The Effects of Patch Loads On Thin-Walled Steel Silos: M. Gillie, J.M. Rotter
The Effects of Patch Loads On Thin-Walled Steel Silos: M. Gillie, J.M. Rotter
The Effects of Patch Loads On Thin-Walled Steel Silos: M. Gillie, J.M. Rotter
www.elsevier.com/locate/tws
Abstract
This paper reports on the results of a detailed parametric study into the effects of patch
loads on the stresses in thin-walled circular steel silos. Firstly, an analysis of the effects of a
typical patch load on the stresses in a silo wall is presented. The results show that the stresses
set up are complex and that they could potentially lead to failure of the silo by either elastic
buckling or plastic collapse. A parametric study is then conducted which examines the effects
of varying the circumferential width of the patch load, the vertical extent of the load, the point
of application of the load and the pressure distribution within the load. The results show that
the circumferential width of the applied load and the pressure distribution both strongly affect
the form and magnitude of the stresses produced in the silo wall. The magnitudes of stresses
in the silo wall were found to vary almost linearly with the vertical extent of the patch load.
2002 Elsevier Science Ltd. All rights reserved.
Keywords: Silos; Patch loads; Shells; Finite element analysis; Design
1. Introduction
The traditional method of silo design assumes that all the loads from the stored
bulk solid are symmetrical with respect to the silo centreline. Recently, a number
of studies have shown this is not the case [13]. Because it is not yet possible to
predict with accuracy what form the asymmetric components of loads might take,
the concept of a patch load was introduced into the German DIN standard [4] for
Correponding author.
E-mail address: mgillie@eng.ed.ac.uk (M. Gillie).
0263-8231/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 2 6 3 - 8 2 3 1 ( 0 2 ) 0 0 0 2 8 - 9
836
Nomenclature
d
F
h
mx
mq
mxq
nx
nq
nxq
p
p0
qxn
q?n
R
t
y
a
q
seq
sx
sq
txq
txn
tqn
silo pressures. A patch load is a local load that can be applied at any point on the
silo wall. They are normally taken to act as pairs of opposing loads acting on opposite
sides of the silo at a given level. While these substitute loads have been devised to
produce appropriate bending resistance in the walls of concrete silos, their relevance
to metal silos remains unclear and the task of defining appropriate patch load forms,
sizes and amplitudes has not yet been undertaken.
Concrete silos respond to asymmetric pressures predominately by local circumferential bending that produces local vertical cracks, so the patch load method is well
suited to their design. The controlling failure criteria for steel silos are not the same
as those for concrete silos (meridional membrane stress buckling as opposed to bending stress cracking dominate design considerations) and their structural response also
differs. Because of these differences, it is at present rather uncertain how patch loads
should be applied to thin metal silos, and a very simplified model has been used in
draft Eurocodes [5,6].
837
The aim of this study is to explore how the position, size and form of a patch
load on an unstiffened metal silo affects the stresses set up in the silo wall. It is an
extension of, and complimentary to, an earlier study by Rotter [2] and is concerned
with the stresses induced in the walls of thin-walled silos. The stresses caused by
patch loads are examined in the light of potential failure modes for such metal silos.
The peak patch pressure is typically defined as a multiple of the Janssen filling
pressure at the same level. As a consequence, a patch located at the top of the silo
is unlikely to significantly affect the structural behaviour, as the Janssen pressure is
small here. Similarly, patch pressures near the bottom of the silo induce rather small
meridional stresses in the wall and will also have little effect.
838
ied to the patch loads means that the results can be applied to a wide range of
problems and are not limited to one silo or load arrangement.
2.1. Reference geometry
The reference silo (Fig. 1) was chosen to have a dimensionless height, h, of 2000,
a radius, R, of 500 and a thickness, t, of 1. This gave an aspect ratio of 2, and a
radius to thickness ratio of 500; values typical of a large silo designed for economy.
The wall thickness was assumed to be constant over the whole silo. The base boundary condition of the silo was taken to be pinned (meridian free to rotate but otherwise
fixed) and the top boundary condition to correspond to a stiff ring that was free to
translate as a whole but was restrained from out-of-round deformations. This arrangement is shown in Fig. 1. Youngs modulas was taken as 207 GPa and Poissons
ratio as 0.3.
