Sei sulla pagina 1di 12

Powder Technology.

19 (1978) 243 - 254 243


@ Elsevier Sequoia S-A.. Lausanne - Printed in the Netherlands

Stress Distribution in Hoppers

R. M. HORNE* and R. M. NEDDEmlAN


Department of Chemical Engineering. Cambridge Uniuersity. Pembroke Street. Cambridge CBB 3RA (Gt. Britain)
(Received July 18.1977; in revised form September 19.1977)

SUMhL4RY method is extended to cover both active and


passive two-dimensional hoppers (i.e. bunkers
The method of characteristics, used in a with converging sides).
previous paper [l] to calculate limiting stress A ‘knowledge of the stress distribution in
distributions in granular material in two- hoppers is important in their design. Perhaps
dimensional vertical-sided bins, is extended the most common design procedure is that of
to similar problems in two-dimensional Jenike [2], and his method of predicting the
hoppers. Under normal conditions, a continu- minimum size of outlet to prevent arching
ous solution arises in the active case and a dis- requires an accurate estimate of the stresses.
continuous one in the passive case. He assumes the distribution near the apex is
The e_xistence of Jenike’s radial stress field of the radial stress field form, that is to say
at the apex [Z] is considered and comparisons that the magnitudes of the stress components
are made with the exact solution developed are directly proportional to the radial coordi-
here to investigate the distance over which the nate so that p/r and 6 are functions of angular
radial stress field is valid. It is found that in coordinate only. Walker’s modification of his
the active state the exact solution approaches method [3] employs a stress field based on
the radial stress field very slowly, so that the the differential slice analysis first used by
latter is valid only in the immediate vicinity Janssen [4] for bins. Both these analyses re-
of the apex. In the passive case a saw-tooth quire assumptions additional to those used to
stress distribution is found, unlike the con- set up the limiting, static equilibrium equa-
tinuous distribution predicted by the radial tions for which the method of characteristics
stress field- The latter does, however, give a provides an exact solution. It would thus
good average value- seem worthwhile comparing results by the
Wall stresses given by the present method various methods to see how much in error the
are compared with those predicted by the approximate solutions are.
approximate methods of Walker [3], Walters Some method of characteristics calcula-
[lo] and Enstad [ 111 _ Substantial agreement tions in passive hoppers have been carried out
is found in the active state, but the discon- by Johanson and Jenike 153 and Johanson
tinuous solutions found in the passive state [6], although few details are given. They
differ markedly from the continuous soiution chose, however, to adjust artificially the
predicted by the appro_ximate methods- boundary conditions near the top surface so
that &continuities in the stress field are not
generated. Their results for this restrictive set
1. INTRODUCTION of conditions showed that the stresses con-
verged to those given by the radial stress field
In a previous paper [l], the analysis of well before the apex.
stresses in two-dimensional bins of granular
material was carried out by the method of
characteristics. In the present paper, the
2. OUTLLNE OF SOLUTION

*present address: Agricultural Division, Imperial A brief account of the derivation of the
Chemical Industries Ltd., Billingham, Gt. Britain. characteristic equations and details of the
244

