USIMINAS: \ 6 MAR 1971 90 / SEPTEM BER 1963

Centro de IntormaC oes Tecnicas

Pub\. para uso e guarda de:

~~=:...:..=-=--~~~'--'
U . . . . ' ')
t ~._.:._-_

I:HBUOTECb

WELDING RESEARCH COUNCIL

BULLETIN

A Critical Study of the Solutions for the Asymmetric Bending of

Spherical Shells
Solutions for the Stresses at Nozzles in Pressure Vessels
Stress Concentration Factors for the Stresses at Nozzle Intersections
in Pressure Vessels

F. A. Leckie and R. K. Penny

These Bulletins contain final reports from projects sponsored

by the Welding Research Council, important papers presented before
engineering societies and other reports of current interest
which are not published in WELDING RESEARCH.
Bulletin No. 90 .... $1.50 per copy
 ____ Bulletin Series _ _ _ _ _ _ _ _ _ _ __
1 Steel Compositions and Specifications, by Charles M. Parker, March 1949. $0.50.
2 The Nature of the Arc, by J. D. Cobine, April 1949. $0.50.
3 Burn-Off Characteristics of Steel Welding Electrodes, by D. C. Martin, P. J. Rieppel and C. B. Voldrich, May 1949. $0.50.
4 Testing Pressure Vessels, by F. G. Tatnall; Effect of Welding on Pressure Vessel Steels, by A. F. Scotch brook, L. Eriv, R. D. Stout
and B. G. Johnston, February 1950. $0.50.
5 How Plastic Deformation Influences Design and Forming of Metal Parts, by John R. Low, Jr.; Brittle Fracture in Mild Steel, by
J. S. Hoggart, May 1950. $1.00.
6 Stress Corrosion Cracking of Stainless Steels, by Helmar Nathorst; Part I. Practical Experiences; Part II. An Investigation
of the Suitability of the U-bend Specimen, October 1950. $1.00.
7 Instrumentation for the Evaluation of the Stability of the Welding Arc, by L. P. Winsor, L. McDonald Schetky and Robert A. Wyant,
November 1950. $1.00.
8 Tests of Columns Under Combined Thrust and Moment, by Lynn S. Beedle, Joseph A. Ready and Bruce G. Johnston, December
1950. $1.00.
9 Copper in Stainless Steels, by Helmut Thielsch, August 1951. $1.00.
10 Thermal Fatigue and Thermal Shock, by Helmut Thielsch, April 1952. $1.00.
11 Investigations of Effect of Fabrication Operations Upon Pressure Vessel Steels, by S. S. Tor and R. D. Stout, May 1952. $1.00.
12 An Analytical Study of Aluminum Welding, by C. O. Smith, E. R. Funk and H. Udin, June 1952. $1.00.
13 A New High-Yield Strength Alloy Steel for Welded Structures, by L. C. Bibber, J. M. Hodge, R. C. Altman and W. D. Doty, July
1952. $1.00.
14 Welding Processes and Procedures Employed in Joining Stainless Steels, by Helmut Thielsch, September 1952. $2.00.
15 Slag-Metal Interaction in Arc Welding, by Nils Christensen and John Chipman, January 1953. $1.00.
16 Third Technical Progress Report of the Ship Structure Committee; and Research Under the Ship Structure Committee, by
Captain E. A. Wright, Finn Jonassen and H. G. Acker, November 1953. $2.00.
17 A Critical Survey of Brittle Failure in Carbon Plate Steel Structures Other Than Ships, by M. E. Shank, January 1954. $2.00.
18 Testing of Stainless-Steel Weldments, by Helmut Thielsch, October 1954. $1.00.
19 Review of Welded Ship Failures, by Harold G. Acker, November 1954. $1.00.
20 High-Strength, Low-Alloy Steels; Analytical Chemistry Fosters Progress in Steelmaking; Steel Quality, by Charles M. Parker,
January 1955. $1.00.
Listing continued on inside back coyer

Published January, February, April, May, July, September, October and November at 20th & Northampton Sts., Easton, Pa. by the
WELDING RESEARCH COUNCIL of the Engineering Foundation, 345 East 47th St., New York 17, N. Y. Application to mail at
second·class postage rates is pending at Easton, Pa. © Copyright 1963 by the Welding Research Council.
ACritical Study of the Solutions for the
Asymmetric Bending of Spherical Shells

by F. A. Leckie and R. K. Penny

SYNOPSIS. The objects of the present investigation are to

find:
1. Solutions of the spherical shell equations which are
valid for all values of the colatitude and for any degree of
asymmetry.
2. Simplified solutions suitable for hand calculations.
3. Solutions for concentrated loads applied to the shell,
suitable for use in the influence line method.
The ranges in which the various solutions are valid are
assessed by comparing edge influence coefficients and, as a
result, recommendations, which are useful for design
purposes, have been made.
NI

Fig. l-Coordinate system, displacements and stress resultants

Nomenclature
a = spherical shell radius
h = shell thickness Several authors 2, 3 have used the Langer method
E = Young's modulus
K = extensional rigidity to find approximate asymptotic solutions of the
MJ,M 2 ,
M 12 , M21 = stress couples differential equations for shells subjected to axi-
NJ,N2 , symmetric loading. The great merit of such solu-
N 12, N' l = stress resultants tions is that they are valid throughout the shell and
QJ, Q. = resultant ("effective") shear forces
Ul, U2 = displacements tangential to middle surface are not restricted to a certain range of the colatitude
VJ, V 2 = transverse stress resultants h.4 More recently Galletly and Radoks have
W = displacement normal to middle surface
~J, ~. = coordinates on middle surface shown that the accuracy of edge influence co-
v = Poisson's ratio
Ih = rotation efficients obtained from the Langer solutions com-
pare very closely with those obtained by integrating
K

( ).
= [3(1 - v') jI/. (ita)'h ; characteristic of shell the differential equations numerically.
= partial derivative w.r.t., ~1 Asymptotic solutions for the asymmetric loading
( )' = partial derivative w.r.t., ~2
of the sphere have already been given by Havers.!
I ntrod uction These solutions can best be found by means of the
Dingle methodS, 6 and it then becomes apparent
The differential equations describing the bending that the comparison equation used is not satis-
action of a spherical shell loaded at its edge were factory in the region where h is small. Aas
first derived by Havers.l He also demonstrated Jacobsen7 proposed a simplification of the Havers'
that the four solutions for the deflections obtained solutions which greatly reduced the burden of
from the membrane equations satisfy exactly the calculation and eliminated the use of special tables.!
eighth order bending equations.! The remaining The Reissner8 solutions on the other hand, are
four solutions are of the damped oscillatory type suitable for small values of ~l only and it appears
and are governed by the differential equations, that up to the present no investigation has been
'" R". made into the ranges in which the various solutions
R + R cot ~1 + sm
-.--
h
=F 2ZK2R + R
2
= 0 (1) are valid.
The aims of the present paper are threefold.
F. A. LECKIE is Fellow of Pembroke College and Lecturer at the Uni- The first is to produce solutions of the edge bending
versity of Cambridge, Cambridge, England, and R. K. PENNY was formerly equations which are valid for any degree of asym-
Research Engineer, Central Electricity Research Laboratories, Leatherhead,
England, and is now Senior Research Associate, Advanced Materials metry and all h. The second is to establish the
Laboratory, Pratt and Whitney Aircraft Co., North Haven, Conn. ranges in which simple solutions, convenient for
This paper was originally presented at the World Conference on Shell
Structures, October 1962, and will appear in the Proceedings of that Con- design use, are valid. The third is to produce
ference. The authors gratefully acknowledge the permission of the National
Academy of Sciences to reprint the paper in this Bulletin. singular solutions for concentrated radial and

Spherical Shells 1
tangential forces as well as for a concentrated 1
moment. With the availability of these singular a( v' .
(<PI ker nS - <P2 kei nS) (7)
sm ~l
solutions it is possible to apply the influence line
technique demonstrated by Kenedi. 9 In these expressions <PI and <P2 are the real and
imaginary parts of <p.; also ber .s, etc., are real.
In finding the expressions for the deflections and
The Langer Asymptotic Solution lU stress resultants (Appendix I) both F = F n cos
Separation of variables in (1) is achieved by the n~2 = (Rn + Rn) cos n~2 and G = G" cos nh =
substitution R = Rn cos n~2 to give: i(Rn - Rn) cos n~2 are required. The expressions
for Fn and G n are then:
Rn + Rn coql + (1 - sm
~)Rn
6 2
'f' Fn = AlaI + A 2a2 + Aaaa + A(a(
Gn = A l a2 - A2al + Aaa4 - A(aa (8)
where
The Langer asymptotic solutions of (2) have
already been given 11 for n = 0 and n = 1, and this = (AI - A 2)i, A2 = Al + A 2,
Al
method appears to be the most promising way of Aa = (13 1 - B2)i, A( = 13 1 + B2
extending the solutions to higher values of n. In order to compute the values of ai it was
Making the substitutionS t = sin ~t!2, and setting necessary to make use of a digital computer.
R n = y Iv'cos 2 6/2 sin 6/2, the following normal- For small values of n the imaginary parts of (4)
ized differential equation in y is obtained: and (5) can be neglected, and then:
d 2y [8,,2 { . (1 - n2)(2 - t2) <Pn = v'~l and s = v'2""h
dt 2 + 1 - t2 - J + 8,,2(1 _ t2) + The expressions for a I are then considerably
simplified to give
~} + !(!4 - n )Jy = 0 2
(3)
Tt.
~smh a2 -_~~
2,,2, t2
al = -.- . bel nS, -.- . ber lIS,
Finding the asymptotic solution of this equation smh
and then transforming back to R" and ~I gives the
aa=~.h ·keins,a(= -'~.ker"s (9)
solution, sm h "sm ~l
These solutions are referred to as the Simplified
Asymptotic Solutions.
where When n = 0 or 1 and when s > 10 then the ex-
pressions (9) can be shown to take the form
eKtl eKE.
al = v' . sin "~h a2 = v' . cos "~h
smh sm ~l
_ e -KEI e -,-Ktl
a3 = v' . sin "~h a4 = _/ . cos "~l (10)
sm h vsm ~l
which, of course, are the well-known Hetenyi12
and
solutions.
+ tan ~l)r/' d~l
2
s = v'2"ft'{1 - i(n ::- 1) (2 2
8 Constant Edge Angle Solutions
(5) The value of the functions (7) can be obtained
only if a digital computer is available, and conse-
Separating In(i';'s) , Y n(i';'s) and CI and C2 into
quently there is much to be said for investigating
their real and imaginary components the solution
solutions, (of the Aas Jacobsen type), which involve
can be written:
only hand computations.
Rn = Ala2 + Al ial + BI a( + BI iaa Substituting Rn = ylv'sin ~l in (2) gives
Rn = A 2a2 - if2 ial + B2 a4 - B2 ia a (6)
where y + { I -. 4n2
'f'
}
2i,,2 Y = 0 (11)
4 sm 2
~l '
1
.
v'smh (<PI bei n S + <P2 ber lI S) ; The essence of Aas Jacobsen's idea is that the
edge bending solutions die out very rapidly and
1 consequently sin 2h can be replaced by sin2o~1,
where O~l is the value of the colatitude at the edge
of the shell. However, sin -2 ~l changes rapidly
1 when h is small and some form of correction would
as = .
v'sm ~l
(<PI kei S
lI + <P2 ker nS) ;
then be expected.

