Vous êtes sur la page 1sur 11

Elastic predictions of

pressures in conical silo


hoppers
J. Y. Ooi and J. M. Rotter
Department of Ovil Engineering, University of Edinburgh, UK
(Received to be supplied)

Theoretical techniques for predicting the pressures on the wails of


conical silo hoppers have generally depended on the assumption that
the mass of material within the hopper is in a plastic state of stress.
The fact that this imposes considerable restrictions on the stress
history and strain state has almost always been ignored. Moreover,
few investigators have undertaken checks to verify the assumption.
In this study, the alternative simple assumption is made that the
material stored within the hopper is in an elastic state. The study is
heuristic in character, as it is not claimed that an elastic state =does
exist within the hopper, though the assumption is not unreasonable
for the initial filling condition. Nevertheless, a number of interesting
discoveries are made about the stress distribution within the stored
mass, and concerning the influence of various p a r a ~ t e r s on the hop-
per wall pressures. The analysis is conducted using a finite e l e ~ n t
analysis which includes the effects of hopper wall friction and hopper
wall flexibility, and these influences are investigated.

Keywords: bulk solids, hoppers, loads, pressures, silos, theory

Many attempts have been made to produce theories weight, and stored in a conical hopper with frictional
which delrme the distribution of pressure on the walls of sliding contact on the wall. The findings of this study
conical silo hoppers. The majority of these have adopted form an important but simple result, which can be used
the assumption that the mass of material within the hop- as a reference when the predictions of more complex
per is in a plastic state of stress. This assumption is so models are being examined. It should also be noted that
regularly made that it is sometimes not even stated, but the assumption of linear elastic stress states within the
the use of failure properties, such as the angle of internal hopper has been made before 2, but additional assump-
friction, indicates that the assumption is being made. tions were always made, so that the calculations were
The fact that this plastic assumption imposes con- not rigorous.
siderable restrictions on the stress history and deforma- The present analysis is conducted using a finite ele-
tions of the mass has almost always been ignored. merit analysis which includes the effects of hopper wall
Moreover, few investigators have undertaken checks to friction and hopper wall flexibility, and these influences
verify the assumption. The properties of dry granular are investigated.
solids under stress states which do not involve failure It is claimed that the findings of the study have
have also been explored only rarely and incidentally. relevance to hopper pressures for a number of reasons.
These are poor reasons for assuming that the material Firstly, the study clarifies the pattern of pressures which
within the hopper is in a plastic state, especially for the can be expected when a very different constitutive model
initial filling condition, which has been shown to be the from the plastic assumption is used. This suggests that
critical state for the structural design of mass flow steel the major feature of known pressure distributions is not
hoppers ~. the assumption of a plastic state, but the predominance
In this study, the alternative simple assumption is of equilibrium requirements on the basic form of the
made that the material stored within the hopper is in an distribution which arises from an assuml~on of unifor-
elastic state. The study is hcm'istic in character. It is not mity for the solid.
claimed that an elastic state does exist within the hopper, Second, it is shown that if gravity is applied to the
or that the initial strains of different parts of the solid are whole elastic mass at once, the initial filling cot~tion is
unimportant. The study is undertaken to explore the far from plastic. This same conclusion could be drawn
stress states which occur in a homogeneous linear elastic from the work of Waiters 3, in which an atmrapt to use
mass of material without initial strains, subject to self a plastic assumption for the filling condition led to

0141-0296/91/01002-11
© 1991 Butterworth-Heinemann Ltd
2 Eng. Struct. 1991, Vol 13, January
Elastic predictions of pressures in~onical~si~t~oppers: J. Y. Ooi and J. M. Rotter

pressures which were clearly wrong 4. By contras~;41~ R .......


widely-used initial filling theory of Walker 5 assumed
only that the principal stresses were in the vertical and Ozt
horizontal directions, together with sliding on the hop-
per surface, and frustum slice equilibrium. The first of
these assumptions leads to an elastic state of stress for
all but very rough steep hoppers. Stored
Third, if it is accepted that the initial ffiling state is not Solid,
plastic, then theories like those of Walker s , or the more Density
general version of M c ~ 6 might be adopted. Both of
these imply an elastic state of stress, but both make the
assumption that the ratio of normal wall pressure to
mean vertical pressure is constant throughout the
hopper. The analysis is thus not rigorous, since this
assumption is not open to verification. In the present
investigation, the uniformity of the lateral pressure ratio
will also be examined.
Fourth, the critical pressure distribution for hopper
structural design is generally initial filling. Under these
Conditions, the elastic solution for hopper pressures pro-
vides a useful comparison with the results of ex-
periments and a useful alternative to Walker's 5 theory, Z
which is commonly used for this purpose. It shows that
a rigoro-_s satisfaction of equilibrium throughout the
stored mass leads to slightly different pressure distribu-
_i_
tions from those of the simple theory.
(a)
Finite element formulation
The conical hopper and the contained stored solid are
each treated as homogeneous isotropic linear elastic
materials without initial strains. This very elementary
and certainly invalid assumption is for heuristic pur-
poses only. The stored solid and the silo wall are
modelled as axisymmetric bodies using the finite ele-
ment method 7. The twelve noded cubic isoparametric
element with nine point Gaussian integration is used.
A contact element is introduced between the stored
solid and the hopper wall to model the wall friction
characteristic. It allows frictional sliding to occur only
tangentially along the surface between the stored solid
and the wall. This finite element has been described fully
elsewhere s. The calculation is nonlinear and is con-
ducted iterafively until the frictional shear at every point
is less than or within a close tolerance (typically 1%) of
the maximum sustainable friction.
The characteristic geometry and a typical mesh
representing a radial section through the hopper are
shown in Figure Ia and Ib.