Passive state:
6 =Tr-(w -&)/Z-Q (2b)
where &,, is the angle of wall friction and
sin w = sin o,/sin 0, where 0 is the angle of
internal friction.
As explained in the previous paper Cl], the
divergence or convergence of characteristics
from the top comer determines whether a
continuous or discontinuous solution results
there_ For a bin (0 = 0) in the active state,
0 =oP = ;r/2 < 8, and the characteristics diverge
(or are parallel). When the hopper half-angle,
fi, is increased and consequently 8 W is de-
creased beyond the value for which 8 top = 0 =,
a discontinuous solution results. As the wall
is rotated still further, the slope of the dis-
continuity decreases_ However, this decrease
is less than the increase in Z2 and eventually
the wall and the discontinuity become coin-
cident_ For larger Q than this, the caIculated
discontinuity position is outside the hopper
and a limiting stress solution cannot be found.
Sokolovskii [7] has determined the ranges of
solution and they are given by:
Continuous:
Fig. 1. (2) Tso-dimensional hopperof half-anglefl; cl< I? G (w -&&?)/2 (3)
(b) Mohr circle for streses at 2 point on the right-
hand wall. Points 1 2nd 2 represent the wall plane Discontinucusr
stresses in the active 2nd pasive states respectively.
Points IV and 2V refer to the corresponding vertical (w - 0,)/2 < R G ir/2 - (0 + &)/2 (4)
planes.
Non-limiting:
numericaI method of soIution are given in our ir/2 - (0 + &)/2 < 52 (5)
earlier paper [l] _ A fuller description of most
aspects is contained in Sokolovskii’s book [7]. It is worth noting that for a fully rough-
The top surface and centre-line boundary walled hopper (0, = Q), the discontinuous
conditions for the hopper are the same as range vanishes.
those in the corresponding bin problem. The A similar analysis for passive hoppers can
only difference between the probIems then be carried out. For the bin (a = 0), BLop=
lies in the wall boundary conditions. For the is > 8, and a discontinuous solution exists
system of axes shown, the right-hand wall of escept in the smooth-wall case. Thus there
the hopper iIIustrated in Fig. l(a) is defined by is no finite range of 9, and SL for the con-
tinuous solution and we need consider only
x=ytan9 (1) the upper limit on fi for discontinuous
where ,O_is the hopper half-angle. From a solutions. Using the same argument as for
MO-hr circl? (Fig_ l(b)) for the stresses at a the active case, there is the possibility that
point on the right-hand wall, the anticlock- the wall and the discontinuity become coin-
wise rotation, 0, of the major principal stress cident- However, there is another way for the
direction from the x axis is given, for a discontinuous solution at the top comer to
cohesionless blohr-Coulomb material, by fail. Since the slope of the discontinuity falls
as SL increases, it might eventually happen
Active St&e:
that the discontinuity and the top surface
e =ir/2+(w -&)/z-n (2a) become coincident. For a Iarger SLthan this,
245

the calculated position of the discontinuity


is above the top surface - once again a
limiting solution cannot exist. The smaller of
the two limits is obviously the applicable one.
For the passive hopper it is the second
criterion which gives the required limit. The
ranges of solution are then given by:
Continuous:
Q-0, o,=o (6)
Discontinuous:
OG K2s n/2-(&J +4,)/2 (7)
Non-limiting:
0. > z/2 -(w + 9,)/2 (8)

It is found that if a limiting solution exists


at the top comer, then it exists everywhere
in the hopper, at least in the case where 0,
ow and ,O_ are constant throughout the
bunker. This conclusion is not generally true
if a non-horizontal top surface is considered. Fig. 2. Pattern of characteristics and discontinuities in
and a detailed analysis of solutions under this a passive hopper with wall stress plotted alongside;
condition has been carried out by Home [S] . L-2= IQ”, 0 = 30’, Qw = 25”. Heavy lines are discon-
We have now established that the only tinuities; thin lines are characteristics.
limiting stress solution in the passive hopper
is of the discontinuous type, as in the passive y tan ,O__ Plotted alongside is the variation of
bin problem [l] _ In the active case, the stress wall normal stress with distance along the
field may be continuous or discontinuous_ wall. It can be seen from the pattern of dis-
However, the latter which occurs at larger ~2 continuities that it is impossible to extend the
values will be seen later to correspond to numerical calculation to the apex, although it
stresses becoming infinite at the apex - an can be extended to within any specified
unrealistic result. Thus only the continuous distance from the apex. That is, of course, a
solution type, as occurred in the bin problem, limitation only in theoretical considerations
need be considered_ since the outlet on an actual hopper will be
Because of the tendency of material to some distance above its virtual apex.
move downwards and inwards in a hopper It can also be seen both from the diagg
which is discharging, and to a lesser extent on and from the actual numerical solution that,
fll;kg, due to consolidation as filling below the first couple of discontinuities, the
proceeds, the active state is unlikely ever to shapes of corresponding discontinuities do
be formed- Nevertheless, it is important to not change much with distance from the
study the solutions for active hoppers because apex, Le. ) along any radius from tile apex, the
of the light this throws upon the relationship discontinuities have the same slope. This
between the fully plastic solution derived by suggests that it might be possible to analyse
the method of characteristics and other solu- the stress nattem as a series of repeating units,
tions such as the radial stress field at the apex. the stresses and size of each being propor-
Figure 2 shows a typical pattern of charac- tional to the distance from the apex_
t&stics and &continuities for material in the
passive state_ It is very similar to that for the
passive bin [l] and is calculated in the same 3. COMPARISON WITH RADIAL STRESS FIELD
manner, the only difference being that the
boundary condition 6 = constant on x = The full stress equations are a pair of non-
constant is replaced by 8 = constant on x = linear partial differential equations of a
236