2 Spherical Shells
 Let y be the solution of the differential equation
(~ + oci» cos (~~1 + ocp)}
{1
2
y + ~ 4n _ 2iK2} 5' =
4 sm2 oh
0 cot ~1) cos
- -2- (~h + ocp) -
The solution of this equation is
y = A 1e(a + i/3Hl + B 1e-(a + if3)~l (~ + oci» sin (~~1 + ocp)}
in which
a + p,cp. - cot h).
-2- sm (~h + ocp) -
a=
(~ + oci» cos (~~1 + ocp)}
~= K ~_ ( 4n2. - 1 ) + J1 + ( 4n2. - 1 )2 e-(a~l+.</»{ ( t ~)
(~h + ocp) -
8K2 sm2 oh ., 8/(2 s1n 2 O~1 0- 4 = V sin h - a + P,ci> - cO
2
1 COS
Now let y = fy and substitute this expression and
its derivatives into (11) to give the following differ- (~ + oci» sin (~h + ocp)} (15)
ential equation in f: These expressions are particularly convenient for
l + 2iYly - f (1 - 4
4n2)(_1_
sin2 O~1
__
sin2 h
1_) = 0 calculating the edge values because at h = O~l both
cp and ci> are zero, and furth~rmore, the factors in
front of the curly brackets can, for convenience, be
(12)
set equal to unity.
The first term of this equation is neglected on the
grounds that f is expected to be a slowly varying
function of h and further because y/5' = O(K). The Influence Coefficients and the Die-Away Effect
solution for f is then The edge influence coefficients obtained for a
In f = 4n2. ~
8 sm oh
1Xy/y JO~1r {si~2 O~l
sm2 ~1
1} 1
_ dtl
spherical cap using the at, a2 solutions given by
(7), (9) and (15) were compared by forming the
flexibility matrix:
and hence for y we obtain:
y =
where
lLe(a~1 + 1'</» +i (f3~1 + 8</» + iile-(a~l + 1'</» - i (f3~1 + 8</»
[~~]
-oMI
K
p, = The edge influence coefficients were determined
for a shell for which K was equal to 12, this repre-
senting a common practical value. The values
chosen for the harmonics were n = 0, 5, 10 and 20
while the edge angle varied in the range 0 0 < oh ~
90 0 • The calculated results for the coefficients
- 1}dt1 = sin2 oh(cot oh - fn, f12 and f21 are shown in Figs. 2 (a), (b) and (c),
the full lines indicating the results obtained from
cot h) + (oh - ~1) (14) the solutions of (7) which are regarded as the yard-
Repeating the process for the complex conjugate ji stick for measuring the accuracy of the solutions of
(9) and (12). Where the results coincide only the
the expressio~ for the a, of (8) are:
full line is shown.
1 . From the graphs it can be seen that the Constant
al = _/. I: e(a~l+p.</» sin (~~1 + (lCp),
vsm <;1 Edge Angle Solutions give satisfactory results for
1 . all oh provided n;::: 5. Even when n = 0 the
a2 = . e(a~l+p.</» cos (~h + ocp) results are satisfactory for the limits 10 0 < oh ~ 90 0 •
Vsm h The Simplified Asymptotic So]utions are valid for
-1 . all O~1 provided n ~ 5. For higher values of the
• e-(aHp.</» sin (~~l + ocp), harmonic these solutions are valid provided that
Vsm ~1
O~l is small enough, but the range of th decreases as
1 .
a4 = cos . e-(a~l+p.</» (~~1 + ocp) n increases.
Vsm ~1 The Constant Edge Angle Solutions reveal very
The derivatives of these functions are clearly the character of the edge bending solutions.
They are of the usual damped oscillatory nature, the
value of a in (12) giving the rate of decay and ~ the
rate of oscillation. It can be seen from (12) that as

Spherical Shells 3
 30 400
~--~------------ n =0 ~w K
oMI
,ASYMPTOTIC
SOLUTION
n =5 n =0
n : 5
20 n = 10
n : 10

20
10

100 n ,; 20

10 20 30 40 50 60 70 80 90
O~l ( DEGREES) ~ m ~ ~ 50 ~ ro 00 ~

O~l ( DEGREES)

Figure 2(a)
Figure 2(b)

sin oh a
n increases and oh decreases the value of a increases n<--X- (16)
3 h
while f3 decreases so that the solutions become more
heavily damped and less oscillatory. The effect Of course, this formula can only be used as a rough
of the correction terms in (9) is to cause the al and guide but, nevertheless, should guard against un-
a2 solutions to be more heavily damped and less necessary computation by taking too many Fourier
oscillatory while the effect on the as and a4 solutions components.
is exactly the opposite.
In order to check the ability of the Constant The Singularities for Concentrated Forces and
Moments
Edge Angle Solution to predict the stresses away
from the shell edge a comparison was made between (a) CONCENTRATED RADIAL LOAD. The mem-
the results obtained when this solution and the brane stress resultants arising from the application
asymptotic solution were applied to a spherical cap of a concentrated radial force P are (Fig. 4 (a)):
for which O~I = 25 0 , K = 12 and n = 0, 5, 10 and 20. P
The agreement was to the same degree of accuracy N, = -N2 = - - - - -
. 271'"a sin2 ~I
as that obtained for the edge influence coefficients.
It was stated earlier that the effect of increasing while the horizontal displacement 0 and the rotation
81 are
n and decreasing O~I is to increase the decay rate of
the solutions. If the decay rate of the solutions is (1 + II)P
0= and 81 = 0
so high that the shell is unaffected by edge bending 271'"Eh sin ~I
outside a distance of say twice or three times the
The displacements and stress resultants arising
shell thickness from the edge, then their value is
from the bending solutions are found by substituting
very questionable because the usual assumptions
in Appendix I, the expressions (cf. Appendix II).
made in shell theory do not cater for such be-
haviour.
For large values of nl K sin oh the value of a
F = ~~. h s- 2K2
271'"K sm h
~ kei s)
(ker
obtained from (12) becomes a = nisin oh. Con-
G = 2-K J .h ((kei s + 2~ ker s)
P
sidedng only the exponential variation, because the (17)
amount of oscillation is so small, gives: 71'" 1s1n ~I K2
(b) CONCENTRATED MOMENT. When a concen-
y = yoe-a.p and y= yoae-a.p
trated moment M is applied to the shell (Fig. 4 (b))
where if; = oh - ~I. the deflections arising from the membrane solution
are:
If If; (Fig. 3), which is given by the intersection
?f the tangent to the curve at O~I with the if; axis, UI -
_ M (In 1- cos ~I 2cos h) cos
----- - --- ~2
IS regarded as a measure of the penetration width, 471'"(1 - II)Ka 1 + cos h sin2 ~I
then:
_ 1 sin oh
U2 = _ M (_2_ + 2+
471'"(1 - II)Ka sin2 h
if;=-=~-
a n
cos h In 11 +- cos
cos
. sm
h
~I •
~2
Assuming a depth of penetration of at least 3h (say)
then If;a > 3h. Hence, using the expression for If;, w = 0

4 Spherical Shell~
 10,000
a K oe
--'
oM,
8,000
n= 0
n=5

o ~14-_--:-~_~~I
30 40
O~l (DEGREES)
50 60 70 80 90
~ = ANGLE OF PENETRATION
Figure 3
Figure 2(c)

and the corresponding stress resultants are

H (_2_ +2 +
47r(1 - p)K sin 2 ~1

1 - cos ~1) .
cos ~1 In 1
+ cos ~1 sm ~2
H ( I-cosh
M(l - p2) w =
47r(1 - p)K sin ~1 In 1 + cos h
• 3 t cos ~2,
4 K 4 tra sIn ~1
2 cos
. h) cos ~2
- M - M(l - p2) cos h . smh
M 12 - 21 - - 4 4 • 3 t sIn ~2
K tra sIn ~1
and the corresponding stress resultants are
The displacements and stress resultants arising
from the bending solutions are found, by substitut-
ing in Appendix I, the expressions (cf. Appendix II)
_ MK ~~ , H(l - cos h) .
F - _ /2 K - . - X ker s . sm ~2
v 7rasm h 7ra 8In 3 h
G= - MK ~~
-- X k·'
eIS (18)
V27rKa sin ~1
(c) CONCENTRATED TANGENTIAL FORCE. When
a concentrated tangential force H is applied to the - M _ -H(l - p2) cos h .
M 12 - 21 - 4 4 • 3 t sIn ~2
shell (Fig. 4 (c)) the deflections arising from the 7rK sIn ~1

membrane solution are

The displacements and stress results arising from

Ul
H
= 47r(1 _ p)K
(1n 11 -+ cosh
cos h 2cos h)
- sin2 ~1 cos b
the bending solutions are found, by substituting in
Appendix I, the expressions (cf. Appendix II)

p
M
H

(a) ( b) (c)
Fig. 4-Concentrated forces applied to the shell

Spherical Shells 5
 F = H(1 + v) ~T k" Technique,9 together with the singular solutions
~-X eiS
2v2K1I"Ka sin h provided in Section 4, is recommended.