Classical tlumries on hopper p r e s s u r e s


Many theories have been proposed for the distribution of
pressures in conical silo hoppers. A brief description is
given here of the more commonly quoted theories for
hopper pressures and their underlying assumptions.
More complete descriptions may be found elsewhere 4,9.

Walker theory ~
The differential slice method, originally used by
Janssen l° for .cylindrical silos, was also adopted by
Walker s, Waiters 3 and Enstad H to derive theories to (b)
describe presmte distributions in conical hoppers.
Figure I Characteristic geometry, a, hopper geometry; b, typical
Under static or initial filling conditions, Walker 5 finite element mesh

Eng. Struct. 1991, Vol. 13, January 3


Elastic predictions of pressures in conical silo hoppers: J. Y. Ooi and J. M. Rotter

assumed that the major principal stress is vertical. This Generalized version of/Walker theory
assumption leads to a hydrostatic distribution of vertical
pressures The two above theories of Walker 5 for filling and flow
appear to be very different. It also appears that the filling
a: = 3,(H - z) + a,, (1) condition is independent of the failure properties of the
material, whilst that for flow appears to depend on the
in which az is the mean vertical stress at height z above angle of internal friction. This difference clouds the
the hopper apex, H is the hopper height and 3' is the truth that the filling theory is a limiting case of the flow
stored solid density. The mean vertical stress in the theory, and that the relation for pressures derived for
stored solid at the cylinder/hopper transition is given by flow conditions has more general application.
If it is assumed that the ratio of wall pressure p to
OZI .
Assuming that the wall friction is fully mobilized, the mean vertical stress az is constant at a single value of
normal wall pressure distribution is given by FD throughout the hopper, then equilibrium of the
material in the hopper leads to equation (4). If the total
equilibrium of the hopper is also considered, then it is
Pi =Fioz (2)
found that the pressures on the hopper wall must satisfy
in which the condition
n = 2 { FDI~otB + FD - I } (lO)
F,- (3)
1 + #cotB Equations (5) and (6) are then no longer needed, and the
only assumptions are equilibrium and the value of the
Here, /z is the coefficient of wall friction and/3 is the wall pressure ratio, FD. The latter may be derived as by
hopper half angle (i.e. the angle between the hopper wall Walker's flow theory (equations (7) and (9)) for the
and the vertical). This stress distribution in the solid is material in a state of plastic failure (thus involving the
often referred to as a 'peaked' stress field. angle of internal friction), or by assuming that no shear
Under flow conditions, Walker 5 examined a cylin- occurs on vertical planes within the solid, which leads to
drical horizontal elemental slice and adopted the equation (3), which in turn gives n = 0 through equation
assumption that material adjacent to the hopper wall is (10), leading to the Walker filling solution of equa-
at passive failure. This leads to tions (1) and (2). Other means may be employed to
obtain relations for FD, so these are described here as
modifications of Walker's theory.
a. - - + azt (4)
" n-1 Waiters theory and other modifications o f Walker's
treatment
in which
Waiters 3 extended Walker's analysis to allow for the
non,uniformity of the vertical stresses on a horizontal
2BD slice of the stored solid, assuming that the entire mass
n = - - (5)
tan B is in a state of plastic failure. He assumed that an active
stress state pertains on initial filling and a passive stress
state during flow. To simplify the equations, it was also
B= sin ~ sin(2B + 2ey) (6) necessary to assume that the vertical shear varies
1 - sin ~ cos(2B + 2~/) linearly with radius. These assumptions lead to the
Walker equations (4), (9) and (t0) but With D given by

2e/= ---r + tan-'# - c ° s - l ( - # ) (7) D = cos~(1 + sin2~o) 4- 2x/(sin2~o - sin2~) (11)


2 sin ~x/(1 + #2) cos~[(1 + sin2~o) 4- 2ysin~o]

in which B is the ratio of vertical shear stress to vertical in which


direct stress at the wall. The distribution factor D, which
is the ratio of the vertical stress at the wall to the mean 2
vertical stress at any given level, is assumed by Walker Y = 3cc (1 - (1 - c) 3/2) (12)
to be unity. The predicted normal wall pressure is found
as
tan T/~2
c= (13)
py = F:Do~ (8) \tan~/

in which
~7=tan_t[ sin(2e+2B)sin~ ] (14)
F: = 1 + sin ~ cos 2~: (9) 1 + cos(2e + 2fl) sin ,p
1 - sin ~ cos(2B + 2e/)

The resulting stress distribution is often referred to as an


'arched' stress field.
2E = ~
* + = - , . + cos'e: ,"--
|\sin~q(1 + #2)
) (15)

4 Eng. Struct. 1991, Vol 13, January


Elastic predictions of pressures M conical silo hoppers: J. Y. Ooi and J. M. Rotter

