

beams 1, 7, 13 and 19, as shown in Fig. 8.
4.2.3. Stress determinations
Stresses in the radial beams were measured using linear strain
gauges (350 ohms) located on four beams opposite one another
(Fig. 9). These gauges transmitted an electrical signal proportional
to the deformation produced. The magnitude of the stress was then
determined by multiplying this value by the modulus of elasticity
of the steel beams.
Gauges R1G1T, R1G1B, R2G1T, R2G1B, R3G1T, R3G1B, R4G1T
and R4G1B were located approximately 25 cm from a beam intersection node, while R1G2T, R1G2B, R2G2T, R2G2B, R3G2T, R3G2B,
R4G2T and R4G2B were placed about halfway along a vein, on the
upper and lower anges of the beams as shown in Fig. 10. This
allowed the distribution of normal stress over the cross section
of the beam to be determined, and traction and compression
stresses to be identied.

. Ramrez-Gmez et al. / Computers and Electronics in Agriculture 106 (2014) 5665

59

Fig. 7. Location of the eximeters for measuring vertical displacement.

4.2.4. Dataloggers
The readings from the strain gauges were recorded using an
ESAM Traveller 1 24-channel datalogger. The readings from the
eximeters were recorded using a DATA- TAKER DT50 datalogger.
5. Results and discussion
Fig. 4. Loading points.

5.1. Nomenclature used in the expression of the results

Placing the strain gauges on the structure required the galvanised surface at the location points rst be sand-papered away and
cleaned with acetone. The gauges were then stuck in place using a
cyanocrylate adhesive.

The present results show the change in normal stress (in N/

mm2) or vertical displacement (in mm), with respect to time t
(in s), at dened points in the structure. A positive normal stress

Fig. 5. System used to apply loads to the structure, and anchoring of the slings.

Fig. 6. Positions of the dynamometers.

60

. Ramrez-Gmez et al. / Computers and Electronics in Agriculture 106 (2014) 5665

Fig. 8. Positioning of displacement gauges for measuring horizontal displacement at the bottom of the radial beams.

The subindex i in Ri varies from 1 to 4 since four radial beams

were chosen for mounting the instruments, between which there
was an angular separation of 90. The subindex j in Gj can take
the values of:
1. Gauges on the radial beam nearer to the tension plates.
2. Gauges on the radial beam nearer to the top ring stiffener.

Fig. 9. Position of the strain gauges on the radial beams.

Finally, the position PL of the strain gauge at each position Gj

can take two possible values: that at T (Top), where the gauge is
on the upper beam ange, or that at B when it is on the bottom
ange.
The curves showing the change in normal stress obtained by the
3-D beam model are identied with the nomenclature PF_Gj_PL,
where PF refers to the PowerFrame software used in the calculations, and Gj and PL are those described above. In the results
obtained using the 3-D beam model, no Ri value is provided since
the distribution of the loads applied was symmetrical (the results
would therefore be the same for all the radial beams).
The legend FLk is used to describe the vertical displacements
recorded, where k indicates the position of the eximeter (Fig. 6).
The displacement curves obtained with the 3D-beam model are
identied with the abbreviation PF (FLk).
5.2. Comparison of experimental results with those obtained using the
3-D beam model

Fig. 10. Arrangement of the strain gauges on the radial beams. The blue points
show the location of the gauges on the upper and lower beam anges.

value reects traction stress while a negative value reects

compression stress.
The curves for the normal stresses measured in the radial beams
are reected following the format RiGj_PL, where Ri refers to the
radial beam in question, Gj marks the position on the radial beam
where the normal stress was measured, and PL indicates the
position (upper or lower ange) of the strain gauge at point Gj.

The normal stress and vertical displacement values obtained

using the 3-D model disagreed considerably with the values
returned by the experimental measurement of these variables.
Fig. 11 shows the model predicted a much lower normal stress
for the upper ange (PF_G1_T) and lower ange (PF_G1_B) at position G1 on radial beam 1, than measured experimentally. For
example, assay repetition E1 (the rst of seven) returned a maximum normal compression stress of 76 N/mm2 for the lower ange,
while the model predicted only 36.5 N/mm2.
Similar ndings were made with respect to the vertical
displacements of the roof structure. The mean maximum vertical
displacement actually measured was about 21 mm (Fig. 12), while
the model only predicted 8 mm.
The experimentally measured radial displacements at the
bottom of the radial beams were also quite different to those predicted by the model. Table 1 shows the displacements measured at
the four positions; the mean radial displacement measured was
11.6 mm, while the model only predicted a radial displacement
of 3 mm.

