You are on page 1of 8

ECCS International Conference "Design, Inspection, Maintenance and Operation of

Cylindrical Steel Tanks and Pipelines", Prague, Czech Republic, 810 October, 2003

MODELING AND ANALYSIS OF


DISCRETELY SUPPORTED THINWALLED SILO SHELLS WITH
STRINGER STIFFENERS AT THE SUPPORTS
Siegfried Wallner1, Werner Guggenberger2
1Ingenieurbro Platzer, Mnzgrabenstrasse 58, 8010 Graz, Austria
Institute for Steel, Timber and Shell Structures, Graz University of Technology, Lessingstr. 25, 8010 Graz, Austria

ABSTRACT
There is a great variety of possibilities of discrete support of thinwalled metal silo structures depending on
the heaviness of the structure in response to the silo loads which are to be transferred to these supports. In the
present paper a compound structure is investigated which consists of a cylindrical shell (= barrel, upper part
of a silo structure) with uniformly spaced discrete stringer stiffeners attached to the shell which extend across
the full height of the cylindrical barrel. An analytical analysis model in developed based on modied Fluegge
shell equations in conjunction with double Fourier expansion of all displacement and stress components.
Special emphasis is put on the consistent and correct modelling of the longitudinal stiffener, including second
order stability effects, and the exible connection to the underlying shell. The adequate stiffener modelling is
crucial since mostly it is this component which is prone to premature yielding and buckling failure in the
region of the most strongly stressed crosssection. Parametric studies are carried out on the basis of linear
elastic stress and displacement analyses. The effects of stiffener dimensions (crosssectional area, moment of
inertia), the exibility of the connections (between shell segments and between the shell and the stiffener) and
different loading cases of introduction of the support force are studied. The results are discussed and
compared and lead to new insight into the load carrying behaviour of these type of compound structures.
Comparative analyses with ABAQUS demonstrate the wellworking of the developed analytical procedure.

INTRODUCTION
There is a great variety of possibilities of discrete support of thinwalled metal silo structures depending on
the heaviness of the structure in response to the silo loads which are to be transferred to these supports. In the
following a compound structure is investigated which consists of a cylindrical shell (= barrel, upper part of a
silo structure) with uniformly spaced discrete stringer stiffeners (mostly 4 to 8, sometimes up to 12). The
stringer stiffeners are attached to the shell wall by welding or bolting and extend over the full height of the
cylindrical barrel. On the one hand, the stringer stiffeners provide crosssectional area for the introduction of
the vertical support forces and distributing them over the height of the cylindrical barrel, on the other hand
they serve as effective stiffening measures against buckling of the silo shell (Fig. 1.a).
For the purpose of analytical structural analysis such complex structures may be subdivided into separate
substructures. The dominant loadcarrying system for the vertical (silo) loading down to the discrete supports

is represented by the isolated system which consists of the cylindrical barrel and the stringer stiffeners above
the supports. In this modelling, the annular base ring plate and the silo roof are mostly represented by simply
supported boundary conditions, simulating ring stiffeners which are rigid in their plane. The hopper itself is
not explicitely present in the model but it is represented by uniform axisymmetric meridional ring forces
which are acting at the transition joint and directing to the hopper apex. Traditionally, structural modelling
was often aimed at creating mathematical models of the structure under consideration, i.e. of the cylinder
stiffener system in particular, which are mechanical meaningful and at the same time analytically computable.
Two models are discussed in the following, rst a simple structural model and then a more rened structural
model.

hopper

qx = F/L

barrel
silo column
h

barrel

stringer stiffener

v=w=0
stringer stiffener

v=w=0

roof

v=w=0

v=w=0

Fig. 1. Modelling of the combined subsystem consisting of barrel and stringer stiffeners at the supports, a.
overall silo structure, b. continuous elastic shellstiffener system, c. membrane shell (traditional modelling)