G=
H(1 + v) ~-~-J
~-X
k '
ers Acknowledgment
2V2nKa sin t1 The contribution to the paper of one of the
In these expressions F and G die out very rapidly authors (R. K. P.), is published by permission ofthe
so that V~dsin ~1 may be taken to be unity. Director, Central Electricity Research Laboratories.
References
Recommendations
1. Havers, A., "Asymptotische Biegetheorie der Unbelasteten Kugel-
From the results of this investigation the follow- schale," lngr-Arch., 6, 282-312 (1935).
2. Clark, R. A., "On the Theory of Thin Elastic Toroidal Shells," Jnl.
ing recommendations, which should prove useful in Math. Phys., 29,146-178 (1950).
3. Naghdi, P. M., and de Silva, C. W., "Deformation of Elastic Ellip-
design, can be made: soidal Shells of Revolution," Proc. 2nd U. S. Congr. Appl. Meeh., 333-343
1. For 0 ~ n ~ 5 with oh < 20 0 (say), the Simpli- (1954).
4. Leckie, F. A., "Asymptotic Solutions for the Spherical Shell Subjected
fied Asymptotic Solutions can be used. Tables of to Axially Symmetric Loading," Symp. on Nuclear Reactor Containment
Buildings and Pressure Vessels, Butterwortbs, London, 1960, 286-297.
values of beins, etc., for 0 ~ n ~ 5 are readily 5. Galletly, G. D., and Radok, J. R. M., "On the Accuracy of Some Shell
available. 13 Solutions," Trans. ASME, &r. E: Jnl. Appl. Meeh., 81, 577-583 (1959).
6. Dingle, R. B., "The Method of Comparison Equations in the Solution
2. For all other cases the Constant Edge Angle of Second Order Differential Equations," Appl. Sci. Res., B5, 345-367 (1956).
7. Aas Jacobsen, A., "Beitrag zur Theorie der Kugelschale auf Einzel-
Solutions can be used. stiitzen," lng-Arch., 8, 275-300 (1937).
8. Reiasner, E., "Stresses and Small Displacements of Shallow Spherical
3. As a guide to the limit to the degree of asym- Shells," Jnl. Math. Phys., 25, 80-85 (1946).
metry which should be considered, the following rule 9. Kenedi, R. M., "The Influence Line Method of Shell Analysis,"
Symp. Nuclear Reactor Containment Buildings and Pressure Vessels,
may be used: Butterwortbs, London 1960, 164-173.
10. Langer, R. E., "On the Asymptotic Solutions of Ordinary Differential

sin O~l
n<3 X
(ah-) Equations with Reference to the Stokes' Phenomenon about a Singular
Point," Trans. Am. Math. Soc., 37, 397--416 (1935).
11. Leckie, F. A., "Localised Loads Applied to Spherical Shells," Jnl.
Meeh. Eng. &i., 3, 111-118 (1961).
12. Hetenyi, M., "Spherical Shells Subjected to Axial Symmetric Bend-
4. For situations which do not lend themselves ing," Bull. Int. Assoc. Bridge Eng., 5,173-184 (1938).
13. McLachlan, N. W., "Bessel Functions for Engineers," 2nd ed., Oxford,
to a Fourier analysis, use of the Influence Line 1955.

Appendix I

The expressions for the deflections and stress a a 1·

resultants are given in terms of F and G when F = - N'2 = KN21 = -.- (F - Fcot ~I)'
K sm ~I
R + R, G = i(R - R) and Rand R are the complex
conjugate solutions of the differential equations, F + (1 2- v) (0 cot ~J + G" + G)
K2 sin2 ~1
R"
R + R cot h + (1 2iK2)R + SIn
-.- 2 ~I o (1 - v) • G"
vF
2K
2 (G cot h + ~.
sm ~1
+ vG)
. . R"
-;-2 ~1
R + R cot h + (1 + 2iK2)R + SIn o
The expressions are (Fig. 1):

V)U1 = F, (1 - V)U2 =
F'
2: 2 (F - F cot ~1) ,}
(1
sin ~I a . a F'
2 - VI = F, K V 2 = -. L
(1 v )w = -2K2G + F K SIn <;1

. .
(1 v 2)a(JI = -2K2G - vF a a {. (1 - v)·
. 2 (G - G cot k1)
. ,,}
j(Ql = K F 2 K2 SIn h
a F"
KN1, = -F cot h where
sin2 h
a . F" F = (AlaI + A2a2 + A 3a3 + A4(4) COS nb
K
N2 = F cot 6 + sm2
-.- ~1
+F - 2K2G
G = (A 1 a2 - A2a1 + A a4 - A4(3) COS nh
g

6 Spherical Shells
 Appendix II

Solution for a Concentrated Radial Force and for moments about the axis given by h = oh,
In order to determine the singularity which arises ~2 = 7r/2:
when a concentrated radial force is applied to the
shell, the combined action of the membrane solu- .£271" (a VI cos O~J sin oh cos ~2 + aNI sin2 oh cos ~2 -
tions given previously and the edge bending solutions MI cos ~2 + MI2 cos oh sin h)d~2 = 0
must be considered. The edge bending solutions are,
It follows that only the membrane solutions can
F = Aaaa + A 4a4 provide the required external resultants and conse-
G = Aaa4 - A4aa quently the contributions from the membrane solu-
tions are those given previously.
where The edge bending solutions are
aa = _/~~I kei sand a4 = ~~ ker s FI = Aaaa + A 4a4 and GI = Aaa4 - A4 a a
"sm ~I 1sm h
where
Only the aa and a4 solutions need be considered

~
since it is they which can become singular at the -~-J- • ~T
- . - keII sand a4 = - . - kerl s
origin. At the origin aa and a4 take the form sm kJ sm h
which in the vicinity of the origin take the form

l ~~
- 2K \h
+ K
2
h In h}
Combining both the membrane and edge bending
solutions it is found that the singular expressions
for the horizontal deflection b and rotation OJ in the
vicinity of the origin are:
;K {- t + K2 h In h}
N ow if continuity of the shell is to be maintained
o= (1 + v)P
----'------'---- at the pole, the following conditions must be
27rEh~1 satisfied:
2K2Aa VA4 (i) Horizontal continuity - U2 = UI
-~I+ -
~I
(ii) Vertical continuity w cos h - UI sin ~I = 0
However since the shell is continuous at the pole In the vicinity of the pole the membrane solu-
o= 0, and from symmetry OI = O. Applying these tions for a concentrated moment become
two conditions gives M (
UI = 47r(1 _ v)Ka 2 In h - PI ;
2)_
v (1 - v 2)P P
A= - - A4 and A4 = = --
2K2 27rEh 27rK
and hence the expressions for F and G of (17) follow.
U2 = 47r(1 W v)Ka C~I + 2 In h)
w=O
Combining these results with those obtained
from the edge bending solutions in the continuity
Solution for Concentrated Moment and conditions gives
Tangential Force
Solutions involving the first harmonic are the only Aa = 2:~a (1 + 2~2) ,A4 = 2:~a
ones which can give rise to an external moment and
tangential shear force. Furthermore it is possible or neglecting 1/2K2 against unity,
to show that the edge bending solutions for n = 1 KM
are self equilibrating since, by themselves, they Aa = A4 = 27rKa
satisfy the equilibrium condition for horizontal
forces: An identical procedure for the tangential force
gives
f:7I" (VI sin oh cos ~2 - NI cos oh cos ~2 + (1 + v)H
Aa = -A4 =
NI2 sin h)d~2 = 0 4K7rKa

Spherical Shells 7
 Solutions for the Stresses at Nozzles in Pressure Vessels
by R. K. Penny and F. A. Leckie

SYNOPSIS. The methods of calculation which are presented Past theoretical efforts 1-3 have usually been
apply to the penetration of a spherical pressure vessel by a
radial, circular cylindrical nozzle. The nozzle may protrude limited in their range of validity by the use of
into the vessel and allowance is made for the case when local shallow shell theory. Most of this work has been
reinforcement in the form of a pad is present; no restriction devoted to axially symmetrical loadings in the
is placed on the size of the nozzle in relation to that of the
sphere. Solutions for a variety of loadings are given; these pressure vessel or nozzle. Work by Bijlaard 4
include internal pressure and nozzle thrust, moment and (and a similar version of this by Bailey and Hicks 4)
shear. For large diameter nozzles (in relation to the vessel on stresses arising from nozzle moments is the only
diameter), it is shown that theories for axisymmetric load-
ings may also be used for asymmetric loadings effort available on asymmetric loadings and it is
limited by its use of shallow shell theory.
Difficulties in assessing the regions in which shal-
Nomenclature
low shell theories are applicable can be overcome
D ET / (1 - p2); extensional rigidity
by using the theory developed by Havers. 6 This is
Dc ET3/12(1 - p2); flexural rigidity of cylindrical valid for all values of the colatitude cf> and if the
shell differential equations are integrated using the
E Young's modulus
F flexibility matrix Langer technique,7 asymptotic solutions, which
H edge force per unit circumference retain this advantage, are achieved.
k [3(1 - p2) p/4(r/t)'/2; cylindrical shell factor
M moment stress resultant per unit circumference. It has been demonstrated,S, 9 how powerful, yet
M bending moment applied to nozzle simple, these solutions are in axially symmetric
m [3(1 - p2) P/2; function of Poisson's ratio
n 0, 1; degree of asymmetry bending problems and also for singular solutions in
p axial loading on nozzle per unit circumference spherical shells subjected to concentrated forces
p (1 + p)/m; function of Poisson's ratio
Q shear stress resultant per unit circumference and couples. Further development of this work
q internal pressure appears in an accompanying paper 10 where solutions
R, r radii of spherical and cylindrical shells, respec-
tively for any degree of asymmetry are derived. I t is on
shear force applied to nozzle this latter work that the spherical shell theory used
T, T', thicknesses of spherical shell, pad and cylindri-
t cal shell, respectively in the present paper is based.
u,v,w tangential and radial displacements, respec- I t is shown that, by elimiriating body movements,
tively
x force vector the asymmetric (n = 1) solution procedure for the
y displacement vector determination of stresses at nozzle intersections is
a spherical shell functions.
il edge displacement of shell identical with that for the axisymmetric case (n =
C/>, 8 shell coordinates 0). As a result it has been possible to write a
p Poisson's ratio
K [3(1 - p2)]1/4(R/T)I/2; spherical shell factor general solution which can be used for both axisym-
X shell rotation metric and asymmetric loadings.
Suffices
c,p, s pertaining to cylinder, pad and sphere, respec- Description of Structure and Loadings
tively
o denoting shell edge We shall deal with structures composed of thin
8, c/> pertaining to circumferential and meridional spherical and cylindrical shell elements and in par-
planes, respectively.
ticular with the practical case of a radially disposed
I ntrod uction nozzle in a spherical pressure vessel. I t is assumed
that the nozzle may be treated as semi-infinite
Openings in pressure vessels are usually inevitable whether it extends into or from the pressure vessel;
and constitute a major source of weakness, the correction factors to allow for the length of the
nature of which is by no means well understood. cylinders could easily be incorporated if this were
Elastic solutions, which are capable of predicting thought to be necessary. The pressure vessel
stresses based on shell theories, help to some extent may be reinforced by a uniformly thickened region
in assessing the strength of pressure vessels and it away from the nozzle. The geometrical param-
is the purpose of this paper to add to the present eters defining the structure under consideration
state of knowledge in this field. are shown in Fig. 1. There are four basic struc-

8 Pressure Vessel Nozzles

 sure vessel design, will be considered in detail.
These are:
(a) Constant internal pressure, q, applied by a
weightless medium.
(b) An axial loading, P (per unit circumference),
applied to the exterior part of the nozzle.
(c) A lateral shear force, S, applied to the ex-
terior part of the nozzle.
(d) A bending moment, M, applied to the ex-
terior part of the nozzle.
(a) and (b) belong to the axisymmetric (n = 0)
class, while (c) and (d) are categorized by n = 1.
In an cases the effects of the weight of the shells are
ignored.

