You are on page 1of 29

Powder Technology,7 (1973)215-243 215

0 Elsevter Sequom SA , Lausanne - Pnnted m The Netherlands

An Analytical Hopper Design Method for Cohesive Powders

P. T. STAINFORTH and R C. ASHLEY

Imperzol Chemrcollndustnes Ltd, PlastzcsDzczszon,Wetwyn Garden Czty, Here (Gt. Brrtom)

(RecerveI March 14, 1972, IIIrevmzd form September 5, 1972)

Thx method overconzes two practzcal dzflcultzes zn the Jenzke graphical solutzon. that of nzakzng azz acczcrate
estimatzon oftlze materzal eflectzue angle of zzzternalfrictzon S (whzch we slzow to he a functzon offow factor.
ffS,), and then of extrapolating the powder failure fzazcrion FF cm-tie to obtain a correct zntersectiozz with the
hopper flow factor line.
We find that FF and all the inter-related nzaterzal izzternal frzctzon angles of simple powders are themselzzes
functions of spec$c tension T, (tensile stress T dzvided by zts associated steady-state zzornzal stress G,), the
critical zialue of which may be calculated from

where n and K are powderflow property constants detennined experimentally; Y IS the yield factor, a constant
combining the other tl<o and lizzking yield stressf at tlzefree surface of the powder arch dzzectl_b with tezzszle stress
T; and ffS is a flow factor calculated at the outlet stresses developed during hopper chargzzzg, and which assumes
urch failzire along tlze plane of the principal shear stress This critical value of T, thezz determines tlze nzaximunz
stable arch span I3 from
6 (,-IV,
B = nzYc”=
K(l-I- l/Q”“_p//T, 1
in whzch 6 and .B are tlze suggested strazght-lzzze constants of the time-consoiidated locus, azzd c azzd 7 are those
of the proposed tenszle stress-bulk density, p, relationship, T= co’, all of which are determined experzmentally.
Tlzis solutiozz fully takes zzzto account tlze effect of material compresszbility azzd hopper surcharge pressures,
and ensures an answer on the conservatzue side. A computer progz-amnze (HOPACALC) has been wrzttezz izz
Fortran to hazzdle the calculatiozzs.

INTROiJUCTION

The basis of design of hoppers and storage bunkers for powders and granulated bulk solids IS essentially
an analysis of pressures developed in the material, both 111the static (or elastic) state and under dynamic (or
plastic) conditions These stresses are controlled both by wall geometry and by the flow properties of the
material stored, and determine whether the powder will flow or not flow through an openmg in the hopper
bottom The ability of the powder to form an “arch” or “rat-hole” above the outlet is also closely connected
with the mode of flow, which can be in one of two forms : mass flow, in which slip occurs at the bunker walls
as well as throughout the material itself, and core flow, in which discharge is confined to a central funnel in
the hopper.
Janssen’ in 1895 was the first to analyse stresses in tall cylindrrcal grain srlos, and showed that below a
certa:m depth of fill the ratio of horizontal pressure pr to vertical pV was a constant k,, which he found to
216 P. T. STAINFORTH, R. C ASHLEY

depend on material flow properties. He also showed that these pressures were well represented by the
relationships

Pv =

where Q is the exponential function (1 - e-J”f’D)


J = 4 k, tan 4 (for circular or square-sectloned bins)
p = the material bulk density
D = the silo diameter
U = the verticai depth of fill
4 = the angle of friction developed between the material and the wall.
This relatlonshlp has remamed the basis of design for vertical-sided silos until the present day. In 1929
Ketchurn’ applied the same principles to hopper design for free-flowing granular materials, but his method
did not hold for cohesive powders, and little further progress was made until 1961 when Jemke3 published
his major work.
In this he identiiied the two modes of flow and showed that the ideal hopper slope B was linked with both
material flow properties and a hopper “flow factor" ff which he derived theoretically from an analysis of the
stresses developed m the region of the outlet durmg flow. Published flow-factor charts enable engineers
to select a smtable hopper angle 0 to handle a material of particular flow properties, together with a cor-
responding fl value from which to calculate the greatest potential arch span.
Flow property characteristics are determined experimentally from a series of tests m a shear ccl , which
are described in detail elsewhere 3 4 . These shear tests produce a family of yield loci, from which corres Jondmg
failure function values of major principal stress X1 and unconfined yield stress f are deduced. When plotted
graphically. an intersection of the failure function FF with the hopper flow factor fl line gives a crmcal stress
below which the material will arch and not flow, and above which unlimited flow can take place. The limiting
arch span B is then given by

(1)
where f, is the critical value of the yield stress at the material free surface, p, is the powder density at the
corresponding critical value of maJor principal stress, and nz= 2 for a conical or pyramidal hopper and unity
for a wedge hopper.
While the Jenike method is simple in theory, its practlcal application presents a number of difficulties The
most obvious of these is that the experimental FF points are often far removed from the calculated _tTline
and, owing to limitations in the graphical method of data interpretation, there is usually considerable
scatter m these points, making extrapolation back to the critical intersection a matter of guesswork.
Furthermore, calculation of the Jenike flow factor depends upon the inter-relationship of the hopper slope,
the wall friction angle &, and the angle of effective frlctlon 6 of the powder. Angle 4 is usually taken to
be constant, but 6 1s found to increase with decreasing major principal stress and must be extrapolated for
an appropriate value at the critical point. A wrong extrapolation, however, has a considerable effect upon
the correct estimation of the flow factor, and 1s a major source of error.
Our method seeks to overcome these difficulties by establishing the correct mathematical relationships
between material flow properties of snnple powders and hopper geometry, and takes as its starting point
rhe Warren Springs equation6 of the yield locus
7 ” G-I-T
0c =-
T
From here, we build on our previous work5, which described a computer technique for handling experi-
mental data As this new analysis lends itself to computation, we have written another complementary
program m FORTRAN called HOPACALC to handle the calculations, and present a worked example
alongside a solution by the traditional graphical method.
The full derivation of our equations is to be found in the Appendix, so we have included a minimum of
mathematics in the text for the sake of readability_
P. T. STAINFORTH, R C. ASHLEY