STRUCTURAL MODELLING CONSIDERATIONS


Traditionally, simplied structural modelling is applied in the present situation, i.e. by uncoupling the stringer
stiffeners from the underlying cylindrical shell. Uniformly distributed meridional contact shear forces are
assumed which fulll global equilibrium but neglect kinematical compatibility (Schnell, 1955 and Rotter,
1985 and 1987a, b) (Fig. 1.c). In the ideal case of uniformly distributed vertical wall shear loading, the shell is
stressed by shear forces only, which are linearly varying in circumferential direction and constant down the
meridian. The stringer stiffeners exhibit constant transverse shear forces, zero bending moments and linearly
varying normal forces down the stiffener axis.
Similar simplied equilibrium modelling has also been used by Greiner (1984) in the context of analysing silo
shells with partial stringer stiffeners above the supports. Thereby the semimembrane cylindrical shell theory
was applied for analysing the uncoupled cylindrical shell problem with partial meridional line loading, resulting in an ordinary linear differential equation of fourth order in meridional direction (Greiner, Oery, 1984).
In the present paper, a more rened structural modelling approach of the coupled cylinderstiffener system
has been adopted. The interaction of the shell wall and the stringer stiffeners are accurately taken into
account, i.e. in the sense of linear elasticity theory. This is thought to be important for rational judgement of
the carrying capacity of the stringer stiffeners since plastic interaction between bending moment and normal
force takes place, including second order effects (Fig. 1.b). A comprehensive and more elaborate treatment of
this topic can be found in Wallner (2002).

barrel

U shell

W shell

radial joint

stringer stiffener

flexible joint

W beam, C

barrel

bolted
connection

U beam, C

axial joint

stringer stiffener

C II

Fig. 2. Flexible joint between stringer stiffener and barrel, a. bolted connection and structural idealization, b.
exible joint in axial direction and c. exible joint in radial direction

hplate
d
F stringer

F plate

F skirt

Fig. 3. Indroduction of support forces into the shellstiffener system by 3 different mechanisms, a. at the end
crosssection of the stringer stiffener, b. by the triangular insert plate, c. directly at the lower edge of the skirt

CYLINDRICAL SHELL EQUATIONS


The derivation of the set of basic equations of the circular cylindrical shell starts out from the general strain
expressions of the threedimensional continuum in orthogonal curvilinear (cylindrical) coordinates. The shell
kinematics is rst described in terms of 3 displacements and 2 rotations of the shell midsurface in a classical
way. By subsequently imposing the normality condition, the shell strain parameters are obtained in dependency of the 3 displacements and represented in concise differential operator notation (Eqn 1). The material stiffness matrix D for orthotropic elastic material behaviour is shown in Eqn 2. The "thinshell" assumptions have
the effect of complete uncoupling of the membrane and bending stiffness parts. The equilibrium operator mat (Eqn 3) is "adjoint" or "workconjugate" to the kinematical operator matrix B. This relationship is dicrix B
tated by rigorous rules based on the principle of virtual work. This important fact is stateoftheart knowledge today, but it has not been sufciently recognized in the past (Basar et al, 1985 and 2000) (Linder, 2001).
x

e =

x
x R

Bu

( )

( )

0
1 ( ) ( )
= -- V
R 0
0
( )

( ) ( )

x R

( ) 2( )

(1)

nx
n

n =

n x
mx R

De =

m R
m x R

Dx

S x

Kx R K R

K R K R

K x R

(2)

x R
R

x R

nx

T n + q = ---1
B
R

( )
0

( )

( ) ( )

( )

( )

( )

( ) 2( )

n x
mx R

qx
+ q =

m R

(3)

qz

m x R

Starting from the equilibrium equations, the fundamental set of three deformationbased partial differential
equations is obtained by successively inserting the constitutive and the kinematical relationships. These steps
of derivation are indicated below again using matrix differential operator notation (Eqn 4).
T n + q = 0 and n = D e and e = B u
B

T D B u + q =
B

K0 u + q = 0

(4)

Concluding, the comparison with Flgges classical equations (Flgge, 1973) reveals discrepancies of his
kinematic and constitutive operators which is not satisfactory from the conceptual point of view. This is due to
the erroneous denitions of the membrane shear strains, torsional curvature and the circumferential bending
curvature adopted in Flgges work. Therefore the resulting material stiffness matrix becomes unsymmetric.
However, the nal set of differential equations is correct (Eqn 4), since the errors cancel out accordingly
(Wallner, 2002).

Second order effects


Second order effects are caused by initial membrane stresses nx,init and n,init in axial and circumferential
directions. Additional deviating forces evolve since these membrane stresses act in the directions of the
deformed shell elements rather than the undeformed ones, thus contributing to the overall equilibrium of an
innitesimal shell element. These additional forces are summarized in the "geometrical stiffness (differential
operator) matrix" Kg which mainly consists of two parts relating to the respective initial membrane normal
stress components (Eqn 5). Second order effects may be destabilizing (axial compression nx,init) or stabilizing (silo pressure dependent n,init). By combination with the usual rst order stiffness (differential operator)
matrix K0 (Eqn 4) the complete second order stiffness (differential operator) matrix of the shell Kshell results.

x, init
K g = ------------
2

( )

( )

( )

n , init
-
+ -------------2
R

( )

( )

( )

with K shell = K 0 + K g

(5)

( ) ( ) ( )

Also in this case Flgges derivation (Flgge, 1932) exhibits a slight defect. This time an offdiagonal term in
the geometrical stiffness matrix is missing leading to unsymmetry (Eqn 5, greyshaded term) (Wallner, 2002).

