You are on page 1of 15

A simple and accurate solution for calculating stresses in

conical shells
Weicheng Cui
a,
*
, Junhou Pei
a
, Wei Zhang
b
a
School of Naval Architecture and Ocean Engineering, Shanghai Jiao Tong University, Jiao Tong, Shanghai 200030,
People's Republic of China
b
China Ship Scientic Research Center, P.O. Box 116, Wuxi, Jiangsu 214082, People's Republic of China
Received 19 January 2000; accepted 17 April 2000
Abstract
In this paper, new variable transformation formulas are introduced to solve the basic governing dierential equa-
tions for conical shells. By performing magnitude order analysis and neglecting the quantities with haR magnitude
order, the basic governing dierential equations for conical shells are transformed into a second-order dierential
equation with complex constant coecients. By solving this second-order dierential equation, a simple and accurate
solution for conical shells is derived. The present solution is simpler than the exact solution because it does not use
Bessel's functions, and also more accurate than the equivalent cylinder solution. Numerical examples are given to il-
lustrate this conclusion. The simple and accurate solution provides a quick means for analyzing stresses in conical
shells. 2000 Elsevier Science Ltd. All rights reserved.
Keywords: Conical shell; Stress analysis; Exact solution; Approximate solution; Simple and accurate solution
1. Introduction
Axisymmetric pressure hulls, such as cone and cylinder shells are important congurations which have been widely
used in engineering structures, such as submarine and submersible pressure hulls [1]. Calculation of stresses in these
structures is a necessity in design. For most of the axisymmetric thin-walled shells, exact solutions can be found but they
are very complex except for cylinders [2,3]. Special functions, such as Bessel functions, Thomson functions and Hankal
functions which are not very familiar to many engineers need to be used. Therefore, these exact solutions have not
found wide applications in practice. Finite element methods have been recommended in Ref. [1] but in terms of e-
ciency, this may not be so favorable, especially in preliminary design. Furthermore, in comparison with the analytical
solutions, nite element methods provide less insight in understanding the fundamental mechanical behaviour of the
structural parameters. Therefore, analytical solutions are still sought by some researchers. For the general axisymmetric
structures, there are two dierent approaches in this aspect. One is to use the concept of equivalent cylindrical shell, that
is, to dene an equivalent cylinder for the general axisymmetric shell and then apply the cylinder solution. This has been
used in Ref. [4] for submarine pressure hull design. However, the accuracy of these solutions has been found to be
poorer than the exact solutions. Therefore, these are called approximate solutions in this paper. The other is to in-
troduce some new types of variable transformations and apply the magnitude order analysis to simplify the governing
dierential equation, simple and accurate solutions can also be derived.
Computers and Structures 79 (2001) 265279
www.elsevier.com/locate/compstruc
*
Corresponding author. Tel.: +86-21-62933182; fax: +86-21-62933160.
0045-7949/01/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved.
PII: S0045- 7949( 00) 00139- 5
In Ref. [3], such a simple and accurate solution for the general shell of revolution was presented based on the
variable transformation formulas they introduced. The solution was expressed only by elementary functions such as
exponential functions and triangular functions. Conical shells are the shells of revolution, however, when we applied
their variable transformation formulas to conical shells, it was found that these formulas are singular. In order to
overcome this problem, a new type of variable transformations will be introduced in this paper. By using these new
variable transformation formulas together with the basic governing dierential equations, a simple and accurate
solution which has the same accuracy as the exact solution will be derived. The main purpose of this paper is to present
such a solution. In order to compare the present solution with the exact solution and the approximate solution for
conical shells, the latter two solutions are briey introduced in Sections 2 and 3, respectively. In Section 5, a numerical
example is used to compare the present solution with both exact solution and the approximate solution.
2. The exact solution for conical shells
Let 2a represent the vertex angle of a conical shell, S represents the coordinate along the generator of the conical
shell from the vertex of the conical shell (Fig. 1). For conical shells, the principal radii of curvature are R
1
= and
R
2
= S tana. By substituting these conditions to the governing dierential equations for the general shell of revolution
[2,3], the following governing dierential equations for conical shells can be obtained:
L U ( ) = Ehc
1
Y
L c
1
( ) =
U
D
Y
(1)
where
U = N
1
R
2
Y (2)
Nomenclature
a half vertex angle of the conical shell
u complement of the half vertex angle
h thickness of the conical shell
E Young's modulus of elasticity
l Poisson's ratio
D bending stiness
S coordinate along the generator of the conical shell
R
2
principal radius of curvature of the conical shell
w displacement in the radial direction
N
1
transverse shear force
T
1
stress resultant in the generator direction
T
2
stress resultant in radial direction
T
+
1
membrane stress resultant in the generator direction
T
+
2
membrane stress resultant in radial direction
M
1
bending moment in the generator direction
M
2
bending moment in radial direction
D
x
displacement along the circumferential direction
r
1
stress in the generator direction
r
2
stress in circumferential direction
q
1
face load per unit area along the generator direction
q
n
face load per unit area along the radial direction
q dead weight per unit area
P concentrated load
l slant length of the conical shell
r radius of parallel circular of the conical shell
266 W. Cui et al. / Computers and Structures 79 (2001) 265279
c
1
=
dw
dS
Y
L ( ) = tana S
d
2
( )
dS
2
_

d ( )
dS

( )
S
_
Y
(3)
where L( ) is the Laplace operator; N
1
, the transverse shear force; w, the displacement in the radial direction; h, the
thickness of the conical shell, E, Youngs modulus of elasticity, and D, the bending stiness which is
D =
Eh
3
12(1 l
2
)
Y (4)
where l is Poissons ratio. Substituting the rst formula of Eq. (1) into the second formula of Eq. (1), we have
LL U ( ) k
4
U = 0Y (5)
where
k
4
=
Eh
D
Y (6)
Eq. (5) can be split into two conjugate dierential equations of second order:
L(U) ik
2
U = 0X (7)
Using the second formula of Eq. (3) and U = N
1
R
2
, R
2
= S tana, Eq. (7) can be written as
d
2
N
1
S ( )
dS
2

