Beruflich Dokumente
Kultur Dokumente
Jim Durack has a background in both academic and practicing engineering. He has worked with Aurecon
(formerly Connell Wagner) for the last eight years and is currently on sabbatical leave completing an engineering
doctorate through University of Southern Queensland. Jim is a specialist in the design of industrial structures
for bulk materials handling facilities. He has also worked internationally on the design and assessment of
large inverted cone silos as used primarily in the cement production industry. In 2007 ABHR published Jim’s
article on the implications for such silos of the then new Eurocode for silo loads, EN1991-4. Since then he has
presented three further conference papers in this field. Both Jim and Professor Tranberg have been involved in
the development of the USQ coursework based, Online delivered, Master of Engineering that has a significant
emphasis on industrial and marine structures.
In preparation for a planned course on silo design, Jim has extended his reinforced concrete silo research to a
speculative investigation of a steel solution for large inverted cone silos. His knowledge and experience with
concrete silos gives him fresh insight into this complex field. Jim considers that there may be a fundamentally new
way of approaching steel silo design that is informed by the Eurocode but not limited by it. This paper provides
a relatively informal presentation of his ideas. He hopes that these may challenge steel silo designers to think
outside the square.
1. Context and background to research have similar lessons for the designers of steel silos as he has found
Silos of one form and another are fundamental to the success of it to have for reinforced concrete silos. The following represent
most industrial projects. Ideally a silo will do its job with minimal some key findings of this research:
maintenance for many years. When a silo suffers some form of fail- • For the design of steel silos, the Eurocode for silo loads EN1991
ure this may cause massive disruption of plant operations. There - 4 is supplemented by the general design code EN1993 - 1 - 6
is at least anecdotal evidence that silos suffer severe structural Strength and Stability of Shell Structures and the silo specific
failures more frequently than any other structure type with the EN1993 - 4 – 1 Design of Seel Structures - Silos.
possible exception of retaining walls. Such structural failures are • There is limited contemporary academic interest in the
still likely to be much less common than operating problems such analysis and design of reinforced concrete silos, but a huge
as flow interruptions. The operating performance of silos should body of academic literature relating to various aspects of the
continue to be of prime concern to plant operators but those who structural behaviour of steel silos. Much of the literature is
are “unlucky” enough to experience the structural collapse of a highly specialised in nature and of limited direct relevance
key process silo will recognise the fundamental importance of to practicing designers.
structural reliability. • Professor Michael Rotter is apparently the most highly pub-
Fundamental to structural design is a rational assessment of lished academic working in the steel silo field. He has played a
the loads that are likely to develop during the lifetime of a pro- leading role in the development of all three Eurocodes relevant
posed structure. In 1895 Janssen developed the Janssen equation to silo design.
that for the first time provided a rational method for assessment of • Professor Rotter has published a text “Guide for the Economic
one aspect of the loads exerted by stored material on a silo wall. Design of Circular Metal Silos”. This text is specifically directed
It has always been acknowledged that “real silo loads” are more to practicing silo designers. First published in 2001 it repre-
complex than those predicted by this equation. “Traditional” code sents a general guide to steel silo design plus specific guidance
methods for silo loads such as those of the ACI code and the now related to the application of the three silo Eurocodes.
superseded DIN code, used fairly simplistic “fudge factors” to • Durack has reported that for reinforced concrete silos, the Eu-
scale up the Janssen equation loads to account for the unknown rocode for silo loading is more conservative than earlier codes
and perhaps unknowable complexities of real silo loading. by a factor in excess of 1.5. No such comparative review for
The Eurocode for silo load estimation EN1991 – 4 released in steel silos has been identified in the published literature.
draft form in 1995 is still based on the Janssen equation but speci- • Rotter’s text and the codes it supports are typical of code speci-
fies a fundamentally different loading model. Research undertaken fied design methods. They allow a designer to undertake a de-
by Durack and others generally concludes that while there remain sign assessment of a proposed steel silo but they provide only
numerous unanswered questions, there is good evidence to sug- limited assistance in developing the form of understanding that
gest that the Eurocode loading model is substantially correct. For a designer needs in order to exercise responsible creativity in
reinforced concrete silos, Durack has shown the Eurocode to be achieving a reliable and economic design solution for a particu-
considerably more onerous than earlier codes. More importantly, it lar steel silo application.
forces the designer to consider important structural issues that are
not evident when using earlier code defined loading models. Du- 2. Focus of the current research
rack recommends that any contemporary structural designer of re- To address the last of the previous dot point items, a combination
inforced concrete silos should give consideration to the Eurocode of finite element modelling and relatively simplistic first principles
even if the final silo design is not strictly compliant with this code. analysis has been undertaken in order to explore structural phe-
Durack has recently undertaken scholarship and finite ele- nomena exhibited by steel silos in response to the Eurocode load
ment modelling to investigate whether the Eurocode load may and related analysis and design issues.