REFINED ANALYSIS MODEL


A rened analysis model is developed for the combined cylinderstiffener system by consideration of the
elastic exibility of the cylinderstiffener connection. The boundary condition at the circumferential edges of
the shell are assumed simply supported (V = W = 0). A classical nite strip approach with Fourier expansion
in meridional direction and analytical solution in circumferential direction would be straightforward. Since
the stringer stiffeners are equally spaced around the circumference the particular shell problem can alternatively also be solved by Fourier double series expansions, by inserting displacement and load expansions
(Eqn 6) into the deformationbased equilibrium equations (Eqn 4). This is the approach adopted in this paper.
U =
V =
W =

U
n

cos
n

mn cos m

sin
n

mn sin m

qx =

x mn cos m

mn sin m

z mn cos m

and

q =

sin
n

mn cos m

qz =

cos
n
sin
n

with

nR
n = ------------------L

(6)

sin
n

The shell is loaded by the statically indeterminate meridional and radial connection forces which act along the
discrete joint meridians. For each meridional wave number n, appropriate circumferential expansions of these
concentrated unit loadings qx,n=1 and qz,n=1 have to be carried out which correspond to cyclic DiracDelta
circumferential distributions. Keeping the meridional wave numbers n at xed values and adopting the meridional unit load case qx,n=1 and qz,n=0, the contributions of the resulting deformations Umn and Wmn, due to
the Fourier expansion in circumferential direction, are added together (Eqn 8), thus yielding three exibility
coefcients. The procedure is repeated for the complementary radial unit load case qx,n=0 and qz,n=1 and
yields three further exibility coefcients. Assembling the exibility coefcients properly and discarding the
coefcients relating to circumferential actions yields the reduced 2by2 exibility matrix of the shell joint.
Subsequent inversion yields the discrete stiffness matrix of the shell joint, valid for wave number n

shell, n =
C

K shell, m, n
1

q x, m, n

0
0

shell, n, red 1
shell, n = C
K

(7)

q z, m, n

The global analysis model consists of three parts, the upright cylinder, the equally spaced stringer stiffeners
and the elastic exible connections in meridional and radial directions (Fig. 2). The exibility of the circumferential connections of the shell segments themselves may be approximately taken into account in a smeared
manner, by assuming equivalent orthotropic membrane stiffnesses (Eqn 2). Second order effects are taken
into account for the shell and the stiffener by assuming equivalent uniform distributions of the initial normal
stresses in the shell and of the initial normal force in the stiffener. These equivalent initial stress values are
chosen according to approximate mean values, obtained by prior rst order analysis of the system. The
stringer stiffener is modelled as Euler Bernoulli beam with excentric axis of reference (C), the continuous
displacement parameters being Ubeam,C and Wbeam,C. The coupled system of ordinary differential equations
is then solved by Fourier expansion in meridional direction yielding a linear twobytwo stiffness relationship between the displacement amplitudes and the connection force amplitudes, valid for wave number n.
C II

n = K
shell, n + K
beam, n + K
C =
K

C II

C II

C
shell
C II

C
C II

beam
C

U shell

C II
C

C II

C
C II

n =
and U

W shell

(8)

U beam, C
W beam, C

For each meridional wave number n a global system stiffness matrix can be assembled. For rigid connection
of the stringer stiffener and the shell the twobytwo stiffness matrices of these two members are directly
added together yielding the twobytwo system stiffness matrix. In the case of exible connections there are
four displacement unknowns since the meridional and radial displacement amplitudes of the shell and the
stiffener are independent, thus permitting relative displacements (Eqn 8). The spring stiffnesses of the elastic
connection are C II and C in longitudinal and transverse directions respectively. The fourbyfour stiffness
matrix of the elastic exible connection is independent of the wave number n. The assembly of the stiffness
matrices of the individual components yields the global fourbyfour system stiffness matrix. The system is
loaded by selfequilibrating meridional forces which depend on the type of support (Fig. 3) and are represented by Fourier series expansions. The summation over the wave number n completes the analysis. A computer
programme has been developed (Matlab, 1997) which serves as a tool for carrying out comprehensive parametric studies. Comparisons with Abaqus (Hibbit et al, 2002) show the accuracy of the developed procedure.