I I I I I 1 8 1 I
-_---_-_ -_

cl POLYETHYLENE GRINDIVXL 20
I PVC154
0 f C I COCOS
x CALGON FLAKE
I Cl CaC03(A)
PVC155
PLKAD MB10501
PTFEl437
PTFEIOOI

Hyperbolic
-U

Power function

P
E, !,
O-30 15 2D 25 33
1 35
I 40
1 45
I 5fJ 55 60 FF

FIN 1 Sm 6/.=F relattonshlp for a range of coheswe materials

Fig 2 C and f_Mohr circle diagram for typxal powder yw.ld locus

Substituting the expressions of eqns (5) and (6) for sin 6 and tan i., eqn. (9) may be arranged as

(1 + l/73(2-“)/” = ” - (Y-W-l/T,)2
KZ C(il+2Y-FF)+(~z-2)1/T, 1 (10)
and solved for l/T, by successive approximation knowing K, Y, n and FF. This allows the critical value of T,
to be calculated by the method descrrbed later
It is worth noting that while T, assumes a value of infinity when G,_ is zero, its largest practical value is
given by
2
T,= (11)
Y(l -sin #lI)
220 P. T STAINFORTH, R C ASHLEY

I 18 DAY TIC LO
0 INSTANTANEOUS
WAL _ FRICTION
ASS11MED TIC P

To D 1000 1SDD ZOCJ 2500 3500 kgld

Fig 4 Instantaneous and I3-day tnne-consohdated yield km-PVC’/55

TABLE I(a)

Instantaneous yteld Ion charactenstlcs for PVC/55

l)=37275. n=21137, K=38089.T,=88172. b=49.04. jJ=O5562

T, f - FF %
&??G, ;;g/m2, ;,ZJ,m? &12, (b/m’) FIi‘-

3979 63 2262 42 108 48 0 02725 6908 43 224444 3 0780 14 63 30 46 97 85


3362 40 1943 23 92 26 0 02743 5938 81 1908 86 3.111 1540 3143 98 97
235106 1378 82 66 29 002309 4185 92 137155 3 052 15 61 31 58 99 59
170409 100458 4252 0 02 347 3043 52 100988 3 3175 15 70 3200 99 72
1338 94 794 99 38 65 0 02387 240044 799.58 3002 15 80 32 18 99.71

TABLE I(b)
l&day txmc-consohdared yield loa charactenstxs for PVC/55

@=3727= n=21187, K=3808?. 7,=88172, 6=5393, fi=O6117.

r, f FF %-
&?/m’) k, FIT

3362 403 212.193 108 48 0 0323 6200 04 224444 2 762 1609 33 74 98 94


2865 34* 1805 46 92 26 0 0322 5260 06 1908 86 2756 1607 33 71 9897
235106 148251 75 80 0 0322 4322 43 1568 41 2.756 1608 33 73 99-24
2039 14* 129158 66 29 0 0325 3747.11 1371.55 2 732 16 15 33.86 99.59
1455 45* 934 355 48 52 0 0333 269257 1003 88 2682 16 35 3427 9972
1338 94 884 36 47 01 0 0351 2523 75 97276 2594 1676 35 12 9981
1074 7of 71908 38 64 00360 2040 82 799 58 2 552 16 9.5 35 52 9971

* Extrapolated pomts from correspondmg instantaneous locus


xk
P. T STAINFORTH, R. C ASHLN

THE STRONG-WALL FLOW FACTOR fl,

The Jenike3 theory of arch farlure assumes that when a material flows en masse in a hopper and is brought
to rest, consolidation of the contents is brought about by stresses developed during flow, and the resulting
stress field in the region of the arch just before flow is re-started is the same as it was before it was arrested_
Blanchard and Walker4*’ have confirmed that the flow stress field often persists from one partial with-
drawal to the next. But now both they and Jenike’ recognise that a totally different stress system is developed
during filling, which gives rise to much greater pressures (and therefore much stronger obstructions to flow)
in the region of the outlet than catered for by the original analysis. This new stress system is elastic and static,
as distinct from the plastic and dynamic nature of the flow stress system, and is similar to that developed
in the cylindrical trunk. Jenike and Johansons have analysed this radial stress field, for which they propose
the numerical solution of

2 = (rn+2) S tan Q + p sin 8


2

and
d4 m+~~21r,) -sin~cos~[(m-l)cotB+(~) s*nO] (17)
dB= -(m+2)+cos‘4 +($)cos@]

to determine the static stress field function S in terms of 8 and I#L S is defined in a similar manner to that
for the dynamic stress field (which will now be denoted by S ), and again the intensity of stress in the outlet
region is assumed to increase linearly from the hopper vertex along a ray of length T inclined 8 to the vertical
axis of the hopper. Factor k2 is similar to Janssen’s’ k, for the trunk, and is determined from
k z l+v(m-l+&)
2
1+s+v(m-1-&)
HOPPER DESIGN FOR COHESIVE POWDERS
223

Fzg 6(a) Detimtron of stresses static stress system

Fig. 6(b). Definmon of stresses. dynamic stress system

where v is a notional Poisson’s ratio for the bulk material given by


I-cosisind
Y=
2 ’
and

i = arcsin (sin +/sin 6) - 4

The coefficient ofcompressibility, E, is deemed to be constant for any given material at the pressures con-
cerned, and is found expernnentally during shear cell testing from the bulk density characteristic, thus:
224 P. T STAINFORTH, R C ASHLEY

E = log(l-Po/P”)--og(1-PoIP’)
log (,“/P”) - log (G’lP’)
(19)
where p. is the bulk density datum at FF = 1 already referred to, and p’, p” are bulk densities at test pressures
C’ and G” respectively.
Since a numerical solution requires a starting value for So to be guessed reiteratively until the correct
intersection at 0 and 4 is found, we prefer to make the substitution p= S tan 4 and transform eqns (17) into
a smgle second-order differential equation thus :

d2nu
@ + (m-l)cotfI$ - + (m-l)cosec26
1 sin 0 (20)

which lends itself to an exact solution of p m terms of 8.


