Вы находитесь на странице: 1из 22

DESIGN AND ANALYSIS OF SILOS

FOR FRICTION FORCES


By Demetres Briassoulis 1 a n d James Curtis 2
No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.

ABSTRACT: The effect of axial friction forces on the calculated hoop stress of
an unstiffened silo, or of the top unstiffened portion of a silo is analyzed. Ig-
noring the beam-column interaction between axial friction forces and radial
pressure due to the grain may result in an error of 20% in some cases. This
results in a variable actual factor of safety which depends on the geometric
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.

characteristics of the silo. An instability analysis is carried out which sets limits
on the maximum unstiffened length of a silo. This is done on a probabilistic
basis and an algorithm is suggested for the design of an unstiffened silo or a
silo stiffened by rings.

INTRODUCTION

The design of the wall of a silo is based primarily on the lateral load
applied by the grain while the presence of the axial friction forces due
to the grain is ignored. The effect of the axial friction forces may, how-
ever, in some cases be significant.
J. Struct. Eng. 1985.111:1377-1398.

This paper deals with the analytical solution of the deflection in the
radial direction of the wall of a cylindrical unstiffened silo under static
loading due to grain. The objective here is not the derivation of a closed
form solution for the deflection, but, rather, is the investigation of the
effect of the axial forces due to friction from the grain on the hoop stresses.
To obtain this objective, a Fourier series solution for the deflection of
the wall is used. The analytical solution and the investigation of the in-
teraction between lateral and axial forces is limited to the linear elastic
region. For this reason, a complete analysis of instability of an unstif-
fened silo under the axial friction forces is carried out. This not only sets
the limits to the validity of the interaction analysis, but also establishes
maximum unstiffened length of a silo. Wind loading effects are not in-
cluded in this paper.

BASIC ASSUMPTIONS

Assume an unstiffened silo of height, L, with uniform thickness, t,


level full of grain, and pinned boundary conditions at the top and bot-
tom of the silo wall. This assumption was made primarily to make it
possible to use a Fourier sine series solution. In addition, many practical
cases can be simulated by such boundary conditions. Furthermore, as
will be shown later, the results of the paper apply to more cases of
boundary conditions than the previously specified one.
'Visiting Asst. Prof., Dept. of Agricultural Engrg., Univ. of Illinois, Urbana,
111.2
Prof., Dept. of Agricultural Engrg., Univ. of Illinois, Urbana, 111.
Note.Discussion open until November 1, 1985. To extend the closing date
one month, a written request must be filed with the ASCE Manager of Journals.
The manuscript for this paper was submitted for review and possible publication
on August 9, 1984. This paper is part of the Journal of Structural Engineering,
Vol. Ill, No. 6, June, 1985. ASCE, ISSN 0733-9445/85/0006-1377/$01.00. Paper
No. 19815.
1377
/

No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.

/
/
/
/
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.

/
/
/
/
//
/ /}////// //////// / > /'"rrrrr

- *
-+T
FIG. 1.Cylindrical Silo
J. Struct. Eng. 1985.111:1377-1398.

Imagine a unit width strip of wall cut out from the wall (Fig. 1). As-
sume axisymmetric loading conditions (static loading). Then on this strip-
beam the following forces act:

1. Lateral pressure, p(x), due to the grain (consider only the gravity
field), (Pa).
2. Axial friction forces, q{x), due to the friction of the grain on the
wall, (Pa).
3. The hoop forces, T(x), acting on the strip from the rest of the cy-
lindrical wall, (N).

Then, the total lateral force on the unit width strip-beam per unit of
length, p(x), is (2)
p(x) = p(x)rA9 - T(x)AQ (1)
Also T(x) = a e t = E^t (2)
in which the circumferential strain is
w
*e = - (3)

in which w = radial deflection measured as shown in Fig. 2. From Eqs.


1-3
Ewt
p(x) = p(x)rA9 A0 (4)
r
For the axial friction forces per unit length, q(x), the following relation-
ship is valid:
q(x) = q(x)rAQ = p(x)rA9(x' (5)
1378
T
No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.

W(x)
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.

FIG. 2.Forces Acting on Unit Width Strip-Beam per Unit of Length

in which \L' = tan 4>' = the coefficient of friction of grain on the bin
walls; and 4>' = angle of friction of grain on the bin walls.
For the lateral pressure, p(x), the Janssen's formula is assumed to be
valid (5)
mR
J. Struct. Eng. 1985.111:1377-1398.

v(x) = {l-e-^x,K) = v{l (6) ' )

in which m = weight of grain per unit of volume in kg/m 3 times the


acceleration of gravity in m/s 2 , (N/m 3 ); R = hydraulic radius of bin (=
r/2 for cylinder), (tn); k = ratio of lateral to vertical pressure of grain
(nondimensional); v = mR/|x'; and p = k\i'/R.