So that the results are generally applicable, they are presented in dimensionless
form by dividing by the circumferential membrane stress (CMS) that should be
expected to arise from application of the patch pressure according to simple membrane theory(sq pr / t). Thus with R/t=500 and a peak value of dimensionless pressure of 0.002 at some point in the patch load, a peak circumferential membrane stress
of unity would result. All pressure areas were rectangular in shape, as specified in
DIN 1005 [9] and Eurocodes 1 and 3 [5,6,10].
The pressure was always taken to act outwards in the analyses though both
increases and decreases in pressure may be important in design[1]. There is good
evidence [1,3] that an area of increased pressure on a silo wall is balanced by an area
of decreased pressure elsewhere. This study can be used to address these situations by
using the principle of superposition to add patch loads acting in different locations
and directions.
839
P P0cos
pq
p(zy)
cos
a
d
q
q
d
d
for a and y zy
2
2
2
2
(1)
The form of the pressure variation within the patch led to a different total applied
force, as noted above, but the simple theoretical hoop stress at the centre of patch
was the same. Graphs illustrating these two types of patch load are shown in Fig. 2.
Fig. 2.
Patch pressure distributions. (a) Uniform pressure distribution. (b) Bell-shaped distribution.
840
nx mx
,
t t2 / 4
txq
nxq mxq
qxn
2 , txn ,
t
t /4
t
sq
nq mq
t t2 / 4
txn
qqn
t
(2)
841
It is thus possible to substitute values into dimensionless results from this investigation to obtain a low lower-bound to the probable plastic collapse strength.
Fig. 3.
842
Fig. 4.
base whilst above the patch load the deflection is free to increase somewhat because
the distance to the top of the silo (where out-of-round deflections are prevented) is
relatively large.
4.2. Stresses
The dimensionless stresses developing in the wall are shown in Figs. 5 and 6,
where the most significant stresses are the circumferential bending stress (CBS), the
meridional membrane stress (CMS) and, consequently, the von Mises stresses. Fig.
5 shows that there are two areas in which stresses may be high. These are along
meridians near the patch (q=15) edge and along the zero meridian (q=0). Membrane stresses peak along the zero meridian whilst bending stresses peak at the
patch edges.
Fig. 5. Stresses against theta at y/R=1 (patch centre) for a typical load case.
Fig. 6.
843
Stresses against meridional co-ordinate at the patch centre for a typical load case.
At the patch centre the CBS shows a sharp negative peak just outside the edge
of the patch load at a co-ordinate of 20 (Fig. 7). This peak arises from the abrupt
start of the patch load causing the shell to bend sharply at this point (c.f. Fig. 3),
resulting in high compression on the outer surface of the shell, or negative bending.
The behaviour of the CBS near the centre of the patch load is more complex. The
CBS is high throughout the whole region of the patch but slightly lower at the patch
centre than at around 10. This dip in bending at the patch centre is the result of
two opposing trends. The silo wall is responding to the patch load firstly as a point
load. This sort of response results in high positive bending over a small region at
the centre of the patch as the flexible wall is forced to bend around the load. This
Fig. 7.
844
Fig. 8.
behaviour can be clearly seen in Fig. 7 when the CBS curve for a load of 10 is
examined. The patch load is wide enough however for this simple point load response
to be complicated by the silo beginning to respond to the load as a distributed load.
A distributed load does not result in high positive bending at the patch centre because
the uniform pressure within the region of the load means that silo wall is not required
to bend in this area. The curve for a load of 90 width in Fig. 7 illustrates this
behaviour. These two effects are both present with a load of the 40 width and so
the local dip in bending is observed at the patch centre.
It can be seen in Figs. 812 that away from the patch centre the maximum CBS
occurs along the zero meridian and no longer at the patch edges. This is because
Fig. 9.
845
Fig. 10. Stresses against theta at y/R=1.2 for a typical load case.
Fig. 11. Stresses against theta at y/R=1.6 for a typical load case.
the bending stresses away from the region of the patch load result more from the
deformed shape of the silo rather than from the disruption caused by the abrupt
change in the load intensity at the patch edges.