The research focuses on a 12 500 tonne capacity cement stor- 4. Two dimensional structural phenomena
age inverted cone silo with a diameter of 18 metres and a storage Prior to considering a three dimensional silo it is informative to
chamber height of 40 metres. This silo is not typical of most steel consider the behaviour of a two dimensional ring extracted from
silos that are likely to use a concentric discharge system and to such a silo. Figure 3 represents a 1 metre high ring extracted
be smaller and more slender. The case study silo has been chosen from the case study silo with the loading corresponding to that
because of the author’s interest in reinforced concrete silos of at a height of 15 metres above the base of the silo.
these proportions. If the thickness of the 1 metre high ring is set to 25 mm then
It should be noted that applicability of the Eurocode for silo it is obvious that these loads are very large. Despite this, we may
loading specifically excludes the “inverted cone” discharge system expect that the uniform outward pressure should not create a
that is common in the cement industry. This discharge system uses difficulty because this can only cause a uniform increase in the
an inverted cone at the base of the silo that sheds the stored mate- circumference of the ring without inducing any bending. The
rial to sequentially operated discharge points located around the axial hoop tension can be calculated using the simple hoop ten-
silo perimeter. Despite this, the research is generally relevant to sion equation:
the design of a steel silo of any sort. T = Uniform pressure x Ring height x Radius (1)
= 123 kPa x 1 metre x 9 metres
3. The case study silo and its Eurocode loading = 1107 kN
The basic details of the case study silo are shown in Figure The corresponding axial hoop stress around the circumfer-
1. This also gives a three dimensional representation of the ence is then:
Eurocode loading. fa = 1107 x 103 / (1000 x 25)
In accordance with the Eurocode EN1991 – 4 Clause 5.2.4.3 = 44.3 MPa
and with z measured downwards from the top of the silo, the The differential pressure of +/- 85 kPa over the 30° (4.71 me-
loadings phse and phse – phce shown in Figure 2, that vary both cir- tre arc length) flow zone and the two adjacent shoulder zone
cumferentially and with height were input as follows: would obviously cause a problem if the load had to be carried
by bending action in the flexible 25 mm thick plate. Fortunately
phse = 150.94 x (1 – 2.7813 ^ (-z / 14.56)) kPa this is not the case.
The hoop tension equation is commonly applied to uni-
phse– phce = 150.94 x (1 – 2.7813 ^ (-z / 14.56)) – 40.4 x form pressure situations but may also be applied to situations
(1 – 2.7813 ^(-z / 3.9)) kPa where the pressure varies around the perimeter. It may be put
in the form:
T(θ) = p(θ) x r(θ) (for unit ring height) (2)
Where T(θ), p(θ) and r(θ) represent the values of the hoop
tension, radial pressure and ring radius at any location θ around
the perimeter of the ring. Equation (2) may be algebraically rear-
ranged as:
r(θ) = T(θ) / p(θ)
It may readily be shown by FE modelling or consideration of
a small differential arc element of the ring, that for loading of the
type shown in Figure 2, the hoop tension T(θ) remains constant
at the value given by equation (1). Thus, we may write:
r(θ) = T / p(θ)
And substituting for T from Equation (1)
r(θ) = (Uniform pressure / Pressure at position θ)
Figure 1 Basic silo details and a graphic illustration of the internal pressure on the silo walls. x Initial silo radius (3)
Equation (3) may be read as follows:
Provided the ring changes its shape in response to the load
such that at any angular location θ around the ring circumfer-
ence, the ring radius equals the ratio of the uniform pressure
to the pressure at θ, times the initial silo radius, then the load
will be carried by hoop tension action rather than by bending.
This shape may be referred to as the “compliant shape”.