basically hyperbolic type [I, 2,7] _ Tf, how- (a) Active state
ever, these equations are expressed in polar Solutions for an active hopper with half-
coordinates (r, 5) and the assumption is made angle 5”, internal angle of friction 30” and
that the average stress p is directly propor- various wall friction angles have been deter-
tional to r, so that p = ~rq(E) and 6 = e(E), a mined. The variation of normal stress on the
pair of ordin&T differential equations results. wall with depth is shown in Fig. 3. The
These can readily be sohred [Z, 71 and the solution is continuous for @, 3 9.7”. For
resulting solution is known as the Radial 9, > 9.7”, stresses tend to zero at the apex-
Stress FieId. The solution of the ordinary For 0, = 9_7”, however, all characteristics
equations for the RSF requires fewer bound- are straight and the pattern is merely an
ary conditions than the solution of the full extension of the top surface Rankine zone.
equations and hence the solution cannot be Therefore the wall stress increases linearly
generally correct. In particular, the extensive with depth and has a finite value at the apex.
tabuiations of the RSF solutions given by For 0, < 9_7”, the solution is discontinuous
Jenike [2] require only a knowledge of 8 (0) and, although no precise calculations were
and 0 (C2); information about the stresses carried out, it was established that the stresses
applied to the top surface is not used. The increase with depth and tend to infinity at
analogous situation in ordinary differential the apex.
equations is perhaps more familiar_ Fre- Although Jenike is not concerned with
quently a particular integral can be written active hoppers, radial stress fields may be
down by inspection. Though this is an exact determined for the apex in the same way as
solution to the equations, it wiIl not neces- he determines them for passive hoppers.
sarily satisfy the boundary conditions and a However, they exist only for 0, > 21.78”,
complementary function is also required for at which value q becomes infinite everywhere.
a complete solution_ Similarly, the RSF is a These values obtained for wall stress are
solution to the equations but is not neces- plotted on Fig. 3 for comparison. There is
sarily the solution to the problem_ obvious agreement over the lowest 10% of
Jenike [23 argues that the applied stresses the hopper in the fully rough-wall case i@, =
on the top surfaces only have a local effect 30”), but for & = 28”, the exact solution
and that the full solution rapidly tends to appears to line up with the radial stress field
the RSF. The accuracy of this is considered (RSF) result only very close to the apex_
below. This can be seen more clearly by looking at

Fig_ 3_ Distribution of wall stress with depth in an active hopper with various 9, values; Q = 0, Q = 30”, n = 5O_
247

numerical values: dpjdh,, = 0_86r; at h/h, =


0.05, the exact solution gives dp/& = 0.707,
while at h/h, = 0.01, we find dp/dh = 0.81~
and p/h = O.SZy, where h is height above the
apex and ho isthe height of the free surface.
A limiting value ciose to 0.867 seems there-
fore quite likely. For Q, = 22”. there appears
from the figure to be no agreement between
the two results. Examination of numerical
values shows that the slope of the curve at
h/ho = 0.001 is still only one-seventh of the
RSF slope. However, the slope is changing
rapidly enough for there to be the possibility
Fig. 4. Wall and centre-line stress distribution in a
that a value close to the RSF slope might
passivehopper; Q = 0, o = 30";9, = 25",52 = 10".
eventually be achieved.
Thus it seems that the radial stress field
may well provide an asymptote for the
hopper stress field at the apex. However,
the distance over which this asymptote
approximates the solution well is negligibly
small, except when the wall is fully rough,
and even then only over one-tenth of the
hopper depth.
Referring to the figure again, we see that,
for & = loo, the wall stress curve reaches
the apex with infinite slope. If the RSF and
exact solutions agree everywhere, then we
expect this infinite slope to apply for all
0, < 21.78”. Although the curve for 0, =
21” may not appear to attain such a slope,
once again the actual numerical values show Fig_ 5. Comparison of wall stresses predicted by
an extremely large rate of increase in slope >xrious analyses in a passive hopper; Q = 0, fi = 40’,
near the apex, and an infinite slope is per- 0 = 30”. 0, = 20”.
fectly feasible_ Further support for this is
given later in Section 4, where Walters’ one-quarter of the way for the steeper one.
analytic distribution of stresses approxi- Consequently there are far fewer jumps in the
mates closely the exact solution over a stress distribution for the shallower hopper.
finite region near the apex and predicts an Since the average slope of the d&continuities
infinite slope at the apex for much the same is much the same in each case, this result is
range of 9, _ predominantly the geometrical effect of the
different hopper half-angles_ Now it can be
observed that, below the first discontinuity,
Figure 2 shows the pattern of characteris- the wall stress falls with a nearly constant
tics and discontinuities in a hopper for which average slope to the apex in both cases. The
Q = 10”. @ = 30” and 4, = 25”_ Figure 4 effect of the different geometries then is to
shows the stress distribution on the wall and make this average slope much higher and the
on the centre-line in this case. Similarly proportion of the hopper over which it
Fig. 5 shows the wall distribution for the case applies much lower for the shallow than the
Q = 40”, Q = 30” and 0, = 20”. steep hopper.
In both cases, the stresses tend to zero at Radial stress fields have been generated for
the apex. The major difference between the these two cases and for several others for
plots is that for the shallow hopper the fiit which methods of characteristic solutions
discontinuity reaches the wall three-quarters have been generated. Since the RSF gives a
of the way down the hopper compared with linear distribution with depth, it obviously
248

