See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/315648807

Calculation of Shells of Revolution with Arbitrary Meridian Oscillations

Chapter · March 2017

DOI: 10.1007/978-3-319-53363-6_17

CITATIONS READS

10 234

1 author:

Tatiana Vladimirovna Zinovieva

Institute for Problems in Mechanical Engineering Russian Academy of Sciences, Saint-Petersburg, Russia
38 PUBLICATIONS   125 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Mechanics of drill string View project

Stress-strain state and stability of expansion bellows View project

All content following this page was uploaded by Tatiana Vladimirovna Zinovieva on 19 October 2017.

The user has requested enhancement of the downloaded file.

Calculation of Shells of Revolution with
arbitrary meridian Oscillations

Tatiana V. Zinovieva1

Abstract The system of the equations and its numerical solution for free and
forced oscillations of shells of revolution with arbitrary meridian is obtained. A
variant of the classical theory of shells developed on the basis of Lagrangian
mechanics is used. The natural frequencies and amplitudes of oscillations of shells
with various meridians are defined by the finite difference method.

Key words Elastic shells, linear shell theory, forced oscillations, natural
frequencies, finite difference method.

Introduction

Calculation of oscillations of thin-walled structures is an actual task of modern

engineering; many technological objects perceiving dynamical load are shell of
revolution shaped – e.g. cases of steam generators or power unit turbines, aircraft
fuselages, pipelines, tanks etc. Of interest here is a determination of oscillation
amplitudes for such shells under the influence of stated forces as well as finding
out natural frequencies.
As a rule, the full system of equations for shells is obtained from equations
presented in third-dimensional theory of elasticity; for that, certain simplifying
assumptions are added; it led to creation of different shell theories [1-9]. However,
the use of direct approach to shells as deformable surfaces and the use of modern
analytical technologies gives a possibility to develop an exact theory.
The given paper is based on a variant of classical theory of shells as
surfaces with material normals; it is developed both on the basis of Lagrange
mechanics and virtual work principle [10-13]; the use of this theory in different
applications is described in articles [14-16].
The system of equations and its numerical solution for forced and free
oscillations of shells of revolution with arbitrary meridian is developed. Presented
data generalize results of study [17], where calculations of statics for shells of
revolution under the influence of axis-symmetrical loads were described.

Equations of Thin Shell Theory

The shell is considered as a material surface, which particles have five

degrees of freedom: three translations and two rotations. The movement is

1
Peter the Great Saint-Petersburg Рolytechnic University, St.-Petersburg, Russia
e-mail: tatiana.zinovieva@gmail.com
determined by the vector of small displacements u and the vector of small
rotation θ in tangent plane. For convenience's sake we introduce here the vector of
change of normal toward the shell: ϕ ≡ θ × n = nɺ (the dot denotes a small
increment during the deformation). The expression of work of external distributed
moment is used to introduce generalized force corresponding to ϕ : m ⋅ θ = m× ⋅ ϕ,
m× ≡ m × n .
The degree of surface deformation can be determined by tensors

ε = ( ∇u ) ⊥ , κ = − ( ∇ϕ )⊥ + b ⋅ ∇uT , ( b ≡ −∇n ) .
S
(1)

From this point on (...) ⊥ denotes tensor constituent in tangent plane; icons
(...) ,(...)T denote symmetrization and transposition. According to Kirchhoff's
S

kinematic hypothesis, the rotation is connected with the displacement

ϕ = −∇u ⋅ n. (2)

Virtual work principle allows creating the whole system of equations,

where characteristics of internal forces inside of the shell are introduced as
Lagrange multipliers: τ and µ are symmetrical tensors of stresses and moments,
Q is a shearing force vector; all they lay in tangent plane. The variational
formulation serves as a basis for derivation of balance equations for forces and
moments:

∇ ⋅ ( τ + µ ⋅ b + Qn ) + q = 0 , ( ∇ ⋅ µ )⊥ + Q = m × , (3)

and boundary conditions on a contour in general form:

 P 0 − ν ⋅ (T + Qn ) + ∂ l ( A ⋅ ln)  ⋅ δu − A ⋅ νn ⋅ ∂ ν δu = 0 , A ≡ M 0 × n + ν ⋅ µ , (4)

where q and m × are an external distributed force and the moment on the surface,
P 0 and M 0 are those on the contour. In total we have five equilibrium equations
in components and four scalar conditions on the shell contour.
The force and the moment are applied (from the side of ν ) to the section of
internal contour with the length of dl and the normal of ν

dF = ν ⋅ (T + Qn ) dl , dM = ν ⋅ ( µ × n ) dl , (T ≡ τ + µ ⋅ b ) .