1
S
d N
1
S ( )
dS

_

1
S
2

ik
2
S tana
_
N
1
S ( ) = 0X (8)
It is a pair of conjugate dierential equations of second order. The solutions to Eq. (8) are also a pair of conjugate
complex functions. As long as the solution to any one of the Eq. (8) is obtained, then the real part and the imaginary
part of the solution are derived. Using these two parts, we can nd the general solution for dierential equation (5).
Introducing the following transformations
g = x

i
_
= 2k

1
tana
_

S
_
i
_
Y
x = 2

3 1 l
2
( )
4
_

2
htana
_

S
_
Y
(9)
Fig. 1. A conical shell with parameters dened.
W. Cui et al. / Computers and Structures 79 (2001) 265279 267
and using any one of two equations in Eq. (8), we have
d
2
N
1
S ( )
dg
2

1
g
d N
1
S ( )
dg
1
_

4
g
2
_
N
1
S ( ) = 0 (10)
which is a Bessel equation. The solution to this equation is
N
1
S = C
1
J
2
g ( ) C
2
H
(1)
2
g ( )X (11)
In Eq. (11), J
2
(g) is the rst type of Bessel functions of second order; H
(1)
2
(g), the rst type of Hankal functions of second
order, and C
1
, C
2
are complex constant coecients. The functions mentioned above can be expressed by means of
Thomson functions and their rst derivatives [5],
J
2
g ( ) =
2
x
bei
/
x
_
berx
_
i
2
x
ber
/
x
_
beix
_
Y
H
(1)
2
g ( ) =
2
p
2
x
ker
/
x
_
keix
_
i
2
p
2
x
kei
/
x
_
kerx
_
Y
(12)
where the symbol (
/
) represents the derivative of the function. If we dene
C
1
= A
1
iA
2
Y
C
2
=
p
2
B
2
( iB
1
)Y
(13)
and substitute Eqs. (12) and (13) into Eq. (11) and select the real part, we will have
N
1
=
1
S
A
1
berx
_ _

2
x
bei
/
x
_
A
2
beix
_

2
x
ber
/
x
_
B
1
kerx
_

2
x
kei
/
x
_
B
2
keix
_

2
x
ker
/
x
__
Y (14)
where A
1
, A
2
, B
1
, B
2
are unknown constants which can be determined from boundary conditions. In many applications
the conical shell is long enough to make it possible to neglect the eect of one edge on another edge. In order to simplify
the calculation, we can choose the terms with B
1
and B
2
for edge AA of Fig. 1 and the terms with A
1
and A
2
for edge BB
of Fig. 1.
From the denition of U = N
1
R
2
, U can also be obtained from Eq. (14). Using the rst formula of Eq. (1) and the
second formula of Eq. (3), we can obtain
c
1
=
L U ( )
Eh
=
tan
2
a
Eh
S
d
2
N
1
S ( )
dS
2
_

d N
1
S ( )
dS
N
1
_
X (15)
After obtaining U and c
1
, the calculation of the stress resultants T
1
and T
2
, bending moments M
1
and M
2
, and dis-
placement D
x
along circumferential direction are performed using the basic denitions [2,3]:
T
1
=
U
R
2
cosu
sinu
= N
1
tanaY
T
2
=
dU
dS
=
d N
1
S ( )
dS
tanaY
M
1
= D
dc
1
dS
_

lcosu
R
2
sinu
c
1
_
= D
dc
1
dS
_

l
S
c
1
_
Y
M
2
= D
cosu
R
2
sinu
c
1
_
l
dc
1
dS
_
= D
c
1
S
_
l
dc
1
dS
_
Y
D
x
= e
h
S sina =
S tanasina
Eh
_

d N
1
S ( )
dS
lN
1
_
X
(16)
And the nal results for exact solution are expressed as follows:
268 W. Cui et al. / Computers and Structures 79 (2001) 265279
N
1
=

3(1 l
2
)
_
htana
8
x
2
A
1
berx
_ _

2
x
bei
/
x
_
A
2
beix
_

2
x
ber
/
x
_
B
1
kerx
_

2
x
kei
/
x
_
B
2
keix
_

2
x
ker
/
x
__
Y
c
1
=

3(1 l
2
)
_
2tana
Eh
2
A
1
_ _
beix
2
x
ber
/
x
_
A
2
berx
_

2
x
bei
/
x
_
B
1
_
keix
2
x
ker
/
x
_
B
2
kerx
_

2
x
kei
/
x
__
Y
T
1
=

3(1 l
2
)
_
h
8
x
2
A
1
berx
_ _

2
x
bei
/
x
_
A
2
beix
_

2
x
ber
/
x
_
B
1
kerx
_

2
x
kei
/
x
_
B
2
keix
_

2
x
ker
/
x
__
Y
T
2
= 4

3(1 l
2
)
_
h
A
1
_ _

2
x
2
berx
1
x
ber
/
x
4
x
3
bei
/
x
_
A
2
_

2
x
2
beix
4
x
3
ber
/
x
1
x
bei
/
x
_
B
1
_

2
x
2
kerx
1
x
ker
/
x
4
x
3
kei
/
x
_
B
2
_

2
x
2
keix
4
x
3
ker
/
x
1
x
kei
/
x
__
Y
M
1
= A
1
4(1
_
_
l)
beix
x
2
8(1 l)
ber
/
x
x
3