PRACTICAL EXAMPLE
The simplied model (Fig.1.c) is motivated by rigid-plastic considerations (lower bound static limit theorem).
However, it is easily comprehensible that this model is also exactly valid in the purely elastic case if the
crosssectional area A of the stringer stiffener tends to innity. Therefore, the investigation of the effect of the
crosssectional stiffness EA is of primary interest. It turns out that major bending is introduced into the stiffener caused by this axial exibility. Additional second order effects and the plastic momentnormalforce interaction may cause premature failure of the stringer which has to be considered in engineering design. The
system parameters are as follows (Figs 1.a and 3): R = 500 cm, t = 1 cm (i.e. R/t = 500), L/R = 2, hplate/L =
0.1, d/R = 0.157; E = 21000 kN/cm2, v = 0.3. Stringer stiffener: HEB 300 with excentricity e = h/2 = 15 cm;
number of supports nsup = 4. The results are nondimensionalized adopting the following reference values:
N 0 = F ; M 0 = F ( h + t ) 2 ; n x0 = ( F n sup ) ( 2R )

and n x0 = F ( 2L )

(9)

In the rst study the crosssectional area of the stiffener is varied. The excentricity and the bending stiffness
remain unchanged at the same time. The nondimensional area parameter adopts the values 0.1, 0.2, 1, 20;
the comparison value for the HEB 300 amounts to = nsup x A/(2R x t) = 0.2. The results are presented in
the diagrams Figs 4.ad. The solid lines for = 20 show practically the results pertaining to the limiting case
of the mentioned membrane model. The decrease of the axial stiffness causes decreasing normal forces in the
stiffener and increasing axial normal stresses in the shell. For = 0.1 the stiffener normal force has halvened
and about half of the axial load is carried by the shell (Figs 4.a, c). This is accompanied by increasing inward
directed bending of the stiffener (Fig. 4.b). At the same time, with decreasing , the shear membrane forces
start to develop a pronounced peak in the zone close to the lower boundary. This local membrane phenomenon is not straightforward to deal with in practical engineering design (Fig. 4.d) (prEN 199341, 1999).
In the second study the effect of three different modes of introduction of the support forces is investigated, i.e.
the stringer mode, plate mode and skirt mode according to Fig. 3, at xed value of the crosssectional
area parameter ( = 0.2). For all three modes the same magnitude of the support force F is applied thus
permitting direct comparison of the effectiveness of load introduction. The results are presented in Figs 4.e, f
and show the nondimensional normal force and bending moment distributions in the stringer stiffener.
Support force introduction by the "plate" and "skirt" mechanisms has two effects: Firstly, normal forces are
reduced due to the indirect load introduction at the lower part of the shell (Fig. 4.e). Secondly, bending
moments are even increased in the region above the support (< 0.2L) but they are almost zero in the remaining upper part of the stringer. A clear illustration of the mechanical behaviour of the coupled system is provided by an analogy model consisting of a composite beam with a exible elastic interface (Wallner, 2002).

0,8
0,7

0,8
0,7

0,4

0,6

X/L

0,5

= 0.1
= 0.2
= 1.0
= 20

0,9

e
an
br
em
lm
ea
id

0,6

X/L

= 0.1
= 0.2
= 1.0
= 20

0,9

0,5
0,4

0,3

0,3

0,2

0,2
0,1

0,1

0
0

-0,2

-0,4

-0,6

-0,8

0
0

-1

-0,1

-0,2

= 0.1
= 0.2
= 1.0
= 20

0,9
0,8
0,7

0,8
0,7

-0,5

0,6

X/L

X/L

-0,4

= 0.1
= 0.2
= 1.0
= 20

0,9

0,6
0,5

0,5

0,4

0,4

0,3

0,3

0,2

0,2
0,1

0,1

0
0

-1

-2

-3

-4

0
0

-5

-2

-4

0.8
0.8

Fplate

em
an
br

= 0.2

0.6
0.6

Fskirt
Fplate

0.5
0.5
0.4
0.4

400
0.4
300
0.3

0.3
0.3

200

0.2
0.2

0.2

0.1

100

0.1

0.0 0

-10

Fstringer
0.7
0.7

lm

500
0.5

0.9
0.9

Fskirt

ea

600
0.6

= 0.2
id

700
0.7

Fstringer

X/L []

800
0.8

-8

1.0
1.0

1000
1.0
900
0.9

-6

nx // nxfi,ref
nx0
nxfi

nx
nx // nx,0
nx0

X/L []

-0,3

M / M,0

N / N,0

-0,1 -0,2 -0,3 -0,4 -0,5 -0,6 -0,7 -0,8 -0,9

N / N,0

-1

0.1
0.0

0.00,4

0,32 0,24 0,16 0,08 0

-0,08-0,16-0,24-0,32-0,40

M / M,0

Fig. 4. Nondimensional a. normal force, b. bending moment in the stringer stiffner, c. meridional normal
stress and d. shear membrane stress in the cylindrical shell for varying area of the stiffener crosssection;
e. normal force and f. bending moment in the stringer stiffener for varying introduction of support forces.