61

. Ramrez-Gmez et al. / Computers and Electronics in Agriculture 106 (2014) 5665

Fig. 11. Comparison of normal stress values measured experimentally and predicted by the model, for all the radial beams at position G1 in one of the assays repetitions.

Fig. 12. Comparison of vertical displacement values measured experimentally and predicted by the model.

These results show the tension plate/radial beam joints (Fig. 13)
were less rigid than the values contemplated by the model. The reason for this lies in making the joint via the folding of the tension plate
and the use of bolts of insufcient resistance. This led to the opening
of a radial aperture at the base of the radial beams that was larger
than envisaged. This reduction in rigidity also led to the radial beams
having to stand greater forces since the ratio of the rigidity of the
radial beam to the tension plate became greater than expected. Thus,
the normal stresses and vertical displacements measured experimentally were greater than those predicted by the model.
5.3. Modication of the numerical model and comparisons of the
stresses and displacements predicted by the modied model with those
experimentally obtained
The observations made on the behaviour of the tension plate/
radial beam joints led to the modication of the model. To prevent
confusion, the original model is hereinafter referred to as Model 1,
and the modied model as Model 2.

Table 1
Radial displacement measured experimentally (C1C4) and obtained using the model.
Experimental measurements (mm)

Model (mm)

C1

C2

C3

C4

Mean

12.98

10.60

11.22

11.49

11.57

The authors only used the results obtained for the rst test in
order to adjust the stiffness of the bar simulating the tension plate.
The stiffness of the bar simulating the tension plate was reduced
until the radial aperture of the silo matched with that experimentally measured. The numerical model that considered this reduced
value of the stiffness for the tension plate was Model 2. Then, the
different load patterns corresponding to each repetition were
applied in Model 2, and the new normal stresses and vertical
displacements provided by this modied numerical model were
then calculated for every assay and compared to those actually
measured. The results are shown below.

62

. Ramrez-Gmez et al. / Computers and Electronics in Agriculture 106 (2014) 5665

Table 2 compares the maximum normal stress values recorded

at position T (top) for any of the radial beams with those predicted
by Model 2 for all repetitions (E01E07). The difference between
the experimental and numerical mean values was less than 10%.

Fig. 13. Detailed view of the joint between a tension plate and radial beam seen
from above.

5.3.1. Normal stresses at position G1 on the radial beams

Fig. 14 shows the normal stress results predicted by Model 2 to
agree with those actually measured by strain gauges R1G1T,
R1G1B, R2G1T, R2G1B, R3G1T, R3G1B, R4G1T and R4G1B (situated
on both anges of radial beams 4, 10, 16 and 22) in one of the assay
repetitions. In all the radial beams, the normal stress in the upper
ange was one of compression, while in the lower ange it was one
of traction. In addition, the normal stress of traction was always
smaller than the normal stress of compression at any single position and at any moment in time. Thus, the pertinent position on
the radial beam was subject to a positive bending moment and
normal compression stress.
For either of the anges (T, top; B, bottom), and at any instant t,
the small differences between the normal stresses measured in the
four radial beams must be due to the forces applied at each of the
54 loading points not being exactly the same. Even so, the differences between the minimum and maximum stresses recorded
were always less than 20%.
The steps in the curves in Fig. 15 provided by the experimental
measurements correspond to the loading steps. With each new
load, an initial peak appeared in the stress recorded. This is due
to the sling tightening. When this procedure ended, the stress fell
slightly and stabilised around a constant value representing the
normal stress for the load analysed. This effect was also seen in
the results for position G2, and in the measurement of the vertical
displacements.

