You are on page 1of 17

Comparison of various methods used in

the analysis of silos without wall friction


M. Khouri
Department of Civil Engineering, Lebanese University, Lebanon
Abstract
The use of circular silos with independent hopper structure in the process of
storing grain and other granular material has become very extensive nowadays.
The structural analysis of silos is of a complex nature and requires sophisticated
mathematical techniques for proper calculation. Janssen, Airy and Reimbert
suggested different material lateral pressure approaches that can be followed in
the process of calculating silo wall forces. This paper compares the above three
methods to the water behaviour approach and makes use of the finite element
method to check the closed form techniques. Also, average values of the three
approaches were determined for various silo dimensions and graphed for the use
of engineers and designers.
1 Introduction and background
Different types of silos have been developed but the most used are circular ones.
Recently, the use of a totally independent structure of hopper within the silo was
found to be more economical due to the isolation and simplicity in the process of
construction and design. Silos depend on many factors such as the stored
material, the type of silo, the construction material properties, the height and
diameter of the silo, wall thickness, the angle of the hopper, etc.
Many investigators have studied circular silos. Timoshenko et al. [1,2]
presented the differential equation that can be used in solving silo problems. The
equation calculates the deformation shape of the wall for any pressure function
by isolating a vertical strip in the wall and using symmetry. In 1895, Janssen [3]
suggested a pressure application approach, while Airy [5] addressed his pressure
application approach in 1897; Reimbert and Reimbert [4] in 1976 suggested a
pressure application approach, which is similar in nature to the other two
2005 WIT Press WIT Transactions on Modelling and Simulation, Vol 41,
www.witpress.com, ISSN 1743-355X (on-line)
Computational Methods and Experimental Measurements XII 425
methods. These material pressure application approaches can be used in
Timoshenko's equation to calculate silo wall forces. Safarian and Harris [6]
attempted to do an evaluation of the methods (Janssen, Airy, and Reimbert).
Also Thonier [7] presented the strip analysis, which by itself, is a very interesting
and useful approach. Many engineers including Safarian and Harris [6],
Reimbert and Reimbert [4], and Fintel [8] have also worked on the analysis of
silos, which led to the setting up of the ACI Committee 313, which is the
Committee on Construction of Concrete Bins, Silos, and Bunkers for Storing
Granular Materials [9].
Note that the angle of the hopper is not considered in this analysis and the
flow of the material within the silo is assumed to be uniform (mass-flow) [10]
where the outlet is considered to be sufficiently large to allow the flow without
creating a stagnant solid in the area where the hopper meets the silo wall. In
addition, the wall friction between the material and the silo wall is not taken into
account in the differential equation, keeping in mind that the methods that are
being compared consider friction in the determination of the lateral pressure.
(Wall friction evaluation will be presented in a future paper following this).
This study is divided into three parts, the first part of this work evaluates
Janssen, Reimbert and Airys static lateral pressure application approaches,
compares them to each other and to the water behavior pressure application
approach. In the second part, the silo is modeled using the finite element method
to check the mathematical closed form techniques. The third part however, deals
with finding average values for Airy, Janssen, and Reimbert for various silos
with various heights and diameters, and various wall thicknesses; the average
values are then graphed to allow the designing engineer to simply use them.
It is important to note that the following work is limited to static analysis and
silos also need to be checked for wind and seismic loads.
2 Comparing the approaches
We begin by setting up the differential equation, perform the analysis for each
approach, and compare the results for circular silos with both hopper and the
wall structure being independent. (See Figure 1.)


Figure 1: Silo geometry.
2005 WIT Press WIT Transactions on Modelling and Simulation, Vol 41,
www.witpress.com, ISSN 1743-355X (on-line)
426 Computational Methods and Experimental Measurements XII
The wall of the silo is divided into two parts:
1. The upper part where the lateral pressure p is applied between ho and
(ho

+

h1).
2. The lower part where no lateral pressure is applied between the base and ho
Timoshenko gives the differential equation to calculate the deformation shape
of the wall by isolating a vertical strip in the wall and using symmetry.
The differential equation given by Timoshenko and Gere [1] is:
DE
p
y 4 y
4 (4)
= +
(1)
in which y is the lateral displacement, is the elastic length given by,
( )
4
2 2
2
ep R
1 3