The positive signs above refer to the initial , ~ g ~ s u r e s m hoppers by considering the equilibrium of a
(active) condition and the negative sign~ to the fl~W fall[tufa Of the hopper bounded by two non-concentric
(passive) state. The angle e for the passive condition is circular arcs. The stresses within the arch are assumed
the same as ~( in Walker theory. to vary only in direction, not in magnitude. Expressed
If Waiters Itheory is expressed instead in terms of the in terms of the mean stress in the arch a, Enstad's solu-
parameter B, then it should be noted that equation (5) tion for pressures during flow is given by
must be adjusted to
"rYs + a, - (20)
2( BD ) (16) ° = X- - 1 I X-1
n= ~tan/3 + D - 1
in which s is the meridional distance up the cone from
It has been noted 4 that Waiters' solution yields the apex, s, is the meridional distance measured to the
unrealistically high stresses for the initial filling condi- top of the hopper and at is the mean vertical arch stress
tion unless the hopper half angle is very small: that is, acting at the transition due to cylindrical surcharge. The
unless terms X and Y are given by

1
/3 < -=- (T - 12e;) (17) 2 sin ~ 1 + sin(2ef +/3) ] (21)
2 X = 1 - sin ~0 sinB

in which ei is the value of e for initial filling. This y=


limitation implies that the active solution is only useful
for very steep hoppers. The passive solution, which is 2 sin/3 [ 1 - cos(eI +/3)] + sine/sin2(ef +/3)
assumed to occur during discharge, is identical to that of (1 - sin ~o)sin3(e/+
Walker if the distribution factor D is taken as 1.0.
Unfortunately, the assumption that the shear stress on (22)
any horizontal plane increases linearly with radius can
be shown n to be invalid if the entire mass is in a in which e! is given by the passive version of equation
passive state of failure and the hopper half angle/3 is (15). The normal wall pressures on the hopper can be
greater than expressed in terms of the mean arch stress a as in equa-
tion (2).

/3~ = ~ - tan-l/~ + c°s-~ sin~x/(1 + #2) p = (1 + sin ~ocos 2¢f)a (23)

k should be noted that Enstad's analytical expression is


(18) for a cylinder-hopper combination. For a hopper without
a cylindrical surcharge, it has the uncertain disadvantage
To overcome this limitation, Home and Neddermn._n12 that the top boundary surface has an unrealistic spherical
suggested that, for hoppers shallower than /3m,x, the form.
value of D should be taken as the value for a bopper of
half angle/3u"
More recently, McLean ~ suggested that equation (4), Jenike and Johanson theory
which Walker derived as the general solution of the
hopper equifibrium equation when the lateral pressure The most commonly quoted theory for the pressure
ratio is comnant, should be used to represent the pres- distribution in a hopper is probably that proposed by
sure distribotion for initial filling. Based on the Jenike Jenike and his co-workers 2'~-~6. Based on the assump-
and Johamon 2 radial stress field solution, he recom- tion that the pressure distribution near the apex of the
mended that a value of the parameter n should be hopper is in the form of a radial stress field, Jenike et
deduced as al solved the equilibrium equations without considering
the top boundary condition.
For a hopper without a cylindrical silo surcharge
2/~ above it, a triangular pressure distribution is assumed
n = ~ (19)
tan/3 such that the radial stress field is cut off by a second
linear variation of stress with magr.itudes decreasing
which corresponds to FD = 1. He suggested that n towards the free surface at the hopper top. For the static
should be about 3 for typical conical hoppers, though (or initial filling) condition, this radial stress field
this simplification is not adopted here. The value of n depends on the hopper half angle/3, the wall friction
given by equation (19) differs significantly from the coefficient ~t and the elastic properties of the stored
value n = 0 which leads to m e Walker filling equations solid. For the flow (or discharge) condition, the radial
(equations ( 1 ) - ( 3 ) ) . stress field depends on the hopper half angle/3, the wall
friction c6efficient ~, and the effective angle of internal
Enstad theory friction of the stored solid ~,. Vertical equilibrium of the
hopper requires that the triangular pressure distribution
Enstad II derived another approximate theory for has a unique peak pressure, pp.

Eng. Struct. 1991, Vol. 13, January 5


Elastic predictions of pressures in conical silo hoppers: J. Y. Ooi and J. M. Rotter

3,R
Pp = (24)
(1 + z/H)(tan/3 +/~)

in which the peak pressure is taken to occur at height z,


and the whole hopper is of height H.

More recent treatments


Other methods of analysis of stresses in hoppers have
also been used. The method of characteristics17 is one
way of eliminating assumptions concerning the stress
distribution. It was used by Johanson 13, Home and
Nedderman 12 and Wilms TM. Arnold et al 4'19"2° have pro-
posed modified versions of Jenike's treatment which are
easier to use. Finite element methods have also been
developed to predict the hopper pressure distribution.
Haussler and Eibl 2~ and Link and Elwi 22 have described
finite element analyses of mass flow hoppers during
discharge, using complex nonlinear material characteri-
zations.
By contrast, this paper presents a relatively simple
finite element analysis in which the wall and the stored t3
solid are both treated as linear elastic materials, con- ,w=0.3
nected by a frictional sliding contact surface.