5.3.2. Normal stresses measured at position G2 on the radial beams

Fig. 15 shows the stresses recorded in one of the assay repetitions by strain gauges R1G2T, R1G2B, R2G2T, R2G2B, R3G2T,
R3G2B, R4G2T and R4G2B, situated on both anges of radial beams
4, 10, 16 and 22 (see Fig. 8), and the results predicted by Model 2.
The experimental results show that, for all the beams, the stress on
the upper ange was one of compression, while that on the lower
ange was one of traction. In addition, the absolute value of the
normal stress for the lower ange was always smaller than that
associated with the upper ange at any time t and at any position.
This implies that the position G2 on the radial beams is subject to a
positive bending moment and a normal compression stress.
Unlike that recorded for the normal stresses at position G1, signicant differences appeared between the radial beams at position
G2. For example, for the last load step in repetition E1, and with
respect to the lower ange of the beam R2, a stress of 26.5 N/
mm2 was obtained, while for R4 this gure was 7.5 N/mm2. With
respect to the upper anges of the same two beams, the values
were more similar at 42.4 N/mm2 and 30.6 N/mm2 respectively.
However, the maximum normal stresses predicted for both anges
by Model 2 were around 5.1 N/mm2, which are much smaller than
those actually recorded.
The discrepancy between the results returned by the four strain
gauges and those obtained with Model 2 are due to the position
selected for the location of the G2 gauges, which was very close
to a point in which a change of sign in the moments curve occurs
(Fig. 15). Consequently, a small variation in position would lead to
large differences in terms of stress values. In addition, the heterogeneity in the application of the loads, or the existence of small
imprecisions during the construction of the structure, could be to
blame. The maximum stress values reached at position G2 on
any radial beam, and for any gauge location or repetition of the
assay, were very much lower than those obtained for position
G1. Thus, when calculating the dimensions of the radial beams
required for similar structures, only the stress values recorded at
position G1 should be contemplated.
All the loads were applied over the silo roof by following
the vertical direction. Thus, any radial beam is simultaneously

Fig. 14. Comparison between the normal stress measured experimentally and predicted by Model 2 for the radial beams at position G1 in one of the assay repetitions.

63

. Ramrez-Gmez et al. / Computers and Electronics in Agriculture 106 (2014) 5665

Fig. 15. Comparison or normal stresses measured experimentally and predicted by Model 2 for all the radial beams at position G2.

Table 2
Maximum normal stress obtained at position G1 in the experimental assays and those predicted by Model 2.

r (N/mm2)

Experimental
Model 2
Difference

Repetition. Position G1

Mean

E01

E02

E03

E04

E05

E06

E07

76.5
79.1
3.40%

83.1
84.8
2.05%

85.1
93
9.28%

85.1
92.3
8.46%

77
93.9
21.95%

87.6
86
1.83%

86
96.5
12.21%

82.91
89.37
7.79%

Fig. 16. Curve for the moments of a radial beam in one of the assay repetitions predicted by Model 2. [in kN m].

subjected to compression forces and bending moments. However,

the real position where strain gauges G2 were placed is close to
a point where it appears a change in the sign of bending moments
(Fig. 16). So, a small variation in the point where numerical or
experimental results are obtained can produce discrepancies
between both sets of results. Thus, the numerical results obtained
for strain gauges G2 (Fig. 15) show that both anges are under
compression stresses because the value for the bending moment
is almost zero in the beam point where results are analysed. Therefore, normal stresses are caused mainly by the normal force, which
is of compression for any radial beam due to the type of loads
applied to the roof structure.

5.3.3. Vertical displacements

Fig. 17 shows the vertical displacements obtained in one of the
assay repetitions for each of the positions occupied by the three
eximeters. Also shown is the curve for the vertical displacements
for different loads obtained via Model 2. The latter curve adequately predicts the experimental results. Certainly, the maximum
values predicted by the model are of the same magnitude as the
displacements actually measured.
Over the rst load steps, hardly any difference was seen in the
values returned by the three eximeters. However, when the loads
were near maximum, some differences were noted. Fleximeter FL2
measured smaller vertical displacements than either FL1 or FL3.

64

. Ramrez-Gmez et al. / Computers and Electronics in Agriculture 106 (2014) 5665

Fig. 17. Vertical displacements measured experimentally and obtained by Model 2 in one of the assay repetitions.