For the isotropic material, the elasticity ratios are represented as follows

τ = C1εa + C2 ε, µ = D1κa + D2 κ , a ≡ ∇r , ε ≡ tr ( ε ) , κ ≡ tr ( κ ) , (5)

C1 = Ehν (1 − ν ) , C2 = Eh (1 + ν ) , D1 = C1h 12 , D2 = C2 h 12 .
2 2 2

Ratios C1 − D2 are taken as those applied in Kirchhoff plate; E is the

elasticity module of the material used for the shell, h is its thickness, ν is Poisson
ratio. There are no elasticity ratios for shearing force vector Q in classical theory;
instead, we have a relation (2).
The System of Equations for Shells of Revolution

Let's consider the shell with the surface formed by rotation of the meridian about
an axis x [10]. The meridian is set through the dependence of cylindrical
coordinates on an arc coordinate x = x ( s ), ρ = ρ( s ) , its placement on the surface
is determined by the angle θ (see Fig. 1).

Fig. 1 Revolution surface

Radius vector of surface points is given by the equality

r ( θ, s ) = x ( s ) i + ρ ( s ) eρ ( θ ) , eρ = j cos ( θ ) + k sin ( θ ) ,

where i , j , k are axial-oriented unitary vectors of Cartesian system, x , y , z ,

respectively.
The unitary vector for the line tangent to a parallel:
eθ = − j sin ( θ ) + k cos ( θ ) = eρ′ . For unit vectors of the line tangent and the normal
to the meridian we have

t = ∂r ∂s = x′ ( s ) i + ρ′ ( s ) eρ ( θ) , x′ = cos ψ , ρ′ = sin ψ ,
∂t ∂s = ωn , ω ≡ ψ′ ( s ) , n ≡ − i sin ψ + eρ cos ψ ,

here ω is a curvature of the meridian, ρ −1 is that for the parallel.

Shell displacement vector has three components: u = uϑ eϑ + ut t + un n .
Formulas (1) and (2) are to determine vector for the turn and deformation tensors

εθ = ρ−1 ( ∂ θuθ + uρ ) , uρ ≡ ut sin ψ + un cos ψ , εt = ∂ s ut − ωun ,

1
εθt = εtθ = ρ−1 ( ∂ θut − uθ sin ψ ) + ∂ s uθ  ,
2
ϕ θ = ρ −1 ( uθ cos ψ − ∂ θun ) , ϕt = −ωut − ∂ s un ,
 κ θ = −ρ−1 ( ∂ θϕθ + ϕt sin ψ ) − ρ−2 cos ψ ( ∂θuθ + uρ ) , κ t = ω ( ∂ s ut − ωun ) − ∂ s ϕt ,
κθt = −ρ−1 ( ∂ θϕt − ϕθ sin ψ + ∂ s uθ cos ψ ) , κtθ = ωρ−1 ( ∂ θut − uθ sin ψ ) − ∂ s ϕθ . (6)

We have relations for forces and moments from (5):

µ θ = ( D1 + D2 ) κ θ + D1κ t , µt = ( D1 + D2 ) κ t + D1κ θ ,
µ θt = D2 κ θt , µ tθ = D2 κ tθ ,
Tθ = ( C1 + C2 ) εθ + C1ε t − ρ −1µ θ cos ψ , Tt = ( C1 + C2 ) εt + C1εθ + ωµt ,
Tθt = C2 ε θt + ωµ θt , Ttθ = C2 εθt − ρ −1µtθ cos ψ. (7)