2
x
bei
/
x
_
A
2
_
4(1 l)
berx
x
2
8(1 l)
bei
/
x
x
3

2
x
ber
/
x
_
B
1
4(1
_
l)
keix
x
2
8(1 l)
ker
/
x
x
3

2
x
kei
/
x
_
B
2
_
4(1 l)
kerx
x
2
8(1 l)
kei
/
x
x
3

2
x
ker
/
x
__
Y
M
2
= A
1
_
_
4(1 l)
beix
x
2
8(1 l)
ber
/
x
x
3

2l
x
bei
/
x
_
A
2
4(1
_
l)
berx
x
2
8(1 l)
bei
/
x
x
3

2l
x
ber
/
x
_
B
1
_
4(1 l)
keix
x
2
8(1 l)
ker
/
x
x
3

2l
x
kei
/
x
_
B
2
4(1
_
l)
kerx
x
2
8(1 l)
kei
/
x
x
3

2l
x
ker
/
x
__
Y
D
x
=
tana sina
Eh
A
1
(1
_
_
l)berx 2(1 l)
bei
/
x
x

x
2
ber
/
x
_
A
2
(1
_
l)beix 2(1 l)
ber
/
x
x

x
2
bei
/
x
_
B
1
(1
_
l) kerx 2(1 l)
kei
/
x
x

x
2
ker
/
x
_
B
2
(1
_
l)keix 2(1 l)
ker
/
x
x

x
2
kei
/
x
__
X
(17)
It is important to select the calculation method of Thomson functions and their derivatives for obtaining stable nu-
merical results. In this paper, Thomson functions and their derivatives are calculated by means of recursive formulas for
Thomson functions and their derivatives [5]. Based on the recursive formulas for Thomson functions and their de-
rivatives, it is easy to calculate Thomson functions of any order and their derivatives of any order if we know Thomson
functions of zero order and their derivatives of rst order. Thomson functions of zero order and their derivatives of rst
order are calculated using the polynomial approximation formulas given in Ref. [5].
After obtaining the stress resultants T
1
and T
2
and bending moments M
1
and M
2
the stress calculations are per-
formed using the following expressions:
r
1
=
T
1
T
+
1
h

6M
1
h
2
Y
r
2
=
T
2
T
+
2
h

6M
2
h
2
Y
(18)
where T
+
1
, T
+
2
are the stress resultants for membrane solution which are given as follows [2,3]:
T
+
1
=
1
S
_
(q
1
_
q
n
tana)S dS C
_
Y
T
+
2
= q
n
S tanaY
(19)
W. Cui et al. / Computers and Structures 79 (2001) 265279 269
where q
1
, q
n
are the face load per unit area along the S direction and its normal direction, C is a constant to be de-
termined from the membrane boundary condition.
It is necessary to point out that the exact solution for conical shells given above is of an accuracy which is the same
as the thin-walled shell theory. Due to the various assumptions made in thin-walled shell theory, the error magnitude
order of thin-walled shell theory is haR, so the error magnitude order of the exact solution for conical shells given here is
also haR.
3. The approximate solution for conical shells
Considering the complexity of the exact solution for the shell of revolution, the approximate solutions based on the
concept of an equivalent cylindrical shell are generally used for solving problems on the shell of revolution [3,4]. Let the
values of u, S, r, R
2
= rasinu at the large end be u
0
, S
0
, r
0
, R
20
= r
0
asinu
0
, the cylindrical shell which the thickness is
the same as the shell of revolution and the radius is [R
20
[ is dened as the equivalent cylindrical shell. For the equivalent
cylindrical shell, the governing dierential equation for the axial symmetrical case becomes
d
4
c
1
(cylinder)
dS
4

Eh
DR
2
20
c
1
(cylinder) = 0X (20)
The error order between the governing dierential equation (20) and the governing dierential equation of the general
shell of revolution for the axial symmetrical case is

haR
_
. For Eq. (20), the solution can easily be found [24]. The
solutions for the conical shell as shown in Fig. 1 can be expressed as follows:
Ehc
1
=
d
dS
(Ehw)Y U = DR
2
d
3
w
dS
3
Y
T
1
= Dtana
d
3
w
dS
3
Y T
2
=
1
R
2
EhwY M
1
= D
d
2
w
dS
2
Y M
2
= lD
d
2
w
dS
2
Y
N
1
= D
d
3
w
dS
3
Y Q
x
=
D
cosa
d
3
w
dS
3
Y D
x
= wcosaY
(21)
where
Ehw = C
1
h(bS
/
) C
2
f(bS
/
) C
3
h(bS
/
) C
4
f(bS
/
)Y
d
dS
(Ehw) = b[C
1
u(bS
/
) C
2
w(bS
/
) C
3
u(bS
/
) C
4
w(bS
/
)[Y
d
2
dS
2
(Ehw) = 2b
2
[C
1
f(bS
/
) C
2
h(bS
/
) C
3
f(bS
/
) C
4
h(bS
/
)[Y
d
3
dS
3
(Ehw) = 2b
3
[C
1
w(bS
/
) C
2
u(bS
/
) C
3
w(bS
/
) C
4
u(bS
/
)[Y
(22)
h(bS
/
) = e
bS
/
cosbS
/
Y h(bS
/
) = e
bS
/
cosbS
/
Y
f(bS
/
) = e
bS
/
sinbS
/
Y f(bS
/
) = e
bS
/
sinbS
/
Y
u(bS
/
) = e
bS
/
( cosbS
/
sinbS
/
)Y u(bS
/
) = e
bS
/
( cosbS
/
sinbS
/
)Y
w(bS
/
) = e
bS
/
( cosbS
/
sinbS
/
)Y w(bS
/
) = e
bS
/
( cosbS
/
sinbS
/
)X
(23)
The unknown constants C
1
, C
2
, C
3
and C
4
in the expressions given above are determined from boundary conditions. As
we know the stress resultants T
1
and T
2
and bending moments M
1
and M
2
, the stress calculations are made using Eq.
(18). Due to the error introduced in the basic governing dierential equation, the accuracy of this solution is of a
magnitude of order of

haR
_
.
4. The simple and accurate solution for conical shells
The approximate solution for conical shells is simple but the accuracy is low. The accuracy of the exact solution for
conical shells is high but is very complex. Is it possible to derive a solution for conical shells which possesses both
270 W. Cui et al. / Computers and Structures 79 (2001) 265279
accuracy and simplicity? The answer is positive. In Ref. [3], by introducing the following variable transformation
formulas:
y = ~ c
1