Table 3
Maximum vertical displacement measured by eximeters FL1, FL2 and FL3, and
results predicted by Model 2.
dy (mm)

Experimental
Model 2
Difference

Repetition

Mean

E1

E2

E4

E5

E7

21.5
17
20.93%

25.3
20
20.95%

22.5
20
11.11%

22.1
21
4.98%

22.4
20.8
7.14%

22.76
19.76
13.18%

This might be explained in that FL2 was located on the ring stiffener at the top of the roof structure, while the other two were
located on the underside of two radial beams. Small differences
were also seen between the measurements made by FL1 and FL3.
These were due to the fact that the force applied at the 54 loading
points was not strictly the same. This would have led to some
heterogeneity in the loads applied in each area of the roof, and
therefore some variation in the vertical displacement registered.
Table 3 shows the maximum vertical displacement recorded
(for the eximeters as a whole) and the value predicted by Model
2 for all the assay repetitions. The results for repetitions E3 and E6
are not shown given that problems occurred with the instruments
during these assays. The measured vertical displacements are
slightly greater than those predicted by Model 2; the mean values
differed by 13%.
It should be remembered that the maximum eximeterprovided displacement value was contemplated; this would be
the value most similar to that predicted by the model. Further,
the values measured by the eximeters were always 12 mm
greater than the real value, due to the vertical compression of
the wall sheeting on which the roof rests.

The present experimental results did not agree with those predicted by the original model (Model 1). This was due to the joints
between the tension plates and the radial beams, in the area of
contact with the silo wall being less rigid than predicted. This
shows the need to use more conservative models in the design of
these structures that take into account possible difculties in their
actual mounting. This is especially important taking into account
that it is not common to try to validate the model used to design
the silo roofs before they are mounted.
The observations made on the behaviour of the tension plate/
radial beam joints led to the modication of the rigidity of the bars
simulating the tension plates in the original model. With respect to
the maximum stress affecting the radial beams and the roof
displacement values, Model 2 provided results much more in line
with those collected experimentally. The stress values for the
radial beams predicted by Model 2 were a mean 7.8% higher than
those recorded experimentally, while the vertical displacement
values were 13.2% smaller.
The present results highlight the importance of the tension
plates in the behaviour of a silo roof structure. On the other hand,
different silo designs may have different joints and stiffness may
vary signicantly. The accuracy of the results obtained with a
computational model greatly depends on a correct simulation of
the stiffness of the joints.
Acknowledgements
The authors thank SYMAGA for funding this work, Prof. Jrgen
Nielsen of the Danish Building Research Institute, Aalborg University, for assistance in the design of the experiment, and Prof. Jos
Miguel Gonzalo-Magro for his interesting discussion.
Appendix A. Supplementary material

6. Conclusions
Performing experiments on steel structures at full scale is
difcult and costly. During the design of the present experiment,
decisions had to be made regarding simplications of the experiment that would still allow the behaviour of the structure to be
determined and problems or anomalies not predicted by the
original model to be detected.

Supplementary data associated with this article can be found, in

the online version, at http://dx.doi.org/10.1016/j.compag.2014.05.
008.
References
Ayuga, F., 2008. Some unresolved problems in the design of steel cylindrical silos.
In: Chen, J.F., Ooi, J.Y., Teng, J.G. (Eds.), International Conference on Structures