ANALYSIS OF WALL STRIP-BEAM

To formulate mathematically the problem of the wall strip-beam, con-


sider a section of the strip as shown in Fig. 3.
The equilibrium equations for the section AB shown in deflected po-
sition are

Vo - V(. ,-f Jo
PWZ = O. (7)

?({)
W(x)^
H
Cut

wall '
W(x)
strip-
beam

FIG. 3.Section of Wall Strip-Beam (Deflected Position)

1379
Q - Q(. qiOd^O, Q(x) = Q, ?(9d. (8)
Jo Jo

M + I | p(Q<% - | ;(8<?(dS - M M + xV(x) + w(x)Q(x) = 0 . (9)


No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.

Jo Jo
differentiate
dV(x)
= ~p(x). (10)
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.

dx
dQ{x)
= cj(x) ., (11)
dx
dM(%) dV(x) dw(x) _
*?(*) - w(x)q(x) - ^dx + V(x) + x ^
dx + ^
dx Q(x)

+ w(x) =0 (12)
dx
J. Struct. Eng. 1985.111:1377-1398.

Substitute Eqs. 10 and 11 into Eq. 12


dV(x) dQ(x) dM(x) ^ T7. . ^ dV(x) dW(x)
-x w{x) + V(x) + x 7 + - ^ Q(x)
dx dx dx dx dx
, , dQ(x) dM(x) dw{x)
+ w(x)^-+ L (13)
dx = 0 or r-
dx = V(x) + Q(x) dx ""
dw d$(x) M
Assume = (j>(x) = w" = w
dx dx EI
in which (x) = the slope of the beam at x. Then Eq. 13 becomes

-f(E^=-V(*)-0(*)^ (14)

differentiate
dw dV(x)
^ ( E l a O + f- Q(x) = p(x). (15)
dx dx Jx~ dx
The boundary conditions at A give
Q(0) = Qo = 0
Then Eq. 8 becomes with substitution of Eq. 5

Q(x)=\ m<%=\ mrM\i'dZ or Q(x) = rAO^' I p(# (16)


Jo Jo
Jo Jo
The moment of inertia for the strip-beam is given by
r&Qt3
I'- (17)
ll
1380
Substitute Eqs, 4.. 16 and 17 into Eq, 15
rAGf3 . d fff ~\dw\
dw\ Etl
EtAQ
E vf + rAQix' w = p(x)rAQ (18)
12 dx
12a' f
No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.

12 12
w v+ p{x)w + p(<%
' w\ ' r't* Et3
(19)

Using Eq. 6
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.

v(l - e-fi)di =v di + - e'^d(-pO = vx + - e" pt + C


o Jo P P

So - ^ i . = vx + - <rp* + C; Q(0) = 0 = - d Q=-- (20)


rAG(x' p p

Thus v ( l - e _ p 5 ) ^ = vx + - ( e ~ p * - l )
h P
Then Eq. 19 becomes
J. Struct. Eng. 1985.111:1377-1398.

12M.' , 12 12v
w +- vx + - (e""* - 1) zu" + v(l - e-|Mr)u>'} + 2 a> = 3 (1 - e~")
P J r r Et
W21)
12u/v 12u'v 12 12v
zvw + 4- x - - (1 - e"p
*) w" + -rrr- x
Ef3 P Ef3 (1 - e~> )w' +r
^
w=
Ef3
(1 - e^"*)

px
or K),0 + A x - - (1 - e" ) w" + A ( l - e-px)w' + Bw = C(l - e~ p *)... (22)
P
in which A = 12u'v/Ef 3 ; B = 12/r2t2; a n d C = 12v/Et3. Eq. 22 is the
governing differential equation for the wall strip-beam u n d e r the lateral
and axial (friction) static loading d u e to the grain. This equation, h o w -
ever, is extremely difficult to solve analytically a n d obtain the exact closed
formed solution for the deflection w(x). The main source of the difficulty
is the presence of the distributed nonuniform axial friction forces. This
is evident from Eqs. 15 a n d 18. If Q(x) = constant (thrust), Eq. 15 yields
a zero coefficient for w' a n d Eq. 18 is simply

ErA6f3 . EfA8
w'v + Qw" + w = p(x)rA8 (23)

But what could be an appropriate expression for a "constant thrust" such


that an analysis of the effect of the actual axial friction forces can be
reasonable? The answer to this question is the selection of a large enough
thrust that it represents a conservative axial force in relation to the actual
condition.
Fig. 4 shows the actual distribution of the friction forces o n the strip-
beam. This distribution represents a total axial force (from Eqs. 6 a n d
16)

Q a'v (1 - e-*x)dx rAG; Q = | u'v (1


_ -pL rA6. (24)
Jo
1381
No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.

fl(x)
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.

rrnrf

FIG. 4.Friction Force Distribution

This total axial force is not applied as a constant force throughout the
strip beam but it is the maximum axial force applied at the lower end
of the beam. Thus, it is conservative to "apply" this total axial force to
the full length of the beam. How conservative such an assumption is
J. Struct. Eng. 1985.111:1377-1398.

