§2.

13 Reduction for Ring

In the event that the shell is very short, then the solution can be simplified. When
the loading is self-equilibrating, the behavior is that of a beam (without foundation).
When a resultant force and/or moment per unit length of the circumference exists, the
behavior is that of a ring. A derivation can be found in many texts; here it is shown that
the shell equations do yield the correct result. The approximation is that the ring cross
section remains undistorted and the only significant stress is the component in the
circumferential direction. For simplicity, a short flat plate element of length L, with L
<< r, is considered. The constitutive relations (2.8.15-16) give

N θ = E t εθ
(2.13.1)

while the equilibrium equation (2.4.10) gives

N θ = r pH
(2.13.2)

The quantities can be integrated over the element. When E and p H are constant, the
result is

f = LpH = ELt h = EA h
r2 r2 (2.13.3)

where f is the total radial force per unit length in the circumferential direction, and A is
the cross-sectional area in the meridional plane of the element. A moment loading of the
flat-plate element is provided by Q. From (2.8.17-18) is obtained

3
Et
M θ = κθ
12 (2.13.4)
while (2.4.12) gives

M θ =–rQ (2.13.5)

Together these give

3
EtL EI
m =–L Q = χ = 2
χ
2
12 r r (2.13.6)

in which m is the externally applied moment per unit length of the circumference. For
the general cross section of the ring, I is the area moment of inertia about the axis
passing through the centroid and parallel to the r - axis. The results may be placed into
the form of a stiffness matrix:

46
 EI 0
rm = r χ
•
rf 0 EA h
r (2.13.7)

47
 Chapter 3

Cylindrical Shell

The cylinder is a very fundamental shape which is extensively utilized for pipes,
pressure vessels, liquid storage tanks, and as the basic structural element in tubular frame
structures, including bicycles, off-shore drilling platforms, and the long bones of the
skeleton in vertebrates.

§3.1 Equations and Analogy with Beam on an Elastic Foundation

The cylinder is relatively easy to fabricate and analyze, since the geometry is
rather simple:

ϕ =π/2
r = constant (3.1.1-2)

The Novozhilov equation (2.11.11) for the homogeneous, isotropic shell with constant
thickness reduces to an equation with constant coefficients:

2
– i c r d H + H = i c d r 2 pH – ν r V
d s2 ds (3.1.3)

while the coefficient matrix of the equation (2.11.22) reduces to:

0 1 0 0

0 0 0 Εt
A= r

1 0 0 0
r Ε t c2
0 0 –1 0
(3.1.4)

while the load vector is:

0
r pH - ν V
P =
0
0 (3.1.5)

These equations have exactly the same form as the equations for a beam on an elastic
foundation (Winkler foundation ) with the following replacements:
48
 Cylindrical Beam on Elastic
Shell Foundation

2
Etc EI Beam bending stiffness per
unit width
2
Et /r k Foundation stiffness modulus
pH – ν V / r pH Pressure load (3.1.6-8)

The meridional bending stiffness is the same in beam and shell, except for the
dependence on Poisson's ratio for the shell. The load term differs only by the Poisson
contraction due to the axial resultant in the shell. The significant feature is the
correspondence of the elastic foundation stiffness and the term in the shell equation
which can be identified as the circumferential ring stiffness. For the beam on a
foundation, much of the load is carried by the foundation; for the shell, much of the load
is carried by the important circumferential stress. The resulting behavior can be quite
unlike that of a beam supported only at a few points. By straight-forward elimination, the
matrix equation can be reduced to the scalar equation with the normal displacement as the
dependent variable, which is the customary form for the beam equation:

d 2 E I d 2 h + k h = pH
d s2 d s2 (3.1.9)

or, for the cylindrical shell,

d 2 E t c 2 d 2 h + E t h = pH – ν V / r
d s2 d s2 r2 (3.1.10)

The solution of these equations can be obtained by standard techniques. First, the
equations are linear and nonhomogeneous, so the principle of superposition can be used,
by which the total solution can be divided into a particular solution and a
complementary solution. The particular solution is any solution of the entire equation (
i.e., with the right-hand-side load term). The complementary solution is a solution of the
homogeneous equation ( i.e., without the load term), with a sufficient number of
independent, arbitrary constants with which the boundary conditions can be satisfied.
Closed-form solutions for some special cases are treated in the next sections.