. Ramrez-Gmez et al. / Computers and Electronics in Agriculture 106 (2014) 5665

and Granular Solids From Scientic Principles to Engineering Applications.
The Royal society of Edinburgh, Scotland, UK, 12 July, 2008, pp. 123133.
Briassoulis, D., 2000. Finite element analysis of a cylindrical silo shell under
unsymmetrical pressure distributions. Comput. Struct. 78 (13), 271281.
Briassoulis, D., Pecknold, D.A., 1987. Behavior of empty steel grain silos under wind
loading: Part 2: The stiffened conical roof shell. Eng. Struct. 10, 5764.
Couto, A., Ruiz, A., Aguado, P.J., 2012. Design and instrumentation of a mid-size test
station for measuring static and dynamic pressures in silos under different
conditions Part I: Description. Comput. Electron. Agr. 85, 164173.
Eibl, J., Hassler, U., 1984. Numerical investigations in discharging silos. J. Eng.
Mech. 110 (6), 957971.
EN 1991-1-3, 2003. Eurocode 1 Actions on Structures Part 13: General Actions
Snow Loads. Brussels, CEN.
EN 1991-1-4, 2005. Eurocode 1: Actions on Structures Part 14: General Actions
Wind Actions. Brussels, CEN.
Foutch, D.A., Goel, S.C., Roeder, C.W., 1987. Seismic testing of full-scale steel
building. Part I. J. Struct. Eng. ASCE 113 (11), 21112129.
Gallego, E., Rombach, G.A., Neumann, F., Ayuga, F., 2010. Simulations of granular
ow in silos with different nite element programs: ANSYS vs SILO. Trans.
ASABE 53 (3), 819829.
Gonzlez-Montellano, C., Gallego, E., Ramrez-Gmez, A., Ayuga, F., 2012. Three
dimensional discrete element models for simulating the lling and emptying of
silos: analysis of numerical results. Comput. Chem. Eng. 40, 2232.
Guaita, M., Couto, A., Ayuga, F., 2003. Numerical simulation of wall pressure during
discharge of granular material from cylindrical silos with eccentric hoppers.
Biosyst. Eng. 85 (1), 101109.
Hrtl, J., Ooi, J.Y., Rotter, J.M., Wojcik, M., Ding, S., Enstad, G.G., 2008. The inuence
of a cone-in-cone insert on ow pattern and wall pressure in a full-scale silo.
Chem. Eng. Res. Des. 86, 370378.
Jofriet, J.C., Lelievre, B., Fwa, T.F., 1977. Friction model for nite element analyses of
silos. Trans. ASAE 20 (4), 735744.
Kim, S.E., Kang, K.W., Lee, D.H., 2003. Full-scale testing of space steel frame
subjected to proportional loads. Eng. Struct. 25, 6979.

65

Meng, Q., Jofriet, J.C., Negi, S.C., 1997. Finite element analysis of bulk solids ow:
Part 1, Development of a model based on a secant constitutive relationship. J.
Agric. Eng. Res. 67 (2), 141150.
Nielsen, J., 1998. Pressures from owing granular solids in silos. Philos. T. Roy. Soc. A
356, 26672684.
Ooi, J.Y., Rotter, J.M., 1990. Wall pressures in squat steel silos from simple nite
element analysis. Comput. Struct. 37 (4), 361374.
Pingue, F., Petrazzuoli, S.M., Obrizzo, F., Tammaro, U., De Martino, P., Zuccaro, G.,
2011. Monitoring system of buildings with high vulnerability in presence of
slow ground deformations (The Campi Flegrei, Italy, case). Measurement 44,
16281644.
Portela, G., Godoy, L.A., 2005. Wind pressures and buckling of cylindrical steel tanks
with a conical roof. J. Constr. Steel Res. 61 (6), 786807.
Ramirez, A., Nielsen, J., Ayuga, F., 2010a. On the use of plate-type normal pressure
cells in silos. Part 1: Calibration and evaluation. Comput. Electron. Agr. 71, 7176.
Ramirez, A., Nielsen, J., Ayuga, F., 2010b. On the use of plate-type normal pressure
cells in silos. Part 2: Validation for pressure measurements. Comput. Electron.
Agr. 71, 6470.
Ramirez, A., Nielsen, J., Ayuga, F., 2010c. Pressure measurements in steel silos with
eccentric hoppers. Powder Technol. 201, 720.
Ruiz, A., Couto, A., Aguado, P.J., 2012. Design and instrumentation of a mid-size test
station for measuring static and dynamic pressures in silos under different
conditions Part II: Construction and validation. Comput. Electron. Agr. 85,
174187.
Teng, J.G., Lin, X., Rotter, J.M., Ding, X.L., 2005. Analysis of geometric imperfections
in full-scale welded steel silos. Eng. Struct. 27, 938950.
Vidal, P., Gallego, E., Guaita, M., Ayuga, F., 2006. Simulation of the lling pressures of
cylindrical steel silos with concentric and eccentric hoppers using 3dimensional nite element models. Trans. ASABE 49 (6), 18811895.
Xue, W., Liu, S., 2009. Design optimization and experimental study on beam string
structures. J. Constr. Steel Res. 65, 7080.
Yang, J., Liu, Q., 2012. An experimental study into exural behaviour of sigma
purlins attached with roof sheets. Eng. Struct. 45, 481495.