can be shown as follows: In the case of the actual distribution of the


friction forces consider the axial "resultant" acting on the center of grav-
ity of the distributed forces, Fig. 5(a). The same force is applied now to
the upper end of the strip-beam, Fig. 5(b).
As Fig. 5 indicates, if the axial force, Q, is applied at the top of the
strip-beam, the additional moment on the beam due to axial load is greater
than if the axial force is applied at a distance aL from the bottom, (a <
1). This means that it is conservative to assume that Q is applied on the
top boundary of the strip instead of the actual point of application of
the resultant friction force. Of course, this is a qualitative type of anal-
ysis. However, the purpose here is not to determine the exact portion
of the deflection due to the axial friction forces, but rather to investigate

assumed
C.G.of
distributed
axial forces
aL

rffr
(b)

FIG. 5.Actual (a) and Conservative (b) Axial Resultant Forces on Strip-Beam
Deflected by Lateral Pressures

1382
No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.

Pin)
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.

for givn = L, t . m . / i

error %

FIG. 6.Qualitative Deflected Shape and Error Distribution for Deflection under
Effect of Qb Compared to Zero Axial Force
J. Struct. Eng. 1985.111:1377-1398.

whether or not this deflection is a significant portion of the total de-


flection of the strip-beam.
To this end, the case of Fig. 5(b) will be assumed to apply. This case
will be referred to as Q = Qb (Fig. 6) in a later section when the effect
of the axial friction forces on the wall is considered. Under this condition
of the pinned-pinned end beam under thrust and lateral loading p(x),
the expected maximum deflection is greater than the actual one. Thus
the upper bound case is investigated, conservatively.

FOURIER SERIES SOLUTION

To investigate the effect of Q on the strip-beam deflection, two solu-


tions should be compared. One with Q = 0 and another one with Q
given by Eq. 24. Rewrite Eq. 23 as
Y1Q 12 12
wm + =-.vf + -r-,w = Mx) (25)
ErA0f3 r2t2 Et3^ ; ;

First represent p(x) as a Fourier sine series


A MTTX

pw = I, P" sin~r (26)


2 f mrx
in which pn = - I p(x) sin dx (27)
L J0 L
This representation of p(x) is not a "good" one, since p(L) = 0. How-
ever, the Fourier series solution of Eq. 25 is convenient and it is used
for the purpose of comparison, even if it may require many terms for
accurate calculation of the deflections.
Assume now
1383
^ mtx
w(x) = 2,nSin (28)

substitute Eq. 27 into Eq. 25


No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.

TT \ 12Q nir\ 12 HTTX 12 ,--, mrx


Es L ' ErA6t 3
VT/ 2 2
rf
sin =
~ m ^ " ~T 3 p sin (29)

Then, equating coefficients


Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.

Et3Vn
=
(30)
'MIT 12Q /nir\ + J2_
ErABf3 \L J r2t2
or, substituting Eq. 27 into Eq. 30

12 2 f rmx
;_ P(*) sin dx
(31)
J. Struct. Eng. 1985.111:1377-1398.

nir 12Q
L]
V - ErAQt ^V LV 3

Then, for Q = 0, substituting Eq. 31 into Eq. 28

24
. p(x) sin dx
if 3 ! j 0
3
MTTX
w\ '(*) = 1 y sin- (32)
/HTT\
+ 2 2
T/ r f
For Q given by Eq. 24, Eq. 28 becomes

24 HTTX
p(x) sin dx
mrx
w< w = =iS/ I T \ /Mir\ 12(x v pL 12
sin
L
(33)
L--(l -e~ )
"it3- P 7?
In Eqs. 32 a n d 33, the integral of the numerator for p(x) given by Eq. 6,
is
L
px
ftTTX vL
v(l - e ) sin dx = (1 - cos wn)
JOo L . nir

vnn
-pi
ve MTT
cos mr
"IT (34)
2 , |"P KIT
p2
P + ( T +

Eqs. 32 and 33 are the solutions of the deflection of the wall strip-beam
1384
with axial thrust equal to zero and to the resultant of the friction forces,
respectively.

INSTABILITY ANALYSIS

Eq. 33 indicates that instability may arise if the denominator becomes


No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.

zero. The buckling of the strip-beam is obtained when


4 / x 2 ,
wr\ /ntry 12|x v L 12
(1 -e-t> (35)
Et3 r2t2
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.

or jxv L - - ( l
_ p-pi - I'^VIi! fA^i ^ (36)
12 mr/ r
put in dimensionless form

JJL'V L - - (1 - e-pL)
L P J _ /nirV tl
(37)
Et " V L / 12+ n-n
Divide by Vt2/12r2;
J. Struct. Eng. 1985.111:1377-1398.

|X V I - - (1 - e-pL) 12r2
P mr\ 2 2 12r2
EtV? 12

12rj tr Vl2
2 (38)
Vmr/ r fz2 Vij Vl2 rt
Let fr/L2VT2 = 1/X2, then
1
L - - (1 - e_pL) 'Vl2
(XV
P + =/(X) (39)
Et2 T U '
To obtain the minimum L for Eq. 36 to be satisfied (critical L), differ-
entiate Eq. 39 with respect to X2 and set the result equal to zero and
solve for n
2 i,2 _ 2
(40)
X4 (nir) n Tt .