Comparison for a typical mass flow rigid


hopper after initial tilting
Example hopper
An analysis of a typical rigid mass flow hopper with
gravity loading alone (no cylindrical surcharge) is
presented here first. The hopper wall was built-in at the
top and a horizontal top surface was used for the stored
solid.
Young's modulus for the hopper wall was taken as
2.0 x 105 MPa and that for the stored solid as 50 MPa, Figure2 Yielded zones for~ = 30 ° and • = 40 °
with a Poisson's ratio of 0.3 for both materials. A wall
friction coefficient of 0.3 was adopted. The radius at the
top of the hopper R was related to the wall thickness t,
as R/t = 100. This rather thick wall ensures that wall
The predicted principal stress field is shown in Figure
deformation is minimal and the case resembles a rigid
3. Near the apex, the major principal stress trajectories
hopper. These values of the governing parameters are are close to being vertical and approximate the pattern
taken as the reference values in the later parametric of a 'peaked' stress field, as suggested by most writers.
study. However, the present elastic analysis also suggests that
there is sufficient eonstdidation during filling for the
Elastic and plastic stress states major principal stress trajectories to approach an
Some existing theories 3'23 are based on the assumption 'arched' stress field in much of the hopper.
that a state of active failure pertains on initial filling, Normally this would only be expected during
with the major principal stress trajectories approxi- flow 2'3'5. Thus, the present finding is in contrast to the
mately vertical. The hypothesis that the material is at assumption used by most previous theoretical studies.
failure was investigated first in the present study using Nevertheless, there is considerable exper~nental
a simple Mohr-Coulomb failure criterion. evidence for pressure distributions of the form predicted
The zones in the stored solid in which the stresses are here occurring on initial filling 25'~. These experiments
predicted by the present analysis to exceed the yield lend credence to the present analysis. Thus, existing
criterion are shown in Figure 2. Two different values of solutions which characterize the entire bulk solid as in
the effective angle of internal friction ~ (30 ° and 40 °) an active or a plastic state (e.g. Walker's flow treatment,
are used. Only a very small portion of the stored solid Walt=rs' analyses), or with a hydrostatic vertical stress
is found to exceed the yidd condition, especially if the field (Walker's static treatmemt) cannot be expected to
effective angle of internal friction is large. It should be give accurate predictions of the wall pressures for initial
noted that most bulk solids have an effective angle of in- filling.
temal friction in excess of 30 .24. Thus the present The phenomenon of a changing stress field was also
analysis is seen to be valid for the larger part of the investigated by examining the ratio of wall pressure to
stored solid. mean vertical stress, FD, at different heights in the

6 Eng. Struct. 1991, Vol 13, January


Elastic predictions of pressures in conical sik) hoppers: J. Y. Ooi and J. M. Rotter

0.8

J: 0 6 -

"E
==

02
X

,I t , t
0.6 O.B 1.0 1.2 1/, 1.6
Latera( Pressure Ratio (p/o=)

Figure 4 Lateral pressure ratio distribution for the example


hopper

1.0

Es=SOMPavs=0.3 0.8
Ew=2xlOSMPa vw=0.3
R/t=IO0 o Normal WaU Pressures '~
~=0.3 £ O.6
~ a
in Eontact Etement
U Pressures deduced ~o
=20 °

c5

0.2

I I
0.2 O~ 0.6 0,8
Dimensionless Normal Wall Pressure (P/1R)

Figure 5 Comparison of normal well pressures

Figure 3 Principal stress fields

hopper. The result is presented in Figure 4. Near the The pressure distributions predicted by the present finite
bottom, the value of FD is as low as 0.87. It then rises element formulation appear to be less susceptible to
steadily for much of the hopper, reaching 1.0 at mid- spurious variations within each element than those of
height, but increasing rapidly towards the top, where similar studies (Compare, for example pressure distribu-
1.6 is achieved. This variation compares with the tions calculated by Mahmoud and Abdel Sayed 27, and
constant values adopted in applications of Waiker Link and Elwi22).
theory: FD ffi 0.55 for Walker filling theory, 1.29 for The wall pressure distribution obtained from the pre-
Walker flow theory, 2.07 for Waiters flow theory, and sent analysis is compared in Figure 6 with several pre-
1.0 for McLean's proposai. vious solutions, assuming the effective angle of internal
friction ~, is 30 °. The present predictions lie near
Pressure distributions McLean's theory near the hopper bottom, but closer to
Walker's flow theory towards the hopper top. None of
In Figure 5 normal wail pressures deduced from the wail the existing analytical solutions provides a really close
membrane stresses using the membrane theory of shells fit to the elastic prediction, but the distributions are very
are compared with the normal wail pressures in the con- similar in form. This same pattern can be observed for
tact elements. The two pre~*.,lre distributions should be a wide range of hopper angles, wail frictions and hopper
shr,ilar except aear the top boundary where shell bend- flexibilities.
ing phemmmm affect the wall deformations and wall Towards the hopper top (zlH > 0.9), the present
s~, which in turn affect the derived wall pressures. results are quite close to the Walker and Waiters flow
The close match between the two pressure distributions theories. This should be expected, as the stress field in
indicates that equilibrium is being satisfied properly. the stored solid is not far from the 'arched' form

Eng. Struct. 1991, Vol. 13, January 7


Elastic predictions of pressures in conical silo hoppers: d. Y. Ooi and d. M. Rotter

I0
'l ~ -----7-"~--'--~, '
~ ,_~ - ¢:30 o
I

013 --~x ...... WalkerSt~t,~


~ , . " % \ ", ',.~ " W~tk~rF*o~
~o~\ ', ] .... Waiters Flow
/ ~ /
i,,.. '., .... Enstad
~= 06