= , and DE is the flexural rigidity given by
( )
2
3
- 1 12
ep
DE

= ,
where ep is the wall thickness, R is the average radius, and is Poissons ratio.
The solution of the differential equation (1) for the upper part h1 of the silo
without the right term is:
| | | |
x x
h
y e A1.cos( x) B1.sin( x) e A2.cos( x) B2.sin( x)

= + + + (2)
If the particular solution of equation (1) is y
p
, the general solution of the
differential equation (1) for the upper part h1 of the silo is:
p h g
y y y + =
The general solution of the differential equation (1) for the lower part ho
where there is no grain is:
| | | |
x x
h
y e C1.cos( x) D1.sin( x) e C2.cos( x) D2.sin( x)

= + + + (3)
A1, A2, B1, B2, C1, C2, D1, and D2, are constants.
Note that for the water behavior and Janssen approaches the method for
finding the particular solution is direct, while in case of Airy and Reimbert
approaches the use of the method of reduction of order is needed.
2.1 Boundary conditions
y
1
(x) is the general solution of the differential equation for the upper part of the
silo h1, and y
2
(x) is the general solution of the differential equation for the lower
part of the silo ho.
Due to the existence of the cover at the top of the silo, it is assumed to be
pinned or simply supported on its circumference:
For x = h1 + ho y
1
(x) = 0 and y
1
(x) = 0
At the bottom, the silo is fixed:
For x = 0 y
2
(x) = 0 and y
2
(x) = 0
At the point where the upper part of the silo meet the lower part, the
function and its derivatives are continuous and are represented as
follows:
For x=ho y
1
(x) = y
2
(x) y
1
(x) = y
2
(x)
y
1
(x) = y
2
(x) y
1
(x) = y
2
(x)
2005 WIT Press WIT Transactions on Modelling and Simulation, Vol 41,
www.witpress.com, ISSN 1743-355X (on-line)
Computational Methods and Experimental Measurements XII 427
The problem now is to solve the eight equations for eight unknowns given the
above boundary conditions.
2.2 Water behavior approach
In this method, the distribution of the lateral pressure p applied to the wall is
assumed to be linear (or triangular) with depth x, and can be represented as
follows:
|
.
|

\
|
+
=
h1
x ho h1
p p
o
, in which x is taken from the bottom of the
silo, is the weight per unit volume, k
o
is the Rankine coefficient for active earth
pressure, and p
o
is the pressure of the stored material at depth h
1
below surface of
stored material, h1 k p
o o
= , where k
o
is given by
sin 1
sin 1
k
o
+

= , where is
the angle of internal friction for the stored material, and h1 is the height of the
storage zone.
The solution of the differential equation (1) for the upper part h1 of the silo
without the right hand side term is obtained from equation (2), and the particular
solution of equation (1) is:
fp1(x)
h
x ho h1
.
KK
p
y
1
0
P

+
= , in which .DE 4.
R
E.ep
KK
4
2
= = , and
E is the elastic modulus. The general solution of the differential equation (1) for
the lower part ho of the silo is obtained from equation (3).
The resolution of the differential equation is made for function ff (presented
later) which is the deflection multiplied by the elastic modulus; this allows the
calculations independent of the silo material. The following analysis was done
using MathCAD 2000 and can be summarized as follows:
The boundary conditions give M * Constant vector = VV, where M is an 8x8
matrix corresponding to the application of the boundary condition, and the vector
VV is:
VV
fp1 ho h1 + ( )
d2fp ho h1 + ( )
1
2
2

0
0
fp1 ho ( )
d1f ho ( )
1

d2fp ho ( )
1
2
2

d3fp ho ( )
1
2
3

\
|
|
|
|
|
|
|
|
|
|
|
|
|
|
.
:=
where,
2005 WIT Press WIT Transactions on Modelling and Simulation, Vol 41,
www.witpress.com, ISSN 1743-355X (on-line)
428 Computational Methods and Experimental Measurements XII
d1f x ( )
x
fp1 x ( )
d
d
:=

d2fp x ( )
2
x
fp1 x ( )
d
d
2
:=

d3fp x ( )
3
x
fp1 x ( )
d
d
3
:=

By solving this system of 8 equations, the values of the eight constants of
integration AA, AA1, BB, BB1, CC, CC1, DD, and DD1 can be determined.
Each function is divided into two intervals where u [0, ho] and x [ho, ho +
h1]. The functions ff (x) and gg (u), which represent (Deflection * E) are:



The wall moment in the vertical direction is:

M(x) = DE
.
{d2fp(x) + (2-
2
) [ff(x)]}, and

M(u) = DE
.
(2-
2
)
.
[gg(u)],

The wall shear in the vertical direction is:

V(x) = DE
.
{d3fp(x) + 2
.

3

.
[ff(x)]}

V(u) = DE
.
2
.

3

.
[ff(x)]

The force N in the circumferential direction is:
N(x) = ff(x)
.
[ep / R]
N(u) = gg(x)
.
[ep / R]
2.3 Janssens approach
Janssens method is based on equilibrium of a thin horizontal layer of stored
materials and the equation for horizontal pressure is:
(

=

h
R
.Y .k
1
h
0 1
e 1

.R
p in which, Y is the depth, ko is the Rankine
coefficient for active earth pressure, R
h
is the Hydraulic radius, and 1 is the
angle of friction of the stored material against the wall.
The procedure will be the same as for the water behavior approach. However
the particular solution fp1(x) is:
fp1 x ( )
1
KK
1 4

4

4
4
4
+
e
h1 ho + x ( )

(
(
(

:=

2005 WIT Press WIT Transactions on Modelling and Simulation, Vol 41,
www.witpress.com, ISSN 1743-355X (on-line)
Computational Methods and Experimental Measurements XII 429
with
2
R
R
R
ko
.
R
ko
1.
h
h h
= = =
Having the particular solution, the constants of integration and the functions
that represent deflection*E, moment, shear and force N in the circumferential
direction can be determined similar to the water behavior approach.
2.4 Airys approach
The Airys equation was derived by considering static equilibrium of wedge-
shaped portions of the stored material above the plane of rupture. Airys
equation for silos leads to lateral pressure at depth Y:
( )
(
(
(
(

+ +
+

+
=
1 1
2
1
. 1
R
Y
1
1

R . . 2
p

in which, 1 is the angle
of friction of the stored material against the wall, is the internal friction of the
material = tan (), and is the angle of internal friction for the stored material.
The procedure is similar to the water behavior approach. However, the
particular solution fp1(x) is calculated as a linear differential equation with
constant coefficients using the method of reduction of order [11].
The differential equation can be written using the operator techniques in
factored form as:
) ( ). ).( ).( ).( .(
4 3 2 1 0
x R y m L m L m L m L a = , (4)
where m
1
, m
2
, m
3
, m
4
are constants, L is the linear operator, and a
0
is the
coefficient of the highest derivative.
This method yields the general solution if all arbitrary constants are kept,
while if arbitrary constants are omitted it yields a particular solution [11].
In this case a
0
= 1, m
1
= -l(1+i), m
2
= l(1-i), m
3
= l(-1+i), and m
4
= l(1+i).
To solve the first order linear differential equation, use the general solution
for the equation = + q(x) p(x).y(x) (x) y'


+

=
pdx pdx pdx
c.e dx Q.e . e y , where c is an arbitrary constant (take c = 0),
q x ( ) 1 1

1 x +
|

\
|
.
2

(
(

:=
, where,

1
1
1 + ( )

