You are on page 1of 15

NUMERICAL INVESTIGATIONS

ON DISCHARGING SILOS

By Ulrich Haussler 1 and Josef Eibl 2

ABSTRACT: A numerical method to simulate discharging processes in mass-flow


silos is presented. The essential point is to formulate the appropriate consti-
Downloaded from ascelibrary.org by Universite de Liege on 06/01/22. Copyright ASCE. For personal use only; all rights reserved.

tutive law for a granular bulk material, which covers solid-like as well as fluid-
like behavior during discharging. An elastic-plastic law is chosen for the former
one, which is completed with a simple first approach .for fluid-like behavior.
As large and fast deformations occur, geometric nonlinearities and mass prop-
erties of the bulk material are considered with respect to an Eulerian frame of
reference. The complete set of field equations is numerically solved by the finite
element method spatially and by the finite difference method in time. Due to
the nature of the finite element method a broad variety of boundary conditions
can be studied. The method provides transient velocity and stress fields within
the bulk material for a first period of discharging. Remarkable stress redistri-
butions with strong increases of wall pressures are computed.

INTRODUCTION

While silo structures have become bigger during the last 50 yr, heavy
damages have occurred. This is mainly d u e to the poor k n o w l e d g e of
silo pressure especially during the p h a s e of discharging.
In the last 80 yr silo research in this field w a s mainly g o v e r n e d by
experiments. In addition a n u m b e r of pressure theories have b e e n de-
veloped, which however are not capable of describing the complex bulk
material behavior in a qualitatively a n d quantitatively satisfying m a n n e r .
While the behavior of the material at rest m a y be approximately de-
scribed by the well-known Janssen theory, it is doubtful, w h e t h e r closed
analytical solutions for discharging m a y be gained even in the future.
Trying to overcome these difficulties the authors have d e v e l o p e d a
numerical m e t h o d to compute silo pressures during charging a n d dis-
charging, which is presented in the following.
Using the well-established theory of continuum mechanics in a n Eu-
lerian frame of reference one has to start with 3 equations of d y n a m i c
equilibrium (Eq. 30); 6 constitutive relations (Eq. 12); a n d 9 kinematic
relations:

d = -(Vv+vvr) (1)

w = - (Vv - Vv T ) (2)

which form 18 equations altogether for the 2 x 6 c o m p o n e n t s of a, d,


and the 2 x 3 components of w, v. From these three groups of equations
'Dipl.-Ing., Univ. of Karlsruhe, Karlsruhe, West Germany.
2
Prof., Dr.-Ing., Univ. of Karlsruhe, Karlsruhe, West Germany.
Note.—Discussion open until November 1, 1984. To extend the closing date
one month, a written request must be filed with the ASCE Manager of Technical
and Professional Publications. The manuscript for this paper was submitted for
review and possible publication on May 17, 1983. This paper is part of the Journal
of Engineering Mechanics, Vol. 110, No. 6, June,'1984. ©ASCE, ISSN 0733-9399/
84/0006-0957/$01.00. Paper No. 18914.

957

J. Eng. Mech., 1984, 110(6): 957-971


the constitutive law and its appropriate formulation implies the essential
problem. As granular bulk material may in some aspects be regarded as
solid and in other as fluid, an incremental viscoplastic constitutive law
is used to describe the material behavior.

VISCOPLASTIC CONSTITUTIVE LAW


Downloaded from ascelibrary.org by Universite de Liege on 06/01/22. Copyright ASCE. For personal use only; all rights reserved.