The CMS, or hoop stress, rises smoothly to a peak value of just over one at the
patch centre. A membrane theory would give this value as one. Except near the silo
base (see below) the CMS is tensile forces and can therefore be resisted with ease.
The value of compressive CMS is of importance when designing against buckling
failure. Figs. 812 show that in the circumferential direction the CMS has a negative
peak at 25 and 35, depending on the location being considered, and rises smoothly
to a positive peak at the patch centre. The maximum CMS does not occur in the
region of the patch load but at the base of the silo due to lever arm action of the
load on the silo wall. This maximum occurs on the zero meridian as can seen from
Fig. 5.
846
Fig. 12.
The meridional bending stress (MBS) can be seen in Figs. 5 and 6 to maintain a
value of around one-third of that of the CBS. The MBS arises mostly from the CBS,
producing stresses in the perpendicular direction by Poisson effects and so this close
relationship between the two may be expected. In a similar manner to the CBS, the
maximum MBS moves away from the region of the edge of the patch load to the
zero meridian as positions further from the patch load are considered (Figs. 812).
As the maximum MBS is less than the maximum CBS except the base of the silo,
it is not of great importance throughout most of the silo wall.
Figs. 5 and 812 also show that the membrane and twisting shear stresses are
generally low and always smaller than the largest of the direct stresses. This results
in the von Mises stresses being governed largely by the direct stresses and therefore
peaking in the same regions as these stresses.
847
Fig. 13. Maximum stresses against circumferential patch width. Uniform patch load applied at y/R=1
Fig. 14.
Maximum stresses against circumferential patch width. Uniform patch load applied at y/R=1.4.
pin-point the exact height of load application at which the greatest stress occurs but
the approximate height of y/R=2.05 indicated by Figs. 1315 is consistent with a
value of y/R=2.15 obtained by Rotters earlier study [2].
The abrupt changes in stresses at the edges of uniform patch loads leads to rather
curious curves when the maximum values are extracted for circumferential bending
with larger patch sizes (Figs. 1315) because the sign of the greatest bending changes
as the patch width is increased. However, if the magnitude of the CBS is considered,
it may be seen to follow a similar trend to the von Mises stresses. This suggests
that the most important term in Eq. (2) for these load cases is sqd.
The curves of the maximum compressive CMS rise from zero and peak at about
30 patch width (this peak is very weak for the loads applied at y/R=3.4). The peak
is explained by two opposing trends. As the patch width is increased the force acting
on the silo increases and this results in a greater CMS. On the other hand, the
increased patch width means the meridional force is distributed over a larger portion
848
Fig. 15.
Maximum stresses against circumferential patch width. Uniform patch load applied at y/R=3.4.
of the silo wall and the CMS is thus reduced. For patch loads wider than 30, the
CMS maintains a value that is largely independent of patch size or height of application.
5.2. Bell-shaped patch loads
The peak stresses for a bell-shaped pressure distribution are shown in Figs. 1618.
The shape of the maximum von Mises curves is similar for all three heights of
load application. Rising rapidly from zero the curves peak quite sharply at around
1520 patch size. Beyond this peak the curve reduces asymptotically towards zero.
The peak in the curve can be understood by considering the two ways in which the
silo responds to loading. Initially the response is as if the silo is subject to a point
loadincreasing the size of the load is, in effect, increasing the magnitude of a point
load. Above around 20, however, the silo begins to respond as if subject to a distrib-
Fig. 16.
Maximum stresses against circumferential patch width. Bell-shaped patch load applied at y/R=1.
849
Fig. 17. Maximum stresses against circumferential patch width. Bell-shaped patch load applied at
y/R=1.4.
Fig. 18. Maximum stresses against circumferential patch width. Bell-shaped patch load applied at
y/R=3.4.
uted load. The bending starts to take place over a large circumferential width and
so the CBS (which is the dominant term in the von Mises stresses) is not as large.
At very large patch widths the response is approaching what would be observed with
uniform internal pressure, that is no circumferential bending.