cannot reproduce the effects of the discon- which is necessary to ensure flow under
tinuities. However, it is found from pIots of these conditions.
the results such as Figs. 4 and 5 that, below As Jenike has pointed out, radial stress
the first discontinuity, the RSF provides a field soIutions exist for all +,&, combinations
good average to the oscillating exact solution over the entire range of hopper half-angles
at the wall in all cases. Examination of numer- from 0” to 90”. This is, of course, much
ical values on the centre-line shows that the wider than the range for which solutions to
RSF provides a good average there also_ free surface problems could be found by the
That the radial stress field is in fact ap- method of characteristics. It is natural to ask
proached towards the apex of the hopper whether RSF solutions outside this range
has been proved by Jenike [Z] from an have any significance_ We can begin to answer
analysis of the governing equations and this question by remembering that the radial
argued by Gardner [9] from dimensional stress field is an exact solution to the ful!
analysis cdnsiderations. The existence of equations of equilibrium. If we consider a
discontinuities in the exact solution in the hopper with top surface conditions satisfying
usual passive case is an indication of the the RSF for the appropriate 0 and O,, then
limitation of these arguments. Jenike’s proof the RSF must be a complete solution to the
is made under the assumption of a continuous probIem. Moreover, the solution is unique
stress field in the neighbourhood of the apex since it is constructed entirely using the three
and so ignores the occurrence of discontinu- basic boundary-value problems, the solution
ous fields. Furthermore, the analysis is to each of which is unique f12] _ Thus the
strictly correct only as r + 0. By carrying RSF is the only complete solution for these
out a dimensiomii analysis, we find, as top surface boundary conditions_ This has
Gardner must have found, that for a sur- been confirmed by reproducing the RSF by
chargeless surface, the method of characteristics in several cases,
although it should be added that the solutions
show some numerical instability.
It seems likely therefore that for hoppers
with half-angles outside the range, certain
where position is given in polar coordinates surcharge conditions on the top surface will
(r, f) centred at the apex, and B is some bin produce fully plastic solutions. Furthermore,
dimension (e.g. height of the free surface it is possible that if the top surface conditions
above the apex)_ This shows that we cannot do not allow a fully plastic solution, the
expect that the RSF will hold over any finite resulting rigid regions near the top of the
radius at all. This conclusion can also be hopper will rrct as an effective surcharge
drawn from the occurrence of infinite p/-yr allowing the stressfieldto be plasticevery-
values in certain active cases_ If this were where near the hopper outlet and thus,
true for anq~finite radius, p would become perhaps, leading to flow without stagnant
infinite, which is impossible. regionsthere.Inthiscasethisplasticregion
For passive free-surface hoppers, it is near the outlet could be expected to be
found that except for close to the maximum asymptotic to the RSF. Nevertheless it is
value given by eqn. (7), the RSF is a good unlikely that with the existence of rigid
average to the osciUating solution over at least regions the overall flow which does OCCUT
20% bf the height_ This result constitutes the will be mass flow- Thus the maximum 52
only true justification for the use by Jenike value forwhichafreesurfacesolutionexists
123 of the RSF results in his design procedure possibly represents the limitingvalue of Q
for sizing the hopper outlet_ However, if the or-at Ieastan upper bound on ~2 for which
fully plastic solution applies, the use of the massflowcanoccur.
RSF values wili still lead to error- We can
define a new p/yr by considering the straight
4_ COMPARISON WITH OTHER APPROXIMATE
line through the apex which marks an uIjper
SOLUTIONS
limit to the discontinuous stress distribution_
This is illustrated in Fig. 4_ Using this value The differential slice method of Janssen
in the design, we deduce a larger outlet size 143 was extended by Walker [ 3 ] to analyse
249