Evaluation of S now allows the most important hopper loading pressures to be determined for structural
design purposes. For example, wall normal stress, CT,,in the region of the outlet up to the peak pressure centre
ts given by
2k,
~0 =
( )
l-tk, h--S

as shown in Fig. 6(a). In addition, we are now able to calculate the most conservative value of 6 for material
above the outlet, and so arrive at a practical flow factor for the no-arching condition.
At this point we must consider how flow can br started in a hopper in which a static stress field system
obtains, i.e. where there has been no withdrawal during charging, and where the contents have consolidated
and arched across the outlet under the influence of a near vertical major principal stress. In the chemical
industry it is not uncommon for sdos to be batch operated, and this entails frequent emptying and recharging.
The dynamic stress system developed m the hopper during partial emptying is, therefore, destroyed after
each cycle, and the silo pressure system returns to the static state at each new filling. Since it is important to
ensure that flow can always be re-started under these conditions of operation, it is essential to consider the
much more stringent conditions for arching brought about by these larger static pressures.
In this situation there can be no switch to a dynamic stress system until some flow has taken place, and
clearly thus cannot occur until the obstructing arch has been destroyed Furthermore, the wedging action
of the walls and the stability of the arch ring prevents any slip at the abufments, so collapse cannot be brought
about by a flow factor based on the wall friction characteristic 4 and hopper slope angle 8. The powder arch
must therefore fail within itself, and in fact it does so by crumbling along the plane of the principal shear
stress forming the core flovr channel, inclined at B. to the vertical axis of the hopper. In order to derive a
flow factor which wil! ensure failure of an arch formed by the static stress field system, Jenike’s dynamic
equations must be modified to take into account initial flow along the plane of the principal shear stress,
whtch is also inclined to tne direction of the maJor principal stress, C,, at 45”. Angle q is normally defined
as shown in Fig. 6(b), bu. in this particular case q. describes the obtuse angle between Xi and 8, and so
assumes the specific value of 3n4. The equations are now written as

dS s’ sin 2t70+sin(6,+217,)+(m-l) s’ sin d [cot e,(l+cos 2ao)-sin 2r10]


-=
de0 cos 2~~ - sin 6

and

dV0 -l-[(m-l)S’sm6(1+sinb)(cote0sin2~,+cos7~,-1)]
-zz
de0 2s’ sin d (cos 2i~e- sin 6)

_ [cos e. - sin 6 CDS(e, + 2~1~)+ S (1 -sin* a)]


2s’ sin 6 (cos 21, -sin S)

and solved numericaIly for S’ in term& of &, and q,, using an EnglandlO modified Runge-Kutta sub-routine.
Starting conditions are taken as fJ,=O, q,,= x/2, but Se has to be guessed to obtain the desired intersection
HOPPER DESIGN FOR COHESIVE POWDERS 225

I I I I # I 1
10 1, 7.2 73 14 75 ‘16 17 16 19 20

Flow-factor. If,

Fig_ 7 Strong-wall flow-factor chart (ax]-symmetrical hoppers).

at 9,, and qO = 3nj4. Angles 6 and B. are related, as will be seen later, *nd must be determined at the s .resses
derived from the static stress field system.
The function S now determines the strong-wall flow factor, ff by the relationship
mS’ (I + sin 6)
_K = (23)
2 sin e.
where 28, = $=_ It turns out that strong-wall flow factors are primarily functrons of 6 with minor modification
due to variations in indivrdual flow property characteristics summarised by eL_ This will be evident from
Fig 7 for axi-symmetrical flow channels. The lower part of the chart is bounded by the condition $r= 6
(when II= 1). There is also a practical upper boundary when n is approximately 2 2, which we have also
roughly indicated.
Generally it will be seen that strong-wall flow factors are somewhat larger than Jemke’s, and are related
to hopper geometry and wall friction characteristics only so far as these factors influence the static stress
field system at the outlet. Because of this, they fully take into account the effect of surcharge pressures, a
factor which has been found in practice to contribute to arching in a way that IS not predictable by the
Jenike theory. This matter will be discussed further later

WALL FRICTION AND THE ANGLE OF INCIPIENT MASS FLOW

It may be thought from the foregoing that, having broken the arch internally along the plane of the core
Cow channel, a mass-flow hopper design is no longer important. This is not so for two reasons Firstly, it
is essential to develop flow at the walls as soon as possible to prevent core flow becoming established through-
out the hopper with the attendant danger of “rat-holing” And secondly, once mass flow is established.
pressures in the outlet region are greatly reduced, allowing a greater margin of safety over arching conditions
for subsequent withdrawals. Should any physical change in the material take place during storage, such as
226 P.-l- STAINFORTH, R C ASHLEY

Fig. 8 Z-Mohr cmzle for pant of mctpient mass flow (axI-symmetrical hoppers)

caking due to moisture condensation, the conservatrve strong-wall flow factor will probably be large enough
to cater for most operational contingencies. We now determine the cntical hopper angle, 8,, for the worst
stress conditions in the hopper (i.e. at the centre of peak wall stress) in the following manner.
Consider first of all the c2se of a shzllow hopper with an outlet of sufficient size to allow core flow to
develop within the contained material_ The yield locus for the prevailing outlet stress conditions is 2s shown
in Fig 8 An element of that powder IS subjected to a longrtudinal normal stress cL, a s’rearing rLH and 2
horizontal normal stress G”, g iving rise to a major principal stress Z,. When flow takes place, the element
cleaves along the plane of the principal shearing stress, rmaxrwhich is inclmed at an angle &, to the vertical.
The included angle of the core flow channel is therefore 2&, whrch will be seen to be equal to \jlr_ from the
geometry of the Z-Mohr circle.
Now any point of the Mohr circle diagram represents the normal and shear stress on some plane in
the material. The intersection point of the wall yieid locus with the Mohr circle at W (where the norm21
stress is o, and the shear stress is r&) represents the normal 2nd shear stresses on the plane of the hopper
wall. If now the slope of the hopper IS made just sufficiently steep for failure to occur there as well as within
the material itself, then the principal shear plane of the powder will make an angle 0s with the hopper
wall, which will itself be inclined at an angle 8, with the vertical. The included angle of the hopper -walls for
incipient mass flow is then
HOPPER DESIGN FOR COHESIVE POWDERS 227

2e1 = 2eo+2e2 (24)