For reasons of objectivity


a = & + a-Vf — w-ff (3)
is chosen as measure for stress rate. This is the well-known co-rotational
stress rate due to Jaumann. Also to cover viscous effects an objective
measure for strain acceleration is introduced:
d=d+d-w-wd (4)
This is similar to the second stretching tensor:
(2)
d = d + d-w- w d + d d (5)
defined by Truesdell and Noll (13).
The co-rotational rates a and d can be explained as material deriva-
tives of a and d in a frame of reference which follows the rigid motion
of a material point. They are distinguished from other objective rates by
the fact, that stress and strain-rate invariants remain constant for zero
stress-rate and zero strain-acceleration respectively, e.g.,
£ = 0 o i „ , riB/ lil„ = 0 (6)
Considering Eq. 3, and Eq. 4 a constitutive law in the following gen-
eral form is assumed:
*(d,a> = *,(d) + *„(*) (7)
Herein the Cauchy stress <y is divided into a rate independent part as
and a rate dependent part av:
a = a s + a„ (8)
The co-rotational rates of these parts are determined by
as = H d (9)
and av = G • d (10)
Regarding Eqs. 3, 4, and
3( )
C ) * ^ 7 + V()-v (11)
or
the local stress rate in an Eulerian frame reads
do- 3d
— =H•d + G ( c r w - W C T + Vffv)
dt dt
+ G ' ( d - w - w - d + Vd'v) (12)
Herein the physical properties of the material are represented by the first
958

J. Eng. Mech., 1984, 110(6): 957-971


and second part of the right hand side and the geometric nonlinearities
result in the third and fourth part on the right hand side.
Rate independent constitutive laws for granular materials regarding
large deformations have been proposed, e.g., by Gudehus (4), Nemat-
Nasser et al. (9-10). Based upon former successful calculations with an
elastic-plastic law for small strain given by Lade (7):
Downloaded from ascelibrary.org by Universite de Liege on 06/01/22. Copyright ASCE. For personal use only; all rights reserved.

ACT* = H * - A € J ! (13)
This is adopted in a first approach and modified to meet large displace-
ment requirements. The asterisks in Eq. 13 characterize small strain
measures.
The small strain tensor Ac* consists of an elastic component Ae*, a
plastic contractive component Aec* and a plastic expansive component
Ae*, such that
Ac* = Ae* + Aec* + Ae* (14)
The different components are shown in Fig. 1.
The elastic component is calculated by an isotropic hypoelastic law
with a stress dependent Young's modulus E and a constant Poisson's
ratio v:
Ae* = E-A<r* or Ae(e)i; — Ef/rs ' Ao"( s ) ra (15)

with Eijrs = 7; [(1 + v)8,r • 8/s - v • 8,y • 5re] (16)

The plastic contractive component Acc* is due to the partial collapse of


the grain's structure caused by volumetric compression. It occurs when
a yield condition,
/ c (Ia;,n fff ) - hc(Wc) = 0 (17)
is fulfilled, in which the hardening function hc depends on the plastic
work

Wc= I «r*:de? , (18)

The yield function fc forms a sphere in the principal stress space, the

Elastic
Plastic-contractive
Plastic-expansive

FIG. 1.—Stress-Strain Relation in Triaxlal Compression Test


959

J. Eng. Mech., 1984, 110(6): 957-971


size of which is determined by the value of the hardening function hc.
The plastic contractive component Ae* is calculated by an associated
flow rule:

Ae* = AXC- - ^ (19)


do?
Downloaded from ascelibrary.org by Universite de Liege on 06/01/22. Copyright ASCE. For personal use only; all rights reserved.

The unknown multiplier A\c may be found using the relation


Wc = Wc(fe) (20)
The plastic expansive part is due to a change of the grain structure
caused by increasing deviatoric stresses. It results in volumetric and de-
viatoric strains. The corresponding yield condition is
/,(n.!,ni.:) - K(wp) = o pi)
Different to the plastic contractive component Ae* , the plastic expan-
sive component Ae* is determined by means of a nonassociated flow
rule with a plastic potential g(l„flll„*):