2 R
DE
:=
1
2
+ := 1 1 := 1 2
1 + ( )
2 R
:=

y1 x ( ) e
m1 x
x q x ( ) e
m1 x

d :=

y2 x ( ) e
m2 x
x y1 x ( ) e
m2 x

d :=

2005 WIT Press WIT Transactions on Modelling and Simulation, Vol 41,
www.witpress.com, ISSN 1743-355X (on-line)
430 Computational Methods and Experimental Measurements XII
y3 x ( ) e
m3 x
x y2 x ( ) e
m3 x

d :=

y4 x ( ) e
m4 x
x y3 x ( ) e
m4 x

d :=

Let ) Re(y4(x) z(x) = , (Re means the real part of) where x is the depth
taken from the top to the base of the silo. After making the analytic
transformation fp1 becomes, x). - h1 z(ho fp1(x) + = Having the particular
solution, one can calculate the constants of integration and determine deflections
and forces as before [11].
2.5 Reimberts approach
Reimbert method for computing static pressure considers that at the bottom of
the silo, the lateral pressure becomes asymptotic to the vertical axis. At depth
ho, the lateral pressure reaches a maximum value equal to p
max
.
The procedure is similar to Airys approach where the lateral static pressure at
depth Y given by the Reimbert equations is:
(
(

|
.
|

\
|
+ =
2
max
1
C
Y
1 p p
, in which
1
h
max

.R
p =
with R
h
being the Hydraulic
radius = R/2 for circular silo, and
0 1
.k 4.
D
C=
, with D being the diameter of silo.
2.6 Numerical comparison between the four approaches
In order to compare the four methods, a circular silo is taken with the following
variables:
= 0.75 T/m
3
; ho = 10 m; h1 = 30 m; D =10 m; ep = 0.3 m.
= 0.2; = 30
o
; 1 = 0.3; Total Height = 40 m.


Figure 2: Pressure vs. height for the four methods.
2005 WIT Press WIT Transactions on Modelling and Simulation, Vol 41,
www.witpress.com, ISSN 1743-355X (on-line)
Computational Methods and Experimental Measurements XII 431
Figure 2 shows the difference between the pressure application approaches
for all four methods. The following Figures 36 represent the comparison of
results (deflection, moment, shear and circumferential force) for the four
methods.
Note that, on the graphs, number 1 represents the water behavior approach,
number 2 represents Janssens approach, number 3 represents Airys approach,
and number 4 represents Reimberts approach.


Figure 3: Deflection*E vs. height for the four methods.

Figure 4: Moment vs. height for the four methods.
As can be seen in Figure 4, the moment is small as we move away from the
top of the hopper upwards, while in the same region we have a deflection. This
is due to the fact that the shell membrane forces compensate the deflection
thereby decreasing the flexural forces in the wall of the silo. Note that to
evaluate the deflection, values of the graphs (Deflection*E) should be divided by
the elastic modulus E.
It can be noticed from Figures 26 that Janssen, Reimbert, and Airy method
give comparative results while the water behavior method gives larger values as
compared to the above three methods. So, averages were done for the Reimbert,
2005 WIT Press WIT Transactions on Modelling and Simulation, Vol 41,
www.witpress.com, ISSN 1743-355X (on-line)
432 Computational Methods and Experimental Measurements XII
Airy and Janssen methods only excluding the water behavior approach since it
tends to exaggerate the values.


Figure 5: Shear vs. height for the four methods.

Figure 6: Force N in the circumferential direction vs. Height for the four
methods.
In comparing the three methods (Janssen, Reimbert, Airy), Reimberts gives
values little greater than the other two. In addition, it seems that Airy method
does not give reasonable values at the top of the silo due to the fact that his
pressure equation does not properly represent the behavior in that region, as can
be seen in Figure 2.
3 Comparison with the finite element method
In order to verify the validity of the results obtained by solving the differential
equation, analysis of the same silo studied above was done using the finite
element method (Effel Program). See the finite element model in Figure 7. The
lateral pressure considered was as defined by the water behavior approach and
2005 WIT Press WIT Transactions on Modelling and Simulation, Vol 41,
www.witpress.com, ISSN 1743-355X (on-line)
Computational Methods and Experimental Measurements XII 433
the results of the deflection, moment, shear and circumferential force obtained by
the finite element method were compared to the results obtained by solving the
differential equation.

Figure 7: The finite element model.

Figure 8: Deflection vs. Height (with E = 2000000 T/m).