The other stresses behave in a similar way to when a uniform load is applied. As
for the uniform load, the maximum value of CBS rises and falls as the point of
application of the load is changed, with a peak occurring when a load is applied at
around y/R=2.05 dimensionless height. The CMS displays a weak peak that occurs
at somewhat larger patch sizes than for uniform loads. This peak occurs as larger
patch width loads are applied higher up the silo wall.
850
Fig. 19.
Fig. 20. Stresses against vertical extent of patch. Uniform patch load.
851
the bottom of the load becoming proportionally nearer to the base of the silo as the
vertical extent is increased, resulting in an asymmetry that disrupts the otherwise
simple behaviour.
7. Conclusions
The results presented in this paper show that a patch load has the potential to
produce significant von Mises stresses and compressive membrane stresses within
the wall of a steel silo. This suggests that patch loads could cause structural failure
by either plastic collapse or elastic buckling. Clearly more information is needed
about effects of size, position and form of patch loads on the stresses produced
in silos.
In addition, the following general points can be concluded from the results of
the investigation:
The response of a thin-walled steel silo to a patch load is complex. As well as
the most important stresses mentioned above, there are many lesser effects and
many interactions between the various stresses.
The form of the pressure distribution in the patch load is important. Different
pressure distributions lead to different variations in stresses and so to different
requirements for silo wall thickness
The circumferential width of the patch load has a strong effect on the stresses
within the silo. Broadly speaking, two regimes can be identified: a narrow patch
load to which the silo responds as if it were a point load and a wide patch load
to which the silo responds as if it were a distributed load. The dividing line
between theses two regimes is not clear-cut and there is some interaction for
medium sized loads. The different behaviours can be seen in Fig. 6.
For wide uniform patch loads the maximum values of the bending stresses set
up in the silo are largely independent of the loads width and point of application.
Stresses in the silo increase approximately in proportion to the vertical extent of
the patch load.
References
[1] Chen JF, Rotter JM, Ooi JY. Statistical inference of unsymmetrical silo pressures from comprehensive wall strain measurements. Thin Walled Struct. 1998;31:11736.
[2] Rotter JM. Patch loads and their use in metal silo design. Research Report R96-13, Department of
Civil and Environmental Engineering, University of Edinburgh; 1996.
[3] Ooi JY, Rotter JM, Pham L. Systematic and random features of measured pressures on full-scale
silo walls. Engng Struct. 1990;12(2):7487.
[4] DIN 18800. Stahlbauten: stabilita tsfalle, schalenbeulen, DIN 18800 Part 4 Deutsches Institut fu r
Normung. Berlin; November 1990.
[5] Eurocode 1: Basis of design and actions on structuresPart 4: Actions in silos and tanks. CEN:
Brussels; 1996.
852
[6] Eurocode 3: Design of steel structuresPart 4-1: Silos. Brussels: CEN; 1999.
[7] Rotter JM, Pham L, Nielsen J. On the specification of loads for the structural design of bins and
silos. In: Proceedings of Second International Conference on Bulk Materials Storage Handling and
Transportation; 1986 July; Institution of Engineers, Australia, Wollongong; 1986. p. 2417.
[8] Rotter JM. The FELASH suite for the finite element analysis of thin axisymmetric shells. In Proceedings of the Fourth International Conference on Civil and Structural Engineering Computing; London;
1989. p. 3238.
[9] DIN 1005. Design loads for buildings. DIN 1005 Part 6, Deutsches Insitut fur Normung. Berlin;
May 1987.
[10] Eurocode 3: Design of steel structuresPart 1-6: Strength and stability of shell structures. Brussels:
CEN; ENV 1993, draft 1997.
[11] Rotter JM. Bending theory of shells for bins and silos. Trans. Mech. Engng, Instit. Engnrs, Australia
1987;ME12(3):14759.
[12] Rotter JM. Flow and pressures in silo structural integrity assessments. In: Proceedings of the International Symposium: Reliable Flow of Particulate Solids III; Aug 1999; Porsgrunn, Norway; 1999.
p. 28192.
[13] Teng JG, Rotter JM. Plastic collapse of restrained steel silo hoppers. J. Construct. Steel Res.
1989;14:13958.
[14] Save MA, Massonnet CE. Plastic analysis and design of plates shells and disks. North-Holland, 1972.