Ae = A (22)
* V:S
OO s

Again A\ p can be found by a relation


Wp = Wp(fp) (23)
While fc may infinitely expand in the principal stress space, plastic ex-
pansive hardening is limited by a condition,
/„ =* i (24)
so that in principal stress space the yield surface fp has a limiting surface
T). During unloading and neutral loading both fc and fp remain at their
highest reached position.
With Eqs. 14, 15, 19 and 22 the elastic-plastic strain increment is given
by

Ae* = E.Aa s * + A X c ' f ^ + A \Pp - f ^ (25)


do* do*
This equation has to be inverted to give the elastic-plastic constitutive
tensor H*, resulting in a rather lengthly expression, which is not re-
peated here. The different material functions, respectively, material pa-
rameters fc, fp, g, hc, hp, Wc(fc), Wp(fp), E, and v, have been experi-
mentally determined by Lade (7). By means of this constitutive law, triaxial
tests with sand can be described well, (Fig. 2). For different granular
materials corresponding constitutive parameters may be found by the
same test procedure as performed by Lade.
The constitutive law given for small displacements Eq. 13 (25) is trans-
ferred for application on large displacements as follows. Taking the cur-
rent state as reference state, the materially fixed Lagrangean frame and
the spatially fixed Eulerian frame coincide, while the Lagrangean frame
has a velocity relative to the Eulerian frame. Now it seems physically
reasonable to relate stress- and strain-increments
960

J. Eng. Mech., 1984, 110(6): 957-971


Downloaded from ascelibrary.org by Universite de Liege on 06/01/22. Copyright ASCE. For personal use only; all rights reserved.

e,iv.l E( [V.I

r
e,[%

Loose sand

( — computed , •

FIG. 2.—Results of Computed and Measured Trlaxlal Behavior

Aa*
a* = Lim- k (26a)
<-»o Af
Ac*
d* = L i m — - . (26b)
!->o At
respectively, to the Lagrangean frame. Transforming those to the Eu-
lerian frame one gets
ffs-»as + (Tj'W-wo,s • (27a)
while d*^>d ••• • •••• (27b)
retains its form.
Transferring also H* from a Lagrangean frame to H in an Eulerian
frame one has to remember that, beginning with x(t = 0) = X(t = 0),

x(t) = X(t = 0) + vdt (28)

due the particle's movement in an Eulerian frame. As Eq. 28 can be


numerically integrated for every time step, the history of motion for every
material point may be traced.
The viscous behavior of fast flowing granular materials, which is rep-
resented by Eq. 10, was first examined by Bagnold (1), rate dependent
constitutive laws were proposed by Goodman/Cowin (3), Gudehus/Ko-
lymbas (5), and Savage (11), for example. For reasons of simplicity a
form analogous to the incompressible Newtonian fluid is chosen:
6\, = G • d or &rv)ij = 'Gijrs ' *-* rs (29a)

with G^ = 2li(8fr-8/,--oy (29b)

961

J. Eng. Mech., 1984, 110(6): 957-971


This introduces a new material parameter p,, called viscosity number
in the following. Unfortunately there are nearly no experimental or the-
oretical investigations concerning this viscosity number. Therefore it has
to be estimated comparing computed with measured flow velocities in
mass flow silos, the details of which will be explained later.
With these final assumptions a complete viscoplastic constitutive law
Downloaded from ascelibrary.org by Universite de Liege on 06/01/22. Copyright ASCE. For personal use only; all rights reserved.

in an Eulerian frame of reference is available.