stresses in hoppers. He considered a slice with Walters [lo] considered the more accurate
vertical sides as shown in Fig. 6(a) and derived differential slice with sides along the hopper
the expression: as shown in Fig. 6(b). For 2D wedges, his
expression is the same as Walker’s except
that:

cc BD
-+D-1 (12)
where tans2
BD Walters also suggests the use of a different
CC- value of D from Walker.
tans2
(11)
Recently, Enstad [ll] has devised another
D, the distribution factor, is the ratio of the approximate theory for stresses in hoppers.
vertical stress at the wall to the average ver- He considers a differential slice bounded by
tical stress across the hopper; B is the ratio of differential segments of the walls of the
shear stress to vertical normal stress at the hopper and two non-concentric circular arcs
wall and co is the average vertical stress at h = as shown in Fig. 6(c)_ Each arc represents the
ho. For the special case of C = 1, the analysis line of action of the major principal stress.
yields the equation The arc is located from the knowledge of 0
and hence the tangent to the arc at the wall
and centre-line. Along each arc, p is con-
(a,,,), = rh In (tc) + a0 k
sidered constant_ Enstad derives the following
expression for p:

7Yr Y yro r -?
p=x-_l (
+ PO-x-_l >I >6 (13)

where r is the distance along the wall from the


apexandOiriro,
X = sin o {l + sin(2d + Q)/sin R}/(l -sin 6)
y=W+W sin Q + sin d sin(d + 52))
(1 -sin 0)sin2(d f ~2)
d = (w f ON.)/2

and p. is the value of p at r = I-,.


The use of this analysis for a hopper alone
(as compared with a bin-hopper combina-
tion) has the difficulty that the elementary
slice forces the top surface to be the arc of a
circle, which may be rather unrealistic- The
discrepancy between the arc and the hori-
zontal at the same r is largest for large fi and
Fig. 6_ Differentialslicesused in varioushopperstress 0, close to 9. It must be pointed out that
L%lZlYseS. Enstad develops the analysis for a bin-hopper
combination which does not involve having
Some features of these expressions can to take the free surface as a circular arc.
easily be shown: Before the analysis of Walker or Walters
(1) For C I 0, the stress at the apex is zero, can be used, a value of D has to be decided
and for 0 < C < 2 the gradient of the stress upon. Walker suggests taking D = I for the
there becomes infinite (from the positive passive case_ He does not consider the fuhy
side)_ active case, but by analogy with the passive
(2) For C < 0, the stress at the apex be- case we will refer to the D = 1 active case as
comes infinite with infinite gradient. Walker’s analysis. Under this value of D,
Walters’ solution yields the same results, since
the Same value of C is given by eqns_ (11) and
(12). However, Walters assumes that, on any
horizontal pIane. uu = constant and T,,, =
(_x/_K,)(~,,),\. _)where x = 0 at the centre-line_
These are the conditions which hold at great
depth in a bin, and so his D values are the
same as those occurring there. except that 0,
is replaced by a new parameter n where:

tan 17 = (~X>L./(~Xz)~- (14)


These assumptions become unworkable for
the passive problem if -O_ lies outside the
range:
2_0_ d Q - Q,~ f cos-l (sin p&in 0)
which, using G, reduces to the more usefu:
form
2-C f X/2 f 0 -Gi -6, (15)
The maximum value of-O_ is given when 17 = o_
For -CLoutside the range, insistence that the
shear stress is proportional to distance from
the centre-Iine means the hlohr circie at the
centre-line must be that for the active state,
which is impossible in a passive problem.
The restrictiveness of (15) is apparent for
the esample of a fully rough wall, since we
finda G 0,i.e. thehopper cannotopenup
at ah_ Walters suggests (15) may be a limit for
mass flow, but this disagrees with all esperi-
mental findings_ Moreover, this is merely a
limit beyond which the assumptions cannot
hold, and although they can be shown to be
valid at great depth in a bin, there is no theo-
retical reason why they should hold in a
hopper-