Figure 9: Moment vs. height.
2005 WIT Press WIT Transactions on Modelling and Simulation, Vol 41,
www.witpress.com, ISSN 1743-355X (on-line)
434 Computational Methods and Experimental Measurements XII
The two methods are very close to each other as expected (See Figures 811).
A little difference exists in the graphs due to the fact that the membrane forces in
the vertical direction in the silo wall are assumed to be zero and are not included
in the differential equation, while in the finite element method, these forces exist
and tend to decrease the deflection and consequently the forces.
Note that in Figures 811, number 1 represents results of the differential
equation while number 2 represents results of the finite element method. Both
methods are done for the water behavior approach.


Figure 10: Shear vs. height.


Figure 11: Force N in the circumferential direction vs. height.
4 Average results for different silo dimensions
After doing a comparison between all approaches, average values for Janssen,
Airy and Reimberts are calculated for various silos, with various heights and
diameters and wall thicknesses. Note that only silos (and not bunkers) are
analyzed, and the ratio H/D is kept greater than 1.5 [6].
2005 WIT Press WIT Transactions on Modelling and Simulation, Vol 41,
www.witpress.com, ISSN 1743-355X (on-line)
Computational Methods and Experimental Measurements XII 435
Table 1: Dimensions of silos used in the analyses.
D
(m)
h
o(m)
h
1(m)
Notation on graphs
5 10 10 1
5 10 20 2
5 10 30 3
10 10 20 4
10 10 30 5
10 10 40 6
10 10 50 7
15 10 25 8
15 10 30 9
15 10 40 10
15 10 50 11


Figure 12: Average deflection * E vs. height for silos with different
dimensions. (ep = 30 cm).

Figure 13: Average moment vs. height for silos with different dimensions. (ep
= 30 cm).
2005 WIT Press WIT Transactions on Modelling and Simulation, Vol 41,
www.witpress.com, ISSN 1743-355X (on-line)
436 Computational Methods and Experimental Measurements XII

Figure 14: Average shear vs. height for silos with different dimensions.
(ep = 30 cm).

Figure 15: Average force N in the circumferential direction vs. height for silos
with different dimensions. (ep = 30cm).
The dimensions (presented in Table 1) of silos chosen in this analysis satisfy
this condition and cover significant number of silo used. Other uncalculated
dimensions can be interpolated as required by the designing engineer.
Therefore, given a silo with specific dimensions, one can determine from the
curves presented in Figures 1219 the shear, moment, deflection and the
membrane force N in the circumferential direction; this can be done directly or
by interpolation between the curves.
Note that in the above graphs, the material stored in the silo has a unit weight
g = 0.75T/m
3
. To determine shear, moment, deflection or force N in the
circumferential direction for a stored material with a different unit weight, the
2005 WIT Press WIT Transactions on Modelling and Simulation, Vol 41,
www.witpress.com, ISSN 1743-355X (on-line)
Computational Methods and Experimental Measurements XII 437
values determined from the graphs need to be multiplied by (
material
/0.75) since
values are proportional to the unit weight.


Figure 16: Average deflection * E vs. height for silos with different
dimensions. (ep = 20 cm).

Figure 17: Average moment vs. height for silos with different dimensions.
(ep = 20 cm).
5 Conclusion
This work is divided into three parts, the first deals with the comparison between
all approaches (Janssen, Airy, Reimbert and water behavior), while the second
part was done to check the closed form solution of the differential equation
where finite element method was utilized to do this checking.
2005 WIT Press WIT Transactions on Modelling and Simulation, Vol 41,
www.witpress.com, ISSN 1743-355X (on-line)
438 Computational Methods and Experimental Measurements XII

Figure 18: Average shear vs. height for silos with different dimensions.
(ep = 20 cm).