EQUATIONS OF EQUILIBRIUM AND THEIR SPATIAL


AND TEMPORAL DISCRETIZATION

Assuming that the velocity field fulfills the kinematic boundary con-
ditions the equations of dynamic equilibrium can be formulated in an
Eulerian frame of reference by means of the principle of virtual velocities:

§dT-<rdV + 8vT-(v-b)prfy- 8v T -ldS (30)


v Jv JST

According to Eq. 11 the acceleration is given by


dv
v = — + Vv • v (31)
dt
Within the frame of an F.E. approach the velocity field is approximated
by
v = N(x) • a(f) (32)
from which d = B(x) • a(£) (33)
N = matrix of form functions and B = matrix of its partial derivations,
both only depending on x; while a is the vector of nodal velocities only
depending on time. In a first approach plane triangular elements with
three nodes, a linear velocity distribution and therefore a constant stress
and strain rate within an element are used. Gradients are computed by
interpolating between nodal values of stress and strain rate.
With Eqs. 30 H- 33 and the abbreviations:

M = I N T 'Nprfy (34)
Jv

Mc= NT-VvNpdV (35)


Jv

i = BT-(rdV •-...;:• (36)


V

p = I NT-bpdV+ NT-ldS (37)

we get from Eq. 30:


962

J. Eng. Mech., 1984, 110(6): 957-971


M • — + Mc • a + r = p (38a)
dt
da
whence — = M _ 1 • (p - M c - a - r) (38b)
dt
Assuming the mass per unit volume p to be constant in space and
Downloaded from ascelibrary.org by Universite de Liege on 06/01/22. Copyright ASCE. For personal use only; all rights reserved.

time, the mass matrix M is also constant. The convective mass matrix
M c and the nodal forces r depend on the velocity field, M c due to the
velocity gradient and r due to the constitutive law. Eq. 38b represents a
system of nonlinear ordinary differential equations of first order in the
unknown nodal velocities a depending on time t, which can be numer-
ically treated by the finite-difference method.
A set of first order differential equations,
3a
- = fM) (39)
dt
may be approximately solved by
a"+1 = a" + At [(1 - a) • f" + a • f"+1] (40)
n
in which a" = a{t ) ..<..' (41a)
n
with t = n-At, At = const .....(41b)
£" = f(a",t") (41c)
and a is an integration parameter. With a = 1 and using abbreviations
analogous to Eq. 41, we may derive from Eq. 38b, considering Eqs. 39
and 40:
1 \ 1
- • M + M? +1 • a"+1 + r" +1 = P n + 1 + — M • a" (42)
At / At \ ;
Assuming the tangential stiffness matrix K = dr/da to be positive def-
inite and neglecting the convective mass term, the integration scheme
Eq. 42 is unconditionally stable, i.e. truncation errors are not accumu-
lated for arbitrary time steps At.
In treating Eq. 42, the main problem is to calculate the nodal forces
r"+1. Eq. 36 leads to

r" + 1 = BT-a"+1dV (44)


v
whereby the stresses have to be determined by means of Eq. 12, which
may be abbreviated as
So- 3d
, x
— = Hd + G ht, T=-(aw - w a + VCTV)
dt dt
+ G • (d • w - w d + Vd • v) (45)
Herein H and T depend on stress and velocity, while G is constant. Again
using a one-step method analogous to Eq. 40 this equation may be ap-
proximated by
963

J. Eng. Mech., 1984, 110(6): 957-971


<r"+1 = a" + Af • (H" +1 • d"+1 + T" + 1 ) + G • (d n+1 - d") (46)
so that Eq. 44 can be written as
r"+1 = t" + At-Ksn+1 • a"+1 + At-rnn+1 + C• (a"+1 - a") (47)

Herein K,"+1 =. Br-H"+1'BdV (48)


Downloaded from ascelibrary.org by Universite de Liege on 06/01/22. Copyright ASCE. For personal use only; all rights reserved.

is the static tangential stiffness matrix, which contains the physical non-
linearities due to the elastic-plastic material behavior, while

C= BTGBdV (49)
Jv
represents the viscous material behavior and

rfJ= B r -T" +1 dV (50)


Jv
gives the geometric nonlinear part of the nodal forces.
So Eq. 42 may finally be written as

f — M + Mc"+1 + C + At-K"+A-a"+1 + Af-rf1

= p " + 1 _ r " + f—--M + c l - a " (51)