From Frg. 8 it will be seen that
28, = i.,+ arcsin (sin I.,/& 6,) - n/2 = $Q_ (2%
and
20, = n/2 - 4, - arcsin (sin &,/sin 6,) (25)

leading to

28, = (A,-$,)+arcsin(sm &/sin a,)-arcsm(sm +,/sin 6,) (27)


If now the hopper is made even steeper so that 20; < 201, point W on the Mohr circle rises to W’, making
s& ~5,~ (the minimum shear stress required for slip), and flow ram the hopper is wholly withm the mass
flow region. A shallower hopper than 28, puts W” Selow W, so that rFOc T,~, and slip at the walls cannot
occur. If the angle of wall friction 4 becomes so large that point W rises to P, then angle e2 disappears,
and slip occurs along the plane of the principal shear stress Conical hoppers steeper than the core flow
channel, &JO, therefore give rise to plug-flow condmons of discharge similar to that occurring m the cylmdrical
trunk, and from the engineering point of view are rarely economical In this situation recourse is usually
made to wedge or chisel-hoppered bins, in which the conditions for flow at the walls are not nearly so exacting
For this latter type of hopper the material is deemed to flow with the direction of its major principal
stress very close to the hopper wall, so that there is little change m the stress system pattern from the initial
static state to the dynamic. In the Mohr circle diagram of Fig 8, angle Bz is now given by
28, = rr/2+arcsin(sin 4,/sin a)-&, (28)
which on combination with eqn. (25) gives
28, = (A,-&)+arcsin(sm &/sin 6,)+arcsm(sm +=/sin 6,) (29)
the only difference between the above and eqn. (27) being the change in sign of the last arcs n involvmg &,_
For transition hoppers, however, where the flow pattern is less easily predrctcd, it IS scfer to use twice
the angle given by eqn. (27) for axi-symmetrical flow conditions This double angle does in fact correspond
approximately to the dotted line of Jenike’s charts for plane flow, so, for the sake of a conservative design,
we too recommend its adoption for all types of plane flow hopper.
Because Jenike selects his hopper angle from outlet stress condmons, it has been customary tc oren te
at point W’ on the Mohr circle of Fig. 8 well above W by making
2e = 28, - 100 (30,
This ensures that small changes in wall friction characteristic do not carry W ?_iL3the core flow Lone, and
also caters for any change in the angle of incipient mass flow that might occur higher up m the hopper as
the peak pressure centre is neared. This is because angles i. and 6 both decrease with increasmg stress,
thereby demanding a steeper hopper angle for incipient mass flow than that re *uired at the lower stresses
of the outlet. For this reason we prefer to determine the angle of incipient mass flow at the peak pressure
centre to ensure that the worst condition has been covered, and to obviate zny danger of core flow and a
sizable rat-hole developing at that point. This is especially important with the more cohesive powders, which
also exhibit angles of wall friction thaL are not constant with increasing stress. These latter, we find, have a
wall friction characteristic that is best sattsfied by the straight-lme relationship
rti = a+aG, (31)
where the constant a is indicative of the residual wall adhesion found in the majority of the powders tested.
This equation gives

4, = arctan ;+a
L 1 (32)
HOPPER DESIGN FOR COHESIVE POWDERS 229

The appropriate value of S, s or 3 now determines the correspondmg internal frictton angles at the outlet
for the requrred strong-wall flow factor, and agam this IS brought about by first calculating the ttme-con-
solidated value of T, in the arch rmg from
cl,v [I”’ v(-)‘n+Il(V-
n 1=-
1) SS
2 sin 0 [
KVi’“-_I3(V-
6 1
1) (r-i)‘Y
(38)

where V= (1 + l/T,). The computation is usually started by putting ff = 1.5 in eqn. (10) to obtain an Initial
value of Tg_ From this a value of B may be calculated for use in eqn. (38) above to determine successively
T,, i, 6, IjlL,s’, and a better value of fl for the second and subsequent iterations.
It will be noted that while the “flow” value of T: IS used in the calculation for B, the static stress T, is
used to calculate the internal friction angles which control the strong-wall flow factor ffs, and subsequent
iterations are always referred back to the initial stress system values. This is necessary to ensure the correct
flow factor for arch failure, and to prevent it from dwindling with rising 6 to an unsafe value on switch-over
from static to dynamic conditrons, as indicated by Frg 1.
Because flow factors have been pegged to static outlet pressures, they are directly influenced by the pressure
drstrrbution pattern developed during hopper charging Incompressible materials, which generally produce
a peak pressure centre at or near the outlet, are hkely to have their controlling angle 6 in the range of 40-50”,
which is in accord with Jenike’s work on iron ores, coals and anthracites_ Compressible materials, on the
other hand, will have their peak pressure centres at or near the transition, which could lead to a high outlet
6 and too low a Jenike flow factor for safety as shown in Fig. 9. In this latter case use of the larger strong-wall
flow factor restores the situation, as will be seen from the example that follows.

THE COMPUTER SOLUTION “HOPACALC”

While strong-wall flow factors may be presented in the form of easily read, single charts for both axt-
symmetric and plane flow, we cannot get over the drtliculty of obtaming correct values of 3 and G/Lat static
230 P. T STAINFORTH, R. C ASHLEY