The system in components is enclosed with balance equations (3):

ρ −1 ∂ θTθ + (Tθt + Ttθ ) sin ψ  + ∂ sTtθ + ρ −1Qθ cos ψ + qθ − βuɺɺθ = 0 ,

ρ−1 ( Tt − Tθ ) sin ψ + ∂ θTθt  + ∂ sTt − ωQt + qt − βuɺɺt = 0 ,
−ρ−1Tθ cos ψ + ωTt + ρ−1 ( ∂ θQθ + Qt sin ψ ) + ∂ sQt + qn −βuɺɺn = 0 ,
ρ −1  ∂ θµθ + (µ θt + µtθ ) sin ψ  + ∂ sµtθ + Qθ = mθ× .
ρ −1  (µt − µθ ) sin ψ + ∂ θµθt  + ∂ sµt + Qt = mt× , (8)

where β is the mass of the shell per unit area.

Four scalar conditions on a shell contour follows from (4). In the case of
rigid fixing: uθ = ut = un = 0 , γ θ ≡ ∂ s un = 0 . On a free contour with the normal of
ν = t , the stretching force Tt = Pt 0 , bending moment µt = M θ0 and two
combinations connecting a torque with force components are set up

Fθ ≡ Ttθ − ρ −1 cos ψµtθ = Pθ0 + ρ−1 cos ψM t0 ,

Fn ≡ Qt − ρ −1∂ θµtθ = Pn0 + ρ −1∂ θ M t0 .

Forced and Free Oscillations

Equations (6) – (8) can be reduced to a system

∂ sY ( θ, s, t ) = f (Y ) , Y = ( uθ , ut , un , γ θ , Fθ , Tt , Fn , µ t )
T
(9)

for the column of unknown values set at shell edges.

Let's consider forced oscillations under the applied periodical load

 qt   qt ( s ) 
 qθ   qθ ( s )   q  =  q s  cos nθ sin Ωt ,
 m ×  =  m × s  sin nθ sin Ωt ,  n   n( )
 θ   θ ( )  m×   m× ( s ) 
 t   t 
where n is an integer and Ω is the loading frequency. We search for the solution
for this system (9) in analogical form, uθ , Fθ proportional to sin nθ and unknown
ut , un , γ θ , Tt , Fn , µ t to cos nθ .
For amplitudes we'll obtain a system

C2
uθ′ = ρ −1 sin ψuθ + G,
C2 + 4 D2ρ−2 cos2 ψ
4 D2 −2
ρ cos ψ ( −nωut + ρ−1n sin ψun − n γθ ) ,
2
G≡ Fθ + ρ −1nut +
C2 C2
1
ut′ = εt + ωun , εt =
C1 + C2
( Tt − C1 εθ − ωµt ) , εθ = ρ−1 ( nuθ + sin ψut + cos ψun ) ,

1
un′ = γθ , γθ′ = −ω′ut − ω2un − 2ωεt + κ t , κ t = ( µt − D1κ θ ) ,
D1 + D2
κθ = −ρ−1 ( nϕθ + sin ψϕt ) − ρ−2 cos ψ ( nuθ + sin ψut + cos ψun ) ,
ϕθ = ρ−1 ( cos ψuθ + nun ) , ϕt = −ωut − γθ .

Remaining four equations follow from balance equations (8):

Fθ′ = ρ −2 cos ψ  −nµθ + ( µθt + µt θ ) sin ψ  −

−ρ−1  −nTθ + ( Tθt + Fθ − ωµtθ ) sin ψ  − qθ − ρ −1 cos ψmθ× − βΩ2uθ ,
Tt′= ω ( Fn + ρ −1nµtθ ) − ρ −1 (Tt − Tθ ) sin ψ + nTθt  − qt − βΩ2 ut ,
Fn′ = ρ−2 sin ψn ( µθt + µtθ ) + ρ−1 cos ψTθ − ωTt − ρ−1 sin ψFn −
−ρ−2 n 2µθ − qn − ρ −1nmθ× − βΩ2 un ,
µ′t = −ρ −1 ( µt − µθ ) sin ψ + n ( µθt + µt θ ) − Fn + mt× . (10)