BOUNDARY AND INITIAL CONDITIONS—FURTHER CONDITIONS

The rather complex material behavior at the silo wall is approximated


by the following simple boundary conditions:
o» = p; (52)
», = 0 if o-f < tan 4> •CT„, otherwise CT, = tan $ • u„ (53)
in which v„, u„ are the components of velocity and stress normal to the
boundary, vt, <r, the tangential components and tan <$> is the coefficient
of friction between boundary wall and bulk material. The term

N T 'l" + 1 dS (54)
JSi

in Eq. 37 has to be calculated by iteration, as at time t = tn+1 the stress


component ann+l is unknown. The boundary condition at the outlet is <r„
= ff( = 0.
Because the results of the charging process are the initial conditions
for discharging, charging has to be computed too. Without going into
details it may be stated that after some modifications the charging prob-
lem can ajso be solved by the method described. Simulating this process
by incrementally loading the silo, the viscous and mass terms in Eq. 51
have to be omitted and the boundary conditions at the outlet must be
964

J. Eng. Mech., 1984, 110(6): 957-971


changed from a„ = a, = 0 to the ones given in Eq. 52 and Eq. 53.
In computing stresses it further has to be considered that no tensile
stresses are allowed and that the limit condition of Eq. 24 imposed by
the constitutive law must be fulfilled, i.e. the deviatoric stresses cannot
exceed a certain limiting value. The computed stresses have to be mod-
ified, if they violate one of these conditions.
Downloaded from ascelibrary.org by Universite de Liege on 06/01/22. Copyright ASCE. For personal use only; all rights reserved.

SOLUTION METHOD

Assuming velocities and stresses to be known for the time t = tn, Eqs.
51 and 46 allow the computation of the velocities and stresses for the
time t = tn+1. Eq. 51 is nonlinear, as M" +1 depends on v" +1 , r" +1 on v" +1
and <r"+1, as well as Ks on <r"+1.
The numerical computations simplify essentially, if we set o-"+1 — a"
when calculating K"+1 and r,"+1. This is admissible for small time steps.
Nevertheless the nonlinearities referring to v" +1 necessitate an iterative
solution method. At a time step f"+1 Eq. 51 may be written as

«|i(a"+1) = p" + 1 - — M• (a n+1 - a") - Mnc+1 • a"+1 - r"+1 = 0 (55)

For an approximate solution 'a"+1 the term t|i('a"+1) represents an error


vector. Roots of Eq. 55 may be found by a Newton-Raphson scheme:

— a = ; a" +1 ) -»K'a"+1) (56fl)


,9a /
or '+1Aa"+1 1 +1
= - J " • t|i('a" ) (56b)
1
For convenience J = — • M + M? + C + At • K? (57)

is chosen and kept constant through each iteration cycle until the pro-
cess is stopped, when |«|i| reaches a given small value. The iteration
starts with °a"+1 = 0.

RESULTS

As the static pressure distribution in a charged silo is of general in-


terest and as this state determines the initial values for the problem of
discharging, the silo at rest was also studied. Fig. 3 shows computed
principal stress fields and wall pressures for two plane strain silos out
of a series of investigated geometries. These values are compared with
experimental results given by Motzkus (8). Further comparisons with
experiments and German Standards for different silo shapes and also
rotational symmetric cases are given by Eibl et al. (2). However as the
static behavior of a silo is not the subject of this paper, in the following
the main attention will be concentrated on silo discharging.
Fig. 4 shows the FE-mesh for an arbitrarily chosen silo geometry, in
which symmetry and plane strain conditions are assumed.
Fig. 5 gives the velocity field at the beginning of discharging.
After a short time mass flow develops within the whole silo area, which
965

J. Eng. Mech., 1984, 110(6): 957-971


Downloaded from ascelibrary.org by Universite de Liege on 06/01/22. Copyright ASCE. For personal use only; all rights reserved.