§3.2 Particular Solution

When the properties and the right-hand-side terms are constant, the exact
particular solution is just:

49
 2
r
h = p H– ν V / r
Et (3.2.1)

This includes the two important special cases of a pure axial load and a pressure load.

§3.2.1 Axial Force

For only an axial force P acting on the shell, the results for all the quantities are:

V = Ns= P / 2 π r
Nθ = 0
ν P
h =–
2πEt
P
v = s – s0
2π rEt (3.2.2-5)

and the remaining quantities are all zero:

χ , Ms , Mθ (3.2.6)

This is simple axial stress in the shell. The only difference between this and the
elementary solution for a rod in tension or compression is the shell equations yield the
radial displacement due to the Poisson's ratio (3.2.4).

§3.2.2 Pressure

The next important case is uniform pressure loading, described by:

p H = p = constant ps = 0 (3.2.7)

There are situations in which no axial force is present. However, the more typical case is
indicated in Fig. 3.1, in which the ends of the vessel are closed. For the present, consider
the ends to consist of rigid plates. The internal pressure will act on the end plates with a
net force of

F = π r 2p (3.2.8)

and the solution (3.2.1) gives the results:

50
 V =N s =p r/2
Nθ=pr
2
pr ν
h = 1–
Et 2
pr 1
v = –ν s –s0
Et 2 (3.2.9-12)

The rotation and bending moments (3.2.6) are all zero in this case as well. For both cases
of axial force and pressure, the quantities of interest are constant, except for the axial
displacement which increases linearly with the axial coordinate s. The value s0 is an
arbitrary constant which determines the rigid body axial displacement.

Axis
Rigid end plate

Deformed cylinder wall - without

consideration of end constraint

p
H

Figure 3.1– Cylindrical pressure vessel with rigid end plates. The
simple particular solution gives a uniform expansion of the vessel.

Note that the axial force is carried by purely axial stress, just as for a beam.
Unlike a flat beam or plate with discrete supports, the pressure load is also carried by
tangential stress, which is possible due to the curvature. For the beam on the elastic
foundation (3.1.9), the uniform transverse load is carried by the foundation; the bending
is zero. The curvature thus may be thought of as providing the effect of a continuous
foundation.

The solutions (3.2.2-5) and ( 3.2.9-12) are exact solutions of the equation and
therefore satisfy the definition of particular solutions. However, the only arbitrary
constant is the rigid body displacement, so it is not possible to satisfy end conditions
withe these solutions alone. For example, the rigid plate shown in Fig. 3.1 requires that
the rotation and radial displacement be zero at both ends. It is clear that the solutions
51
(3.2.2-5) and (3.2.9-12) do not satisfy these conditions. The solution must be
"completed" with the complementary solution.

§3.2.3 Comparison with exact solutions

A comparison of the approximate shell solution with an exact three-dimensional

solution provides a verification of the procedure. Rigorous error bounds tend to be either
difficult to obtain and/or too conservative to provide quantitative guidance. The uniform
axial stress in §3.2.1 is an exact solution, but rather trivial. More interesting is the
solution for pressure loading of a thick cylinder, which is a well-known problem of plane
strain. The thin shell solution from §3.2.2 used in the stress calculation (2.9.2) gives the
stresses in the axial and longitudinal directions of

σ s = r pH/ 2 t σ θ = r pH/ t
(3.2.13)