(a) Actiue state (b) a> == Lz”, x2%-- L” d&l w/s


In the active state. values of D at great Fig_ 7_ Distribution of wall stress in an active bin for
depth in a bin are close to l_ Hence there will various Is,. values; Q = 0. fi = 5”. 0 = 30”.
be little difference between the results of
WaIker and WaItem Plotted on Fig_ 7(a) are D is exactly 1 and the WaIker result follows
some results by the method of characteristics that of the method of characteristics. For all
for the waII stress distribution on the 5O other &, there is a distribution of D with
hopper discussed in Section 3. Also plotted depth which could be determined numerically
are the corresponding results by WaIker’s from the exact solution as was done for the
metbod_Itis seenthattheseresultsfallbelow active bin. However, some information can be
the exact solutions over the lower half of the obtained about D without doing this_ We have
hopper. Agreement, in generaI, is good already shown that the radia.I stress 6eM
although it is worse for large 0, than for approximates well the walI stresses near the
smah. This is expected since, for 0, = 9.7”, apex. It is a simple matter to caIcu1at.e D
the exact stress distribution is an extension of during the generation of the radial stress field,
the Rankine zone_ Under these circumstances, and this must then be the same as the exact D
251

near the apex. If this value of D is used in the


Walters analysis, the predicted stresses near
the apex must also agree with the exact solu-
tion. Stresses by this method are shown along
with the D = 1 results in Fig_ 7(b) for 0, =
28“ and 22”: It can be seen that the agree-
ment is excellent, not just over the very small
distance for which the radial stress field holds,
but over most of the hopper_ This indicates
that, although the absolute values of the RSF
are not correct over this region, the integrated
effect of the relative distribution of stresses
across a horizontal plane is nearly correct and Fig. 8. Comparison of wall stresses predicted by
hence the value of D calculated is the same as vakous anal&s in a passive hopper; 4 = 0.0 = 30”,
I& = 25”, fi = 10”.
that for the exact solution. Agreement is
poorer over the top half of the hopper, but
this is to be expected from the use of any Figures 5 and 8 show the exact wall stress
constant value of D since D is known to vary distributions for the two hoppers considered
near the top surface. in Section 3. The results by Walker’s method
We can also calculate the distribution using (i-e_ D = 1) are also plotted. In both cases,
the value of D suggested by Walters. For 0, = agreement is poor everywhere in the hopper.
28”, D = 0.8712 compared with the RSF Stresses in the lower section are larger than
value of 0.8688. For 0, = 22”, D = 0.9637 the average of the method of characteristics
compared with 0.9634. Thus the Walters dis- solution and are larger even than the locus of
tribution of wall stress can be expected to be the peaks of the discontinuous pattern. For
barely distinguishable from that using the the shallower hopper (Fig- 5), C as given by
RSF value of D, and therefore also to agree eqn. (12) is 0.255. This is in the range 0 <
well with the exact results over the lower part C < 1 for which the stress distribution tends
of the hopper. This is confirmed in Fig. 7(b), to zero at the apex with an infinite slope_ This
where the Walters curve is coincident with the is in complete disagreement with the radial
corresponding DRsF curve in the 0, = 22” stress field result shown. Before examining
and 28” cases. For 0, = 21”, there is no RSF the results from Walters’ equation with
and therefore no Da,, curve, but the Walters various D values, let us consider the results so
curve which reaches the apex with infinite far for an even shallower hopper, Le. X2 = 55”,
slope is once again in good agreement with Q = 30”, f&, = 25”. Under these conditions,
the exact solution over the lower part. fl is greater than the maximum value (r/2 -
One point for comparison is the maximum (w + 9,)/2) of the range (7) and no fully
value of &, for which an infinite stress plastic solution can be found for a horizontal
gradient is predicted at the apex. For the RSF stress-free surface. Since the above maximum
it was found to be 9, = 2X78”_ For the value of !EI coincides with the condition
Walters/Walker methods it is given by solving (T._)_. = 0, and since (T.&. = 0 leads to a zero
the equation C = 1. For D = 1, this gives value for Walker’s C, the above hopper condi-
0, = 2X55”, and for the Walters value of tions give C < 0, and Walker’s stress distribu-
D (O-97), 6, = 21.69”. There is thus good tion tends to infinity rather than zero at the
agreement_ apex. This unrealistic result and the RSF
solution for the same conditions are plotted
(b) Passive state in Fig. 9.
It must be emphasised at the outset that For the three hoppers under discussion, we
since the analyses of Walker, Walters and can de&&D values from their radial stress
Enstad give smooth wall stress distributions, fields. As explained in the active case, the
they can never approximate the discontinu- Walters distributions derived using these
ous exact solution well. The best they can be values of D will agree with the RSF near the
expected to do is give good average values apex and hence give a good average to the
over one section or other of the hopper- discontinuous wall stress over at least that
252