Substitute Eq. 40 into Eq. 39

o-P L 1 -Vl2
(X V (1 -
+ =2. (41)
e2 HIT/ \mrl
independent of n. Then, L critical is given from Eqs. 39 and 24, by
Q 1 2Ef2 Et2
= \L V L- - - ( 1 - - e-pL) (42)
rA0 P rVl2 rV3
1385
The right hand part of Eq. 42 gives the classical critical load for sym-
metrical buckling of a cylindrical shell under the action of uniform axial
compression neglecting the Poisson's ratio (7). Eq. 42 may be rewritten
to include Poisson's ratio, |x, according to the classical theory
Et2
No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.

|A V -(l-e^L) (43)
P rV3(l - |x 2 )'
Dividing by t one obtains
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.

Q Et TT2E (t\
frA6 rV3(l - (JL ) = K C12(1
- |xj2)- \L.
2 7 (44)

4V3
in which Kc = 1Z ( 45 )

aci = classical buckling stress for a cylinder under uniform axial


compression; and Kc = the buckling coefficient for moderately long cyl-
inders (1) in which 7 = a factor which has been included to account for
the difference between theoretical and experimental results; the classical
theoretical value is 7 = 1.
J. Struct. Eng. 1985.111:1377-1398.

Z = - V l - |x2 (46)
rt
Thus, the preceding instability analysis is valid for the assumptions tac-
itly made: Cylinder initially perfect, with r/t ratios large enough to be
susceptible to elastic shell-type buckling. The last assumption is valid
for all practical cases of unstiffened silos and for the unstiffened parts
of a ring stiffened silo, as long as (1)

yZ = Jyj I jj Vl-(i 2 > 2.85 K0 (47)


in which Lu = the unsupported length between two rings of the silo; K0
= 1 for simply supported edges, and K0 = 4 for clamped edges.
However, it is well known that actual cylinders may elastically buckle
and fold at stresses which may be as low as 30% of the critical stress
given by Eq. 44. It has been shown that initial imperfections are re-
sponsible for the discrepancy between actual buckling loads and the
classical theory (7). Thus, 7 is less than 1 and it can be expressed as a
best fit curve of Harris data (4) (for 600 < r/t < 5,000), by
/ N. - 0 . 5 4

7 = 11.78 n (48)

Eq. 48 along with Eqs. 45 and 47 will estimate the buckling stress with
a probability of 50%.
The case of farm silos, however, is actually one of a pressurized cyl-
inder with an elastic core. Experimental and theoretical investigations
have shown that the critical buckling stress is increased with the increase
of the internal pressure and the stiffness of the elastic core by reducing
the initial imperfections.
1386
Of _ L _ . _n._
No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved. 6

8 -
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.

0.10 - ^
8
*? '
4 ___

>
0.01
2 4 6 8 2 4 6 8 2 4 6 8
J. Struct. Eng. 1985.111:1377-1398.

0.01 0.10 1.0 10

FIG. 7.Increase in Axial Compressive Elastic Buckling Stress Due to Internal


Pressure Expressed by (2) p = [p(x)/E](r2/t)

Large-deflection theory indicates an increase in the buckling stress with


pressure until a maximum value is reached, but the experimental values
are lower than those predicted by theory. This discrepancy, although
not as large as for unpressurized cylinders, is likewise thought to be a
result of the imperfections of the test specimens (4). Harris, et al. (4)
developed statistically defined curves which give the buckling stress of
cylindrical shells with a knowledge of the cylinder geometry only. Ac-
cording to Harris' data
Et
a = (0.6057 + A7) (49)
r
in which 7 is given by Eq. 48 for a 50% probability; and A7 = the increase
in buckling stress as a function of pressure and geometry only, and is
obtained from Fig. 7.
Bucklin, et al. (3) showed experimentally that internal grain pressure
and the elastic effects of the grain mass would act to stabilize the bin
wall. The ratio of the observed buckling stress to the theoretical classical
buckling stress is given by a best fit equation
Et
ov = 0.6057' (50)

283 14
in which 7 ' = 0.000276 1.0 - 0.00413 ('- ) e<- .8r/<r . . . (51)

1387
and Vo = m Q2 (52)
Eqs. 48 and 51 are best fit equations which predict the loads at which
50% of bins with a given r/t ratio and p0 will buckle. Eq. 51 is considered
No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.

good for design for r/t ratios from 1,000-3,000 (3).


Based on the preceding information obtained from the literature cited,
an upper limit was imposed to the axial critical stress due to friction.
This was obtained by equating Eq. 43 to either Eq. 49 or Eq. 50
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.

Et
(0.6057 + for 600 s - s 5,000 (53fl)
M-'v
x - - (1 - e~") = <
t L p Et
0.605-y' - for 1,000 < - =s 3,300 and p < 0.33 . . . (536)

The data analyzed by Bucklin, et al. (3) were measured for p0 s 0.33.
Eq. 53 reveals a significant behavior of the silo grain system which is
simulated in the equation by a pressurized cylinder under axial compres-
sion.
J. Struct. Eng. 1985.111:1377-1398.