R
2
R
1
sinu
_
Y (24)
n =

3 1 l
2
( )
4
_
_
u
u
0
R
1
[ [

R
2
[ [h
_ duY (25)
y
1
=

R
1
[ [
_

R
2
[ [c
4
_ yY (26)
they are able to derive such a solution for the general shell of revolution, where c is a constant parameter dened in
Appendix A. Their solutions are supposed to be valid for the general shell of revolution. However, for conical shells, the
principal radius of curvature R
1
= , thus the variable transformation formula (24) turns out y = 0, the variable
transformation formula (25) becomes R
1
= times du = 0, while the third variable transformation formula, Eq. (26),
also becomes y
1
=
+
0. Therefore, u cannot be used as a principal curvature coordinate for conical shells. In order to
overcome this singularity, a new set of the variable transformation formulas suitable for conical shells are proposed in
this paper. They are
y =

S
_
~ c
1
Y (27)
n =

3 1 l
2
( )
4
_

htana
_ 2

S
_
_
2

S
0
_
_
Y (28)
y
1
=
1

cS tana
4
_ yY (29)
where c is dened in Eq. (A.18). Following the similar procedure as that used in Ref. [3], the simple and accurate
solution has been re-derived for conical shells. The detailed process of derivation is given in Appendix A and the so-
lution is given as follows:
c
1
=

12 1 l
2
( )
_
Eh
2

R
5
20
4
_

S
4
_ c
1
e
n
cos n (
_
f
1
) c
2
e
n
cos( n f
2
)

Y
U =
Eh
2

12 1 l
2
( )
_
2k
2
Eh

R
20
S
4
_
c
1
e
n
sin n (
_
f
1
) c
2
e
n
sin( n f
2
)

R
5
20
4
_

S
4
_ c
1
e
n
sin n (
_
f
1
) c
2
e
n
sin( n f
2
)

Y
N
1
=

R
5
20
4
_
R
2

S
4
_ c
1
e
n
sin n (
_
f
1
) c
2
e
n
sin( n f
2
)

Y
(30)
T
1
= N
1
tana
=

R
5
20
4
_
R
2

S
4
_ tana c
1
e
n
sin n (
_
f
1
) c
2
e
n
sin( n f
2
)

Y
T
2
=

R
5
20
4
_

S
4
_ c
1
e
n
_ _

3 1 l
2
( )
4
_

Shtana
_ cos n ( f
1
)
1
4S
_

3 1 l
2
( )
4
_

Shtana
_
_
sin n ( f
1
)
_
c
2
e
n

3 1 l
2
( )
4
_

Shtana
_ cos(
_
n f
2
)
1
4S
_

3 1 l
2
( )
4
_

Shtana
_
_
sin( n f
2
)
__
Y
(31)
W. Cui et al. / Computers and Structures 79 (2001) 265279 271
k
1
=

12 1 l
2
( )
_
Eh
2

R
5
20
4
_

S
4
_ c
1
e
n
_ _

3
4
_
1 l
2
( )

Shtana
_ sin n ( f
1
)
1
4S
_

3 1 l
2
( )
4
_

Shtana
_
_
cos n ( f
1
)
_
c
2
e
n

3 1 l
2
( )
4
_

Shtana
_ sin(
_
n f
2
)
1
4S
_

3 1 l
2
( )
4
_

Shtana
_
_
cos( n f
2
)
__
Y
k
2
=

1
_
2 1 l
2
( )
Eh
2
tana
R
2

R
5
20
4
_

S
4
_ c
1
e
n
cos n (
_
f
1
) c
2
e
n
cos( n f
2
)

Y
(32)
M
1
=
h

12 1 l
2
( )
_

R
5
20
4
_

S
4
_ c
1
e
n
_ _

3 1 l
2
( )
4
_

Shtana
_ sin n ( f
1
)
1
4S
_

3 1 l
2
( )
4
_

Shtana
_ l
tana
R
2
_
cos n ( f
1
)
_
c
2
e
n

3 1 l
2
( )
4
_

Shtana
_ sin(
_
n f
2
)
1
4S
_

3 1 l
2
( )
4
_

Shtana
_ l
tana
R
2
_
cos( n f
2
)
__
Y
M
2
=
h

12 1 l
2
( )
_

R
5
20
4
_

S
4
_ c
1
e
n
tana
R
2
_ _ _

l
4S
l

3 1 l
2
( )
4
_

Shtana
_
_
cos n ( f
1
) l

3 1 l
2
( )
4
_

Shtana
_ sin n ( f
1
)
_
c
2
e
n
tana
R
2
_ _

l
4S
l

3 1 l
2
( )
4
_

Shtana
_
_
cos( n f
2
) l

3 1 l
2
( )
4
_

Shtana
_ sin( n f
2
)
__
Y
(33)
e
2
=
1
Eh
T
2
( lT
1
)
=
1
Eh

R
5
20
4
_

S
4
_ c
1
e
n
_ _

3 1 l
2
( )
4
_

Shtana
_ cos n ( f
1
)
1
4S
_

3 1 l
2
( )
4
_

Shtana
_ l
tana
R
2
_
sin n ( f
1
)
_
c
2
e
n

3 1 l
2
( )
4
_

Shtana
_ cos(
_
n f
2
)
1
4S
_

3 1 l
2
( )
4
_

Shtana
_ l
tana
R
2
_
sin( n f
2
)
__
X (34)
The displacement along the circumferential direction D
x
can be expressed as
D
x
= wcosa usina = e
2
R
2
cosa
=
R
2
cosa
Eh