CONCLUSION
A rened model has been presented for the efcient analysis of discretely supported thinwalled silo shells
with stringer stiffeners at the supports. The distinct feature of this model is the acurate representation of the
elastic exible connection of the silo shell with the attached stringer stiffeners. The analysis model is based on
modied Fluegge shell equations. The discretization is carried out by Fourier double series expansions in
meridional and circumferential directions, similar to the nitestrip methodology. The effectivity of the
developed procedure could be veried by successful comparison with Abaqus nite element calculations.
Parametric studies were carried out in which the crosssectional area of the stringer stiffener was varied. The
results were presented in nondimensional form. Finally, different modes of load introduction of practical
relevance were also investigated.
Advanced effects like the exibility of the circumferential bolted connections of the segments of the silo shell,
and stepped wall thickness of these segments in vertical direction may be approximately taken into account in
the present modelling by adopting equivalent mean values of the orthotropic membrane stiffness parameters.
Meridional wall segments in multicell silos may be easily incorporated by introducing additional plane
stress wall segments which exhibit wavenumberdependent coupled stiffness coefcients.

REFERENCES
Basar, Y., Krtzig, W.B. (1985). Mechanik der Flchentragwerke, Vieweg & Sohn, Braunschweig.
Basar, Y., Krtzig, W.B. (2000). Theory of shell structures, VDIFortschrittBerichte Nr. 258, Reihe 18:
Mechanik/Bruchmechanik, 2. Au., VDI Verlag GmbH, Dsseldorf 2001.
Flgge, W. (1932).. Die Stabilitt der Kreiszylinderschale, Ing. Arch., Band III, pp 463506.
Flgge, W. (1973). Stresses in Shells, 2nd. Ed., Springer.
Greiner, R. (1984). Zur Lngskrafteinleitung in stehende zylindrische Behlter aus Stahl. Der Stahlbau 53 (7),
Ernst & Sohn, Berlin.
Guggenberger, W., Zotter, J., Glnzer, S. (2003). Automated Analysis and Design of ThinWalled Metal Tank
and Silo Structures, Int. ECCS Conference Design, Inspection, Maintenance and Operation of Cylindrical
Steel Tanks and Pipelines, PragueKralupy, 811 Oct. 2003.
Hibbit, Karlsson and Sorensen (2002). ABAQUS Version 6.3, Theory Manual, Pawtucket, Rhode Island.
Linder, Ch. (2001). Theory of general shells of revolution and development of an analogy model for the
efcient computation of axisymmetric edge bending effects, Diploma thesis, Institute of Steel, Timber and
Shell Structures, Graz University of Technology, September 2001, 346 p.
Matlab (1997). MATLAB Users & Reference Manual. Matlab Version 5.2, MathWorks Inc.,Mass., USA.
Oery, H. et al (1984). Beitrag zur Bemessung der Schalen von Metallsilos. Der Stahlbau 53 (8), Ernst & Sohn,
prEN 199341 (1999). Eurocode 3: Design of Steel Structures, Part 4.1 Silos. CEN, European Commitee for
Standardisation, Brussels.
Rotter, J.M. (1985). Analysis and Design of Ring Beams, in: Design of Steel Bins for the Storage of Bulk
Solids, ed. by J.M. Rotter, University of Sydney, March 1985, pp 164183.
Rotter, J.M. (1987a). Membrane Theory of Shells for Bins and Silos, Trans. of Mech. Eng., Inst. of Eng.,
Australia, Vol. ME12, No. 3, Sept. 1987, pp 135147.
Rotter, J.M. (1987b). Bending Theory of Shells for Bins and Silos, Trans. of Mech. Eng., Inst. of Eng.,
Australia, Vol. ME12, No. 3, Sept. 1987, pp 147159.
Schnell, W. (1955). Krafteinleitung in versteifte Kreiszylinderschalen, Zeitschrift f. Flugwissenschaften (3)
Wallner, S. (2002). Modelling and Plastic Carrying Capacity of Discretely Supported Steel Silo Structures
with Longitudinal Stiffeners, Doctoral Thesis, Graz Un iversity of Technology, April 2002.

You might also like