./.'I/~'x. ~L0~US0~.~,,,k~s ..szs-" i


"E " P. P.o . . . . .
/ \-.
,'/ /? ,, \\ .,,
E ///'/ ",....
c3
- . - o - - Present Study l
// ,.
02
1~"/ / ---- w,,te,s Fi0~ i

,¢"~ I I i 1 I
02 0/, 06 08 10 12 14 01 0.2 0.3 04 05 06
Olmens=onless Normal Wall Pressure (p/yR) Dlmensionless Ctrcumferentla[ Stress (NB/IcR2)

Figure 6 Wall pressures for a typical rigid hopper Figure 7 Circumferential wall stresses in a typical rigid hopper

10
throughout much of the hopper (Figure 3), though it is I

far from fully plastic. Presenr Study t/~


" Walker Stahc i"s;~O~ \
Jenike and Johanson's 2 triangular pressure distribu- Watker Flow .,i//~// ~.
08
tion is also shown in Figure 6 for comparison. The peak .... w=t,,~ Flow .~.'-'/~./ ]
pressure in this treatment is dependent on the wall fric- - - - - - M:Le~o ..--" <j'.~" / ]
I
\
tion coefficient #, the hopper angle/9 and the elastic pro- N

=06
perties of the stored solid. The locus of peak pressures
is marked in Figure 6 to indicate the range of possible
distributions. Two alternative values of Jenike and
o O~
Johanson's parameter k are used here to indicate alter-
native distributions which might be deduced from their
analysis based on values they propose. It is interesting to
note that the radial stress field elastic solution using 02

k = 0.9 is closely asymptotic to the present elastic pre-


diction. However, only at the very bottom of the hopper
is this radial stress field valid. If the maximum pressure 02 OL 06 08 10 t2
is chosen to occur at the bottom of the hopper, Jenike Dimensionless Mer~dmnat Stress (6N®sinl3/l(Rz)

and Johanson's solution matches Walker theory exactly Figure 8 Meridional wall stresses in a typical rigid hopper
for initial filling conditions.
Waiters' solution for the initial filling condition is not
included here because the hopper half angle B exceeds
the limiting value give by equation (17), with the im-
plication that the vertical stresses in the solid exceed brane stress signifmantly. This effect is seen in the pre-
hydrostatic values, which seems unlikely. Enstad's solu- sent analysis, but cannot be modelled by classical lmoper
tion predicts slightly larger wall pressures because it pressure theories. Shell bending effects on the meri-
assumes a spherical top surface, dional stress are minimal, as expected 2s.
The meridional stress depends on both the normal
Wall stress distributions pressures and the frictional tractions. At the top of the
hopper all theories must give the same value for the
The normal pressure and the frictional shear exerted by meridional stress, as it is governed by the static global
the stored solid on the wall generate a meridional stress equilibrium of the hopper and its contents. The bopper
and a circumferential stress in the hopper wall. Indeed, top meridional stress is given by
the purpose of predicting hopper pressures is primarily
to predict hopper wall stresses for structural design. It ,yR2
is therefore useful to examine the wall stresses, and to N., - - - (25)
compare the stresses predicted by different theories. 6 sin B
The wall stresses are shown in Figures 7 and 8. Results
are shown for two different effective angles of internal Also shown in Figures 7 and 8 are the corresponding
friction @ (30 ° and 40°). A= _;~expected from the mem- wall str~s predictions of t h e Walker, WaRers and
brane theory of shells, the hopper circumferential stress McLean theories coupled with the ~ theory of
distribmion peaks further from the apex than the n o ~ shells. It should be noted that shell membrane theory
wall pressure distribution (Figure 6). Near the top of the accurately reflects the real ~ in the body of the
hopper, .where a structural discontinuity exists, shell hopper, but the support ~ ~ adjacent to
bending phenomena influence the circumferential mere- the hopper topL Walker s static solution does not

8 Eng. Struct. 1991, Vol 13, January


Elastic predictions of pressures in conical .silo!hoppers: J. Y. Ooi and J. M. Rotter