The left side of the equation is the total axial compressive stress caused
by the grain friction at a depth, x, below the top of the silo (neglecting
the dead load of the structure).
The right side represents the axial stress at which an unstiffened por-
tion of the silo with its lower edge at a depth, x = Lcr, below top and
an unsupported length, Lu, will buckle with a probability of 50% (L =
Lcr for an unstiffened silo or the top unstiffened portion of a stiffened
silo). This L is subjected, however, to the condition described by Eq.
47, i.e., the unstiffened portion has to be in the range of intermediate
length for Eq. 53 to be valid. The thickness, t, to be used in Eq. 53 is
the one of the unstiffened portion. If the thickness is not uniform, the
minimum, t, within the particular portion under consideration should
be used.
The right side probabilistic stress for a given r/t ratio and E depends
on the grain lateral pressure alone. But the friction stress (left side) also
depends on the lateral pressure for a given |x', as shown in Eq. 24. This
is a major point: Compressive stress and lateral pressure of this partic-
ular pressurized cylinder are not independent. Thus, the buckling stress
of the system should not be determined for any arbitrary value of in-
ternal pressure as in other cases (4).
Under the foregoing considerations, the satisfaction of Eq. 53, and thus
the buckling behavior of the silo, depends only on Lcr for given E, r, t,
|JL' , k and m. Eq. 53 is valid as long as Eq. 47 is satisfied for L. The
selection of Lcr as the independent variable is justified as follows.
In any particular case of design, all E, r, t, juu', k and m, as well as L
are given. Then, using these E, r, \i', k and m values in Eq. 53, and
setting first x = Lcrt , the equation is solved with some iterative scheme.
(The thickness, t = tr, to be used in this solution procedure is the min-
imum one within the length, Lcrl). Then, Lcn = the distance from the
top of the silo down to the maximum unstiffened length of the wall for
which a 50% probability exists for buckling (Fig. 8). If Lcrl > L, no stiff-
1388
No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.

Lers
n7T777T7777777777777777r77n77Tn7

FIG. 8.Ring Stiffened Silo

enersvertical or ringsare required; if Lcrt < L, then either a ring stiff-


ener is required at a maximum length, Lcrt , from the top or vertical stiff-
eners should be used to prevent buckling. In the case of a ring stiffener,
stiff enough to make the radial deflections at its position negligible, the
J. Struct. Eng. 1985.111:1377-1398.

iterative procedure can be continued as follows. The minimum thickness


of the next unstiffened portion is substituted into Eq. 53, and the equa-
tion is solved for Lcr2, (Fig. 8). Then, L2 is checked against Lu given by
Eq. 47 for K0 = 1 (conservatively). If Lcr2 - Lcrl = Lu2 satisfies Eq. 47,
then, a second ring stiffener is required at a max length Lcr2 from the
top of the silo. If, however, Lw2 does not satisfy Eq. 47, Eq. 53 is no
more valid, since the portion with length, Lu2, behaves like a short cyl-
inder. In this case, Eq. 53 is modified on the right side (1):

x-(l-e-px)
t L P
12 (Lu\(LU T7 2 E
= 1+- (54)
HTAT'^ 12(1 - \S)) \Lu
in which 7 is given by Eq. 48.
Eq. 54 is based on buckling of an unpressurized cylinder with no elas-
tic core, and as such is conservative. Substituting, x = Lcr2 = Lcrl + Lu2,
in which Lcrl has been calculated from the previous step, the only vari-
able in Eq. 54 is Lul, as it should for a short cylinder. Again, the thick-
ness, t, in Eq. 54 is the minimum one within the Lu considered.
In general then, for the sth unstiffened portion of the silo, Eq. 53 is
solved for Lcrs with t = ts. Then Lus = Lcrs = Lcrs - Lm-x is checked against
Eq. 47. If Eq. 47 is not satisfied, Lcre_i and ts are substituted into Eq. 54
and the maximum Lus is obtained. (In the case of fs_i = ts, Eq. 54 gov-
erns always for s > 1).

CRITICAL LENGTH OF UNSTIFFENED SILO

The solution of Eq. 42 is shown in Fig. 9 for eight different values of


t. With the increase of the ratio, r/t, the critical unstiffened length, Lcr,
1389
No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.
J. Struct. Eng. 1985.111:1377-1398.

8000 3000
r/t RATIO

FIG. 9.-L Critical According to Classical Theory Eq. 42

decreases. This decrease is dramatic for low r/t values which indicates
that reducing the radius, r, or increasing the thickness, t, is beneficial
in terms of the unstiffened length of the silo, only in the range of low
r/t values.
Figs. 10 and 11 show the solution of Eqs. 53a-b, respectively. Clearly
Eq. 53a from Harris' data shows the same general behavior as the clas-
sical theory, while Eq. 53b gives a critical length that is nearly constant
for each t. Fig. 12 presents a combination of the solutions of Eqs. 53a-b
in which Eq. 53b was used in the range 1,000 < r/t < 3,300 for p <
0.33. The two solutions coincide at r/t = 2,000, while 59b yields L much
lower than that given by 53a at r/t = 1,000.
A comparison of the solution of Eq. 53b versus 53a with A7 = 0 (ig-
noring the effect of the internal pressure) is shown in Fig. 13. Here it
can be seen that while the effect of the internal pressure given by Eq.
53b is quite significant in the region of r/t = 2,000, it is zero, however,
in the region of r/t = 1,000. If one considers the fact that Eq. 53b is based
on data from a model silo, that is of a pressurized cylinder with an elas-
tic core (grain) present, while Eq. 53a with A7 = 0 is based on data from
a nonpressurized cylinder, then Fig. 13 reveals that Eq. 53b is not re-
alistic in the range of low r/t values. Thus, Eq. 59a is considered to ap-
ply, and Fig. 10 gives the critical length for which 50% of the unstiffened
1390
100 -
1328)