Basis of Shell Theories Used

Theories for two basic shell shapes, the sphere
and the cylinder are all that are required for this
problem. For these shapes it has been shown, 6, 11
that complete solutions to the governing differen-
tial equations can be obtained by adding the sepa-
rate solutions due to the shell acting as a membrane
in carrying the applied loading and those due to
inextensional and edge bending.
Membrane theories for these shells are well
known 11 and most attention must be given to the
edge bending problems. Hoff12 has given edge
bending solutions in explicit form for the cylindrical
Fig. 1-Schematic diagram of nozzle and vessel geometry
shell. These solutions are based upon the simpli-
fied governing equations derived by Donnen 13 and,
as Hoff points out, there are indications that
stresses and deflections obtained from the edge
ture types which can be dealt with: bending solutions differ very little from the more
refined Fliigge 11 equations. Simple as Hoff's re-
(a) Flush nozzle with uniform thickness pressure sults are, they can be further simplified in the case
vessel. of first harmonic loading (n = 1), to give the same
(b) Protruding nozzle with uniform thickness results as for the axisymmetric case (n = 0), if it
pressure vessel. is realized that the effects of bending in the circum-
(c) Flush nozzle with pad-reinforced pressure ferential planes are negligible compared with those
vessel. in the axial planes. Adequate justification of this
(d) Protruding nozzle with pad-reinforced pres- may be found in Ref. 5. Since the axisymmetric
sure vessel. bending behavior of cylinders is well known 14 little
The flush nozzle can project inward or outward more need be said about the cylindrical shell elements.
from the pressure vessel. Influence coefficients for these are derived in Ap-
As is usual with shell assembly problems the pendix I.
shells are assumed to join at the intersections of the The situation for spherical shells is not so clear
median lines of their thicknesses. This, together since the most commonly used equations governing
with the basic assumption, which is usual in all the asymmetric bending behavior apply only in
shell theories, that initially plane sections remain limited regions of the shell (Reissner 15 and Havers6).
plane after bending, means that the ensuing Solutions to the Reissner equations are restricted
analysis will predict the gross behavior of the struc- to shallow shells and, whereas this restriction does
ture and will not give the microscopic effects due not apply to the Havers theory, the asymptotic
to fillet profiles. This latter effect could be found solutions found by Havers are not valid when the
by a separate analysis if considered sufficiently im- colatitude cP is small. In an accompanying paper10
portant. Loadings which will be treated are those asymptotic solutions involving Kelvin functions of
of the axisymmetric (n = 0), and asymmetric (n = integer order n and real argument have been de-
1), types. Thermal stresses within the same cate- veloped. In comparison with a more rigorous
gories could similarly be derived. Four particular solution given in the same paper it was found that
loading cases, of importance in most forms of pres- the Kelvin function solution was very accurate for

Pressure Vessel Nozzles 9

 and
sOVI cos CPo + sVI sin CPo
= - sWI (1)
(positive directions being as shown in Fig. 2(a».
The total displacements consist of those above
together with the body movements. The body
movements, which for n = 1 amount to the shell
moving parallel to the plane of its edges by an
amount shl and rotating about this place through
an angle Ndb are shown in Fig. 2(b). The total
displacements are thus:
sU = sUI - shl ; sOv = sOvl + sVI
SVI
sX = sXl - b (2)
In precisely the same way we derive for the cylinder
the following:
K cOvl + eVI
~ eVI
I eXI - b (3)

/ The requirements of compatibility dictate that:

= = = = (4)
'2 b. / eU su; eOH sOH; eOv

The first two conditions give, by eliminating the hi

sOo; eX sX

/ terms:
/
Fig. 2-Spherical shell displacements
Similarly by eliminating the VI terms the remaining
conditions yield:
sOvl eOvI
all ranges of cP providing 0 ~ n ~ 5. This work, sXl +T = eXI +b (5)
therefore, forms the basis for the present investiga-
tion where we are only interested in the cases n = 0 Thus the body movements have been eliminated
and n = 1 but wish to maintain freedom in our and the compatibility requirements can be entirely
choice of cpo Solutions of these types have been satisfied by two equations which involve only the
shown8, 9 to reduce to the shallow shell solutions on edge bending and the membrane solutions. This
the one hand and to the Hetenyi form, for large procedure is now identical with the axisymmetric
values of cp, on the other. (n = 0), case.
We can show in a similar manner that our choice
Derivation of the Compatibility and Equilibrium of edge forces can be the same as that for the axisym-
Equations
metric case. Consider a cylindrical shell which is
In discussing the complete solutions to shell loaded by a moment M and shear force S as in Fig.
problems we must include body movements as well 3. These loadings are reacted by edge stress re-
as edge bending and membrane displacements. Let sultants HI, M" 8I, VI as shown. On account of
us consider the n = 1 case and derive first the the equilibrium of the shell parallel to the force S
compatibility equations which must be satisfied in there must be a linear relations between 8 1, HI and
joining two shells together. Consider first the S. Thus only one of the stress resultants 8 1 or HI
spherical shell having displacements (as in Fig. 2), can be regarded as unknown. Similarly, to retain
due to edge bending and membrane solutions: equilibrium of moments, only one of the stress re-
sultants VI or Ml can be regarded as unknown.
For consistency with the axisymmetric case we
choose HI and MI as these unknowns. This
In cylindrical coordinates the displacements of reasoning applies for any shell.
the edge of the shell are completely defined by the
following: Description of the Junction Analysis
The procedure to be followed for the junction
analysis is similar to that of Ref. 2 for symmetric load-
where ings. We consider here only the axisymmetric
sO HI = sWI sin CPo + sVI cos CPo case since by virture of the reasoning given in the

10 Pressure Vessel Nozzles

 M,
, ill
(H-NCOS+.)I
,. .11I N,
/ II
M· I II
•
t-+-..... H, I
I
11
5, I
Fig. 4-Forces acting at shell junctions
Fig. 3-Lateralloadings on cylindrical shell

preceding section the n = 1 case can be similarly Py'1,2 = E[pOh T'pXh p02, T'pX2J
treated. For generality the two joint case involv-
ing cylinder-pad-and pad sphere intersections is .y'2 = E[.02, T'.X2J (8)
considered. By "cylinder" is meant a flush or the remaining terms being self explanatory. Deriva-
protruding nozzle and by "sphere" the unthickened tion of the flexibility matrices is given in Appendix
parts of the pressure vessel; the "pad" refers to I while the membrane displacement vectors are
the thickened portion of the vessel. given in Table 1 for four typical loading cases.
In Fig. 4 the forces acting at joints CD and ® Our compatibility requirements are:
between the elements of the structure are shown. (9)
These consist of self-equlibrating discontinuity
forces (H, M) shown by closed arrows and the and hence by combining eqs (7) we obtain for the
membrane forces (N), reacting the applied loading, unknown force vector:
shown by open arrows. The imposition of the -I
discontinuity forces is necessary to maintain the i0 0
_~~__ l_~___~
junctions in a closed state; their magnitudes are
determined from this requirement. First we form
the displacement/force relations based on the fact o O~ [c~!:J
Y
• 2
- p y 1,2 m

that the total displacements are composed of those o0i sF

due to membrane effects and those caused by edge
bending forces. In general these can be written (10)
symbolically as: where the following additional definition holds:
y = Fx + ym (6) eYl m = cY m - cF[Nlo cos ¢d (11)
where y is the total displacement vector, ym the
membrane displacement vector, F is flexibility Discussion
matrix and x the force vector. In our particular Solution of eq (10) provides the means for de-
problem these relationships, for each of the cylinder, termining the final stress distribution resulting from
pad and the sphere, will be written: any loading condition defined by n = 0 or n = 1,
once the flexibility matrices and the membrane
cYI = cF eXI + cylm displacement vectors appropriate to a given loading
pYI,2 = pF pXI,2 + pYI,2 m are known. Matrix equations for stress resultants
.Y2 = .F .X2 + .Y9 m (7) in the spherical shell components are given in Ap-
pendix IV; those for the cylindrical components are
where the following definitions hold: well known. 14
~IJ
It is worth noting that since the flexibility ma-
eX I' = [(HI - NI cos ¢I), trices are defined only by the geometry of the
structure, the matrix inversion involved in eq (10)
pXI,2' = [ HI, ~\ H ~2}
2, .X2' has only to be performed once for a given structure
subjected to a variety of loadings of a given class.
cYI' = E [co), T'cxd When ¢ ;:: 5 V2/ K, simplified asymptotic solutions

Pressure V essel Nozzles 11

 Table l-Expressions for the Membrane Displacement Vectors
------Membrane displacement vector ym

]
Loading Cylinder Pad Sphere

r]
[''-'l<~'
1. Constant internal (a)
flush [ (1 - K8' Sill </>' ]
p) ;;-
'f!- 1~ ~] + [ k2 [
2~' COB </>1 ~
pressure, q. m R m
cy/'m = (2 p) ;; R sY2 =

(q~ = 1)- m
PYI,2
_
-
0
Kp2
(b) protruding (1 - p);; sin </>.
n=O

cYlplm = ~
2m
cos cf>l[lJ
0

2. Thrust loading on
cylinder, P. [p<.,! T ] sY2
m [
p'" Rr ~in1 </>. T t]
(P~=l). PYI,2
m
p Kp 2
0
!._1_!
n=O R~in</>. T
Pk • .!. ]
cY1P Im = kK.'
-
2m
cos <PI
0
[lJ + 2m T
P T'
[ -- - k
2T
3. Moment applied to (a) flush