outlet pressures without resort to a computer_ Nor can we be certain of an exact intersection of the flow factor
and failure function hnes without gomg deeply into the theoretical analysis. We therefore advocate a fully
computed solution, and so have written a conversational programme called HOPACALC in Fortran for
general termmal use. This programme carries out the calculations by stages ; firstly determining the stresses
of the cylindrical trunk, then the angle of incipient mass flow for the hopper at the point of peak static pressure,
and finally the crrtical outlet dimension for no arching. If practicable, HOPACALC ~111 also produce a
core flow solution for a 60° conical hopper with an outlet large enough to obviate rat-holing All the relevant
static loading stresses are also determmed for structural design purposes.
For a worked example of our method we have chosen a severely cohesi e ground polyvinyl chloride
polymer described here as PVC/55, not only because rt presents a very diffcult bulk handling problem,
but also because rt illustrates the chief differences between Jenike’s graphical so rution and our own analyttcal
method
The “instantaneous” and l&day trme-consolidated yield loci of Fig 4 were found m a series of Jenike
shear tests, the data of which were also processed in the POWDERFLO programme to give the flow property
characterrstlcs of Tables 1 (a) and (b). This mformation was then used both to draw the material failure
functron E-1; the X-d, and the X-p plots of Frg. 9, and to provide the data for HOPACALC. The wall fraction
characteristic for PVC/55 against polyurethane resin-coated steel was determined by a second series of
tests in a Jemke cell, and IS shown as the superrmposed plot of Fig. 4 giving a and CLvalues of 13.39 and
0.4863 respectively_ It IS now required to destgn a conical-hoppered silo of 50 tonnes capacity having an
equal trunk height and dtameter of 4 metres
Reference to Jenikds flow factor charts3 would indicate a hopper half-angle of 14O and ff value of 1.18
for +=27O and 6=61“, the latter being extrapolated from the PVC/55 plot of Frg. 1. Alternatively, had a
near straight line been assumei for the X-6 relationship of Fig. 9 (which would have been quite justifiable
from the expertmental data plot without special knowledge), a 6 of 35O would have resulted, giving ff = 1.54
and a= 12O. Since the difference m outlet size prediction is quite large (50 cm against 75 cm), a designer using
the graphical method would feel safe in assuming the latter to be a conservative answer. Unfortunately,
this material has been known to arch over even greater diameters m similarly steep hoppers-a fact not
explainable by the Jemke theory, and one which prompted the present development of a more exact analytical
approach.
In the computerrsed solution, the calculation proceeds along the lines already discussed : firstly, the evalu-
ation of the trunk pressures and parameters J and Q, then the determinatron of 0, z and S for the hopper,
and finally the calculation of the critical outlet dimension B together with the wall normal pressures at I
and B. The results are as set out in Table 2, but are best presented diagrammatically as in Fig. 10 These
require little comment_
Because a safe value of 0 has been found at peak hopper static pressure the result is one degree steeper
than that selected from Jenike’s charts. Moreover, the safe outlet size for no arching has been increased
to 120 cm, the critical dimension J3 being more m lme with our practical experience with this material_
There is, however, one further point of interest that is worthy of comment. Readers will have noted the
significant difference in the angle of incipient mass flow at the outlet and at the peak pressure centre. With
commonly found compressible materials of a not too cohesive nature, some economic advantage may be
taken of this feature by the employment of two-slope or multi-slope hoppers on large-scale installations.
But, unlike the hyperbolic shape suggested by Richmond’ a and Walker’ for handling incompressible
materials such as crushed rock ores and tine coals, this design would start with its steepest angle at the transi-
tion (where the pressure is highest), maintain this slope to just below the peak pressure centre, and then
switch to the shallower mass flow angle required at the outlet. Such a configuration would prevent core
flow developing in an intermediate zone because mass flow above and below wou!d always destroy a stagnant
area as soon as it tried to form.
While there may be few major points of difference between the analytical method presented and Jemke’s
graphical solution, we consider the former has some practical advantage in highlighting the trunk and
hopper stresses for the purposes of engineering design. Also. the computer solution allows several combina-
tions of hopper diameter and trunk height to be tried until the most satisfactory design is determined. Most
important of all, it removes much of the uncertainty of the material flow propetiy analysis. and restores
232 P. T. STAINFORTH. R. C ASHLEY

+dI,‘;
PK
--_-__-__3_-_____-_____-_---_--_-_ 8848

1
260

PF
866m
1, 1
‘1, p=
559m p $--- ----a 1766

I\ e 12m
I\
‘r B/l 1 ---- 2512

! i
_i_-t____-2 vertex -26
I
200 400 600 800 1000 1200
Wall narmol ~reswre kQ/m2
Fig 10 Dlagrammatlc representatzon of 50-tonne PVC/55 sdo design.

and K, its characteristic ratio ofcohesion, C. to tension, T-and from an rmportant variable, T,, called specific
tension and delined as T divided by its corresponding normal stress D,__The crtttcal value of T, may be cal-
culated from

containing the constants n and K and a fourth matenal constant Y, the yield factor, which links yield strength
at the free surface, f,, directly with tension, T. The last unknown is the flow factor ff,, which is a function
of material flow properties.
(2) In order to ensure that a potential arch breaks under all stress conditions at the hopper outlet, we
propose the employment of a “strong-wall” Ilow factor, to distinguish it from Jenike’s. This is determined
at the outlet stress conditions developed during hopper charging, and assumes that an arch first breaks
along the plane of the principal shear stress and that no slip at the walls takes place until flow begins. By
so doing the effect of trunk surcharge pressure is fully taken into account.
(3) Where the time-consolidation effect is brought about solely by increase in bulk density, the fully
time-consolidated locus can be extrapolated from the continuous flow locus by carrymg out one tirne-
consolidated shear experiment at point (Tm, 8,). All steady-state loci can, m practice, be well represented
by straight line functions

which appear to converge on a focal pomf TA, thereby enablmg the time-consolidated constants 5 and B
to be derived with negligible error.
(4) We confirm that the functional relatiocjhip between tensile strength T and bulk density p,
2-Q cp”

appears to hold good for most bulk solids down to a discontinuity at minimum bulk density p,,. when
HOPPER DESIGN FOR COHESIVE POWDERS 233

fJ= 1. This allows the critical span B of an arch to be expressed in terms of specific tension T, as
5 (>-1)1,
B = mYdJJ
l/T,)““_B//T, 1
the latter being determined from the flow factor relationship given in (1).
(5) This method, which is most conveniently carried out on a computer, removes the uncertamty of extra-
polation present in the existing graphical solution. Such a programme (HOPACALC) has been written irl
Fortran, and allows different ratios of hopper diameter to trunk height to be tried quickly until the most
economIca configuration is obtained.