THICKNESS, cm (in)
No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.
J. Struct. Eng. 1985.111:1377-1398.

r/t RATIO

FIG. 10.I Critical According to Eq. 53a for 50% Probability of Buckling

I00
(328)
"

80
(262)
"

60
(197)
"

THICKNESS, cm (In)
2.34 x
(l,000)\
40 2.22-.~--_
t)3l) (0 8 7 3 ) ^ - - . . - - - __

io623j/~~" ~ - - _ SJj~ ~-~~SZ~LZZ~SI~~


(o.sooj/C """
20 - 0.93' / ~~
166)
0.64 V ^"~
0,32
(0.123)

0
r/t RATIO

FIG. 11.L Critical According to Eq. 536 for 50% Probability of Buckling ( In-
side the Range of Validity of Eq. 536 Outside Range of Validity of Eq. 536)
1391
THICKNESS, cm (in)
No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.
J. Struct. Eng. 1985.111:1377-1398.

r/t RATIO

FIG. 12.L Critical According to Eq. 53b in Range 1,000 s r/t 3,300 (for p <
0.33) and Eq. 53a in Rest Range of r/t and for p > 0.33 ( Eq. 53a in the Regions
where Eq. 53b was Used)

THICKNESS, cm (in)

2000 3000
r/t RATIO

FIG. 13.L Critical According to Eq. 53b ( ) versus that of Eq. 53a, but with
Ar = 0 (); Probability of Buckling 50%
1392
too
(328)
No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.

THICKNESS, cm (rnl
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.

60

<
o
. 40
1
1131)

20
168)
J. Struct. Eng. 1985.111:1377-1398.

1
3000
r/t RATIO

FIG. 14.L Critical According to Eq. 53a for 1 % Probability of Buckling

silos will buckle. For a 1% probability of buckling, 7 is given by Harris,


et al. (3)

(Jj \ -0.54

(55)
and the solution of Eq. 53a is given in Fig. 14. Thus, Figs. 10 and 14, or
Eq. 53a gives the critical length of an unstiffened silo or the max unstiff-
ened top portion of the silo, such that buckling from the wall friction
forces is prevented.
In the case where the height of the silo is greater than Lcr, the pro-
cedure described in the previous section can be applied to determine the
position of the next stiffening rings. However, if the resulting un-
stiffened portions are governed by Eq. 54, it is more likely that the spac-
ing of the ring stiffeners is very small and the structure becomes un-
economical. An example which shows the spacing, Lu2, (Fig. 8) for the
case of constant thickness (fj = t2) is given in Table 1. It can be seen
that the spacing increases with thickness and r/t ratio. The increase of
Lu2 with increase in r/t ratio comes from the fact that Lux decreases with
an increase in r/t, thus the axial force and lateral pressure associated
with Lu2 increases as r/t decreases. Provided that the thickness remains
the same, Lu decreases only slightly with the number of ring stiffeners.
This is due to the fact that Lu obtained by Eq. 54 is in general very small,
so that the increase in the axial stress is rather small from ring to ring.
1393
If, however, the thickness changes, Eq. 53a may control in more than
one unstiffened portion of the silo, and so the effectiveness of the ring
stiffening increases.
Compared to the stringers, ring stiffeners may become more econom-
ical for certain cases. This can be shown only after a complete analysis
No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.

is made for both systems of stiffening for any given case. The stress from
the dead and live load of the roof should also be added in the left hand
part of Eq. 53 in such an analysis.
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.

THE EFFECT OF AXIAL FRICTION FORCES ON THE WALL

The general tendency obtained from comparing the solutions of Eqs.


32 and 33, as shown in Fig. 6, is that the maximum effect from the axial
friction forces occurs at the region of the bottom of the silo (or strip-
beam).
This phenomenon of the error distribution is explained if the deflected
shape of the strip-beam under the lateral pressure is considered without
any axial force applied (Fig. 6). It is obvious that the region with the
maximum deflection from the lateral loading is the most sensitive one
to the addition of any axial force. The effect of the actual friction forces
J. Struct. Eng. 1985.111:1377-1398.