change with the internal friction angle. The two flow


theories and McLean lead to patterns of hopper wall
stress which are very similar to that of the present
analysis, though, for the flow theories, the differences
increase With increasing internal friction.
It is clear that the differences between the predicted {
hopper pressure patterns (Figure 6) are larger than the ~ 0.6 - .=0.5 0.1
differences between the resulting wall stresses (Figures
7 and 8). The question of the 'right' hopper pressure
distribution should therefore not be treated as too serious .~ 06 //'" / /
a matter. Walker flow theory and McLean give the
closest comparison with the present analysis for both
hopper wall stress distributions. 0.2 - j .~,~"" Walker Flow IO.30°1
~NMcLean
-,,--<>--Present Study
Parametric study of mass flow hoppers I
02 O~ 06 08
Dimonsionless CircumferentiaStress
l (Ne/]R2)
The reference hopper parameters
Figure 10 Circumferential w e l l stresses for v a r y i n g wall friction
In the following parametric study of mass flow hoppers coefficients
without cylindrical surcharge, the effects of varying
the wall friction coefficient /~, the hopper half angle
~, the bulk solid elastic modulus E~ and Poisson's
ratio J,~ are explored. The hopper analysed in the
previous section of this paper is taken as a reference ex-
ample in this parametric study. Pressures from Walker 1.0 i i i i "~
flow theory (~, = 30 °) and McLean are also plotted for m
I II We,ker F,ow)..30.1
comparison. Ilk
----.cL.n iF/i
0.8 -..¢,,-,pzO.1 (Present Study) j~.[ I
Effect of the wall friction coefficient --o.-,p=0.3 (Present Studyl ~ / #
,.,..,,o--.p=O.5 )Present Study) ~ / / / /
The coefficient of wall friction /~ was varied to in-
__. 0.6
vestigate its effect on hopper pressures. The dimen-
sionless nornud wall pressure is shown in Figure 9 for ..0.3 %,/
three different values of the wall friction coefficient.
The present elastic analysis predicts that the normal wall
pressures decrease throughout the hopper as the wall
friction coefficient increases. This effect is also seen in
0.2
the Walker and McLean theories.
The maximum normal wall pressure is found to occur
closer to the bopper bottom as the wall friction coeffi- m J t ~ I
cient is reduced. For very smooth hoppers (~, = 0.1), the 02 0t,. 0.6 0.8 1.0 12
maximum pressure is near the bottom, giving a pressure Dimensionless Meridional Stress (6Nesin[~/'fR2)
distribution more like that of Walker filling theory. Figure 11 Meridional wall stresses for v a r y i n g w a l l friciton
McLean's theory moves the locus of this maximum in a coefficients

1.0 I I ~ B ~ / ~ I I ~ I ~ I I ~ R , , . . . ~ I i ' " Walker Flowllo=30 °)


similar manner, though the peak value is different. By
~,~lll~l~C&~ ~ ------ McLean
contrast, Walker flow theory suggests that the locus of
the maximum rises when the wall friction is reduced.
The corresponding dimensionless circumferential and
mcridion~ wall stress resultants are shown in Figures I0
and 11. These stress resultants are important because
they are the real goal of hopper pressure predictions.
'%%%% The Circumferential stresses (Figure 10) are well related
to the pattern of the normal wall pressures, in the
manner expected from shell membrane theory. With in-

0.4
I lil
/ J -
':
,' _
creasing wall friction, the meridional stress (Figure 11)
is predicted to decrease throughout most of the hopper
except near the top where a slight increase occurs.
By contrast with this finding, Walker flow theory
predicts that the meridional stress should increase
0.2 06 0.6 0.6 'tO 1.2 1~ throughout the hopper as the wall friction coefficient in-
Dtmensionless Norm=t Wall Pressure (p/lfRI
creases. Waiters flow theory matches the present finding
Figure 9 Normal wall pressures for varying wall friction in this matter, even though these two classical theories
coefficients only differ in the distribution factor D.

Eng. Struct. 1991, Vol. 13, January 9


Elastic predictions of pressures in conical silo hoppers: J. Y. Ooi and J. M. Rotter

10

=06 .. ..... .
McLean z~
~Present Study \
o

~04 7~/// c
c3

02

, I I
02 Or, 06 08 02 O~ 06 08
Dimensionless Normal Wail Pressure (p/l~R) Dimensionless Normal Watl Pressure (p/l"Ri

Figure 12 N o r m a l w a l l p r e s s u r e s f o r v a r y i n g bulk s o l i d P o i s s o n ' s Figure 13 Normal wall pressures for varying bulk solid-wall
ratio stiffnesses

Effect of Poisson's ratio of stored solid parameter


The Poisson's ratio of the stored solid might be expected
to affect the hopper pressures. Dimensionless normal E~ R
o~ - - (27)
wall pressure distributions are presented in Figure 12 for
-

Ew t
varying values of the bulk solid Poisson's ratio ~,s. The
effect of Poisson's ratio is seen to be very small, but the
This has been widely verified, and is indicated in Figure
maximum normal wall pressure increases slightly as
13 by the three results shown for c~ = 0.25, which
Poisson's ratio increases, and the normal wall pressures
derive from the three combinations E, = 500 MPa,
near the top and near the bottom of the hopper decrease. E w = 2 × 1 0 s M P a , R/t=lO0; E , = 5 0 M P a , Ew=
McLean's prediction again gives a good fit to the present 2x10SMPa, RIt=lO00 and E , = 5 0 M P a , E,,=
results, giving the appearance of being valid for 1, = 0.5. 2 x 10 4 MPa, /tlt = 100. As the value of (z increases
from zero, very little change in the calculated pressures
Effect of stored solid and hopper stiffnesses occurs until ~ reaches about 0.1. Thereafter, the peak of
the distribution begins to occur at lower points in the
Jenike and Johanson 2 suggested that the stored solid hopper, and to increase very slightly. Pressures near the
compressibility should alter the pressure distribution top and bottom of the hopper (z/H < 0.4, z/H > 0.8)
significantly. McLean 6 also believed this, and added correspondingly change slightly.
that a flexible hopper wall would have the same result. These changes in the pressure distribution occur only
The effect of the relative stiffness of the stored solid and in very thin-walled hoppers containing very stiff solids.
the hopper wall were explored by varying the modular The highest practical value of o~is found for a very thin
ratio of the stored solid and the hopper wall EJE,.. steel hopper for which R/t might reach 1000. For c~ to
Both a thick-walled hopper (R/t = 100) and a thin- be as high as 0.1, the bulk solid elastic modulus must
walled hopper (R/t = 1000) were examined so that both then exceed 20 MPa. Adopting equation (26) with a
stiff and flexible hoppers could be investigated. value of ~v = 100, the vertical stress must exceed
There is little experimental data or agreement on 200 kPa for the effects of bulk solid compressibility and
appropriate precise values for the elastic modulus for hopper flexibility to make any difference to the wall
most stored solids. However, it has been suggested 23"29 pressure distribution. Such a b.igh value of the required
that the elastic modulus of the stored solid Es may be vertical stress in a stiff solid with a very light hopper in-
estimated as dicates that the effects of material compressibility and
hopper wall flexibility should be negligible in virtually
E, = Kva~ (26) all practical situations.
Another conclusion may also be drawn from the insen-
in which Kv is termed the modulus contiguity sitivity of the pressures to the assumed stored solid
coefficient 29, and is a constant varying between about modulus. It indicates that equilibrium (and not kine-
70 for dry grains and about 100 for loose dry sand but matic) considerations probably dornirmte the pattern of
up to 200 for dense-packed hard granules. Here the initial fitlmg hopper pressures. The use of a nonlinear
practical range of the bulk solid modulus Es is taken as constitutive model for the bulk solid may therefore make
5 MPa to 500 MPa. little diffea-ence to the calculated hopper pressures.
The dimensionless normal wall pressure distributions
are shown in Figure 13. As would be expected from the Effect of hopper half angle
theory of elasticity, it was found that the wall pressure
distributions are almost identical for fixed values o f the Dimensionless normal wall pressure distributions for a