For the combination of loads defined in Figure 3 there are
three pressure zones and corresponding radii being:
The general area with phse = 123 kPa with:
r(θ) = 123 / 123 x 9 = 9 metres
Figure 2 The decomposition of the Eurocode load into uniform and differential components. The high pressure shoulders with phse + (phse – phce)
= 208 kPa with:
r(θ) = 123 / 208 x 9 = 5.32 metres
The low pressure flow zone with p hse – (p hse – p hce)
= 38 kPa with:
r(θ) = 123 / 38 x 9 = 29.13 metres
These compliant radii are dependent only on the pressure ratios
and are independent of both the thickness of the ring and the actual
magnitude of the pressures. Thus if the pressures are scaled up by
a load factor then there will be no change to the compliant shape
and corresponding radii. Figure 4 represents the FE Prediction of
the compliant shape for the ring subject to the combination of the
Figure 3 Two dimensional ring extracted from the 3D silo model. uniform and differential components of loading shown in Figure 3.
• For feasible steel plate thicknesses and the non uniform Eu- • The hoop tension force varies around the perimeter of the silo
rocode loading, there is a good correlation between the pre- from 968 to 1252 kN/m. The average of these two figures is
dictions of simple hoop tension theory and results obtained very close to the constant hoop tension in the two dimensional
from FE modelling. ring of 1107 kN/m.
• A steel silo is dependent for its equilibrium on moving towards • The maximum and minimum plate stresses (at the surface of
the compliant shape. Linear static analysis that assesses equi- the plate) are 247 MPa and - 163 MPa compared to those for the
librium in the undeformed position (with a perfect circular two dimensional ring of 236 MPa and – 148 MPa.
shape) will not give any meaningful prediction of the actual
behaviour of the silo. All of the previous FE results are based
on geometry non linear analysis.
• The Eurocode loading will cause large radial deformations of
an unstiffened steel silo. These large deformations develop at
quite low load factors but as the load increases they reach a
maximum value beyond which they will not further increase.
• Increasing the wall thickness of a silo will not significantly
reduce the magnitude of radial deformations (until the thick-
ness reaches very large values more typical of a reinforced
concrete silo).
• For a given non uniform load combination, increasing the wall
thickness of a silo will increase the total combined axial and Figure 6 Aspects of the behaviour of the Figure 5 silo focussing on the ring level.
bending stress. A maximum stress in excess of the yield stress
should not be taken as an indication that the silo is approach- 7. Three dimensional case study silo - vertical
ing failure. It simply means that with thicker plates, some de- effects
gree of yield may be necessary to allow the silo to move to the In addition to the horizontal loading on the silo wall, the stored
compliant shape. material also causes vertical loading as the stored material at-
• For low loads a silo will exhibit very low stiffness but as soon tempts to slide down the inner face of the silo wall. This vertical
as it approaches the compliant shape the stiffness will radical- loading is referred to as the wall traction loading and at any depth
ly increase. For this reason it may be difficult to obtain conver- and radial position θ is equal to the horizontal pressure times
gence of an FE solution for the first low increments of loading. the assessed coefficient of friction between the stored material
and the silo wall. As with horizontal loading, traction loading
6. Three dimensional case study silo – has been decomposed into a uniform traction component that is
horizontal effects constant around the perimeter and a differential component that
is downwards in the high pressure shoulder zones and upwards
in the low pressure flow zone.
The load path for these traction loads is straight down the silo
wall. Figure 7 shows the analysis results for the uniform traction
component causing a steady increase in the vertical wall stress
from zero at the top to a maximum of - 62 MPa (compression)
at the base of the 25 thick wall. The differential traction com-
ponent produces a more complex stress variation that includes
both compression and tension areas but with maximum value of
only + 5.4 MPa and – 3.6 MPa this is clearly of secondary concern.
The view on the right of Figure 8 (a) represents the same The situation for non uniform loading is more complex. It
stresses but with the plotting range limited to 50 MPa tension and has already been noted that proper assessment of silo behaviour
100 MPa compression. This shows that high vertical compression requires non linear analysis – but finite element buckling analy-
(and tension) stresses exist high up the wall height. The maxi- sis by definition, is linear and thus cannot directly predict the
mum compression region is the low pressure flow zone. It can be onset of buckling due to non uniform loading with its inherently
demonstrated that this increase in vertical stresses is a result of non linear characteristics. The analytical Equation (10) is strictly
the differential horizontal loading. It is difficult to come up with speaking only applicable to uniform loading but it is informa-
a satisfactory intuitive understanding for this effect. One way of tive to apply it in the flow zone region where, for non uniform
thinking about it is to consider it to be a result of the curved silo loading, the silo moves (almost) to the compliant shape. The FE
wall attempting to carry the differential horizontal loading by model predicts a final silo radius in this region of 24.8 metres.
spaning vertically from the base of the silo up to the roof. The We then have:
very high vertical stresses at the base of the wall correspond to a σcr = 0.6 x 200 x 103 x 25 / 24 800
rigid base to these vertically spanning wall elements giving rise = 121 MPa
to high “negative moments” at the base of the wall. Despite the Figure 9(b) represents the FE buckling prediction, again for
difficulties in understanding the origin of this increase in vertical uniform traction loading only but for a model that is initially de-
stresses, they have to be accepted as being real. formed to the shape illustrated in Figure (6) with the final ge-
ometry non linear, 24.8 metre silo radius in the critical region.