Coefficients in them are determined the similar way, with the use of equations
(6) – (7); those are not shown here due to their complexity.
Note that at n = 0 from (10) we'll obtain a system for amplitudes of axis-
symmetrical oscillations, while at Ω = 0 we'll get equations for shell statics.
ODE system (10) is supplemented by four boundary conditions for each
edge of the shell, or by eight periodicity conditions for closed shell. This boundary
problem is solved through the method of finite difference on s ∈ [0, L] interval.
Differential equations are approximated by difference ones with the constant step
of δ = L / N ; uθ … µt continuous argument s functions are replaced with
( uθ )i … ( µt )i (i = 0,..., N ) grid functions. Numeric scheme allows calculation of
approximate function values in nodes. Functions are then restored by
interpolation.
During system approximation (10) the implicit symmetrical one-step
difference scheme is used with a second-order accuracy [18]:

Yi +1 − Yi fi + f i +1
Y ′ ( s ) = f (Y ) ⇒ = , i = 0,..., N , (11)
δ 2
for that, one additional node with i = N + 1 number is introduced and the solution
is continued out of [ 0, L] definition range for one more δ interval on the right of
the boundary. Therefore, 8 N + 16 values become unknown.
Finite difference equations (11) are the system of 8 ( N + 1) algebraic
equations which is supplemented by difference analogues of eight boundary
conditions in nodes numbered i = 0, N . In total, we have 8 N + 16 equations for
the same number of unknown values and the problem can be solved through
standard algorithms. When it is solved, the values of unknown functions at the
additional node are discarded and not used during interpolation. The presented
scheme was implemented in program Mathematica [19].
In case of free oscillations of shell of revolution in (10) system, all external
loads are set as equal to zero ( q = 0 and m × = 0 ) while the frequency Ω is
unknown and should be determined. Boundary conditions are homogeneous. Using
finite difference method and the formula (11) we'll come to the homogeneous
system of linear algebraic equations for amplitudes:

B ( Ω ) ϒ = 0, ϒ = ({uθ }i ,... {µ t }i ) ,
T
(i = 0,..., N + 1) . (12)

Values Ω at which the system (12) has non-trivial solution are natural frequencies
of the shell, it’ll be possible only upon the condition

Det B ( Ω ) = 0 . (13)

Roots of the equation (13) can be found, e.g. through secant method [20].

Numerical Results

Presented algorithm can be used for the analysis of statics and oscillations of shells
of revolution with an arbitrary meridian. Such calculations can be illustrated by
giving several examples.
Let's consider the problem of static deflection of the shell due to its own
weight; the gravity force is directed perpendicular to rotation axis. In this case, the
distributed load is q = − β g j , where g is the acceleration of gravity; we have in
components for amplitudes at n = 1 qθ = βg , qt = −β g sin ψ, qn = −βg cos ψ.
For circular cylindrical shell with the radius of R and the length of l we
assume that ψ ≡ 0 , the meridian is set by equalities: x( s) = s , ρ( s) = R .
Calculations have been performed for the shell radius of R = 1 m, the
length of l = 20 m and the thickness of h = 0,01 m with both edges rigid fixed.
The shell is made of steel with the elasticity module of E = 210 GPa, Poisson
ratio of ν = 0, 28 , volumetric density of 7800 kg/m3.
Fig. 2a shows the displacement of the shell meridian along y axis.
Deformed shell shape is shown in Fig. 2b; when plotting it, all displacements were
scaled.
It is easy to calculate the displacement of thin-wall cylinder under gravity,
using beam-based model. The boundary problem

y + β1 g = 0,
u IV u y ( 0 ) = u y ( l ) = 0, u′y ( 0) = u′y ( l ) = 0

has its solution

 β1 g 4
uy ( s ) = −
24 EI
( s − 2ls3 + l 2 s 2 ) ,
where β1 is the mass of the beam per unit length and EI is its bending stiffness.
The result is shown at Fig. 2a, displacement difference in comparison with the
modeled by shell is 23%. With the increase of relative shell length this difference
is reduced – for instance, it is 7 % for l = 40 m.