CYli(n~ ~'1)
7rR2 '
[
- II -

o
-
TK"] +---;--
Rrtm- 2k3 COB <PI

m sm' </>1
I m T1
Lk'T-'t
I
T]

(b) protruding
n = 1

4. Shear force applied (a) flush

to cylinder, S. ~~ -PKS' ]
+ ,m =
C~ = 1). (1 cos </>1)
(1 + p) .y [ (1
(1 + cos </>.)
+ p). ~
n = 1 sin2 CP1 sin' </>. T
-P K8 2
(1 + cos </>')
(b) protruding

CYIP/m =
_(2 ~ p) ~~
+ p) ':£ '!::. !!
fJ +
(1 + p)
sin2 cf>2

[ (2
2 t Tr
k' ~_ _ -=-__----,-
2m sin </>1 (1 + cos </>1) GJ

sium on Pressure Vessel Research toward Better Design, Inst. of Mechanical

can be used and under these circumstances the Engineers, London, 1961, 5-20.
flexibility matrices involved for n = 1 are the same 4. Bijlaard, P. P., "Stresses in a Spherical Vessel from External Moments
Acting on a Pipe," Welding Research Council Bulletin No. 49, Part (2),
as those for n = O. This observation leads to con- April 1959.
5. Bailey, R., and Hicks, R., "Localised Loads Applied to a Spherical Pres-
siderable reductions in the calculation of the asym- sure Vessel through a Cylindrical Insert," Jnl. Meeh. Eng. Sci., 2 (4),302
metric loading problem. The implications of this (1960).
6. Havers, A., "Asymtotische Biegetheorie der unbelasteten Kugel-
are much deeper in that for large openings the same schale," Ingr-Arch., VI Band, 1935.
7. Langer, R. E., "On the Asymptotic Solutions of Ordinary Differential
principle will probably apply to the general shell of Equations with Reference to the Stokes Phenomenon about a Singular
Point," Trans. Am. Math. Soc., 37, 397-416 (1935).
revolution. 8. Leckie, F. A., "Asymptotic Solutions for the Spherical Shell Subjected
to Axially Symmetric Loading," Proc. Symposium on Nuclear Reactor Con-
tainment Buildings and Pressure Vessels, Glasgow, May 1960, Butterworths,
1960, 286-297.
Acknowledgment 9. Leckie, F. A., "Localised Loads Applied to Spherical Shells," Jnl.
Mech. Eng. Sci., 3 (2) 111-118 (1961).
The contribution of one of the authors (R. K. P.) 10. Leckie, F. A., and Penny, R. K., "A Critical Study of the Solutions for
the Asymmetric Bending of Spherical Shells," Welding Research Council
to this paper is published by permission of the Bulletin No. 90, September 1963.
11. Fliigge, W., "Stresses in Shells," Springer-Verlag, Berlin, 1960.
Director, Central Electricity Research Laborato- 12. Hoff, N. J., "Boundary-Value Problems of the Thin-Walled Circular
ries. Cylinder," Jnl. Appl. Mech., 24, 243-250 (1954).
13. Donnell, L. H., "Stability of Thin-Walled Tubes under Torsion,"
References National Advisory Committee for Aeronautics, Rep. No. 479, 1933.
14. Timoshenko, S. P., and Woinowsky-Kreiger, S., "Theory of Plates and
1. Bijlaard, P. P., "Stresses in a Spherical Vessel from Radial Loads Shells," 2nd ed., McGraw-Hill Book Co., New York, N. Y., 1959.
Acting on a Pipe," Welding Research Council Bulletin No. 49, Part (I), April 15. Reissuer, E., "Stresses and Small Displacements of Shallow Spherical
1959. Shells," Jnl. Math. Phys., 25, 80-84 (1946).
2. Penny, R. K., "Stress Concentrations at the Junction of a Spherical 16. McLachlan, N. W., "Bessel Functions for Engineers," 2nd ed., Oxford
Pressure Vessel and Cylindrical Duct Caused by Certain Axi-Symmetric at the Clarendon Press, 1955.
Loadings." Proc. Symposium on Nuclear Reactor Containment Buildings 17. Hetenyi, M., "Beams on Elastic Foundations," University of Michigan
and Pressure Vessels, Glasgow, May 1960, Butterworth 1960, 347-368. Press, Ann Arbor, 1946.
3. Rose, R. T., and Thompson, J. M. T., "Calculated Stress Concentra- 18. Lowell,Herman H .. "Tables of the Bessel-Kelvin Functions and their
tions Factors for Nozzles in Spherical Pressure Vessels," Preprint, Sympo- Derivatives," NASA, T.R. R-32, 1959.

12 Pressure Vessel Nozzles

 Appendix I-Derivation of the Flexibility Matrices
For the Spherical Shell (Pad and Sphere) The integration constants are then given by:
Elements
In deriving the flexibility matrices we first state
d = pA-I·XI,2
the relevant equations derived by Havers6 for the where
stress resultants, couples and displacements.
These are as follows:

!i N",
D
= L:
i=1..4
Cj (+
sm
a/
cf>
n
) - a/ n ) cot cf»
(15)

R also the coefficients at,j of pA are given by:

15 Q", eff = ._L: Cia;<n> +
,-1. .4
{
a/ n ) n2
= sin cf> + sin cf>
[(1 - II)
n
-.- .
2
(1 -
2
II) {(aHI(n»)
L: Ci - .- - L: Ct (ai_l(n»)}
-.-
pal,j 2K2p X
smcf> smcf> smcf>
K2 i=l,a ;=2,4
(ajH(n) - ajH(n) cot cf» - a/
n
) cot cf> Jt="'1
R M", { (1 - II)
- - =
D R
L:
i=l,a
Ci a/ n
) +
2K
2 X (j = 1,3)

[ ai+l(n) cot cf> + (1 - sm2

~) cf>
ai+l(n)} + or

(1 - II)
L: Ci
n
{ a/ ) - 2 X
;=~4 2K

u
-n
----.- L: Cja/ n ) pa2,j + (12 -Kp 2
II) [.
aj+l
(n) cot 'I'A. +
(1 II) sm cf> ;=1..4

(j = 1,3)

or

= {a (n) _
j
(1 -
2Kp2
II) [a'_ICn) cot cf>
J
+
i~'4 Ci(a/ +
J} =",.1
n
) 2 K2a t-1(n»} (13)
(
1 - -n2- ) a· len) (j = 2,4)
From eqs (12) we obtain for the force at right sin2 cf> J- '"

angles to the axis of revolution of the sphere, H",: (16)

The a3,j, a4,j coefficients are of precisely the same
~ H", = ~(Q", eff sin cf> - Ncf> cos cf» definitions as for al,j, a2,j, respectively, except that
we put cf> = cf>2 instead of cf>1.
i.e., We now derive the edge displacements:
R
- H", = L: Ci
a/ n )
• +.sm cf>
~{(1 -
2 K2
II)
X
(h + u) = w sin cf> + v cos cf> + u
D ;=1..4 sm cf>

[ ; !taCt(ai+J (n) - aiH(n) cot cf» - ;E4 Ct(at_l(n) -

R (x + ~) w- = v - new cot cf> - v)

= W- nw cot cf> - (1 - n) v (17)

ai_len) cot cf» J- ;=t=.4 Cl a r) cot cf» (14) Substituting for u, v and w from eqs (13), eq (17)
becomes after some manipulation:
If now we substitute for the forces acting at the
edges of the pad which subtends angles cf>l and cf>2 at E(5 H + u) = L
;=1..4
d t {KP2 (Sin m
cf> - ~P) a/ n ) +
sm cf>
the sphere center, we can obtain the integration
constants. Thus we put H", = HI> M",/R = 2Kp4sincf>
PK p2 cos cf>a/ ) }
n
- m X
MdR, at cf> = cf>l' and H", = H 2, M",/R = MdR at
cf> = cf>2 in eqs. (12) and (14); at the same time we ) L: dtai+l(n) - L diai_l(n) t
alter our integration constants to d t = (D / R)c t • /;=I,a ;=2,4 )

Pressure Vessel Nozzles 13

ET'(X +~) = tion (which is as for the pad), it can be shown that
the flexibility matrix sF, for the sphere can be
written:
- (1 + 11)(1 - n) L dic'xi(n)- sF = sB sA-I (22)
i=1. .4

L
i=1. .4
dien cot cpotr) - c'xr» + and the coefficients of sA and sB are:

2Kp2 5. L di(n cot cpotHl(n) - c'xi+l(n»

1,=1,3
L di(n cot if>ott_l(n) - c'xi_l(n» t
i=2,4 l
where
(1 + /I)
m = [3(1 - /1 2) , P = ----
)1/2.
m (18)

The displacements at cp = CPl and cp = CP2 for the

pad are then: (n) + (1 - /I) X
sa2,1 = { ota 2 2
Ks
py1,2' = E[(OH + u)t, T'(X + ~)t' [c'x4(n) cot cp + (1
(OH + U)2, T'(X + ~)J = pFx1,2 (19)
_ { (n) _ (1 - /I) X
sa2,2 - ot4 2 2
Ks
pF is the flexibility matrix for the pad and is given
by: [ c'xa(n) cot cp + (1
(20)
where pA has been previously defined and pB has
the coefficients bi,i given by:

p
b .
1"
= {K 2 (sin if> _ n p ) ot.(n)
p m sin cP 1
+

pKp2 cos if> c'x/ n) -

2Kp 4 sin cP
otHl
(n)} (j = 1,3)
m </>=</>1

or

pb1,i = {K 2 (sin cP _ ~P) ot/ n ) + sb 2,1 = (~) 2 { 2Ks2(n cot cp ot~(n) - c'x4(n» -
p m sm cp
(n cot cp ot3(n) - c'xa(n» - (1 + 11) (1 - n)c'x3(n)} </>=</>'
2Kp4 sin cp }
PKp2 cos cp c'x/ n) + oti_l(n) (j = 2,4)

pb 2,1 = {- (n cot cp ot,<n) - c'x/ n» -

m </>=</>1
sb 2,2 = (~y { -2Ks2(n cot cp ota(n) - c'xa(n» -

+
/1)(1 - n)c'xr) +
(1
(n cot cp ot4(n) - c'x4(n» - (1 + 11)(1 - n)c'x4(n) }</>=</>,
(23)
2Kp2(n cot cp otj+1(n) - c'xH1(n»} </>=</>1 (j = 1,3)
It might be remarked in conclusion that the argu-
or
ments of the Kelvin function involved in the ot;<n)
pb 2,i { - (n cot cp ot/ n) - c'x/ n» - are (V2Kpcp) and (V2Kscp) for the pad and sphere,
(1 + /1)(1 - n)c'xp) - respectively. Discussion of these functions is
given in Appendix II.
2Kp2(n cot cp otj_l(n) - c'xj_l(n»} </>=</>1 (j = 2,4)
For the Cylindrical Shell (Flush and Protruding
(21)
Nozzle)
The ba,j, b4,i coefficients are similarly defined but Dealing first with the flush cylinder (this being
we must put cp = CP2 in those definitions. the basis also for the protruding cylinder results),
For the sphere element the same procedure is it is sufficient to quote the results for the edge
followed except that we consider only the constants bending displacements: 14
d a, d 4 in association with the functions ota(n), ot4(n)
since we are regarding the pad as the only discon-
tinuity in the sphere. Without giving the deriva- o= 2;~c (MI - iHI)
14 Pressure Vessel Nozzles
 sure vessel. Bending is also produced by the ap-
plied loading since the inner part of the cylinder is
unable to take part in reacting this loading. In
Fig. 5 the cylinder is shown separated into parts
inside and out of the vessel. Also shown are the
assumed proportions of discontinuity forces reacted
by these parts. The problem is to determine the
fractions a and b.