10 Eng. Struct. 1991, Vol 13, January


Elastic predictions of pressures in conical silo hoppers: J. Y. Ooi and J. M. Rotter

10 ~rs, the present work suggests that material failure


~ y l i O t ~ :~s~ntial in satisfactory hopper pressure
predictions for the initial filling state. However, it is not
08 suggested here that material failure properties are not
vital in predictions relating to bulk solids flow.
The parametric studies presented here give some in-
sight into the characteristics of these pressure distribu-
\ \ \% \ ~,\ "~/\./[ ,,' tions. An increase in the wall friction coefficient
\ \ \"~ k ~o~,?°" V / r .Z.,-" decreases the wall pressures. Changes in the assumed
= OL, elastic modulus and Poisson's ratio of the stored solid
E ~=60o~ ~ z have very minor effects on the wall pressures. In very
shallow hoppers, the maximum pressure tends to occur
02 towards the bottom. As the hopper is made steeper, the
maximum pressure occurs progressively higher up the
/ -- ~ ---~ McLean
hopper.
02 0, 06 0.8 ,0 More complex material characteristics can easily be
Dimensionless Normal Wall Pressure (p/'llR) introduced into the present model, but the present study
suggests that these may not alter the predictions substan-
Figure 14 Normal wall pressures for varying hopper half angles
tially in many practical cases.

Notation
range of hopper half angles # are shown in Figure 14. B ratio of vertical shear stress to vertical direct
In these calculations, the radius at the top of the hopper stress at the wall
has been kept constant, so that the hopper becomes D ratio of vertical stress at the wall to the mean
shallower as the hopper half angle increases. This results vertical stress in the solid
in a smaller total load in the hopper, and consequently Young's modulus of stored solid, hopper wall
reduced hopper pressures. F ratio of normal pressure to vertical stress at the
The present elastic predictions are remarkably similar wall
in form to those of the Walker flow and McLean theories k Jenike radial stress field parameter
for a wide range of hopper half angles. This indicates n pressure relation exponent
again that equilibrium considerations dominate the pat- N,,N circumferential, meridional membrane stress
tern of hopper pressures. resultant
P normal wall pressure
Conclusions R radius of hopper at the top surface (transition)
S meridional distance up the inclined wall
Theoretical predictions for the pressures on the walls of (origin at apex)
conical hoppers have been presented. These were t thickness of the hopper wall
obtained using a finite element analysis in which the X,Y terms used in Enstad's theory (equation (20))
stored solid was treated as a homogeneous linear elastic z vertical coordinate (origin at apex)
body with frictional sliding against the hopper wall. This
heuristic analysis was undertaken to explore the nature ot hopper-bulk solid relative stiffness parameter
of the pressure distributions arising from an elastic (equation (27))
assumption, but it has been shown to be able to give hopper half angle
many satisfactory predictions of wall pressures for the -y unit weight of stored solid
initial filling state. The initial filling state has previously Kv modulus contiguity coefficient (equation (26))
been shown to be the critical loading case for mass flow ~t wall friction coefficient
steel binsL Poisson's ratio for bulk solid, hopper wall
vs, Vw
Elastic pressure distributions were obtained which oz mean vertical stress in stored solid
relate closely to the modification of Walker theory by 0 mean stress in arch (Enstad's analysis)
McLean s, and slightly less closely to the flow theories effective angle of internal friction
of Walked and Waiters 3, although all these theories
have a very different basis from the present analysis. Subscripts
The pressure distributions were shown to be quite insen-
sitive to the assumed elastic properties of the solid, so f during flow
that uncertainty about the proper value for these i after initial filling or storing
parameters is seen to be unimportant. These considera- t at the cylinder/hopper transition
tions suggest that the form of hopper pressure distribu-
tions is dominated by equilibrium and the assumption of References
homogeneity, and that const;tutive laws may play a
lesser role. 1 Rotter, J. M. "On the significance of switch pressures at the transition
It was found that stress states in most of the stored in elevated steel bins', Prec. 2rid Int. Conf. on BJdk Marls Storage,
bulk solid are not near the Mohr-Coulomb failure sur- Ham~li~g and 7hmsportation, Inst. Eng. Aust. Woliongong, July
1986, lip 8 2 - 8 8
face. Although much of the literature suggests that 2 Jenike, A. W. and Johanson, J. R. 'Bins loads', J. Stract. Div., ASCE
material failure properties determine wall pressures in 1968, 94 (ST4) 1011-41