(а) (b)

Fig. 2 Displacement under gravity a: 1 – cylindrical shell, 2 – beam; deformed

shape of a cylindrical shell b

For cone shell with the length of l we set the normal inclination ψ ≡ const
and the meridian through equities x ( s ) = x0 + s cos ψ , ρ( s ) = y0 + s sin ψ ;
coordinates x0 , y0 determine the position of the initial point of the meridian.
To the shell with parameters of ψ = 20 , x0 = 0 m, y0 = 0,1 m, l = 1 m,
thickness of h = 0,01 m with one edge rigid fixed, the force of qn = 1 MPa,
qθ = 0, 01 MPa was applied; the edge was under the distributed force of Pt 0 = 2
MN/m and the bending moment of M θ0 = 0,1 МN; the material used is steel with
properties described above. Fig. 3 represents calculated displacements and
deformed shape at n = 3 and Ω = 100 rad/s; Fig. 4 shows forces and bending
moment at the inside shell contour.

(а) (b)

Fig. 3 Cone shape displacements a: 1 – circular displacement uθ , 2 – axial one ut ,

3 – normal one un ; deformed shape of the cone shell b
 For shells with the complicated form, it is desirable to set the meridian as a
massive of x and ρ coordinates. Then, having the interpolation function built,
one can calculate the arc coordinate for each pair, using the formula

x
s ( x ) = ∫ 1 + ( ρ′x ) dx,
2

and then make x ( s ) and ρ ( s ) interpolations.

These functions should be smooth, as remaining coefficients of system (10)
are restored upon them:

cos ψ = x′s , sin ψ = ρ′s , ω = ρ′′s ( xs′ ) .

−1

(а) (b)

Fig. 4 Forces in cone shell a: 1 – Fθ , 2 – Tt , 3 – Fn ; bending moment µt b

Example of such calculations is shown below; Fig. 5a demonstrates

wavelike shell before its deformation, while 5b shows it under distributed axial
load of n = 4 applied; one shell edge is rigid fixed, the other is under axial and
normal components of the applied force.

(а) (b)

Fig. 5 Wavelike shell before a and after its deformation b

In technical applications corrugated shells are widely used and it makes

necessary the determination of their stressed state under the load applied and their
natural frequencies.
 The meridian of the corrugated shell is easy to set up as

ρ ( x ) = R0 + r sin Ωx , Ω ≜ 2πm x L , (11)

where R0 and r are forming radii, m is the number of waves along the length of
the shell and x L is its length.
Calculations for steel corrugated shell with the radii of R0 = 0.05 m and
r = 0,01 m, the length of xL = 0,55 m, the thickness of h = 0,001 m and the
number of waves m = 5.5 are performed. Shell edges are rigid fixed.
Natural frequencies found for the shell are presented in Tab. 1. The first
two frequencies correspond to beam-like forms ( n = 1 ), third and fourth ones are
to axis-symmetrical longitudinal-torsional ones ( n = 0 ).

Table 1 Natural frequencies Frequency n=0 n =1 n=2

f1 , Hz 2441 938 2567
f 2 , Hz 2555 2211 2870

The deformation of this corrugated shell under the normal pressure of

qn ( θ, s ) = s 2 cos 2θ МPа that vary over time with 2800 Hz frequency is
considered.

(а) (b)

Fig. 6 Corrugated shell before a and after its deformation b

Fig. 6 and 7 show shell shape before and after its deformation and the
amplitude of internal force factors.

(а) (b)

Fig. 7 Forces in corrugated shell a: 1 – Fθ , 2 – Tt , 3 – Fn ; bending moment µt b

Conclusions

The use of modern variant of classical theory of shells and computer mathematics
allow calculating forced and free oscillations of shells of revolution with the
arbitrary meridian. Obtained equations and developed calculation algorithm can be
used as a basis for the analysis of shells dynamics with arbitrary load applied.

