Professional Documents
Culture Documents
K. R. W ICHMAN
A. G. HOPPER
J. L. MERSHON
MPC – A Council of the WRC, the Materials Properties Council is dedicated to providing
industry with the best technology and the best data that can be obtained on the properties of
materials to help meet today’s most advanced concepts in design and service, life assessment,
fitness-for-service, and reliability and safety.
PVRC – A Council of the WRC, the goal of the Pressure Vessel Research Council is to
encourage, promote and conduct research in the field of pressure vessels and related pressure
equipment technologies, including evaluation of materials, design, fabrication, inspection and
testing.
WRC Bulletins contain final reports from projects sponsored by the Welding Research Council, important
papers presented before engineering societies and other reports of current interest.
No warranty of any kind expressed or implied, respecting of data, analyses, graphs or any other
information provided in this publication is made by the Welding Research Council, and the use of any
such information is at the user’s sole risk.
All rights are reserved and no part of this publication may be reproduced, downloaded, disseminated, or
otherwise transferred in any form or by any means, including photocopying, without the express written
consent of WRC.
ISSN 0043-2326
Library of Congress Catalog Number: 85-647116
The October 2002 Update to the March 1979 Revision of WRC Bulletin 107 includes minor editorial changes for
improvement and readability of several equations, curves and some text. There are NO technical changes.
• The calculation forms (Tables 2, 3 and 5) are improved, particularly to show the "+" and "-" quantities more
definitively.
• The equation for stress in paragraph 3.6.3 is revised to be on one line.
• The parameter definitions on several of the curves (beginning with figure SR-1) are improved and clarified.
• Appendix 8, exponents in Equations 1, 2, 3 and 4 are enlarged for readability.
PVRC thanks Mr. James R. Farr, Honorary Emeritus Member of the Pressure Vessel Research Council, for his
assistance in preparing this update.
NOTE: WRC Bulletins 107 and 297 should be considered (and purchased) as an integral set. In addition,
PVRC Technical Committees are working on a project that is envisioned to culminate in a new publication to
add to the WRC Bulletin 107 and 297 set. The new publication will provide significant new technical information
on local shell stresses from nozzles and attachments.
Greg L. Hollinger
The Pressure Vessel Research Council
FOREWORD
to March 1979 Revision
Welding Research Council Bulletin No. 107 has been Present plans call for a review and possible extension of
one of the most widely used bulletins ever published by curves to parameters which will cover the majority of
WRC. The original bulletin was published in August 1965. openings in nuclear containment vessels and large storage
Since that time, a revised printing was issued in December tanks. Plans are to extend R/T from 300 to 600 and to
1968; a second revised printing was issued in July 1970; a extend d/D range from 0.003 to 0.10 for the new R/T
third revised printing was released in April 1972; and a range, review available test data to establish limits of
June 1977 reprint of the third revised printing was issued. applicability, and develop some guidance for pad
As sometimes happens with publications of this type, reinforcements.
some errors were detected and then corrected in sub-
sequent revised printings. Long range plans arc to review shell theory in general,
and Bijlaard's method in particular. The goal is to extend
In this March 1979 Revision of Bulletin 107, there are the R/t up to 1200 for a d/D up to 0.1. This will include
some additional revisions and clarifications. The for- large deflection theory and other nonlinear effects. In
mulations for calculation of the combined stress intensity, addition, available computer programs will be studied in
S, in Tables 2, 3, and 5 have been clarified. Changes in hope of developing one which will be an appropriate
labels in Figures I C-1, 2C-I, 3C, and 4C have been made supplement to Bijlaard's method. Finally, a review will be
and the calculated stresses for Model "R" in Table A-3 made of limit loads related to large R/T and small d/D.
and Model "C-1" in Table A-4 have been revised
accordingly. The background for the change in labels is J.R. Farr, Chairman
given in a footnote on p. 66. P VRC Design Division
FOREWORD
Several years ago, the Pressure Vessel Research tion of this problem in spheres, aimed at a par-
Committee sponsored an analytical and experi- ticular design problem, and the Pressure Vessel
mental research program aimed at p:t;oviding meth- Research Committee undertook a somewhat sim-
ods of determining the stresses in pressure vessel ilar investigation in cylinders. Results of this
nozzle connections subjected to various forms of work have recently become available emphasizing
external loading. The analytical portion of this the limitations in Bijlaard's data on cylindrical
work was accomplished by Prof. P. P. Bijlaard of shells, particularly as it applies to thin shells over
Cornell University, and was reported in Refer- the "extended range" (page 12 of Reference 10).