R
5
20
4
_

S
4
_ c
1
e
n
_ _

3 1 l
2
( )
4
_

Shtana
_ cos n ( f
1
)
1
4S
_

3 1 l
2
( )
4
_

Shtana
_ l
tana
R
2
_
sin n ( f
1
)
_
c
2
e
n

3 1 l
2
( )
4
_

Shtana
_ cos(
_
n f
2
)
1
4S
_

3 1 l
2
( )
4
_

Shtana
_ l
tana
R
2
_
sin( n f
2
)
__
X (35)
5. Numerical examples
In order to test the present solution proposed in this paper, we calculate the stresses of a conical shell shown in Fig. 2
by means of the various solutions given above. The thickness h of the conical shell is 3 cm, the slant length l is 10 cm and
Fig. 2. A closed conical shell under loading.
272 W. Cui et al. / Computers and Structures 79 (2001) 265279
the half vertex angle a of the conical shell equals 60. The upper end of the conical shell is closed and the lower edge of
conical shell is supported by hinges, in other words, the boundary condition for the lower edge of the conical shell is
simply supported boundary condition. Consequently, the circle at the lowest edge is not movable but is able to turn
round. In the calculation of stresses two loading cases are considered. Loading case 1 is the dead weight q of the conical
shell. Let us assume that the dead weight q per unit area in the middle surface of the conical shell is 1 MPa. Loading case
2 is the concentrated load P. Let P equal 10 000 N. Youngs modulus of elasticity E and Poissons ratio l are
E = 2X1 10
5
MPa and l = 0X3, respectively.
The stress results obtained by all previously given solutions are given in Tables 1 and 2 and Fig. 3. Table 1
makes a comparison between the stress results of the exact solution and of the present solution in loading case 1. It
can be observed from Table 1 that the two stress results agree with each other very well. Table 2 gives the same
comparison in loading case 2. The comparison shows that the two stress results are still in good agreement except
for several small stresses (`10
6
MPa). Because of the calculation error it is possible that very small stresses are
not precise.
Fig. 3 makes comparison of the stress results obtained by the present solution and the approximate solution in
loading case 1. Maximum relative errors are 28X6%, 37X6% and 46X2% for h = 3X0, h = 4X5 and h = 6X0 cm, re-
spectively. It is evident from Fig. 3 that the relative error increases as the thickness of conical shell increases.
6. Conclusions
This paper has presented a simple and accurate solution for conical shells with the same accuracy as the exact
solution. The present solution was obtained by introducing a set of new variable transformations and by neglecting
the terms with the order of magnitude of haR. Dierent from the exact solution which are expressed by special
functions, the present solution for conical shells can be expressed only by the elementary functions, such as expo-
nential functions and triangular functions. Numerical examples have been used to compare the stress results obtained
by the present solution with that from the exact solution and almost identical results have been obtained. The present
solution is also compared with the approximate solution and it is found that quite large error in stresses could be
introduced by using the approximate solution for conical shells. Consequently, it can be concluded that the present
solution for conical shells is simpler compared to the exact solution and more accurate compared to the approximate
solution.
Appendix A
A.1. Derivation of the simple and accurate solution
In the derivation, it is necessary to use the equilibrium equations, geometrical relations and stressstrain relations, so
these equations are rst given as follows:
U = N
1
R
2
= N
1
S tanaY (AX1)
T
1
=
U
R
2
tanaY
T
2
=
dU
dS
Y
e
1
=
1
Eh
U
R
2
tana l
dU
dS
_ _
Y
e
2
=
1
Eh
dU
dS

lU
R
2
tana
_ _
Y
_

_
(AX2)
k
1
=
dc
1
dS
Y
k
2
=
tana
R
2
c
1
Y
M
1
= D k
1
lk
2
( ) = D
dc
1
dS
l
tana
R
2
c
1
_ _
Y
M
2
= D k
2
lk
1
( ) = D
tana
R
2
c
1
l
dc
1
dS
_ _
X
_

_
(AX3)
The governing dierential equations for conical shells in the coordinate system as shown in Fig. 1 have been given in
Eqs. (1)(3).
W. Cui et al. / Computers and Structures 79 (2001) 265279 273
T
a
b
l
e
1
C
o
m
p
a
r
i
s
o
n
o
f
t
h
e
s
t
r
e
s
s
r
e
s
u
l
t
s
o
b
t
a
i
n
e
d
b
y
t
h
e
e
x
a
c
t
s
o
l
u
t
i
o
n
a
n
d
t
h
e
p
r
e
s
e
n
t
s
o
l
u
t
i
o
n
i
n
l
o
a
d
i
n
g
c
a
s
e
1
S
(
c
m
)
1
0
0
0
9
9
0
9
7
0
9
5
0
9
0
0
8
0
0
6
5
0
5
0
0
3
0
0
1
0
0
r
1
1
(
M
P
a
)
E
x
a
c
t