LIST OF SYMBOLS

diameter of rat-hole circumscribing outlet, m


constant of the wall friction characteristic
critical span of arch or dome above outlet, m
constants of the continuous flow locus; constants of the time consohdated locus
cohesion, cohesive strength of powder in ratio K = C/T
constants of the powder density/tensile strength function
maximum diameter or width of hopper, m
hopper diameter at peak pressure centre
effective angle of internal friction
compressibility coeflicient
material failure function, Z/f
unconfined yield stress of powder compact
strong-wall fiow factor
rat-holing function, core flow hopper analysis
head of material in cylindrical trunk, m
compound angle = 4 + arcsm (sm &/sm 6) + 7~; 3x12
half-angles of hopper or flow channel
Janssen’s expression, 4k, tan 4, for cylindrical bins
constant, ratio C/T
Janssen’s function = (1 -sin 6 cos z)/ (1 + sm 6 cos i)
Jenike’s function= [1+~(m-1+~)]/~1+~+1~(m-1--_)]
complex angle of Mohr circle = arcsm (sin 4/sin 8) - 4
kinematic angle of internal friction
tan I, = T&G=
vertical heights above hopper vertex for peak pressure centre
expression [(Y - FF) T, - 1-J
function_ S tan 4
fGnction. sin 8+ kl (cos O-sin 6 tan 4)
factor, 2 for axi-symmetric flow, 1 for plane flow
index of yield locus equation, (r/C)“= (G+ T)/T
exponeniial function (1 -eeJH’“) in Janssen formula
radius of the Z-Mohr circle
length of radial stress ray from hopper vertex
bulk density of powder compact; density at point FF = 1
kinematic angle of wall friction
static angle of internal friction (at free surface)
dynamic angle of internal friction (at steady-state point)
radial stress field functions, static loading conditions
radial stress field function, dynamic loading conditions
23< P. T STAINFORTH, R C ASHLEY

L x2 major and minor principal st -sses


G, GL normal stress, steady-state nc nal stress
0.3 wall normal stress in polar cc xdinates
=7 fLH shear stress, steady-state shear sr::ss
=ro wall shear stress in polar coordinates
T tensile stress of powder compact
t tan 4 = [(Q/G~) + c]
T, specific tension, T divided by corresponding value of rsL
V expression (1-t l/T,)
W function, [l + tan($/2)]/[1- tan(J1/2)]
X expression, 2 (tan 0 -tan + + l/k,)
x function, n (1 -t T,)
YOiY constant, K/n ; and tan y?
Y yield factor, 2rry(Wy + l)/(w - pry)
Z complex function, Sk,/(l+ k2) sin 0
z peak pressure centre height ratio, 1,/l,, also d/D

REFERENCES

1 H A Janssen, Versuche ii Jer Getreldedruck ID Sdozellen, Ver Deut. Ingr. Z ,39(1895) 1045-1049
2 M S Ketchum, The Des gn oJ Walls, Ems and Gram Elecators, McGraw-Hdl, New York, 1929
3 A W Jentke, Gravtty flr#w of bulk sohds, Bull 108,4r~ Expt Sta , Utah State Umv., 1961
4 D. M Walker, A basis ior bunker design, Powder Tech&, 1 (1967) 228-236
5 P. T Stamforth, R C. Ashley and J. N B Morley, Computer analysis of powder flow charactenstics, Powder Technol. 4 (1971)
250-256.
6 R Farley, M D Ashton, D C-H Cheng and F. H H Valentm, Some mvestlgatlons mto the strength and flow of ponders, RheoI
Acta, 4 (1965) 206-218
7 D M Walker, fi rl approximate theory for pressures and archmg m hoppers, Chem Eng SCZ, 21 11966) 975-997
8 A W Jemke and _ R Johanson, Bm loads, J Stnrct Dru , Am Sot Ctuil Engrs,94 (1968) 101 l-1035
9 D M. Walker an1 M H Blanchard, Pressures m expenmental coal hoppers, Chem Eng SCI ,22 (1967) 1713-1745.
10 R England, Error esttmates for Runge-Kutta type solutions to systems of ordmary ddferentlal equatloas, Computer J., 12 (1969)
166-170
11 0 Richmond, GI avlty hopper destgn, Mech Eng . 85 (1963) 46-49.

APPENDIX

Al DERIVATION OF THE I_ AND Z-MOl-%Z ClRCLE RELATIONSHIPS

C’onsider first of all the f_Mohr circle of Fig. 2 By drfferentiaiing the yield locus equation
” a+T
0c
7

=- T
e-1)
developed by Ashton, Cheng. Farley and Valentin6, we obtain the slope at the free surface
dr T
z= tan*== = n(o+T) (A 2)
Also, where the yield locus cuts the r-axis at cr= 0, t = C, therefore the slope, tan JIc,,is given by

The ratio CfT=K 1s shown to be constant in our previous pape?.


HQPPER DESIGN FOR COHESIVE POWDERS 235

But from the geometry of the f_Mohr circle


f-2G
Y=r

Substituting for G and combining the above equation (AZ), v.e obtain
f/2 = my+ r/ny- T (A-4)
and m eqn. (A.l)
7” 7

0 c =- W”

from which

and

(A 5)

Again, substituting eqn. (A-5) for T in (A-4)

f/2= C (y + &-)($)l’-- T (A 6)
giving
f/T= 2[(ny’+ l)(y&~)~‘(~-L)- 1] = y, say _ (A-7)
But

so eqn. (A-7) may be rearranged to give

y = 2nY (WY + 1)
W-ny (A-8)
Since n, y0 and y are a11constants, it follows from eqn. (A-8) that Y is a constant for any particular powder
and, from eqn. (A-7), that f: the yield stress at the free surface, is directly proportional to tensile strength T.
Now considering the IX-Mohr circle, the slope of the yield locus at the steady-state stress point (Q, T&
will be given by
~l_u tan ?b
“IIIpIL= (-4-9)
+L+ n =n(li
Putting x=n(l+ T,), then

tan JIL= tan ;I/x (A-10)


Also, from the geometry of the Z-Mohr circle
Z-R--a,
tan $L = = 1 /tan1 (All)
*L?i >
from which
1
tan$Ltani.=C --
CTL[ l+sin6
1
1 (A.12)
236 P. T. STAINFORTH, R. C. ASHLEY

whence
ZE
0, = (It-sins) 9 + 1) (A-13)
(
also
R2--r& = (Z--R-q)*
which on expansion and rearrangement gives

(A.14)

Hence

($)2 ps] - 2(c) [l+tin6] + sec2i.=0 (A-15)