We have not developed a systematic "thick shell" theory, such as that due to Reissner and
Naghdi and in the recent work of Doxee (1987). Improvement can, however be gained
by using (3.2.10) in the expressions (2.8.15-16) for the strain measures, then using
(2.6.12-13) for the strain, and then (2.8.8-9) for the stress. The load is calculated from
(2.4.15). The result for the circumferential stress due to internal pressure is

t
r pζ r pζ 1–
ID ID 2r
σθ ΙD = σ θ ΟD =
t t t
1+
2r (3.2.14)

while the result for external pressure is

t
r pζ 1+ r pζ
OD 2r OD
σθ ΙD = σθ ΟD =
t t t
1–
2r (3.2.15)

The plane strain elasticity solution for pressure acting on a thick cylinder is

2
b
1– 2
r +ζ
σζ =p
2
b
1– 2
a (3.2.16)

52
 2
b
1+ 2
r +ζ
σθ =p
2
b
1– 2
a (3.2.17)

in which p is the outward pressure acting on the edge r + ζ = a, and the edge r + ζ = b
is free. A comparison of the shell and exact elasticity solutions is shown in Figs. 3.2 and
3.3. The circumferential stress is normalized to the thin shell result (3.2.13). The
behavior is what would be expected. The error in the thin shell solution is about 10
percent when r / t is 10. It is somewhat surprising that the "thick shell" correction does
provide such a substantial increase in accuracy, and is seriously in error only for r / t <
2.

1.4
On ID exact
1.2

1.0
Stress /(pr/t)

Thick shell
0.8

0.6

0.4
On OD exact

0.2
0 2 4 6 8 10 12
r/t

Figure 3.2 - Comparison of exact elasticity solution and the thick

shell correction for the long cylindrical shell with internal pressure.
The circumferential stress on the inside diameter (ID) and on the
outside diameter (OD) is shown divided by the standard thin shell
stress factor.

53
 4

Thick shell
3
Stress / (pr/t)

On ID exact

2 On OD exact

Thick shell
0
0 2 4 6 8 10 12
r/t

Figure 3.3 - Comparison of exact elasticity solution and the thick

shell correction for the long cylindrical shell with external
pressure. The circumferential stress on the inside diameter (ID)
and on the outside diameter (OD) is shown divided by the standard
thin shell stress factor.

54
§3.3 Complementary Solutions

With zero right-hand sides and constant coefficients, the solutions of either the
complex equation of second order (3.1.3), or the real equation of fourth order (3.1.9) or
(3.1.10) is obtained by the standard technique of assuming a solution in exponential form.
Equivalently, the matrix equation can be treated. The homogeneous equation is

dY
– + A•Y = 0
ds (3.3.1)

For the cylinder with constant properties the coefficient matrix (3.1.4) is constant. A
solution may be assumed in the form of a scalar exponential multiplied by a constant
vector:

µ s
Y =ψe (3.3.2)

Substitution into (3.3.1) gives the relation which must be satisfied:

A–µ I •ψ=0 (3.3.3)

where I is the identity matrix. Thus µ must be an eigenvalue of the coefficient matrix
and ψ must be the corresponding eigenvector. In order for a nontrivial (i.e., nonzero)
solution to exist, the determinant of the complete matrix in (3.3.3) must be zero:

A–µ I =0 (3.3.4)

For an n th order system, this yields a polynomial of n th order for µ. If all n roots are
distinct, then each root provides a linearly independent solution of the form (3.3.2). The
general solution is

µ1s µns
Y =C1ψ µ1 e + ... + C n ψ µ n e (3.3.5)

For the cylindrical shell, the system (3.3.1) is of the fourth order, giving a
polynomial of fourth order:

2 4 2
Etc µ + Et/r = 0 (3.3.6)

It is easy to see that this is the polynomial which is also obtained from the scalar equation
(3.1.10). The roots are complex-valued, as indicated in Fig. 3.4. The root in the first
quadrant will be considered as the first root and written without a subscript:

55
 1 i π /4 1+i
µ1 = µ = e =
1 1
cr 2 2c r 2
(3.3.7)

and the corresponding eigenvector is

2 2
−µ rΕ tc
3 2
−µ rΕ tc
ψ( µ ) =
–µ

1
(3.3.8)

The other three roots, as shown in Fig. 3.4, are the negatives and complex conjugates of
the first root:

µ2 = µ µ3 = – µ µ4 = – µ
(3.3.9)

Imaginary

µ µ

Real

µ µ

Figure 3.4 - Locus of roots µ in the complex plane.