3
2
1
9
.
1
6
9

2
1
1
6
.
0
2
2

9
4
8
.
6
2
3
7

7
3
4
.
0
8
1
4

1
7
8
7
.
2
9
7

2
7
8
4
.
7
9
4

2
1
6
0
.
7
3
0

1
6
6
6
.
9
6
7

1
0
0
0
.
0
0
1

3
3
3
.
3
3
3
P
r
e
s
e
n
t

3
2
1
9
.
2
5
7

2
1
1
5
.
1
8
8

9
4
6
.
3
5
2
7

7
3
1
.
2
7
1
5

1
7
8
6
.
0
7
6

2
7
8
5
.
6
3
5

2
1
6
0
.
6
4
5

1
6
6
6
.
9
7
2

1
0
0
0
.
0
0
1

3
3
3
.
3
3
3
r
1
2
(
M
P
a
)
E
x
a
c
t

3
2
1
9
.
1
6
9

4
3
2
7
.
9
2
0

5
4
6
9
.
9
6
1

5
6
1
5
.
1
9
0

4
2
6
3
.
5
0
8

2
5
0
0
.
1
7
6

2
1
7
2
.
5
1
2

1
6
6
6
.
3
6
1

9
9
9
.
9
9
9
4

3
3
3
.
3
3
3
P
r
e
s
e
n
t

3
2
1
9
.
2
5
7

4
3
2
8
.
9
3
6

5
4
7
2
.
4
3
4

5
6
1
8
.
2
1
1

4
2
6
4
.
8
6
6

2
5
4
9
.
3
0
2

2
1
7
2
.
6
0
2

1
6
6
6
.
3
5
6

9
9
9
.
9
9
9
3

3
3
3
.
3
3
3
r
2
1
(
M
P
a
)
E
x
a
c
t

7
7
7
.
9
2
8

1
0
8
3
.
4
7
9

1
9
3
9
.
3
1
9

2
8
3
9
.
8
3
5

4
2
7
6
.
3
5
6

4
1
3
3
.
3
2
3

3
2
4
3
.
2
7
3

2
5
0
0
.
2
1
4

1
5
0
0
.
0
0
1

5
0
0
.
0
0
0
P
r
e
s
e
n
t

7
7
7
.
8
1
8

1
0
8
2
.
7
2
3

1
9
3
8
.
1
4
9

2
8
3
9
.
0
5
8

4
2
7
7
.
2
0
4

4
1
3
3
.
7
0
0

3
2
4
3
.
2
3
5

2
5
0
0
.
2
1
5

1
5
0
0
.
0
0
1

5
0
0
.
0
0
0
r
2
2
(
M
P
a
)
E
x
a
c
t

1
1
5
3
.
5
7
2

2
1
1
6
.
8
2
9

3
6
0
8
.
2
4
9

4
5
3
2
.
5
8
3

5
0
6
8
.
3
6
3

4
0
4
6
.
9
1
5

3
2
4
7
.
6
9
9

2
4
9
9
.
9
9
7

1
5
0
0
.
0
0
1

5
0
0
.
0
0
0
P
r
e
s
e
n
t

1
1
5
3
.
7
3
6

2
1
1
6
.
8
9
6

3
6
0
8
.
7
1
4

4
5
3
3
.
6
6
8

5
0
6
9
.
8
6
7

4
0
4
6
.
7
1
7

3
2
4
7
.
7
2
0

2
4
9
9
.
9
9
4

1
5
0
0
.
0
0
1

5
0
0
.
0
0
0
T
a
b
l
e
2
C
o
m
p
a
r
i
s
o
n
o
f
t
h
e
s
t
r
e
s
s
r
e
s
u
l
t
s
o
b
t
a
i
n
e
d
b
y
t
h
e
e
x
a
c
t
s
o
l
u
t
i
o
n
a
n
d
t
h
e
p
r
e
s
e
n
t
s
o
l
u
t
i
o
n
i
n
l
o
a
d
i
n
g
c
a
s
e
2
S
(
c
m
)
1
0
0
0
9
9
0
9
7
0
9
5
0
9
0
0
8
0
0
6
5
0
5
0
0
3
0
0
1
0
0
r
1
1
(
M
P
a
)
E
x
a
c
t

1
.
2
3
6

1
0

1
.
3
4
6

1
0

1
.
4
7
3

1
0

1
.
5
1
3

1
0

1
.
4
7
3

1
0

1
.
5
2
1

1
0

1
.
8
8
5

1
0

2
.
4
5
0

1
0

4
.
0
8
4

1
0

1
.
2
2
5
P
r
e
s
e
n
t

1
.
2
3
6

1
0

1
.
3
4
6

1
0

1
.
4
7
3

1
0

1
.
5
1
3

1
0

1
.
4
7
3

1
0

1
.
5
2
0

1
0

1
.
8
8
5

1
0

2
.
4
5
0

1
0

4
.
0
8
4

1
0

1
.
2
2
5
r
1
2
(
M
P
a
)
E
x
a
c
t

1
.
2
3
6

1
0

1
.
1
4
3

1
0

1
.
0
5
8

1
0

1
.
0
6
5

1
0

1
.
2
4
5

1
0

1
.
5
4
2

1
0

1
.
8
8
4

1
0

2
.
4
5
0

1
0

4
.
0
8
4

1
0

1
.
2
2
5
P
r
e
s
e
n
t

1
.
2
3
6

1
0

1
.
1
4
3

1
0

1
.
0
5
7

1
0

1
.
0
6
4

1
0

1
.
2
4
5

1
0

1
.
5
4
2

1
0

1
.
8
8
4

1
0

2
.
4
5
0

1
0

4
.
0
8
4

1
0

1
.
2
2
5
r
2
1
(
M
P
a
)
E
x
a
c
t

3
.
8
8
0

1
0

3
.
5
5
3

1
0

2
.
6
7
4

1
0

1
.
7
5
5

1
0

2
.
0
5
5

1
0

3
1
.
2
2
5

1
0

6
.
1
8
1

1
0

5
1
.
9
5
6

1
0

6
1
.
1
0
3

1
0

1
.
3
6
4

1
0

1
2
P
r
e
s
e
n
t

3
.
8
8
0

1
0

3
.
5
5
4

1
0

2
.
6
7
6

1
0

1
.
7
5
6

1
0

2
.
0
4
7

1
0

3
1
.
2
2
8

1
0

6
.
2
1
7

1
0

5
1
.
9
7
6

1
0

6
1
.
1
3
1

1
0

1
.
8
2
6

1
0

1
2
r
2
2
(
M
P
a
)
E
x
a
c
t

3
.
5
3
4

1
0

2
.
6
0
3

1
0

1
.
1
4
1

1
0

1
.
9
9
8

1
0

3
5
.
2
2
2

1
0

3
4
.
3
1
1

1
0

2
.
1
1
4

1
0

3
.
0
9
1

1
0

8
5
.
2
5
1

1
0

5
.
3
9
7

1
0

1
2
P
r
e
s
e
n
t

3
.
5
3
4

1
0

2
.
6
0
3

1
0

1
.
1
4
0

1
0

1
.
9
8
8

1
0

3
5
.
2
3
6

1
0

3
4
.
2
9
3

1
0

2
.
0
9
5

1
0

5
.
2
2
7

1
0

8
5
.
1
9
0

1
0

5
.
7
3
4

1
0

1
2
274 W. Cui et al. / Computers and Structures 79 (2001) 265279
Introducing a complex variable
~ c
1
= c
1
i
U
Eh
2