~-(I-b)H,
-- -- By equating the displacements of the edges of
the cylinders we obtain:
1 r HI kDe
a = 2" + 4k MI - 2rMI Xm
1 k MI k De 3
(I-a) M( b= -
2
+- -
2r HI
+ - 0
r3H1 m
(27)

Om and Xm are the free membrane displacements of

the loaded (outer) cylinder.
The total displacements of the composite cylinder
are thus:
Fig. 5-Forces acting on protruding nozzle

r M Xm k (28)
x = 4kD e I +2 - 2r Om
x = 2{D (2M1
e
- ~HI) (24) Rearranging these in the required form:
Mter some rearrangement these can be cast into
matrix form as foHows: eYI p -- [Eo
ET'x ] -_ Fp e Xl
+ eY p/m (29)

[EO the coefficients of eFp being:

eYI
f -
- ET'x ] _- eFfXI (25)
k3
P P
the superscript on eFf denoting "flush". The coef- c!1.I = - 2m; J1,2 = 0
ficients of the flexibility matrix eFf are then:
2k 3 2k4 R f
c 2.1 P -- O·, (30)
ef 1.1 f -- - - ', ef 1.2 f -- - •-
m m r
also,
T'T 3 T' T R
cf2./ = - 2k2 T't; cf2./ = 4k T 't'-;:
~
(26)
o
For the protruding cylinder we have to find the cyp/m = [ , - 4{T,][E m ] (31)
proportions of bending forces shared by each part _ kT ! ET'
of the cylinder which is inside and out of the pres- 2r 2 Xm

Appendix II-Definition of the a/n) Coefficients and Remarks

on Their Values for Large Openings
Solutions to the governing differential equations These solutions are valid for all values of tj> pro-
for the bending of spherical shells have been shown lo viding 0 :::; n :::; 5. Here we are only interested in
to be: n = 0 and n = 1 values. Tables of Kelvin func-
a}n) = <I> beinz; (X2(n) = <I> bernz tions of zero order are available; 16.18 these functions
(X3(n) = <I> keinz; (X4(n) = <I> kernz (32) are also proportional to the Schleicher functions
where tabulated by HetEmyi. 17 Tabulation of first order
functions is not so frequent but it is possible to
<I> = ~smtj>
. tj> and z = V2Ktj> (33) express these in terms of zero order functions, as
will be shown later.

Pressure Vessel Nozzles 15

 (a) Zero order functions (a/ O»: Derivatives of first order Kelvin functions are as
follows:
<P ber Z
bert z 1 .
<P ker Z ber'lz = - - - - _ r (berz
z ~2
+ belz)
bei t z
ber'tz = - - -
z
_r 1
+ ~2 .
(berz - belz)
where
with corresponding expressions for ker'tz and
bei'z = :/bei z) kei\ z. It then follows that:
al(I) K( a2(0) altO»~ - gal(O)
and
a2(I) -K(a2(O)
+ al (0» - ga2(0)

(35) a3(l) K(a4(0) - a3(0» - ga3(0)

Similarly,
a4(l) -K(a4(0) + a3(0» - ga4(0) (39)

a2(0) = V2K<Pber' z + fa2(0)

where

a3(0) = V2K<Pkei' z + fa3(0) g = ~ (; + cot q))

a4(0) = V2K<Pker' z + fa4(0) (36) It is of interest now to inquire of the form that the
attn) functions take for large values of the argu-
(b) First order functions (a(!):
First order Kelvin functions may be expressed ment z. Using the asymptotic expansions for
in terms of those of zero order as follows: 16 bernz and beinz 16 we find for large z:
z
. 1
herl z = V2 (ber' z- bei'z ) . -e "'2 . ( z
berl z = - belo z "'" V211"z sm V2
1 z
beiJ z = V2 (ber' z + bei' z) (37) "'2
bei1 z = bero z "'" ~1I"Z cos (~ - ~)
with similar expressions for kerlz and kei1z.
Hence, We can similarly show that:
-z
al(I) = _~- (bei' z + bei' z) = ~{(a2(0) + . e "'2 . (z 11")
~2 2K - k elo z "'" - 11" V211"z sm V2 + 8
al (0» - f( a2(0) + al (0» } -z

kei l z = kero z "'" 11" ;;:z cos (.J2 +~) (40)

;K {(a 2(0) - altO»~ - f(a2(0) - al(O»}
In these expansions terms of 0 (1) are neglected by
comparison with terms of O(K) and hence at is
2~ {(a4(0) + a3(0» - f( a4(0) + a3(0»} neglected by comparison with al. Maintaining
this consistency with the flexibility matrices (eqs
16, 21 and 23), it can be seen that these matrices
:K {(a4(0) - a3(0» - f(a4(0) - a 3(0»} will be identical for the cases n = 0 and n = l.
This would be so, of course, only when these expan-
(38) sions are valid, i.e., for z > 10.

Appendix III-The Membrane Results

Although the membrane results are well known, (a) for the spherical shell:
the relevant expressions for four loading types are
given in this Appendix, for completeness, without E(o + u) = qR2 (1 - v) sin q) +
their derivation. 2T

16 Pressure V essel Nozzles

 PR ~in cp] (1 _ v) (1 + v)M 1 reduced to a moment M and a horizontal force S
Tsm cP 1r RT sin2 cP acting at the level corresponding to cP = 0 0 •
(1 + v) S 1 (b) for the cylindrical shell:

1r T (1 + cos cp)
E(o + u)x~o =
qr2
-2t (2 - v) +
r v jJ
v- P - - -
t 1r rt
E (X + ~) = (1 + v) S _1_
b 1r RT sin 2 cP E (X + 0,) = (2 + v) §
b x~O 1r rt
R Psin CPI M
N", =q-
sin 2 cP
+ -1-
-1rR2 sin +
+M S x
2 3 cP qr
N" = - - P
2
-1rr2 + -.-
1rr r
S (1 - cos cp)
1rR sin 3 cP Ne = qr
Ne = -N", + qR N xe = N ex (42)

For both the sphere and cylinder cases the dis-

placements and forces quoted occur at 0 = 0 or
(41)
1r /2 (and vary as cos (J or sin (J elsewhere) for the
It should be noted that in the above expressions M and S loadings. For the q and P loadings they
the lateral forces acting on the cylinder have been are independent of o.

Appendix IV-Matrices for the Stress Distributions

From the definitions of the stress resultants and Table 2-Expressions for the Membrane Stress Vector zm
couples due to edge bending given by Havers 6 and
Membrane
by adding the membrane components, the matrix stress vector,
equations obtained for the relevant forces are: Loading zm

'N", 1. Constant internal pressure, q.

M", (q~ = 1)
R
= pS pA -I XI,2 + zm for the pad (or
.S .A - I X2 + zm for the sphere) 2. Cylinder thrust loading, P. -( sin q"

(43)
(p~ = 1) T sin 2 q,
o
.J
o
( sin q"
where the elements of these matrices are: ---
T sin 2 q,
o .J

(j = 1. . .4)
3. Cylinder moment M ,1'"
(
M
'lrR2
= 1) sin 3 q,
o
o
-1
sin 3 q,
L 0 .J

or 4. Cylinder shear

(! 1) =
S
,
sin q, (1 1 i cos q,) 1
-1
sin q, (1 + cos q,)
L 0 .J

Pressure Vessel Nozzles 17

 (1 - v)n2 (1 - v) cot cf> _
pS3,j a/ n ) {a._l(n)
+ 2 Kp 2sln
' 2
cf> X 2Kp2 J

(aHl(n) - aHl(n) cot cf» (j 1,3)

(
V ~)
+ sm2
cf>
a j +ICn ) } (j 1,3)
or
or
(1 - v)n2
pS3,j a/ n ) - X
2Kp2 sin 2 cf>
(aj_l(n) - aj_l(n) cot cf» (j = 2,4)
(j = 2,4)
n2
- - a Jn ) (44)
sin 2 cf> J

(j = 1,3) The a/ n ) have argument V2Kpcf>.

or For the sphere coefficients we change the prefix
p to s, use K. instead of Kp and a/ n ) of argument
pB4,j
V2 Kscf>.
Expressions for the membrane stress vectors are
(j = 2,4) given in Table 2 for four typical loading cases.