Thus the solution can be written as

µ s µ s
Y =C1ψ µ e +C2ψ µ e
–µ s –µ s
+C3ψ –µ e +C4ψ –µ e
(3.3.10)
56
However, Y must be real, so the complex constants must be related:

C2=C1 C4=C3 (3.3.11)

and the solution can be written as

µ s –µ s
Y = 2 Re C 1 ψ µ e +C3ψ –µ e
(3.3.12)

in which Re denotes the real part. Since the two constants have arbitrary real and
imaginary parts, four real, independent constants are present in the solution. Thus
(3.3.12) provides the complete set of complementary solutions required to satisfy the end
conditions on the cylinder.

The solution (3.3.12) consists of exponentially increasing and exponentially

decreasing terms. For a semi-infinite shell one or the other coefficient must be set equal
to zero for the solution to be bounded. For a shell of finite length L, with 0 ≤ s ≤ L , it is
convenient to adjust the constants by introducing the new constants A and B:

2 C1 = A e µ L 2 C3 = B (3.3.13)

The equation is then

Y = Re A ψ µ eµ (s – L ) + B ψ – µ e– µ s (3.3.14)

In the solution multiplied by B , the exponential term is unity at the edge s = 0, and is
exponentially decreasing with the distance from that edge. Likewise, in the solution
multiplied by A , the exponential term is unity at the edge s = L, and is exponentially
decreasing with the distance from that edge. The envelope of the exponential term is

–µ s s
e = exp –
1
2cr 2
(3.3.15)

Hence in the distance from either edge given by

1
s =δ =π 2cr 2
(3.3.16)
1
= 2.5 r t 2
(For ν = 0.3) (3.3.17)

the solution decreases to about four percent of its value at the edge. We call δ the decay
distance. If the length of the cylinder is greater than δ, the solution is substantially

57
simplified, since the solution from one edge is negligible at the other edge. This means
that the problem of satisfying the boundary conditions uncouples into two equivalent
semi-infinite problems. The conditions at the edge s = 0 are satisfied by the constant B
without regard for the constant A, and, likewise, the conditions at the edge s = L are
satisfied by the constant A without regard for B. If δ < L, the shell is referred to as long,
otherwise short.

It is important to realize that the case of the shell being long is the more common
occurrence. From (3.3.17) the ratio of decay distance to radius can be computed:

1.12 for r / t = 5
δ 0.35 50
=
r 0.11 500
0.035 5000 (3.3.18)

Thus a shell with length greater than the radius will be long; for a very thin shell, a length
rather small in comparison with the radius is sufficient for the shell to be long.

An awareness of this decay distance is very important in numerical calculations.

To capture correctly the behavior of the solution in the edge region a finite element or
finite difference mesh must be small in comparison with the decay distance δ. Typically,
the mesh should be:

1
FEM mesh spacing < 4 ( r t )1/2 (3.3.19)

A prevalent mistake in the use of finite element codes is that the mesh for a thin shell
does not satisfy this condition, with resulting errors in the results which can be large. So
far as we are aware, none of the commercially available, and widely used, finite element
codes provide any warning to the user about this condition. For the static problems under
present consideration, a fine mesh satisfying (3.3.19) is necessary only in the edge region
of width δ. Away from the edge region in the shell interior, a coarse mesh can be used
when the surface loads are reasonably smooth. For problems of dynamics and shell
stability, however, significant activity may occur everywhere in the shell which requires
the mesh (3.3.19). This requires for the thin shell a large number of elements, in
comparison with seemingly similar beam or flat plate problems.