II _ from D > 2 to D < 1 at the changeover


value of Q _ Results for such D values will he
more in error than those for D = 1. -4 possible
compromise is to use, for all SL > filimr the
value of D for SL = -cLri,, Le. for 11= 0. This
leads to a D of 2.6905 compared with RSF
values of 2.32 for the 40” and 2-53 for the
55” hoppers. Thus we expect better agree-
ment in the latter than in the former case.
This is supported by the results in Figs. 5
and 9.
Walker uses his results (Le. D = 1) for
design, in which, like Jenike, he is concerned
with avoiding arching at the outlet. He reabses
that the RSF gives values of D much greater
than 1, but states that “wall stresses are
Fig_ 9. Comparison of wall stresses predicted by reasonably insensitive to great variations of D
various analyses in a passive hopper; Q = 0. ,O_= 55 I.
from unity to infinity”. From the large varia-
0 = 30”. &. = 25”.
tions in shape for various D values, which
Fig_ 5 indica’tes, this statement is untrue if
range for which the RSF gives a good average_ the correct value of C as presented by U’alters
To verify this in practice, the resuIts are is used. Walker’s resu1t.sgive poor approxima-
plotted in Figs_ 5, S and 9_ The agreement is tions to the discontinuous solutions for all
certainly there for the two steeper hoppers. hut very small values of ,O__Walters’ results
For Fig_ 9, however, there is no fully plastic (using his suggested D values with the modi-
result for comparison. Although the gradient fied procedure given above) are much more
of the stress for the IVaIters distribution with accurate. However, the radial stress field
the RSF D value agrees with the RSF gradient solution gives the best average to the discon-
at the apes, these two results quickly diverge. tinuous pattern It could be used in the design
Despite the fact that there is no fully plastic procedure, if the result were increased by a
solution to this problem, if material flows generous factor to simulate the locus of the
from the hopper and the flow channel near discontinuity peaks. However, use of the
the apex estends to the waI!s, the solution exact solution is to be preferred_
must be pIastic in the neighbourhood of the Finally, Enstad’s analysis can be consid-
outlet- Ke can conclude that this plastic solu- ered_ For a particular hopper and particular
tion (or its average if it is discontinuous) wall stress parameters (e.g. o,), the area under
agrees with the RSF only at the apex and the wall stress distribution, which is propor-
therefore not at the likely position of the tional to the weight of material in the hopper.
outlet. is the same whatever method of calculation is
Results can also he obtained using W&es’ used_ However, since Enstad’s analysis
equation and his suggested values of D_ For assumes the top surface of the hopper to he
the 10” hopper, 1) is given by (14) as 2X72”, convex upwards, the weight of material sup-
which gives a value of D of 2.155 compared ported for the same height of material at the
with the RSF value of 2-05. The wall stresses wall (see Fig. 6) is larger in this case than for
in these two cases are consequently in good a horizontal surface. Hence the area under the
agreement, as shown in Fig. 8, and therefore stress curve will be larger, and agreement with
WaRers’ value of D leads to a good average to the exact solution or a good approximation to
the exact distribution over the lower section it cannot be expected_ This is shown in the
of the hopper. For the other two hoppers, three relevant plots - Figs_ 5,s and 9_ The
C, is greater than R,, given by the equality difference increases with increasing Q and o,
in (15). In this case, Walters suggests using (for a given @)_ What can also be seen, how-
the acfive D values for the value of q ob- ever, is that the shape of the curve is in each
tained from (14) Tbis is clearly incorrect case similar to the Walters’ distribution (using
since it means a discontinuous change in D either Walters’ D or the RSF D). Reduction
253