18 Pressure V essel Nozzles

Stress Concentration Factors for the Stresses
at Nozzle Intersections in Pressure Vessels
by F. A. Leckie and R. K. Penny

SYNOPSIS. Calculated stress concentration factors which

occur when a radial circular cylindrical nozzle penetrates a
spherical pressure vessel, are presented in a graphical form
convenient for design purposes.
Four loading cases are considered. These are:
(i) Internal pressure
(ii) Thrust applied to the nozzle
(iii) Moment applied to the nozzle
(iv) Shear force applied to the nozzle

Introduction
R
The purpose of the present paper is to present
the stress concentration factors, which occur at the
intersection of a cylindrical nozzle with a spherical
pressure vessel, in a form convenient for the
designer. We shall deal with structures composed
of thin spherical and cylindrical shell elements and
in particular with the radially disposed nozzle in a
spherical pressure vesseL I t is assumed that the Fig. 1-Shell geometry
nozzle may be treated as semi-infinite whether it
extends into or from the pressure vesseL The is very useful in other design considerations.
spherical portion may be reinforced by a uniformly The stress concentration factors have been calcu-
thickened region away from the nozzle. The lated in terms of the maximum stress which occurs
geometrical parameters defining the structure in the sphere. We have chosen to do this because
under consideration are shown in Fig. 1. the experimental work of Rose 2 and of Taylor,
As is usual with shell assembly problems the Lind and Schweiker 3 suggests that the maximum
shells are assumed to join at the intersections of stresses occur in the sphere and not in the cylinder. *
the median lines of their thicknesses. This to- The graphs presented in this work have been com-
gether with the assumption that plane sections re- piled with a view to convenient design use, but if it
main plane after bending means that the analysis is necessary to have a more detailed knowledge of
will predict the gross behavior of the structure the stress distribution throughout the sphere and
and will not give the microscopic effects due to the cylinders, then the analysis described fully in
fillet profiles. Reference (1) must be completed.
The stress concentration factors have been deter-
mined using classical thin shell elastic theory, Stress Concentration Factors for
employing the solutions presented by Penny and Internal Pressure
Leckie.! The advantage of these solutions is that In calculating the stress concentration factors we
they are not restricted to shallow shells and they assume that a reinforcing pad is necessary and of
can also be used for axisymmetric (pressure and
thrust) as well as asymmetric types of loading * More recent experimental work completed by Taylor on stresses due to
internal pressure and by Leven due to moment loading demonstrates that
(moment and shear). A criticism of the elastic for some tiT ratios (generally when the nozzle wall is relatively thin com-
pared to the spherical shell wall) the maximum stress occurs through the
solutions is that they provide only a limited insight base of the nozzle. Therefore although the graphs presented herein give
the maximum calculated stress in the sphere they do not necessarily give the
into what is a very complex design problem, but maximum stress in the nozzle-sphere combination. It is understood that
nevertheless it has been clearly illustrated by Rose 2 the PYRe plans to investigate this problem further in order to determine
the maximum stresses and stress distribution in both the nozzle and the
that a knowledge of the stress concentration factors sphere.

Pressure Vessel Nozzles 19

such size that the cylinder may be considered to design so that the maximum stresses due to local
penetrate a complete sphere of thickness T'. A disturbances are less than a certain multiple of the
pad may not always be necessary. When it is, membrane stress in the main pressure vessel.
the question of its size is discussed later. Hence, we define the following stress concentration
We define the following stress concentration factor
factor
SCF = Maximum fiber stress/ ~;
SCF' = Maximum fiber stress/::,

where pR/2T' is the membrane stress in a sphere where pR/2T is the membrane stress in the main
of radius R and thickness T', subjected to an inter- vessel due to pressure. At present designers ap-
nal pressure p. pear to favor an SCF = 2.5,2 but in the foregoing
The stress concentration factor SCF' is a function procedure, the designer is free to choose any value
of the dimensionless parameters: of SCF he pleases. The relation between SCF and
SCF' is simply
(i) The radius/thickness ratio of the pad R/T'
(ii) The radius/thickness ratio of the cylinder SCF' = SCF X T'/T
r/t If we assume SCF = 2.5 then the graph of SCF'
(iii) The size of the opening CPl = sin- 1 r/R against T' /T is shown by the full line in Fig. 4.
In order to represent the variation of SCF' with The graphs for SCF = 2.0 and 3.0 are also shown
these three parameters, either one three dimensional dotted in the same figure.
plot or many two dimensional plots are required. The normal design situation is that we are given
However, it is fortunately possible to represent the the dimensions R, T and r and wish to find the
SCF' for all variations of the parameters on a dimensions t and T' (if a pad is necessary).
single sheet of graph paper. We arrive at this Procedure: Assume that R, T and r are given
result by first considering the SCF' which occurs and that we are to design to a SCF = 2.5.
in an unreinforced hole in the pad. It can be First assume that a pad is not necessary so that
shown for small values of cpr, that the SCF' is a T' = T and SCF = SCF'. Then calculate the
function of p = (r/RhIR/T' only. When the value of p, and from the intersection of p and SCF'
SCF' / p curve for the unreinforced hole is plotted = 2.5 in Fig. 2 we find the required ratio tiT'.
for the range of parameters 0.01 ~ r/R ~ 0.4 and This procedure is satisfactory when the opening is
30 ~ R/T' ~ 150 it is found that the curve is in- small and the value of p is less than unity. It can
deed unique except for the highest values of p be seen from Fig. 2 that when p is greater than
when a very slight scatter is observed. unity, the intersection point of p with SCF' = 2.5
This curve is the plot designated by tiT' 0 requires a ratio of tiT > 1 which is unsatisfactory
in Fig. 2. Now in view of this success let UR choose from the design point of view. When this occurs
a new set of parameters which also form a group it is necessary to use a pad.
suitable for defining the geometry of the vessel. When a pad is necessary we then proceed in the
The choice is following manner. Guess a value of T' and hence
calculate p and T' /T. From Fig. 4 find the value
(a) p = ~~ (b)
R
T' (c)
t
T' of SCF', and the intersection of this value of SCF'
with the value of p in Fig. 2 gives the appropriate
We have already demonstrated that when param- value of tiT'. Several guesses may be necessary
eter (c) is zero, the value of SCF' is dependent before dimensions satisfactory to the designer are
only on the value of p, when p is small, and only determined.
very weakly dependent on the value of (b) when p Sometimes it may be necessary to find the SCF
is larger. Now we choose values for (b) and (c) associated with an existing design. We then calcu-
and plot the resulting SCF' / p. curve. Repeating late p and tiT', and from Fig. 2 we find SCF'.
the same process keeping the same value for (c) but Finally in Fig. 4 the intersection of SCF' with T' /T
choosing a different value for (b) it is found that the gives the appropriate value of SCF. In most cases
new SCF' / p curve coincides almost perfectly with an exact determination of SCF is unnecessary and
the first curve. In fact, the curves resulting from then it suffices to know if the intersection point
a particular choice of parameter (c) have the same lies on the "safe" side of SCF = 2.5.
character as those observed when parameter (c) is It can be seen that the combination of Figs. 2
zero. The curves shown in Fig. 2 are for the and 4 offer the designer great flexibility. A simple
flush nozzJe, while those in Fig. 3 apply to the example now follows.
protruding nozzle. Example: Given R = 400 in., T = 4 in. and
r = 60 in. Determine t and T' for an SCF of 2.5.
The Application of the SCF'/p Curves in Design
If T' = T = 4 in. then p = 1.5 and hence a pad
In designing a pressure vessel to withstand inter- is necessary. Choose T' = 6 in., then p = 1.22
nal pressure, it is proposed by Rose 2 to carry out the and SCF' 3.75. The intersection of these

20 Pressure Vesscl N ozzlcs

 12

+, =0 r----- / / /
10 ~ t
l',=0.25-
i
11f
8
~r~ t
'f,=0.5____...

t _
'f' -1.0~
itJ '---- ,..j /

(fMAX.
pR12T'
/ I~ /
V /
6
/
/; I 7
V/ V V
4
v
VV
1/ V/ /
./ V vV /~
------
/ V
,..-- ...........
",
,/
./

",~
v V
.-" i-" ~ .....
2

0.01 0.10 1.0 10.0

p:!.!K
RJT'

Fig. 2-Maximum stress in sphere for internal pressure (flush nozzles)

12
I
t
TI =0 ............
I

19V/
, r
10 I-
25
=' .. ..! -:},=0.
~/
4f=. ~
t
Ti=0.5~K
8
t
V
R
TI = 1.0~ ;" /
~
/ ~/
V
I
/':k' /
/ / V>"" "-
,

~)I
4
V // ~
V Vi-'
/ V

------ ------
,/
/ V /
~ / V~ V
...........
2
., ~ ~ /
- i'-" I-- l- ~
~-

0.01 0.10 1.0 10.0

p=kfi
Fig. 3-Maximum stress in sphere for internal pressure (protruding nozzles)

Pressure Vessel Nozzles 21

 9 S.C.F.=3.0
,7
8

7 / ./ S.C.F.= 2.5

6
./~ ____-;;ii'
S.C.F.= 2.0
/ ~
S.C.F.i 5

4 /~ L.---
----,..- ---- Fig. 4-Variation of S.C.F.' with
T'lT
3 v~--------
2 ---- ---
~ .............
/~----
~.,..

o .2 3,

values in Fig. 2 gives tiT' = 0.50 or t = 3 in. was chosen somewhat as a matter of convenience in
Now try a second choice T' = 5 in. This time making the previous calculations, based on a SCF
we find t = 3.5 in. of 2.5 as the currently suggested design value 2) • In
order to study the effect of the size of the opening
Discussion of Pad Size on pad size, graphs similar to those of Fig. 5 were
The design procedure previously described has produced for a shell with the same dimensions but
been devised on the understanding that the pad with opening size varying between the limits 0.1
is of sufficient size to justify the assumption that < CPl < 0.4. The pad size was again determined
it is semi-infinite. For large openings the stress by the criterion that the SCF should be limited to
distribution in the spherical shell adjacent to the a 4 % increase over the minimum value. The
nozzle, and therefore the size of the truly semi- graph of pad size against size of opening CPl is shown
infinite pad, is a function of (RT') 1/2, while for in Fig. 6, from which it can be seen that the ratio
small openings the attenuation is faster. The of pad size to opening decreases with increasing
practical size of pad required to achieve a given CPl·
acceptable stress level (such as the equivalent of It is difficult to give any general rule about pad
a SCF of 2.5) may vary with the size of the open- size, but of the nine cases studied it was found
ing, and the question now arises as to how large that the graph of Fig. 5 was highly representative
the pad must be to be regarded as "practically semi-
infinite."
Figure 5 illustrates the effect on the SCF of
varying the pad size (CP2 - CPl) of a pressure vessel 4.2
for which CPl = 0.2, RIT = 90, RIT' = 60 and tiT' I I I RI
= 0.5. For CP2 - CPl ~ 0.2 the value of the SCF 4.0 <1>,= 0.2 ~,= 90 T,=60 -
remains unaltered at 2.74 so that pads of this size
can be regarded as semi-infinite. However, as
3.8
f.= 0.75
t
T,=0.5
-r-
(CP2 - CPl) is reduced below 0.2 the value of SCF
rises only very slowly until CP2 - CPl = 0.08. Re-
duction of the pad size below this value causes the 3.6
SCF to rise rapidly to above 4.2. Hence the effect
of a pad size CP2 - CPl = 0.08 is to reduce the SCF u:
~ 3.4
\
from 4.2 to 2.87 while the addition to the pad size
from CP2 - CPl = 0.08 to CP2 - CPl = 0.20 has the
effect of reducing the SCF from 2.87 to 2.74. It is
(f)