Bulk material.- sand


Void ratio: 0,87
Angle of internal friction 41°
Angle of wall friction 29°

FIG. 3.—Computed Principal Stress Fields and Wall Pressures after Charging in
Comparison with Experimental Values

can be divided into two domains. Within the first, bin area, the flow
velocity is approximately constant for all spatial points, i.e. the bulk ma-
terial moves like a rigid body there. This area is bounded to the second,
hopper area, by a line indicated in Fig. 5. From this boundary line on
the bulk material accelerates, with the largest acceleration in the silo cen-
terline and the smallest at the wall. The flow velocity profile in a hori-
zontal plane is given in Fig. 6. Similar shaped profiles exist for all hor-
izontal planes in the hopper area. Flow velocities in fixed spatial points

Discretisation parameters
96 Nodes
192 DoF
140 Elements
Time step At = 0,002 sec

Bulk material: sand


Initial void ratio 0,87
Angle of internal friction 34°
Angle of wall friction 22°

FIG. 4.—-Discretization and Physical Parameters

966

J. Eng. Mech., 1984, 110(6): 957-971


Downloaded from ascelibrary.org by Universite de Liege on 06/01/22. Copyright ASCE. For personal use only; all rights reserved.

i i i . i l l . bin area
i
i
i
i i
i h-t-u
I / i i>-
i i i \ 1111

w
I i i / / //
hopper area
in'/ Hi
^ • pa i_
f = 0,10 sec t = 0,20 sec = 0,30 sec f = 1,00 sec

M s 1,00m/sec (3,28 ft/sec)

FIG. 5.—Velocity Fields

are shown in Fig. 7, where small oscillations can be observed in the bin
area, accompanied by oscillations of vertical stresses. Under steady flow
conditions the largest flow velocity in point 4 is about 4.5 times larger
than the flow velocity in the bin area. Assuming the same mass transfer
in the bin area and in the hopper area, the bulk material has to loosen
up in the hopper area. Its void ratio in point 4 must be about 3.7 times
larger than the void ratio in point 1. This is in contradiction to the as-
sumption p = constant.
Fig. 8 shows the principal stress field. Discharging begins with a re-
orientation of the principal stress directions in the hopper section. The
vertical orientation of the larger principal stress changes into horizontal.
Subsequently the stress level decreases above the outlet and increases
at the transition area between bin and hopper, while the stress level
within the bin area is retained. This stress redistribution is explained in
other silo theories [e.g., Jenike (6), Walker (14), and Walters (15)] as a

t WSI X

N- V
0 0.50 I W t 1.00 IW81 1,33 1^,36)
Distance from symmetry-axis, x, in m { f t ) Discharging time, t , in sec

FIG. 6— Flow Velocities in Horizontal FIG. 7.—Flow Velocities In Fixed Spa-


Plane 0.6 m (1.97 ft) Above Outlet tial Points

967

J. Eng. Mech., 1984, 110(6): 957-971


Downloaded from ascelibrary.org by Universite de Liege on 06/01/22. Copyright ASCE. For personal use only; all rights reserved.

r
I I 11
MM
MM
MM,

\W
t=0sec r = 0,05sec

i 1 8 0,5 Warn1 (72,5 psi)

FIG. 8.—Principal Stress Fields

transition from an active state of stress into a passive state of stress. Wall
pressures are shown in Fig. 9. There is a uniform increase of hopper
pressure during the initial phase of discharging followed by a decrease
near the outlet and a strong increase at the transition between hopper
and bin walls. The latter is known from experimental literature as the
so-called "switch." Vertical stresses in fixed spatial points are shown in
Fig. 10. In the hopper area first they decrease and remain constant sub-
sequently. Within the bin area oscillations at a nearly constant mean
stress level occur, ocrresponding to the already mentioned velocity os-
cillations. Due to changing boundary conditions at the outlet the vertical
stress is reduced, the bulk material is elastically released and oscillations
are initiated.