resultant Qa on the error defined in Fig. 6, would be less pronounced


than from Qb since a greater curvature would be required to obtain the
same maximum deflection on a shorter length of strip-beam. Or, in other
words, more energy is required to deflect the short length of the case
of the friction forces resultant Qa than the full length of the case of the
equal in magnitude axial thrust, Qb. Also, it can be seen that the addi-
tional moment on the deflected shape due to the lateral pressure only
in the region where the errors occur is larger for Qb than for Qa (Fig. 6).
Thus, the error distribution confirms that an upper bound analysis has
been carried out with respect to the maximum deflection.
At this point, a generalization can be made with regard to the bound-
ary conditions. If a fixed-end condition is assumed at the bottom end,
the energy required to deflect the portion aL [Fig. 5(a)] a certain amount
is greater than the one corresponding to the pinned end. This is because
a greater curvature is required for the fixed-end condition than for a
pinned-end condition. As to the top end conditions, it is not expected
that they will affect the results of this analysis. This is because near the
top end, the lateral force, p(x), is very small and the deflection will be
very small. In addition, since the actual friction forces applied at the top
region are insignificant in magnitude, the restraint provided by the hoop
forces dominates and provides a condition similar to the hinged bound-
ary condition.
Thus, the analysis carried out in this paper is an upper bound anal-
ysis, in general, for all. cases of pinned-end or fixed-end bottom condi-
tions and pinned-end or free-end top conditions. However, this analysis
may be a lower bound analysis (unsafe) for the deflection of the wall of
the silo with the bottom end not restrained radially. This is because in
that case, a smaller curvature is required for a certain amount of de-
flection than that with a hinged end. Thus the effect of the thrust on
the smaller than actual deflection may offset the conservative assump-
tion that the thrust is applied at the top of the strip-beam. However, in
1394
TABLE 1.Maximum Spacing Lu2 in cm (ft) for Silos of Constant Thickness
(f, = t2)'
Ratio Thickness, cm (in.)
r/t 0.318 (0.125) 0.635 (0.250) 0.953 (0.375)
(D (2) (3) (4)
No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.

b b
700 44.0 (1.4)
b
800 31.9 (1.0) 46.3 (1.5)
900 17.8 (0.6) 33.4 (1.1) 48.5 (1.6)
1,000 18.S (0.6) 34.8 (1.1) 50.6 (1.7)
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.

2,000 24.1 (0.8) 46.3 (1.5) 68.1 (2.2)


3,000 28.7 (0.9) 52.1 (1.7) 77.5 (2.5)
3,300 29.9 (1.0) 54.7 (1.8) 81.4 (2.7)
4,000 30.6 (1.0) 60.7 (2.0) 89.6 (2.9)
"Parameters used: E = 20,700 MPa (30,000 ksi); |JL' = 0.384; k = 0.5; m = 800
kg/m3 (50 pcf); and M- = 0.3.
b
The L/r ratio in these cases is greater than 10, so that there is no practical
need for additional height.
J. Struct. Eng. 1985.111:1377-1398.

practice there is always some type of restraint of the unstiffened silo


wall, or the unstiffened portion of it, at its lower base.
The analysis of the results for the error distribution (Fig. 6) shows that
the maximum error occurs at the bottom end region of the silo. Fur-
thermore, the deflection increases dramatically just above the bottom
end as shown in Fig. 6, and reaches its maximum value in that region
of the bottom end; then it decreases smoothly to zero at the upper end
of the silo. However, what is important for the design of the lower part
of the silo or an unstiffened region of the silo is only the maximum de-
flection which determines the design hoop stress at that region (Eqs. 2
and 3). Thus, even the error corresponding to the maximum deflection
is not necessarily the maximum one; it is, however, the most important.
This error reflects the "column-beam" interaction of the wall strip-beam
under the action of the lateral pressure plus the axial friction forces.
Both Eqs. 32 and 33 were based on the linear elastic theory, i.e., on
the assumption of small deflections. Thus, they are valid as long as the
height, L, or the unstiffened length of the top unstiffened portion of the
silo is less than Lcrl given from Eq. 42 or Fig. 9. In a real situation, how-
ever, it is Lcrl given by Eq. 53a or Fig. 10 or 14 which determines the
limit in the validity of Eqs. 32 and 33.
Under the preceding considerations, the error corresponding to the
maximum deflection (or maximum hoop stress) was calculated for three
values of tx and for ratios L/r ranging from 1 to the smaller of 10 (prac-
tical limit) or LcrX/r (instability limit). This was done for twelve values of
radii, ranging from 1.52 m (5 ft)-18.29 m (60 ft). The results are shown
in Figs. 15(a-c) (for the parameters given in Table 1, footnote).
The variation of the error in the maximum hoop stress when no axial
force is included compared to the presence of the friction axial force in-
dicates a general pattern. The error increases as the ratio, L/r, and the
radius, r, increases, and as the thickness, t, decreases.
It can be seen that as the radius increases, the error is more sensitive
1395
i fUMJS.BHIl
-18.29 160)
^,16.76 155)
K 7715.24 1501
///13.72 145)
///[IZ.I9 1401


No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.