Solving the quadratic for C/G,


c
-=--- 1
1, I-=J] (A-16)
GL l-sin b [ (
which on combination with (A-13) and squaring gives

[cos%(~+l) - iJ2=1-g
from which

(tanZ+ X
l‘i’--&
I
(tan2j-: 1>+ (‘~o;;;j-j
X
= o

cosz b tai:' I.+


x
x
>z = 2 tan2 It-x-x
x
tan2 /,

cos b = g +2x tan2 i. -x2 tan’ 1.) i


x+tan2i
and
sin 6 = tan i. (x2 + tan’ A)*
(A.17)
x + tan’ 1

or, by substituting x tan I/J,_for tan I.


x sin J/L
sin 6 = (A-18)
cos’ I,/J=
+ x sin’ &_

A-2. DERIVATION OF THE RELATIO~%HIP BETWEEN FF, SIN 6 AND I/TS

Now, substituting eqn. (A-18) and tan +==tan i./x in eqn. (A-13), we obtain

- = 1 +x tan’ JIL[l+ ( l+l/tan2 $,)*I


G,
= 1 t-x tan2~~(l+cosec JI,)
x sin $L
=l-l- (A-19)
l-sin JIL
HOPPER DESIGN FOR COHESIVE POWDERS 237

so by combmmg eqns. (A-8) and (A-19), and putting ff= C/f and (YTJFF- 1 =M
x sin l/l=
M= (A-20)
l-sin I+&
or
sin $t_ = M/(x + M)
which yields the expression on putting tan I/J,_
= L/x

+;(1+2$T (A.21)

Equation (A-18) now transforms into


M2+xM
sm S = (A 22)
M2+2M+x

= (M+1)2+(x-2)(M+l)-(x-l)
(M+1)2+(x- 1)
(x-1)(x-2) + 2(x-1)2
(A-23)
(M+1)’ - (M+1)3 (M+ 1)1

_ (r1-l)(n-2)/T,~+(2n~-33n)/~~+,r~/T, + __ _
(A-24)
(Y FF)3
Smce T, and FF are variables, the sm G/FF(T,) relationship is very complex. However, for very cohesive
powders with n near to 2, reasonably good fits are obtained with a simple hyperbolic. As n approaches unity
the curve becomes less steep and IS often satrsfied by a power function (see Frg. 1) Now in eqn. (A-21) put
(It l/T,)= l”: then
M Y.FF+l--I/ M Y FF+l--V
-= and - ==
I nv ’ L KV””

This leads to
nV
= V(n-2)+2Y- FF+2

1
K2 V2/” n(Y- FF- 1/Ts)2
(A-25)
~ V = FF+n)+(n-2)/T,

giving
(l+l/T,y2-““” = j$[(2y.!&yg$y;),T
s1 (A 26)

which is soluble for l/T,, knowing n, K, Y and FF.

A.3 DETERMINATION OF TRUNK PRESSURES, p, AND P.

The geometry of the Mohr circle relates horizontal and vertical pressures close to the wall with major
and minor principal stresses

3(x1+&) = -$(p,+p,) = cl_ [Z / sin6]


HOPPER DESIGN FOR COHESIVE POWDERS 239

A 4. DETERMINATION OF INTERNAL FRICTION ANGLES AT STATIC OUTLET PRESSURES

By detinrtron of the radial stress field functron, S


C, = prS(l+sin 6)
but from eqn. (A-13), XI 1s aiso given by
z, = o,(l+sin 6)(1+C/x)
therefore

prs = (A) (1+ g> (A-32)

Now
Ifn

Thus

y[KV ““_,y(V-l)] = ;[K2V’2-n)‘“+n(V-l)] (A-33)

But

P = (;)“’ = [c{KV”.&V_l)JI,7and r= &


which transforms eqn. (A-33) into
ES
2 sin 0
~~-~)fi = c1,~
n
K2V(2--n)‘n+n(V-
1
1)
(A-34)

which IS soluble for V= (1+ l/T,) by successrve approxrmatron, entering suitable values for B and 8. This
enables the outlet values of the mternal frictron angles i, and ci to be calculated by the relationships already
given.
Additronally, V may be calculated at any diameter, d, along the radial stress field ray from the hopper
vertex to the peak pressure centre, by substituting d for B. At the peak pressure centre Itself, peak wall pres-
sure, pk. will be given by

(A-35)

where z = d,/D_ Similarly,


5)
(~-36)

AS. DERIVATION OF THE EXPRESSION FOR THE CENTRE OF HOPPER PEAK PRESSURE

The radial stress field assumes a triangular pressure distribution attaining a peak pressure pk at a height
I, above the vertex, and thereafter decaying to a pressure p,, at the transrtion (at I, above the vertex) due to
the surcharge of material in a cylindrical trunk of height H. The ratio I, /I, is put equal to z, as in eqn. (A- 35).
For the general case of a conical hopper surmounted by a cylindrical trunk, the total weight of solids
witbin the bin walls must be supported by the sum of the vertical components of hopper pressure and wall
frrction for static equrlibriurn. That is-
240 P. T_ STAINFORTH. R C. ASHLEY

II 12
?lD p, tan +dH+Xrr tan 0 (A-37)
0 0

where X=2(tan O-tan ++l/k,).


Now, by Janssen’s formula for wall pressures LI cylindrical bms,

p1 = g& (I-e-‘“‘D) (A-38)

where J = 4 k 1 tan @_
k _ l-sin6 cosi
and i = arcsin (sin +/sin S) - Q
* - l+sin6cosi
so total vertical support from the trunk walls is
H

7LD (A-39)
0

Thus the residual load carried by the hopper is

where Q = (1 -e- JH’D).