I = 0,000 [seel

t = 0,100 [sec!
t = 0,200 [sec]

t = 1,000 [sec]

£ IH8
\—|
N/mm1

r\
.«,
2
\\\
w vy
• • —
1

0.UK.5I O.Zln.01 0,3M,5I


Normal wall pressure ! « 111 ,/
-i^,, inN/mm'{psi)
\\
I H ..' -—- •-—
•-—i
2

0.2 0,1*

Discharging time , r , in sec

FIG. 9.—Wall Pressures FIG. 10.—Vertical Pressures in Fixed


Spatial Points
968

J. Eng. Mech., 1984, 110(6): 957-971


Downloaded from ascelibrary.org by Universite de Liege on 06/01/22. Copyright ASCE. For personal use only; all rights reserved.

m m bitid'
t = 0 sec r = 0,1s.

x : elastic : plastic

FIG. 11.—Material Response

Fig. 11 shows the types of material response, in which states of elastic


behavior are distinguished from those of plastic expansive hardening,
i.e. fp < f\, and those fulfilling the limit condition fp = i\. In the latter
plastic state, the deviatoric stresses have reached their largest possible
value.
During charging, a narrow band of limit states arises at the transition
between hopper and bin wall, while the remaining bulk material is in
the plastic hardening range. During the initial phase of discharging the
material mainly shows elastic behavior, followed by a gradual transition
to plastic hardening states in the hopper area, which are terminated by
limit states.
A major uncertainty remains in the choice of the viscosity number \L,
the physical background of which is nearly unexplored. Computations
with variable values of (x were performed. Fig. 12 gives the computed
flow velocity in the bin area and at the outlet versus the viscosity num-
ber for t = 1.0 sec, when steady flow conditions have developed (com-
pare Fig. 7).
According to Schwedes (12) the mean velocity at the outlet can be
assumed as f) — 0.5 • V g - D with the acceleration of gravity g and the
outlet-width D. In the case under consideration v — 2.2 (m/s) and a
value of (JL = 1 • 103 (Nsec/m2) is gained from Fig. 12, which was pro-
visionally used for further computations. Fig. 13 shows that pressures

a * r
^ ^ • ^ ^
„ ii =0.25!0!Nsec/m! IPO-KCI

i* /
f » li = 1,0 • 10!
i |i = 50 • 10'
S
S

IS 1
0 0,1 IKS) 0,2119,01 0,3143,51 0,t(M,0)
--£_.
Normal wall pressure i n N/mm 1 (psi) Viscosity parameter . p , in 10 3 'Nsec/m {Pa sec)

FIG. 12.—Flow Velocity Dependence on FIG. 13.—Wall Pressures for Different


Viscosity Number Values of Viscosity Number
969

J. Eng. Mech., 1984, 110(6): 957-971


nearly remain unchanged, although flow velocities are remarkably in-
fluenced by the viscosity number.
These are the results gained so far. Studies on eccentric outlets, dif-
ferent silo geometries producing plug flow or mass flow and deformable
walls are in progress but have not yet been developed so far that they
could be discussed here.
Future research will concentrate on the improvement of the numerical
Downloaded from ascelibrary.org by Universite de Liege on 06/01/22. Copyright ASCE. For personal use only; all rights reserved.

model and on material research. An experimental verification of the rate


dependent parts of the constitutive law is necessary. Alternative con-
stitutive formulations for the plastic behavior, including cohesion, must
be studied.

SUMMARY AND CONCLUSIONS

A numerical method for computing velocity and stress fields in silos


during charging and discharging is presented. The method incorporates
an elastic-plastic constitutive law for a cohesionless granular material which
is completed with a viscous part. Remarkable stress redistributions are
computed for discharging, where a strong increase of wall pressure,
switch, arises in the transition between hopper and bin walls.
Although the presented method is a first approach to numerical so-
lutions, the computed results agree well with experimentally measured
values for the case of charging, and appear physically reasonable for the
more complicated process of discharging according to the few reliable
measurements which are available.