3.2
\
clearly very uneconomic in this case to provide so '\ 4% INCREASE ON MINIMUM S.C.F.

i
much additional pad for such a small decrease in 3.0
SCF. A much better method is to follow the
design procedure for a value of SCF slightly lower
than the allowable. This value of SCF is then
raised to the allowable by using a smaller pad.
Referring again to Fig. 5 we see that an increase
2.8

2.6
o .04
'" ~

.08
I
I
: .12
--- --- ---
.16 .20 .24
in SCF of 4% of the minimum value of 2.74 allows ( <1>2- <1>,)
us to decrease the pad size from CP2 - CPl = 0.2 to
CP2 - CPl = 0.09 (the specific figure of 4% increase Fig. 5-Variation of S.C.F. with pad size

22 Pressure Vessel N ozzlcs

 compact manner. Instead of presenting this con-
fusing mass of lines, only graphs of the maximum
.5 stresses are shown .
This overlapping occurs because the value of
SCF p at first decreases with increasing tiT', but
~ .4 later reaches a minimum and begins to increase
>&- again. The reason for this is that for small values
I
C\I of tiT' the hoop stress (uu) is higher than the me-
>& .3
ridional stress (uq,). With increasing tiT', Uu be-
comes smaller and u q, larger until u q, begins to
.2 dominate. This changeover takes place when tiT'
.1 .2 .3 .4 is approximately 0.75.
<PI From these graphs, presented for the flush nozzle.
Fig. 6-Variation of pad size with opening in Fig. 7 and for the protruding nozzle in Fig. 8,
it is easy to obtain the maximum stress in a particu-
lar vessel.
of the effect of pad size on SCF and it would, there- By way of an example take R = 400 in., r = 60
fore, appear that, within the range of RIT and in., T' = 5 in. and r = 3.5 in. as being suitable for
r I R ratios considered, a pad size of one half or less design against internal pressure for a flush nozzle.
of the opening is sufficient to limit the stress to a Hence p = 1.34, tiT' = 0.7, RIT' = 80 so that
generally acceptable value. SCF p = 1.75 (from Fig. 7) and Umax = 1.75
vRIT'(PIT') = 15.6PIT'.
The Stress Concentration Factors due to Thrust
Loading The Stress Concentration Factors Due to Moment
and Shear
The design of the nozzle to withstand pressure
having been completed we must now turn to the The stresses which occur when a moment or shear
determination of the stresses which will arise when force is applied to the nozzle are characterized by
the nozzle of the vessel is subjected to additional the fact that they vary according to the cosine or
loads such as thrust, moment and shear. We would the sine of the longitude angle (). The analysis for
expect the pressure loading to be of prime impor- the so-called first harmonic shows great similarity
tance, but nevertheless the additional stresses arising to that for axisymmetric loading and is fully de-
from the short-lived application of these other scribed in reference (1).
loadings can be an important design consideration. In computing the resultant moment and shear
Of these additional loadings the first one we con- force acting on the nozzle we should note that the
sider is that of direct thrust applied to the nozzle. moment M is the total moment due to the applied
Since this loading causes an axisymmetric stress moments and the moment of the shearing forces
distribution the analysis used is very similar to about the point of intersection of the nozzle and the
that employed for the pressure loading. sphere. .
In order to determine the most suitable plot for The aim is again to find stress concentration
presenting the results of this type of loading we factors which are dependent only on the value of
follow the lead of the pressure loading and carry p. For the moment case the required stress con-
out a shallow shell analysis on the unreinforced centration factor is
shell (i.e., for tiT' = 0). In this way we discover
that ifSCF p is defined as SCF M = umax/~, ~~,
SCFp = umax/ ; , ~~, where P = M I 7rr 2, while for the shear loading it is
where Umax is the maximum fiber stress and P is SCFs = Umax / 7r:T'
the thrust per unit circumference of the nozzle
(i.e., total thrust is 2 7rrP) that it is a unique func- The SCFMI p graphs are shown in Figs. 9 and 10
tion of p. Making this plot for the case tiT' = 0, and the SCF sl p graphs are shown in Figs. 11 and
confirms that SCFp is dependent only on p when 12. These graphs can be used in exactly the same
p is small, while for larger values of p is it evident
way as was illustrated in the thrust loading case to
from Fig. 7 that different values of RIT' tend to find the maximum stresses due to the application
produce only a slight scatter. Also from this of moment and shear.
figure we see that similar curves are obtained when
Combined Loadings
tiT' = 0.25.
A difficulty arises in presenting the graphs of From the graphs which have been given, it is
SCF p obtained for values of tiT' = 0.5,0.75 and possible to determine the maximum stress caused
1.0 because they overlap one another in a very by a certain type of loading. For a conservative

Pressure Vessel Nozzles 23

 THRUST LOADING IS P PER. UNIT CIRCUMFERENCE

4
(J'MAX

i-fi 3
j
I I.---
-.~
V ~ ~ ....
-
V
./
-~~
'30 "-~90" 150 } +. . 0
V
/ ~ : ,= 1"0 -.....: ... ~ I-r..
2 r __ 1'30 It. =0.25
~
90.IOU

----- -
f0o- l

-
~~
I--l -
f \i=05
T' .
-= 1=0
~
I--
VV~ I
'"V ~
.. f.-'-
MAXIMUM STRESSES

I
0.01 0.1 1.0 10.0
p=~ft,
Fig. 7-Maximum stress in sphere for thrust loading (flush nozzles)

THRUST LOADING IS P PER UNIT CIRCUMFERENCE

4 L LJ

u MAX.
pI[ I
T'J T'
3
I
V .~
V ~ ...._.
t'iI
1'-.......
[:/' V 30 90
L~-, ~ ..
foo-
171
2
/ ~
~ T' -1"0
l"'- I'--90
~V
} t I 25
150 T': Q
V 7 1:-:--"" f'...30

. t::: vVI--'J..- ~~ ----

~I-' II

MAXIMGM STRESSES
I
[\t,= 0.5
..........
!'.... r-...
.~ ",,10-'
I

10.0
0.01 0.1
P = iff. 1.0

Fig. 8-Maximum stress in sphere for thrust loading (protruding nozzles)

estimate of the stresses occurring under the action of results the following general conclusions can be
of combined loading, one can simply add the maxi- reached.
mum stresses obtained from each of the individual (i) For all cases considered, the maximum of
loadings. This is always conservative because in the two stresses (considered jointly) occurred at
one loading (1'8 is the maximum stress while in another the junction.
(1'", might be the maximum stress. (ii) Generally speaking (1'8 is the maximum
However after having inspected a large number stress except when the hole is reinforced by a cylin-

24 Pressure Vessel Nozzles

 3
" MAX.
PI! .....~.--. ~ ~......
~o t _
IIo
Ty T' 90 150 T'-
WHERE
// t...:::::"._ ...:::::....- ""- I
2
t,: 0.25
-p:-
M
,.. r2

/
V
V
...... "...-

~V V' ~ ~ ~.=0.5
...... ",

I,..-
~j....
lIt.,:1.0
~
1'-"30 ~

-- ';0- 150
r-

~
~
i--""

~ ~ MAXIMUM STRESSES

0.01
- ~
~

0.1 1.0 10.0

p: kft·
Fig. 9-Maximum stress in sphere for moment loading (flush nozzles)

3
....,,-- 1 - - ..
I
~-- ""-...:::
3v
~
V
1}tr o
90 ...... 15
1I
MAX
2
V
V 1/
t _ -. '"'-
'" r.......: 1'--. 150} ~,: O. 25
q

"P..ffl , V ....-
1"- 1.0 "30 90

VT V V~v
"vj....KT': ....... r-....
WHERE
I / ./ ~V
...., 0.5

~~
:/': .,/
15=.M.
. n2 ~ MAXIMUM STRESSES

~
~

0.01 0.1 1.0 10

p: *fl'
Fig. IO-Maximum stress in sphere for moment loading (protruding nozzles)

der with tiT' ~ 0.75 when u¢ tends to be larger. complete the analysis described in reference (1)
(iii) While Uo can be much larger than u¢ for including the stresses in the nozzle.
the lighter reinforcements, in the heavier reinforce- Bibliography
ments u¢ is not significantly higher than uo. 1. Penny, R. K. and Leckie, F. A., "Solutions for the Stresses at Nozzles
Hence it can be said that while adding the in Pressure Vessels," Welding Research Council Bulletin No. 90, September
1963.
maximum stresses is conservative (insofar as the 2. Rose, R. T., "New Design Method for Pressure Vessel Nozzles," The
Engineer, 214, 5556, July 20, 1962 and Welding Res. Abroad, IX, 43-47,
stresses in the sphere are concerned) it should not Feb. 1963.
be excessively so. However if an accurate stress 3. Taylor C. E., Lind, N. C., and Schweiker, J. W., "A Three Dimen-
sional Phot~elsstic Study of Stresses Around Reinforced Outlets in Pres-
distribution is required it will then be necessary to sure Vessels," Welding Research Council Bulletin No. 51, June 1959.

Pressu.re Vessel Nozzles 25

 1\ \
,

20

\ \ \ II Il s
\ \
~,~
\-h'=0.25
\ \ \
15
1\
t
'f'=0.5
~
r=t rI.O \ t

rrMAX.
\ \ r
ShrT' \ \ \
\ \
10
\ \
I\. 1 \
\ \ \
1\ \ \
~ \ ~
5 l\. \ \
l\.
"\..
"-'\ \ \.
""- ~

o
0.01
t
T','"
- 0

0.1
'" "-- r--...: ~
~ r-= ~~
1.0 10.0

p='0
Fig. ll-Maximum stress in sphere for shear loading (flush nozzles)

1\ \
\
20 \
\ \
\ \ 1 L S

~ ~
\ \ I

\ \
15 \ \ \
\ t
r=0.5 ~f'=1.0
\ 1 \
\
10 t r
T,=0.25 1\
\ \
\ \
rrMAX. r\ \ \
,
S'" r T' \ \
\ \ \
5 ~ ~
l\. \ \
'\ \.\

o
0.01
i ,= ~),
0.1
'" , "r--f::::: ~
........
~

1.0 10.0
p=.!.!B-
RJ T'
Fig. 12-Maximum stress in sphere for shear loading (protruding nozzle)

26 Pressure Vessel Nozzles