Figure 19: Average force N in the circumferential direction vs. height for silos
with different dimensions. (ep = 20cm).
In the third part, average values were found for the three methods (Janssen,
Airy, and Reimbert). These average values were done for deflection, moment,
shear, and the circumferential membrane force, and were graphed for silos with
various heights, diameters, and wall thicknesses.
From the above we can conclude the following:
1- Janssen, Airy and Reimbert lateral pressure application approaches
give comparative results, while the water behavior approach gives larger
values than the above three methods especially in critical areas. Due to this
reason, the water behavior method was excluded in finding average values.
2- As silo height increases, forces and deflection increase; as the silo
diameter increases forces and deflection also increase.
3- The Finite Element Analysis is very close to whatever lateral pressure
approach used, as expected.
4- Average values for deflection and forces can be used from the graphs
for the purpose of design. Given a material with a unit weight g that needs to
be stored in a certain silo volume, one can suggest a corresponding height,
2005 WIT Press WIT Transactions on Modelling and Simulation, Vol 41,
www.witpress.com, ISSN 1743-355X (on-line)
Computational Methods and Experimental Measurements XII 439
diameter and wall thickness then determine the wall forces and perform the
design. If the exact dimensions are not available in the graphs, one can
interpolate and determine the corresponding values.
For further research, the author recommends that analysis be made for such
silos with wall friction keeping in mind that including friction will somewhat
complicate the differential equation. Also, additional work needs to be done in
the area of dynamic analysis and design of silos.
Notation
: Weight per unit volume for the stored material.
ho: Height from level 0 to the top of the hopper.
h1: Height from top of the hopper to the top of the silo.
E: Elastic modulus.
DE: Flexural rigidity.
ep: Silo wall thickness.
: Poissons ratio.
: Angle of internal friction for stored material.
ko: Rankine coefficient of active earth pressure the ratio of the horizontal
pressure to the vertical pressure.
D: Diameter of the silo.
R : Average radius.
R
h
: Hydraulic radius.
: Elastic length.
1: Angle of friction (stored material against wall).
: Internal friction.
y: Lateral displacement.
p
o
: Pressure of stored material at depth h1 below surface of stored material.
(Water behavior approach).
p: Lateral pressure.
y
g
: General solution of the differential equation.
y
p
(or fp1): Particular solution of the differential equation.
y
h
: Solution of the differential equation without right term.
Acknowledgements
The author thanks the Lebanese University, Faculty of Engineering, Branch II
for sponsoring part of this on going research. Also, the author thanks the
engineers Bassam Mazloum, Georges Abi Saad and Issam Abou Antoun who
helped in the set up of this paper.
References
[1] Timoshenko, S. and Gere, J., Theory of Elastic Stability, Second
Edition, McGraw-Hill, Mechanical Engineering Series, 1963.
2005 WIT Press WIT Transactions on Modelling and Simulation, Vol 41,
www.witpress.com, ISSN 1743-355X (on-line)
440 Computational Methods and Experimental Measurements XII
[2] Timoshenko, S. and Woinowsky-Krieger, S., Theory of Plates and
Shells, Second Edition, McGraw-Hill, Engineering Mechanics Series,
1959.
[3] Janssen H.A., Versuche Uber Getreidedruck in Silozellen, VDI
Zeitschrift, Dusseldorf, V.39, Aug. 31, 1885, pp 1045-1049.
[4] Reimbert M. and Reimbert A., Silos-Theory and Practice, Trans Tech
Publication, 1
st
Edition, 1976, Claustal, Germany.
[5] Airy W., The Pressure of Grain, Minutes of Proceedings, Institute of
Civil Engineers, London, V 131, 1897, pp 458-465.
[6] Safarian and Harris, Design and Construction of Silos and Bunker,
Published by Van Nostrand Reinhold Company, New York, 1984.
[7] Thonier, H., Conception et Calcul de Structure des Btiments, Presse de
lEcole Nationale des Ponts et Chausses, 1993.
[8] Fintel, M., Handbook of Concrete Engineering, Published by Van
Nostrand Reinhold Company, 2
nd
edition, New York, 1974.
[9] ACI Manual of Concrete Practice, Part 4,Committee313, American
Concrete Institute, Redford Station, Detroit, Michigan, 1995.
[10] Jenike A.W., Elsey P.J., and Wooley R.M., Flow Properties of Bulk
Solids, ASTM, Proceedings, V.60, 1960, pp. 1168-1181.
[11] Spiegel, M.R., Advanced Mathematics for Engineers & Scientists,
Schaums Outline Series, McGraw-Hill Book Company, 1971.
2005 WIT Press WIT Transactions on Modelling and Simulation, Vol 41,
www.witpress.com, ISSN 1743-355X (on-line)
Computational Methods and Experimental Measurements XII 441

You might also like