Reference

1. Berdichevsky VL (2009) Variational principles of continuum mechanics.

V.2. Applications. Springer, Berlin, p 1014. doi: 10.1007/978-3-540-88469-9
2. Birger IA (1992) Rods, plates, shells (Sterzhni, plastinki, obolochki). Nauka,
Moscow, p 392 (in Russian)
3. Blaauwendraad J, Hoefakker JH (2014) Structural Shell Analysis. Springer
Netherlands, p 300. doi: 10.1007/978-94-007-6701-0
4. Chapelle D, Bathe KJ (2011) The finite element analysis of shells -
Fundamentals, Computational Fluid and Solid Mechanics, 2nd ed. Springer,
Berlin, p 410. doi: 10.1007/978-3-642-16408-8
5. Donnell LG (1982) Beams, plates and shells (Balki, plastiny i obolochki).
Nauka, Moscow, p 568 (in Russian)
6. Eremeev VA, Zubov LM (2008) Mechanis of elastic shells (Mekhanika
uprugih obolochek). Nauka, Moscow, p 280 (in Russian)
7. Goldenvejzer AL (1976) Theory of elastic thin shells (Teoriya uprugih tonkih
obolochek). Nauka, Moscow, p 512 (in Russian)
8. Khanh CL (1999) Vibrations of shells and rods. Springer, Berlin, p 423.
doi:10.1007/978-3-642-59911-8
9. Novozhilov VV, Chernykh KF, Mikhajlovskij EM (1991) Linear theory of
thin shells (Linejnaya teoriya tonkih obolochek). Politekhnika, Leningrad,
p 656 (in Russian)
10. Eliseev VV (2003) Mechanics of elastic bodies (Mekhanika uprugih tel) St.
Petersburg State Polytechn. University Publishing House, St Petersburg,
p 336 (in Russian)
11. Eliseev VV (2006) To nonlinear theory of elastic shells (K nelineynoy teorii
uprugih obolochek) St. Petersburg State Polytechn. University Journal
(Nauchno-tekhnicheskie vedomosti SPbGTU) No.3: 35-39 (in Russian)
12. Eliseev VV, Vetyukov YM (2010) Finite deformation of thin shells in the
context of analytical mechanics of material surfaces. Acta Mechanica 209:43-
57. doi: 10.1007/s00707-009-0154-7
13. Eliseev VV, Vetyukov YM (2014) Theory of shells as a product of analytical
technologies in elastic body mechanics. Shell Structures: Theory and
Applications 3: 81-85
14. Filippenko GV (2016) The vibrations of reservoirs and cylindrical supports
of hydro technical constructions partially submerged into the liquid. In: A.
Evgrafov (ed.), Advances in Mechanical Engineering, Lecture Notes in
Mechanical Engineering. Published by Springer International Publishing
Switzerland, p 115–126
15. Eliseev VV, Vetyukov YM, Zinov’eva TV (2011) Divergence of a helicoidal
shell in a pipe with a flowing fluid. Journal of Applied Mechanics and
Technical Physics 52: 450–458. doi: 10.1134/S0021894411030151
16. Yeliseyev VV, Zinovieva TV (2014) Two-dimensional (shell-type) and
three-dimensional models for elastic thin-walled cylinder. PNRPU
Mechanics Bulletin 3: 50-70. doi: 10.15593/perm.mech/2014.3.04
 17. Zinovieva TV (2012) Computational mechanics of elastic shells of revolution
in mechanical engineering calculations. In: Modern engineering: science and
education. Proceedings of second international scientific and practical
conference. Published by State Polytechnic University, SPb, p 335–343
18. Bakhvalov NS, Zhidkov NP, Kobelkov GG (2011) Numerical methods
(Chislennye metody) Binom. Laboratory of knoledges, Moscow, p 640 (In
Russian)
19. Borwein JM, Skerritt MB (2012) An introduction to modern mathematical
computing: with Mathematica. Springer, Vol. XVI, p 224
20. Chapra SC, Canale RP (2014) Numerical Methods for Engineers. McGraw-
Hill Education, New York, p 992

View publication stats