§3.3.1 Edge Stiffness Matrix

The complementary solution (3.3.14) provides the self-equilibrating solutions

necessary to satisfy the edge boundary conditions. It is often convenient first to compute
the relation between the force and displacement quantities on the edges. For linear
problems the relation can be placed in a matrix form. In Fig. 3.5 a cylindrical shell is
shown.

58
 t
Ms h
H

L s

H, h

Ms ,χ

Figure 3.5 - Self-equilibrating edge loads on a cylindrical shell.

At the edge in the positive s - direction, the radial force is in
the same direction as the radial displacement and the moment
is in the same direction as the rotation. At the "back" edge,
the force quantities change direction.

§[Link] Long Shell

When the shell is long L ≥ δ , the effects at each edge are uncoupled. In the
vicinity of the lower edge s = L, the solution from the upper edge is negligible, and the
solution (3.3.14) reduces to

2 2
–µ rΕ tc
rM s 3 2
rH –µ r Ε tc µ (s – L )
Y= = Re A e
χ –µ
h
1
(3.3.20)

in which the arrays have been written out. At the edge, the exponential term is unity.
The relation between the force and displacement quantities is provided by the edge
stiffness matrix:

59
 rMs χ
= K L•
rH L h L (3.3.21)

in which the subscript L denotes values at the edge s = L. From the last two rows of
(3.3.20) and the value of µ from (3.3.7), the edge displacement and rotation can be
obtained in terms of the constant:

h L = Re A
– Re A + Im A
χL =
1
2cr 2 (3.3.22-23)

Solving for the constant gives

1
A = h+i h+χ 2cr 2
L
(3.3.24)

The first two rows of (3.3.20) give the moment and radial force:

rM sL = E t c Im A
Re A + Im A
rΗL =E tc
1
2cr 2
(3.3.25-26)

Substituting in the values of A from (3.3.22) gives the coefficients of the stiffness matrix:

1
2cr 2 1
rMs χ
=E tc •
rH h
L 2 L
1
1
2cr 2
(3.3.27)

In the vicinity of the top edge of the shell s = 0 in Fig 3.5, the term in (3.3.14)
multiplied by the constant B is significant. The algebra is the same with the result for the
upper edge stiffness given by

60
 1
– 2cr 2 1
rMs χ
=E tc •
rH h
0 2 0
1 –
1
2cr 2
(3.3.28)

The diagonal terms are negative because the force and displacement quantities at the
upper edge are in the opposite direction.

Frequently it is convenient to use the inverse flexibility matrices, at the lower

edge s = L :

2
–1
χ 1 2cr
1
2 rMs
= •
h L E tc 1 rH L
–1 2cr 2

(3.3.29)

and at the upper edge s = 0:

2
– –1
χ 1 2cr
1
2 rMs
= •
h E tc 1 rH 0
0
–1 – 2cr 2

(3.3.30)

The details for any prescribed loading of the edges can be worked out from the
solution (3.3.14). In the case of the long shell, the solution significant at the edge s = L
is (3.3.20). The boundary conditions for a radial load and zero moment loading at the
edge are

At s = L , Η =ΗL Μs=0 (3.3.31)

The edge flexibility coefficients (3.3.29) give the radial displacement and rotation:

rHL rHL 1
χL=– hL = 2cr 2
E tc E tc (3.3.32)

The constant A can be evaluated from (3.3.24):

61
 Re A = h L Im A = 0 (3.3.33)

so the distribution in the shell interior obtained from (3.3.20) is

–ξ
h = hL e cosξ
Nθ –ξ
σ θ D= =σ0e cosξ
t
1
2
6M s 3 –ξ
σs B= =– σ0e sinξ
2 2
t
1–ν
(3.3.34-36)

in which the reference stress is the circumferential stress on the edge:

1
E hL H L 2 r 2
σ0= =
r t c (3.3.37)

and a dimensionless edge zone coordinate, which increases with the distance from the
edge, is introduced:

L –s
ξ =
1
2cr 2
(3.3.38)

Note that the reference stress (3.3.37) is equal to the average transverse shear
stress applied to the edge multiplied by a factor of the square root of radius to thickness.
In the example of the cantilevered beam with a shear load (2.8.1), the tangential stress is
of the order of magnitude of the length to thickness larger. So as the beam becomes short
or the shell becomes thick, the assumption that the transverse shear stress is negligible in
comparison with the tangential stress becomes less valid.

The meridional bending stress (3.3.36) is zero at the edge and has a maximum
amplitude at the point ξ = π / 4:

1
σs B 3 2 –π/4 π
= e sin = 1.82 ∗ 0.322 = 0.568
σ0 2 4
max 1–ν (3.3.39)

Thus the maximum bending stress is smaller than the circumferential direct stress.

62
 The results (3.3.34-39) are for the semi-infinite shell. Solutions for the shell of
finite length can be obtained from (3.3.14) which are almost as simple as (3.3.20) if the
load is divided into a parts which are symmetric and nonsymmetric about the center point
of the meridian. Then the exponential functions in (3.3.14) are replaced with suitable
hyperbolic functions. Fig. 3.6 shows the results for symmetric loading of the ends (both
radial loads outward or both radial loads inward). It is clear that the solution for the
semi-infinite shell is quite good when the length is greater than the decay distance,
defined by (3.3.16).

Radial displacement
Shell
Ring
Factors

Semi-infinite
Shell bending stress shell
0
0 1 2
L/decay distance

Figure 3.6 - Effect of shell length for cylinder with symmetric radial
loads on the ends. When the length is greater than the decay distance,
the semi-infinite solution is valid; when the length is smaller than the
decay distance, the elementary ring solution is valid. The factors
consist of the displacement and stress normalized to the values for the
semi-infinite shell.

More details of the transition from semi-infinite shell to ring are shown in Figs.
3.7-9, in which the cylinder radius, thickness, and magnitude of the symmetric end loads
are held the same while the length is varied. For the length equal to 10 decay distances,
Figs. 7a,b show the very localized end deformation and stress. The length equal to the
decay distance in Figs. 8a,b is a transition. For a slight decrease in length in Fig. 9a,b,
the displacement calculated from the ring approximation is not too much in error, but the
significant bending persists. The displacement amplitude in these and the following
figures is greatly exaggerated to show the behavior. In reality, substantial nonlinear
effects are significant when the displacement is larger than the shell thickness.

63
Figure 3.7a - Meridian for cylindrical shell with symmetric radial
end loads (L = 250, r = 100, t = 1, HL = 1000). The self-
equilibrating end loads have a very local effect for this case of a
long shell L / δ = 10.

Figure 3.7b - Stress resultants for cylindrical shell with symmetric

radial end loads (L = 250, r = 100, t = 1, HL = 1000). Significant
variation of the stress occurs within the decay distance.

Figure 3.8a - Meridian for cylindrical shell with symmetric radial

end loads (L = 25, r = 100, t = 1, HL = 1000). This is the case of
length equal to decay distance L / δ = 1. The center of the
meridian experiences only a slight effect of the end loading, as far
as displacement is concerned.

Figure 3.8b - Stress resultants for cylindrical shell with symmetric

radial end loads (L = 25, r = 100, t = 1, HL = 1000). Even though
the radial displacement and circumferential stress are small at the
center, the meridional bending resultant is at a maximum for this
case of length equal to decay distance.

64
Figure 3.9a - Meridian for cylindrical shell with symmetric radial
end loads (L = 15, r = 100, t = 1, HL = 1000). This is the case of
length less than decay distance L / δ = 0.6. A substantial ring
rigid body displacement of the meridian occurs.