Eng. Struct. 1991, Vol. 13, January 11


Elastic predictions of pressures in conical silo hoppers: d. Y. Ooi and d. M. Rotter

3 Waiters, J. K. 'A theoretical analysis of stresses in axially-symmetric Series B 1973, 95 (1), 1 - 1 2


hoppes and bunkers', Chert Engng Sci. 1973, 28 (3) 7 7 9 - 8 9 17 Sokolovski, V. V. Statics of soil media, Butterworths, London, 1960
4 Arnold, P. C,, McLean, A. G. and Roberts, A. W. Bulk Solids- 18 Wilms, H. Spannungsberechnung in Silos mit der Charakeristiken-
Storage, Flow and Handling, TUNRA Bulk Solids Handling metbode, PhD Dissertation, Technical Univ. of Braunschweig, 1983
Research Associates, Univ. of Newcastle, N.S.W., 1978 19 Arnold, P. C. and McLean, A. G. 'An analytical solution for the
5 Walker, D. M. 'An approximate theory for pressures and arching in stress function at the wall of a converging channel', Powder Tech.
hoppers', Chem Engng Sci. 1966, 21, 9 7 5 - 9 7 1976, 13, 255-60
6 McLean, A. G. 'Initial stress fields in converging channels', Bulk 20 Arnold, P. C. and McLean. A. G. "Wall loads in stee] bins'. Steel
Solids Handling 1985, 5 (2) 4 9 - 5 4 Construction, 1978. 12 (31
7 Zienkiewicz, O. C. The finite element method, (3rd edn), McGraw- 21 Haussler, U. and Eibl. J 'Numerical investigations on discharging
Hill, New York, 1977 silos' J Engng Mech.. ASCE 1984, 110 (EM6), 957-71
8 0 o i , J. Y. and Rotter, J. M. 'Wall pressures in squat steel silos from 22 Link, R. A. and Elwi, A. E. "Incipient flow in silos: A numerical ap-
simple finite element analysis', Res Rep., R538, School of Civil and i proach', Structural Engng Rep. 147, Department of Civil Engineer-
Mining Engng, University of Sydney, Feb. 1987 ing, University of Alberta, May 1987
9 Gaylord, E. H. and Gaytord, C. N. Design of steel bins for storage 23 Wilms, H. 'Calculation of stresses in silos by the method of
of bulk solids, Prentice Hall, Englewood Cliffs, 1984 characteristics', Bulk Solids Handling, 1985.5 (2)[19-23
10 Janssen, H. A. 'Versuche uber Getreidedruck in Silozeilen', J 24 Gorenc, B. E., Hogan, T. J. and Rotter J. M. (eds) Guidelines for
Zeitschrift des Vereines Deutscher Ingenieure 1895, 39 (35) the assessment of loads on bulk solids containers, Institution of
1045-1049 Engineers, Australia. 1986
11 Enstad, G. 'On the theory of arching in mass-flow hoppers', 25 Motzkus, U. Belastang von Siloboden und Aaslaufinchtern dutch
Chemical Engng Sci. 1975, 30 (10), 1273-83 kornige Schuttguter. Dr.-Ing Dissertation. Technical University of
12 Home, R. M. and Nedderman, R. M. 'Stress distribution in hop- Braunschweig, 1974
pers', Powder Tech. 1978, 19, 2 4 3 - 5 4 26 Hofmeyr, A. G. S. Pressures in bird. MSc(Eag) Thesis, University
13 Johanson, J. R. 'Stress and velocity fields in the gravity flow of bulk of Witwatersrand, Johannesburg, 1986
solids', J. Appl. Mech., Series E 1964, 31, 499-506 27 Mahmoud, A. A. and Abdel-Sayext, G.. 'Loading on shallow cylin-
14 Jenike, A. W. and Johanson, J. R. 'On the theory of bin loads', J. drical flexible grain bins'. J. Powder Bulk Solids Tech. 1981. 5 (3)
Engng for Industry, ASME, Series B 1969, 91 (2), 339-44 12-19
15 Johanson, J. R. 'Effect of initial pressures on flowability of bins', J. 28 Rotter, J. M. 'Bending theory of shells for bins and silos' Trans.
Engng for Industry, ASME, Series B 1969, 91 (2), 3 9 5 - 9 Mech. Eng., Inst. Eng. Aust. 1987, MEI2 (3) 147-159
16 Jenike, A. W., Johanson, J. R. and Carson, J. W. 'Bin loads- part 29 Draft Australian Standard DR89138, Loads on bulk solids containers.
2: Concepts; pan 3: mass flow bins', J Engngfor Industry, ASME, Standards Association of Australia, Sydney, August 1989

12 Eng. Struct. 1991, Vol 13, January

Vous aimerez peut-être aussi