12 1 l
2
( )
_ _ X (AX4)
Multiplying the rst formula of the Eq. (1) by ia(Eh
2
a

12(1 l
2
)
_
) then adding to the second formula of Eq. (1) yields
L(~ c
1
) i
~ c
1
h

12 1 l
2
( )
_ _ = 0X (AX5)
Letting
T = T
1
T
2
Y M = M
1
M
2
Y (AX6)
and substituting Eqs. (A.2) and (A.3) into Eq. (A.6), we have
T =
dU
dS
_

tana
R
2
U
_
Y
M = 1 ( l)D
dc
1
dS
_

tana
R
2
c
1
_
X
(AX7)
Complex stress resultants are dened as follows:
~
T
1
= T
1
i
Eh
2

12 1 l
2
( )
_ k
2
Y
~
T
2
= T
2
i
Eh
2

12 1 l
2
( )
_ k
1
Y
~
T =
~
T
1

~
T
2
= T i

12 1 l
2
( )
_
h
M
1 l
X
(AX8)
Substituting Eqs. (A.2), (A.3) and (A.7) into Eq. (A.8), the complex stress resultants expressed by complex argument ~ c
1
are obtained,
Fig. 3. The relative error between the present solution and the approximate solution.
W. Cui et al. / Computers and Structures 79 (2001) 265279 275
~
T
1
= i
Eh
2

12 1 l
2
( )
_
tana
R
2
~ c
1
Y
~
T
2
= i
Eh
2

12 1 l
2
( )
_
d~ c
1
dS
Y
~
T = i
Eh
2

12 1 l
2
( )
_
d~ c
1
dS
_

tana
R
2
~ c
1
_
X
(AX9)
Using the second formula of Eqs. (3), and (A.5) can be written as
d
2
~ c
1
dS
2

1
S
d~ c
1
dS

1
S
2
~ c
1
i
1
S tana
1
h

12 1 l
2
( )
_ _ ~ c
1
= 0X (AX10)
In order to eliminate the term with rst-order derivative with respect to S in Eq. (A.10), let us introduce new argument y
which is dened as
y = ~ c
1
exp
1
2
_
1
S
dS
_ _
=

S
_
~ c
1
X (AX11)
The rst-order and second-order derivatives of new argument y with respect to S
dy
dS
=

S
_
d~ c
1
dS

1
2

S
_ ~ c
1
Y
d
2
y
dS
2
=

S
_
d
2
~ c
1
dS
2

1

S
_
d~ c
1
dS

1
4

S
3
_ ~ c
1
X
(AX12)
Substituting Eqs. (A.11) and (A.12) into Eq. (A.10), we have
d
2
y
dS
2

3
4
1
S
2
y i
1
S tana
1
ha

12 1 l
2
( )
_ y = 0X (AX13)
Now let us perform the magnitude order analysis to every term of Eq. (A.13). As far as the general shell of revolution is
concerned, boundary eect will decay rapidly away from the boundary edge. From the point of view of the magnitude
order, calculating the derivative with respect to S every time is equivalent to multiplying by b =

3(1 l
2
)
4
_
a

h R
20
[ [
_ _ _
,
i.e., dividing by

h R
20
[ [
_
. Since the eect of the solution is only conned to local region near the boundary edge, it is
reasonable that the values of the geometrical parameter in the local region, such as S, R
2
, are considered to equal nearly
to the values at the boundary edge S
0
, R
20
. Therefore, for conical shells, S = R
2
atana, the ratio of magnitude order for
every term in Eq. (A.13) becomes
y
h R
20
[ [
X
y
R
20
[ [
2
tan
2
a a
X
y
R
20
[ [h
X
Multiplying every term by h R
20
[ [ yields
y X
htan
2
a
R
20
[ [
y X yX
If the conical shell is not too at, the value of tana will not be very large, then the order of magnitude of the three terms
is
1 X
h
R
20
[ [
X 1X
If we neglect the second term in Eq. (A.13), the error magnitude order will be ha R
20
[ [. Therefore, Eq. (A.13) can be
simplied to
d
2
y
dS
2
i
1
S tana
1
ha

12 1 l
2
( )
_ y = 0X (AX14)
The proceeding equation can be written as
276 W. Cui et al. / Computers and Structures 79 (2001) 265279
d
2
y
dS
2

~
F
2
y = 0Y (AX15)
where
~
F
2
= i
1
S tana
1
h

12 1 l
2
( )
_ _ X (AX16)
Solving Eq. (A.16), we have
~
F =
1 i

2
_
1

cS tana
_ Y (AX17)
where
c =
h

12 1 l
2
( )
_ X (AX18)
Eq. (A.15) is a dierential equation with variable coecient. In order to transform Eq. (A.15) to the dierential
equation with constant coecients we have to introduce the following argument transformations:
n =
1

2
_
_
s
s
0
dS

cS tana
_ =

3 1 l
2
( )
4
_

htana
_ 2

S
_
_
2

S
0
_
_
Y
y
1
= byY
b =
1

cS tana
4
_ =
~
F
f
_ _
1a2
Y
f =
1 i

2
_ X
(AX19)
Using Eq. (A.19), we can calculate the rst-order and the second-order derivatives of n with respect to S and the rst-
order and the second-order derivatives of b with respect to n.
dn
dS
=
b
2

2
_ =
1

2
_
~
F
f
Y
db
dn
=
1

2
_ f
1a2
~
F
3a2
d
~
F
dS
Y
d
2
b
dn
2
= n
3a2
~
F
5a2
d
2
~
F
dS
2
_
_

3
2
~
F
7a2
d
~
F
dS
_ _
2
_
_
X
(AX20)
From the second and the third formula of Eq. (A.19), we have
y =
y
1
b
=
~
F
f
_ _
1a2
y
1
X (AX21)
The derivatives of y with respect to S are
dy
dS
=
1