Figure 3.9b - Stress resultants for cylindrical shell with symmetric

radial end loads (L = 15, r = 100, t = 1, HL = 1000). Although
significant bending persists, the circumferential stress is closer to
the ring value of uniform stress for this case of length slightly less
than decay distance.

65
 The case of loading on the shell ends which is nonsymmetric with respect to the
center point of the meridian produces similar results. Fig. 3.10 shows the variation in the
rotation at the edge and in the maximum meridional bending stress. As in the case of
symmetric loading, the length equal to the decay distance marks the transition between
semi-infinite shell and short ring behavior.

3
Rotation
Factors

Shell
2
Ring
Shell bending
1 stress

0
0 1 2
L/decay distance

Figure 3.10 - Effect of shell length for nonsymmetric radial loads

at each end of the cylinder ( one load outward, one inward ).
Similar to the symmetric load case, the response is that of a ring or
semi-infinite shell depending on whether the length is less than or
greater than the decay distance.

§3.3.3 Center Radial Force

If a line load in the radial direction acts at any point of the meridian of a long
cylinder which is more than one decay distance from either end, then there is little effect
from the ends on the solution in the neighborhood of the load. The shell can be regarded
as two semi-infinite shells joined together. Because of the symmetry, only one of the
semi-infinite shells need be treated. For the upper shell s ≤ 0, the conditions at the edge
are

At s = L, χL = 0 HL = q / 2 (3.3.40)

where q is the intensity of the radial line load. The symmetry requires that the rotation
under the load be zero, and each shell carries half of the total applied load. This
conditions (3.3.40) are mixed, so either the edge stiffness or the inverse flexibility matrix

66
must be partially inverted. For the present system this is an easy matter. Since the
rotation is zero, the stiffness matrix (3.3.27) gives the relation between edge load and
displacement:

rM sL = k 12 h L = E t c h L
2 hL
rΗL = k 22 h L = E t c
1
2cr 2 (3.3.41-42)

so the radial displacement and moment under the load are

1
rHL r 2cr 2
hL = = q
k 22 4E t c
1
k 2cr 2
M sL = 12 H L = q
k 22 4 (3.3.43-44)

The distribution for a long shell is shown in Figs. 3.11a,b.

Figure 3.11a - Meridian for cylindrical shell with a radial line

load at the center (L = 250, r = 100, t = 1, q = 1000, L / δ = 10).
The deformation is localized to a decay distance on either side of
the load.

Figure 3.11b - Meridional stress (SsOD) and circumferential

stress (SthOD) on the outer surface (OD) for a long cylindrical
shell with a radial line load at the center (L = 250, r = 100, t = 1, q
= 1000, L / δ = 10).

67
§3.4 Total Solution for Pressure Vessel with Rigid Ends

The particular solution for surface loading and the complementary solutions for
edge loading of the cylindrical shell are discussed in the preceding sections. In this
section, these are put together for a total solution of a problem of practical interest.

In Fig. 3.1 is shown a cylindrical shell with thick end plates. Because the flat plate
carries the pressure by bending, end plates of the same material as the shell must be
relatively thick in order to be regarded as "rigid". Later the realistic flexibility of the end
plate and other more efficient shapes for the end closure will be analyzed. For the present
assumption of rigid ends, the boundary conditions for the cylinder are

At s = 0, χ 0= 0 h0= 0
At s = L , χ L =0 hL = 0
(3.4.1-2)

The total solution is written as

Y = YP + YC (3.4.3)

where the subscript P indicates the particular solution of section §3.2 and the subscript C
indicated the complementary solution of section §3.3. The total solution must satisfy the
conditions (3.4.1-2):

χ 0=χ 0 C +χ 0 P= 0
h0= h0 C + h0 P = 0
χL=χL C +χL P= 0
hL = hL C + hL P = 0
(3.4.4-5)