Resolving all hopper reactrons and frrctronal forces vertrcally, and calling each element ph, then for a
hopper of total height Z2, the total support for the contents will be given by
J2
X7r tan 8 1.7 ip,dl = Xx tan B 11 p,;dl21 + 11 PNl + s> (P,-Pddl]
0 0 J‘C
Xn tan 13
= 6 [PH(z~+/21,)+PN(21~--121~ -C)l (A-41)
where
pN= pV[sm 0+-k, (cos 0-m 0 tan +)] = NpDQ/J
Substituting zr =1,/l, and 21, tan O=D 111eqn (A 41), we obtain

_:7E tan 0 ‘2
lp,dl = 2f;aye CPH(l+Zl)+PN(2-Z1-Z:)1 (A -42)
6 I 0

Equating eqn. (A-40) with (A-42), and substituting,

F (z” +4-p~(2+2--2) = g [J+~Q tan 61 (A-43)

Now

PH
-= k,S
pD = ZpD, say
Z (1$-k,) sin 0
and

PN =

Thus substituting the above, eqn. (-4-43) yields

z2+z _

giving a positive root for z of


HOPPER DESIGN FOR COHESIVE POWDERS 241

5+Q(6 tan B-2NX) + 1 * -- 1


2’ (A.‘@)
[ x (.I2 - QN) J 2
For the special case of a hopper without surcharge pN=O, Q =0 and eqn. (A.44) yields a similar solutton
to that gven by Jenike and Johanson’:
z = [1/x2+$]* -f (A-45)
Also, when S is very large, which is usually the case when (3> 40°, I IS nearly zero and pK occurs at the
hopper outlet. Equation (A-49) then yields

pH+-2pN = g [J+6Q tan 191


and
pK = g [J+Q(6 tan 6’-2NX)] (4.46)

For the plane flow hopper case, consider a square-sectlon trunk surmounting a wedge hopper with two
vertical sides. The same calculations will also approximate to the special case of a circular trunk surmounting
a transitional or chisel hopper_ In this case Janssen’s hydraulic radius for the trunk is the same for both the
circular and square cross-sections, giving
J = 4k, tan 4 , as before
and
Q = (I-_e--‘H/D)
The total weight of the bin contents is then

(4+5 H
1
D3P 1 (A-47)

and the total frictional support in the trunk is

DQ(fg - 9) (A-48)

The residual load to be carried by the hopper is thus

D3P . (J +4Q tan 0) (A.43


4-J tan B

which must be balanced by the total vertical components of hopper pressure reactions and wall friction.
Thus, for the two sloping sides, vertical support is

The small amount of frictional support on the two vertical hopper ends is neglected for the sake of sim-
plicity. Equation (A-50) now integrates to

which on substituting L = I,/& and I, = D/2 tan 6 as before, and equating with the wetght of contents
carried by the hopper, gives
242 P. T STAINFORTH, R. C. ASHLEY

Substituting once again for PH and PN and simplifying, we obtain


X[z(JZ-ON)+qN] = J+40_ tan 0 (A-51)
which yieIds
J+e(4 tan 6-N-X)
z = (A-52)
x (JZ--QN)

A 6 ANALYTICAL SOLUTION OF THE STATIC STRESS FIELD EQUATIONS

The Jenike-Johanson’ differential equations are

$= (m+Z)Stan@+ (2) sin 8 and


z
d@ - -(m+2) + 1+2~~+m) + (z] cosz&-sin C#J
cos @ km-l) cot 8 + (ZJsin e]
dB-
Substituting p = s tan C$; a = m+2 ; P=(2) ; y= k, 1 1 +m(l-k,)

then
dS
- = S’= rxp+fl sin 0 (A.53)
de

- = p’= yS+/? cos 0-(m-l)p cot 8 (A-W


dt?
Differentiating (A-54) and substituting (A-53) gives
P” + (m - 1) cot 0~’ - (ay + (m- 1) cosec2 ejp = (y - 1) p sin 0 (AX)

(i) m= 1 cuse
Equation (A-55) becomes

sin e (A-56)

Boundary conditions ; B = 0 ; S = $, ; tan 4 = 0


f?=8, S=S , tan(f,=tan$ (data)
Solving (~-56) analytically gives
p = JAt%+C) [exp(E)-exp(-E)]-o sin 8 (A-57)
6
s,+c
s= 2 [exp(E)-exp(-E)]-C cos 0 (ASS)
where
A = 3(2--z)
G B=(y): C=(s); D=(A); E=B,/A
kz
By guessing S, and writing an iterative procedure to adjust So until
FfS = tan $ w-59)
eqn. (A-59) soon converges using a Newton-Raphson technique.
HOPPER DESIGN FOR COHESIVE POWDERS 243

(II) m = 2 cnse
Equation (AX) now becomes
$’ + cot @p’ - (ccxy
+ cosec’ 0) p = /I (7 - 1) sin 19 64-m)

A series solution for p satisfying the boundary conditrons is

p(e)=A [sin@ + (v)-7(e)] + fi(y)-7(0) (A.61)

where A is arbitrary ,&, (p’at O=O) = c-1 and r(6) satisfies

7”+cot B-r’-(ay+cosec20)7 = 8 sm 8 (A-62)

Integrating (A 61) gives


B
.d,=A[,_o,t?+(~)J’;7(B)dt?] ?$$ll7(0)d0 (A 63)
I 0
Integrating (A 62) gives
0 -‘(Gj+r(Q) COt 8-S(l-COS 0)
7(e)de = L (A-64)
f 0 W
and integrating (A-53) gives
B
S(e) = S,+a ‘ude+p(l -cos e) (A-65)
I0
Ehminatmg J”,7 (@de from (A-63) and (A-644)and then substituting (A-61) and (A-65) into ~(0) = S(e) tan 4
at e=B
aA+bS, = c (A-66)
where
n=sin8--n(I--cos8)tan$+(~)[~-7xtan$J~7(@)d8]

b= -tan 6

(A 67)

Equation (16) now yields

S = 2c-aB
0 (A 68)
2b+ ya
a and c contam functions of 7(e). These are obtained by solving the srmultaneous differential equations using
England’s1o Runge-Kutta method :
7’ = p
p’= -p cot 8+(ay+cosec’0)7+8 sin 8 (A-69)
with starting values 7(O) =p(O) = 0 and integrating (A-64) up to 0.
Summarising,
(a) Solve eqns. (A69) to give 7(e) and 7’(8)
(b) Substitute in eqns. (A-64) to give $0 r(Q)de
(c) Substitute these into eqns. (A-67) and (A. 68) to give So and A
(d) Use these values m (A-63) to give PO,u de, and fmally
(e) Use this in (A.65) to give S(@

You might also like