7
/

Jl/ l
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.

i*i2-
/
i
r~^ i
L/r RATIO
(b)

RMHVS. -. I h l 1 8 < 8 g ( 6 0 )
1 0.95 cm
(0.375 tol
XO-I6T6 (551
/ > > 15.24 (50)
/y/*^\l>.lZ (451

1 MoT^Br^^-^r-^^ r L ""
-
J. Struct. Eng. 1985.111:1377-1398.

3.05 (101
TsTisi

FIG. 15.Error in Maximum Hoop Stress Due to Axial Friction Forces Interaction
with Lateral Loading versus Ratio of Silo Height over Radius

to the increase of the total height of the silo (L/r ratio). This is true for
all thicknesses. Also, for a given radius and thickness, the error is, in
general, more sensitive with higher values of the ratio L/r.
The error calculated and shown in Figs. 15(a-c) is valid for both static
and dynamic conditions of loading from the grain. This is because the
dynamic effects are included in the design of the silo walls by using
some empirical multipliers with the static pressures to account for the
overpressures created during the emptying of the silo (6). Thus, both
Eqs. 32 and 33 are multiplied by a constant to obtain the solutions cor-
responding to the dynamic conditions. This, however, does not affect
the error calculated for the maximum hoop stress.
The limit, L crl , shown in Figs. 15(a-c) is for an unstiffened silo or un-
stiffened portion of length, L crt , for which buckling is prevented with a
probability of 50%. The error corresponding to this limit is fairly constant
in the region, L/r 2 1, and it varies only slightly with the thickness. It
can be said that using Lcrl given by Eq. 53a or Fig. 10, the error in the
maximum hoop stress will be between 19% and 24% for ix varying from
1396
0.32 cm (0.125 in.)-0.95 cm (0.375 in.), as long as L/r is between 1 and
10. This limiting error increases to 27% when fj = 0.32 cm (0.125 in.)
and r = 12 m (40 ft), and to 30% when h = 0.64 cm (0.25 in.) and r >
17 m (56 ft), which are not practical cases.
The variability of the error in the maximum hoop stress and its de-
No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.

pendence on ti, r and L/r, indicates that the factor of safety used in the
design of the walls of the silos is actually not constant. Thus, in some
cases it may become too low.
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.

CONCLUSIONS

A closed form solution of the differential equation for the deflection


of the wall of a cylindrical silo is very difficult (if not impossible) to ob-
tain. The source of this difficulty is the presence of the axial friction
forces distributed along the wall of the silo. A conservative solution for
the analysis of the effects of the axial friction forces is obtained by ap-
plying their resultant as a thrust on the wall.
The "beam-column" interaction for the wall of a silo reveals that the
error from ignoring the effects of the axial forces on the calculated max-
imum hoop stresses is in general significant, depending on the thick-
J. Struct. Eng. 1985.111:1377-1398.

ness, radius and height to radius ratio.


This analysis however is valid within the linear elastic range. An anal-
ysis of instability of the cylindrical shell yields the maximum height of
an unstiffened silo or the maximum unstiffened portion of a silo re-
quired to prevent buckling. This suggests an algorithm for the design
of an unstiffened silo or a silo stiffened by rings. However, the effects
of the load of the roof and the wind loading have also to be considered
in such a design procedure.
The error in the maximum hoop stress corresponding to the critical
unstiffened length at which buckling occurs with a 50% probability is
shown to be in the range of 19%-24%. This range of error corresponds
to a wall thickness variation from 0.32 cm (0.125 in.)-0.95 cm (0.375 in.).
The error in the maximum hoop stress shown in Figs. 15(a-c) is the
same for both static and dynamic conditions. Its variation with the pa-
rameters t, r, and L/r causes corresponding variation in the factor of
safety which actually applies when the effects of the axial friction forces
are ignored. Thus, the actual factor of safety is variably less than the
one used in the conventional design of the silo walls.

APPENDIX.REFERENCES

1. Baker, E. H., Kovalevsky, L., and Rish, F. L., Structural Analysis of Shells,
McGraw-Hill Book Co., Inc., New York, N.Y., 1972.
2. Boresi, A. P., Sidebottom, O. M., Seely, F. B., and Smith, J. C , Advanced
Mechanics of Materials, John Wiley and Sons, New York, N.Y., 1978.
3. Bucklin, R. A., Ross, I. J., and White, G. M., "The Influence of Grain Pressure
on the Buckling Loads of Thin Walled Bins," ASAE Paper No. 83-4005, Amer-
ican Society of Agricultural Engineers, St. Joseph, Mich., 1983.
4. Harris, L. A., Suer, H. S., Skene, W. T., and Benjamin, R. J., "The Stability
of Thin-Walled Unstiffened Circular Cylinders Under Axial Compression In-
cluding the Effects of Internal Pressure," Journal of the Aeronautical Sciences,
Vol. 24, No. 8, 1957, pp. 587-596.
1397
5. Ketchum, S. M., The Design of Walls, Bins and Grain Elevators, McGraw-Hill
Book Co., Inc., New York, N.Y., 1919.
6. "Recommended Practice for Design and Construction of Concrete Bins, Silos,
and Bunkers for Storing Granular Materials," AC1313-1977, ACI, Detroit, Mich.
7. Timoshenko, S. P., and Gere, J. M., Theory of Elastic Stability, McGraw-Hill
Book Co., Inc., New York, N.Y., 1961.
No other uses without permission. Copyright (c) 2012. American Society of Civil Engineers. All rights reserved.
Downloaded from ascelibrary.org by D. Craig Brinck on 07/12/12. For personal use only.
J. Struct. Eng. 1985.111:1377-1398.

1398

Вам также может понравиться