When the shell is long, the effect of the edges are considered independently. At the lower
edge, the stiffness matrix (3.3.25) gives immediately the edge values of the resultants of
the complementary solution:

1
2cr 2 1
rMs –χ
=E tc •
rH –h
L C 2 L P
1
1
2cr 2
(3.4.6)

Since the rotation of the particular solution due to pressure is zero and the radial
displacement is given by (3.2.11) the edge bending stress is

68
 2 ν
r M s = E t c – hP = – c r 1– p
2
1
6M s 3 2 ν
σs D= =– 1– σ 0 = – 1.55 σ 0
2 2 2
t 1–ν
(3.4.7-8)

in which the numerical value is for Poisson's ratio of 0.3, and the reference stress is the
circumferential direct stress from (3.2.10):

pr
σ0=
t (3.4.9)

The meridional stress at the inner and outer surfaces at the edge is then

σs OD = σs D+ σ s B = − 1.05 p 0
σs ID = σs D– σ s B = 2.05 p 0
(3.4.10-11)

The behavior in this problem is typical of shells. The great expanse of the shell
carries the pressure load very efficiently, but a price in some bending stress must be paid
in the edge zone. In this example, the maximum stress is at the edge on the inner surface
(3.4.11), which is twice the magnitude of the circumferential stress of the particular
solution. Failure of a shell is often initiated by this peak stress in the edge zone.

For the distribution near the edge since the edge rotation is zero, the constant A
from (3.3.22) is just:

A = – hL P (1 + i ) (3.4.12)

which used in (3.3.20) added to the particular solution (3.2.2-5) gives the following
results for the direct and bending stresses:

69
 σ0
σs D=
2
ν –ξ
σθ D= σ0 1– 1– e cosξ + sinξ
2
1
2
3 ν –ξ
σs Β= – σ 0 1– e cosξ – sinξ
2 2
1–ν
σθ Β= ν σs B
(3.4.13-16)

In Figs. 3.12a,b is shown the results for the deformation of the meridian and the stress.

Figure 3.12a - Meridian in a cylindrical pressure vessel with thick

end plates ( L = 250, r = 100, t = 1, δ / L = 10, p = 10 ). All
bending deformation occurs in the edge regions of width equal to
the decay distance δ.

Figure 3.12b - Stress on the inner and outer surfaces in a

cylindrical pressure vessel with thick end plates ( L = 250, r =
100, t = 1, δ / L = 10, p = 10 ).

§3.5 Symmetric Solution for Short Shell

In the preceding calculations, it has been assumed that the shell is sufficiently
long in comparison with the decay distance, so that the edge bending from one edge is
negligible at the other. Numerical results for the short shell have been discussed in Figs.
3.6, 3.8, 3.9, and 3.10. In this section, the formulas for the edge stiffness of a shell of
any length will be obtained. The problem is simplified by considering symmetry. For
this it is convenient to measure arc length from the center point of the meridian. If a

70
solution is desired for which the radial displacement is an even function of the arc length
s, then the constants in the solution (3.3.12) are chosen to be equal:

C3 = C1 = A
µL
2 Cosh
2 (3.5.1)

and the solution is:

– µ 2 r E t c 2 Cosh µ s
– µ 3 r E t c 2 Sinh µ s
Y = Re A
µL – µ Sinh µ s
Cosh
2
Cosh µ s (3.5.2)

At the edge s = L / 2, the stiffness matrix is:

r Ms = E t c 1 Re z χ
•
rH Im z Re z z 2 h L
(3.5.3)

in which

µL
z = µ Tanh
2 (3.5.4)

the inverse flexibility matrix is:

χ z2 – Re z
= 1 • r Ms
h L E t c Im z – Re z 1 rH L (3.5.5)

For the solution for which the radial displacement is an odd function of the arc length, the
hyperbolic functions Cosh and Sinh are interchanged in the equations (3.5.1-5).

71