2
_ b
dy
1
dn
_

db
dn
y
1
_
Y
d
2
y
dS
2
=
1
2
~
F
f
~
F
f
_ _
1a2
d
2
y
1
dn
2
_
_

d
2
b
dn
2
y
1
_
_
X
(AX22)
By means of the third formula of Eq. (A.20), the second equation of Eq. (A.22) can be written as
d
2
y
dS
2
=
1
2
~
F
f
~
F
f
_ _
1a2
d
2
y
1
dn
2
_
_
_
f
3a2
~
F
5a2
d
2
~
F
dS
2
_
_

3
2
~
F
7a2
d
~
F
dS
_ _
2
_
_
y
_
_
_
X (AX23)
W. Cui et al. / Computers and Structures 79 (2001) 265279 277
Substituting Eqs. (A.21) and (A.23) into Eq. (A.15), we obtain
d
2
y
1
dn
2
1 ( i)
2
1
_

1
2
~
F
2
1
~
F
d
2
~
F
dS
2

3
2
1
~
F
2
d
~
F
dS
_ _
2
_
_
_
_
_

_
y
1
= 0X (AX24)
According to Eqs. (A.17) and (A.18), we know that the argument
~
F only depends on the conguration size and material
properties of the conical shell, therefore, the varying length of
~
F should be R
20
[ [. Based on the reason explained above
the magnitude orders of the derivative of
~
F with respect to S are
1
~
F
d
2
~
F
dS
2
~
1
~
F
d
dS
~
F
R
20
[ [
_ _
~
1
R
20
[ [
2
Y
1
~
F
2
d
~
F
dS
_ _
2
~
1
~
F
2
~
F
R
20
[ [
_ _
2
~
1
R
20
[ [
2
Y
1
2
~
F
2
1
~
F
d
2
~
F
dS
2

3
2
1
~
F
2
d
~
F
dS
_ _
2
_
_
_
_
~
1
2
~
F
2
1
R
20
[ [
2
~
h
R
20
[ [
Y
where the mark ~ expresses magnitude order equivalent. According to the magnitude order analysis given above the
underlined term in Eq. (A.24) can be neglected. The error magnitude order will be haR
2
after neglecting the underlined
term. Thus, Eq. (A.24) is simplied to
d
2
y
1
dn
2
1 ( i)
2
y
1
= 0X (AX25)
Eq. (A.25) is the dierential equation with complex constant and contains only the second-order derivative term and the
zero-order derivative term with respect to n. The solution to this dierential equation is
y
1
= D
/
1
e
1i ( )n
D
/
2
e
1i ( )n
Y (AX26)
where D
/
1
and D
/
2
are unknown complex constants. By using Eq. (A.19), y can be obtained as
y =

S
4
_
D
//
1
e
1i ( )n
_
D
//
2
e
1i ( )n

Y (AX27)
where D
//
1
and D
//
2
are unknown complex constants. The relations between D
/
1
, D
/
2
and D
//
1
, D
//
2
are
D
//
1
=

ctana
4
_
D
/
1
Y
D
//
2
=

ctana
4
_
D
/
2
X
(AX28)
Using Eq. (A.11), we can obtain the complex argument ~ c
1
,
~ c
1
=
1

S
4
_ D
//
1
e
1i ( )n
_
D
//
2
e
1i ( )n

X (AX29)
For simplicity, let
D
//
1
=
2k
2
Eh
c
1
e
if
1

R
20
4
_
Y
D
//
2
=
2k
2
Eh
c
2
e
if
2

R
20
4
_
Y
(AX30)
where
k =

3 1 l
2
( )
4
_

R
20
h
_
X (AX31)
Substituting Eq. (A.30) into Eq. (A.29), we have
~ c
1
=
2k
2
Eh

R
20
S
4
_
c
1
e
if
1
e
1i ( )n
_
c
2
e
if
2
e
1i ( )n

X (AX32)
278 W. Cui et al. / Computers and Structures 79 (2001) 265279
Now, we substitute Eq. (A.32) into Eq. (A.9) and then have
~
T
1
= i
Eh
2

12 1 l
2
( )
_
tana
R
2
~ c
1
Y
= i tana

R
5
20
4
_
R
2

S
4
_ c
1
e
if
1
e
1i ( )n
_
c
2
e
if
2
e
1i ( )n

Y
~
T
2
= i
Eh
2

12 1 l
2
( )
_
d~ c
1
dS
Y
=

R
5
20
4
_

S
4
_
_ _

3 1 l
2
( )
4
_

Shtana
_ i
1
4S
_

3 1 l
2
( )
4
_

Shtana
_
__
c
1
e
if
1
e
1i ( )n

3 1 l
2
( )
4
_

Shtana
_
_
i
1
4S
_

3 1 l
2
( )
4
_

Shtana
_
__
c
2
e
if
2
e
1i ( )n
_
X
(AX33)
Selecting the real part and the imaginary part of Eqs. (A.32) and (A.33), respectively, and using Eqs. (A.1)(A.4) and
(A.8), the stress resultants, bending moments and deformations can be calculated which are given in Section 4, Eqs.
(30)(35).
References
[1] Ross CTF. Pressure vessels under external pressure: statics and dynamics. London: Elsevier; 1990.
[2] Timoshenko S, Krieger SW. Theory of plates and shells. 2nd ed. New York: McGraw-Hill; 1959.
[3] Hwang KC, Xia ZX, Xue MD, Ren WM. Theory of plates and shells. Beijing: Tsinghua University; 1987 [in Chinese].
[4] TOM OMaiedC Crb! b BT, ACT CTaieH HrO O OU, ZrO Odked BC, Gpox-HOCT HOCT] GO OldO OlHO HO! b KO Olrb, CaH HrT T-GeT Tep,ypT, 1994 [in Russian].
[5] Abramowity M, Stegun IA. Handbook of mathematical functions with formulas, graphs and mathematical tables. National bureau
of standards, applied mathematics series. 55, New York: Dover; 1968.
W. Cui et al. / Computers and Structures 79 (2001) 265279 279

You might also like