ACKNOWLEDGMENT

The results presented in this paper were obtained in the course of


research sponsored by the Innenministerium Nordrhein-Westfalen, In-
stitut fiir Bautechnik, and Deutsche Forschungsgemeinschaft.

APPENDIX.—REFERENCES

1. Bagnold, R. A., "Experiments on a Gravity-Free Dispersion of Large Solid


Spheres in a Newtonian Fluid under Shear," Proceedings of the Royal Society
of London, Series A, Vol. 225, 1954, pp. 49-63.
2. Eibl, J., Landahl, H., Haussler, U., Gladen, W., "Zur Frage des Silodrucks,"
Beton- und Stahlbetonbau, Berlin, Vol. 77, No. 4, Apr., 1982, pp. 104-110.
3. Goodman, M. A., Cowin, S. C, "A Continuum Theory for Granular Ma-
terials," Archive for Rational Mechanics and Analysis, Vol. 44, 1971-72, pp. 249-
266.
4. Gudehus, G., "A Continuum Theory for Calculation of Large Deformations
in Soils," Veroffentl. Inst. Bodenmech. Felsmech., Karlsruhe, Germany, Heft 36,
1968.
5. Gudehus, G., Kolymbas, D., "A Constitutive Law of the Rate Type for Soils,"
Proceedings of the 3rd International Conference on Numerical Methods in Geome
chanics, Aachen, Apr. 2-6, 1979, pp. 319-329.
6. Jenike, A. W., Johanson, J. R., "Bins Loads," Journal of the Structural Division,
ASCE, Vol. 93, No. ST4, Apr., 1968, pp. 1011-1041.
7. Lade, P. V., "Elasto-Plastic Stress Strain Theory for Cohesionless Soil with
Curved Yield Surfaces," International Journal of Solids and Structures, Vol. 13,
1977, pp. 1019-1035.
8. Motzkus, U., "Belastung von Siloboden und Auslauftrichtern durch kornige
970

J. Eng. Mech., 1984, 110(6): 957-971


Schuttgiiter," Dissertation, Technical University of Braunschweig, Germany,
1974.
9. Nemat-Nasser, S., Shokooh, A., "On Finite Plastic Flows of Compressible
Materials with Internal Friction," International Journal of Solids and Structures,
Vol. 16, 1980, pp. 495-514.
10. Dorris, J. F., Nemat-Nasser, S., "A Plasticity Model for Flow of Granular
Material under Triaxial Stress States," International Journal of Solids and Struc-
tures, Vol. 18, 1982, pp. 497-531.
Downloaded from ascelibrary.org by Universite de Liege on 06/01/22. Copyright ASCE. For personal use only; all rights reserved.

11. Savage, S. B., "Gravity Flow of Cohesionless Granular Materials in Chutes


and Channels," Journal of Fluid Mechanics, Vol. 92, Part 1, 1979, pp. 53-96.
12. Schwedes, J., "Fliessverhalten von Schuttgiitern in Bunkern," Verlag Chemie,
Weinheim/Bergstrasse, Germany, 1968.
13. Truesdell, C , Noll, W., "The Non-Linear Field Theories of Mechanics," En-
cyclopedia of Physics, Vol. in/3, Springer Verlag, Berlin, Heidelberg, New York,
1965.
14. Walker, D. M., "An Approximate Theory for Pressures and Arching in Hop-
pers," Chemical Engineering Science, Vol. 21, 1966, p. 975.
15. Walters, J. K., "A Theoretical Analysis of Stresses in Axially-Symmetric Hop-
pers and Bunkers," Chemical Engineering Science, Vol. 28, 1973, pp. 779-789.

971

J. Eng. Mech., 1984, 110(6): 957-971

You might also like