of the absolute values of stress by the ratio D ratio of vertical stress at the wall to
of weights of material in the two hoppers the average across the hopper
will obviously give quite a gocd approsima- h height above the apes
tion to the solution. ho height of the top surface
Although, over the lower part of the P average stress
hopper, Enstad’s results are in poor agree- PM-
ment with the radial stress field in these cases
for which o = 30”, Enstad has shown that
E?
r
aeplied stress on top surface
radial distance from some origin
they agree well for larger o_ Arnold and r0 distance from apex to hopper top in
McLean [13, 141 have shown in fact that, Enstad’s analysis
for 0 = 50”, Enstad’s results are closer to horizontal, vertical coordinates
the radial stress field than are Walters’_ terms defined in Enstad’s equation
(13)
weight density
CONCLUSION angular coordinate in polar system
angle defined in \Valters’ equation (14)
By means of the method of characteristics, major principal stress direction
fully plastic stress fields can be found for two- coefficient of friction (= tan p)
dimensional hoppers_ In the active case, solu- normal stress
tions are usually continuous, while in the shear stress
passive case they are usually discontinuous angle of internal friction
Calculations confirm that the radial stress angle of wall friction
field near the apex of the hopper is the sin-i (sin o,./sin o)
asymptote to which the complete solution hopper half-angle
tends. In the active case, the RSF is a good
approximation onIy very close to the apex.
In the passive case, the discontinuous stress REFERENCES
pattern oscillates about the RSF for a signifi-
cant distance from the hopper apex_ 1 R_ hI_ Home and R. AI. Nedderman, Analysis of
Of other approximate passive solutions the stress distribution in two-dimensional bins by
the method of characteristics, Powder Technol.,
available, Walker’s wall stress predictions are
14 (1976) 93.
poor. Walters’ results are much more accurate 2 A. IV_ Jenike, Gravity flow of bulk solids, Utah
in general. although. near the apes, they may Univ. Eng_ Exp_ Stn. Bull_ 108. 1961.
still differ significantly horn the more 3 D. hI_ Walker. _An approximate theory for pres-
rigorous solutions_ sures and arching in hoppers, Chem. Eng. Sci., 21
(1966) 975_
1 H. A. Janssen, Versuche iiber Getreidedruck in
Silozellen. 2. Ver. Dtsch. Ing.. 39 (1895) 1045.
ACKNOWLEDGEMENTS 5 J. R. Johanson and A_ W. J&&e,- Stre& and
velocity fields in gravity flow of bulk solids, Utah
R. M. Home is grateful to the Walter and Univ. I&g_ Esp. Stn. Bull. 116, 1962_
Eliza Hall Travelling Scholarship Fund of the 6 J. R. Johanson, Stress and velocity fields in the
gravity flow of bulk solids, J. Xppl. blech., 31
University of Queensland, and to Peterhouse, (1964) 499.
Cambridge, for the financial support to carry 7 V_ V. Sokolovskii, Statics of Granular Media,
out this research. Pergamon Press, Oxford, 1965.
8 R. &I. Home, Theoretical anaylsis of stress dis-
tributions in bins and hoppers, Dissertation sub-
mitted to Univ. of Cambridge, 1976.
LIST OF SYhIBOLS
9 G. C. Gardner. Central Electricity Res. Lab. Rep.
RD/L/hI/50/1963 cited by J. C. Richards, (Ed.),
a half-width of hopper at the top surface The Storage and Recovery of Particulate Solids,
b characteristic dimension of the hopper Inst. Che& Eng., London, 1966, p_ 106.
B ratio of shear stress to vertical normal 10 J. K_ Walters, A theoretical analysis of stresses in
axially-symmetric hoppers and bunkers, Chem.
stress at the wall Eng_ Sci., 28 (1973) 779_
C BDltan s-2 11 G. Enstad, On the theory of arching in mass-flow
d (0 +9,)/z hoppers, Chem. Eng. Sci., 30 (1976) 1273.
254

12 W. F_ Ames, Non-Linear Partial Differential Equa- converging channel, Powder Technol.. 13 (1976)
tions in Engineering, Academic Press, New York, 255_
1965 14 P. C_ Arnold and A_ G. McLean. Improved ana-
13 P_ C_ Arnold and _4_ G_ McLean. An analytical Iytical flow factors for mass-flow hoppers,
solution for the sires function at the wall of a Powder Techncl.. 15 (1975) 279.

Potrebbero piacerti anche