The FE prediction of 132 MPa buckling stress is very close to
that of Equation (10) with some increase due to averaging of the
0.0 peak stress and the stabilising effect of the adjacent small radius
shoulder regions.
Thus it appears that for this silo, the non uniform Eurocode
RKDOIIORZ]RQH loading is causing:
RVKRXOGHU]RQH • High vertical stresses near the base of the silo
8QLIRUP]RQH
• Vertical compression stresses around 100 MPa reaching well
up the height of the silo (compared to a
D9HUWLFDOVWUHVVHVDWGLIIHUHQWSORWWLQJVFDOHV E&LUFXPIHUHQWLDOYDULDWLRQLQYHUWLFDO • maximum of 62 MPa at the base of the silo for the uniform
VWUHVVDWULQJOHYHO
traction loading only)
Figure 8 Vertical wall stresses due to the full Eurocode loading.
• A substantial reduction in the buckling capacity of the wall in
The magnitude and extent of vertical stresses shown in Fig- the flow zone as evidenced by a reduction in the critical buck-
ure 8 is of great design significance as it has already been identi- ling stress from 319 MPa down to 132 MPa due to the increase
fied that in the flow zone where the compression stresses are in the silo radius as it moves towards the compliant shape.
high, the horizontal pressures cause a local increase in the silo
wall radius as it moves towards the compliant shape. As dis- 8. Working around the problem
cussed below this causes a reduction in the buckling capacity of It appears that the Eurocode load, when applied to steel silos,
the wall in this region. may be considerably more onerous in comparison to earlier
Recall that the Euler critical lateral buckling stress for an ideal codes that considered only uniform loading. It can probably be
axially loaded column with length l and radius of gyration r is: shown that while there have been an unacceptable number of
σcr = 2E / [ (l/r) 2 ] (9) failures of steel silos, there are many that are performing well but
Now a very tall and slender silo could buckle laterally in this do not comply with the requirements of the new loading code.
fashion but it is much more common for a silo to exhibit local It would be easy to start criticising the Eurocode for silo loading
buckling with a buckling shape as illustrated in Figure 9. that could be seen as being unnecessarily conservative.
The author’s research on reinforced concrete silos indicates
that the Eurocode for silo loading probably represents a good
model of actual silo loads. A more creative way around the prob-
lem is to consider whether some modification to the traditional
silo structural system may allow the development of a Eurocode
conforming silo without the need for significant additional cost
while providing significantly enhanced structural reliability.
One possibility is the use of some form of corrugated wall-
ing or of multiple vertical or circumferential stiffeners - or both.
The author considers that this approach may have its applica-
tions in some areas but that in general is unlikely to produce sig-
nificant benefits. The structural efficiency of a steel silo comes
Figure 9 Typical buckling mode shapes for uniform and non uniform loading.
largely from the flexibility of the thin steel wall and its ability
to move to the compliant shape. Stiffeners will inevitably cause
For uniform vertical loading only, applied to a cylinder of high stress concentrations as they attempt to fight against the
radius r the formula to predict the development of the form of flexibility of the wall.
buckling illustrated in Figure 9(a) with Poisson’s ratio of 0.25 is: A second possibility is illustrated in Figure 10 (b). This makes
σcr = 0.6 x E x t / r (10) use of a single relatively massive stiffener at about one third
= 0.6 x 200 x 103 x 25 / 9000 (for the 25 thick wall and height. The idea behind this stiffener started with an acceptance
the original 9 metre silo radius) of the tendency of the silo to attempt to carry some of the non
= 333 MPa uniform load by spaning vertically, resulting in high compression
Figure 9 (a) shows the FE buckling prediction for uniform ver- stresses in the high radius flow zone. The stiffener represents
tical traction loading only applied to the perfect circular silo. The an attempt to provide an internal support for this vertical load
buckling stress prediction of 319 MPa shows excellent correlation path thus increasing its efficiency in carrying some of the non
to Equation (10). uniform components of the load.