ences 1 to 8 inclusive. Development of the Incident to the use of Bijlaard's data for design
theoretical solutions involved a number of simplify- purposes, it has become apparent that design
ing assumptions, including the use of shallow engineers sometimes have difficulty in interpreting
shell theory for spherical vessels and flexible load- or properly applying this work. As a result of
ing surfaces for cylindrical vessels. These cir- such experience, PVRC has felt it desirable that all
cumstances limited the potential usefulncrss of the of Bijlaard's work be summarized in convenient,
results to dt! D 1 ratios of perhaps 0.33 in the case "cook-book" form to facilitate its use by design
of spherical shells and 0.25 in the case of cylindrical engineers. However, before this document could
shells. Since no data were available for the larger be issued, the above mentioned limitations became
diameter ratios, Prof. Bijlaard later supplied data, apparent, presenting an unfortunate dilemma, viz.,
at the urging of the design engineers, for the values the test data indicate that the calculated data are
of (3 = 0.375 and 0.50 (d 1 / D 1 ratios approaching partially inadequate, but the exact nature and
0.60) for cylindrical shells, as listed on page 12 of magnitude of the error is not known, nor is any
Reference 10. In so doing, Prof. Bijlaard in- better analytical treatment of the problem avail-
cluded a specific warning concerning the possible able (for cylinders).
limitations of these data, as follows: "The values Under these circumstances, it was decided that
for these large loading surfaces were computed on the best course was to proceed with issuing the
request of several companies. It should be re- "cook-book," extending Bijlaard's curves as best we
membered, however, that they actually apply to can on the basis of available test data. This de-
flexible loading surfaces and, for radial load, to the cision was based on the premise that all of the
center of the loading surface. It should be under- proposed changes would be toward the conserva-
stood that using these values for the edge of the tive (or "safe") side and that design engineers
attachment, as was recommended for small load- would continue to use Hijlaard's extended range
ing surfaces, may be unconservative." data unless some alternative were offered. The
Following completion of the theoretical work, following paper is therefore presented in the hope
experimental work was undertaken in an effort to that it will facilitate the use of Bijlaard's work by
verify the theory, the results of which were pub- design engineers. Every effort has been made to
lished in References 17 and 18. Whereas this point out any known limitations in the work and
work seemingly provided reasonable verification to explain the exact nature of the changes which
of the theory, it was limited to relatively small d 1/ - have been made to Bijlaard's original curves and
Dt ratios-0.10 in the case of spherical shells and data; however, users are warned that the resulting
0.126 in the case of cylindrical shells. Since vir- work is not necessarily adequate for all cases. It is
tually no data, either analytical or experimental, the hope of the Subcommittee that additional theo-
were available covering the larger diameter ratios, retical work can be undertaken to provide more
the Bureau of Ships sponsored a limited investiga- adequate data on various phases of this problem.
F. S. G. WILLIAMS, Chairman
PVRC Subcommittee on Reinforced
Openings and External Loadings
CONTENTS·
1. Nomenclature 1 4.2.2.3 Rect. Att. Subject to P 9
2. General Equation 2 4.2.2.4 Rect. Att. Subject to
3. Spherical Shells . 2 M. 9
3.1 Sign Convention ·2 4.2.2.5 Rect. Att. Subject to
3.2 Parameters 3 ML 9
3.2.1 Round Attachments 3 4.3 Calculation of Stresses 10
3.2.2 Rectangular Attachments 3 4.3.1 Radial Load . . . . . . . . 10
3.3 Calculation of Stresses 3 4.3.2 Circumfere~tial Moment Load-
3.3.1 Radial load ·3 ing . 11
3.3.2 Moment Loading . 4 4.3.3 LOngitudinal Moment Loading 12
3.3.3 Torsional Moment Loading 4 4.3.4 Torsional Moment Loading. 12
3.3.4 Shear Load 4 4.3.5 Shear Load. 12
3.3.5 Combined Loading (Computa- 4.3.6 Combined Loading (Computa-
tion Sheet). 5 tion Sheet) . 12
3.4 List of N ondimensional Curves . 5 4.4 Nondime1,1sional Curves . 12
3.5 Limitations of Application . 5 4.4.1 List of Nondimensional Curves 13
3.6 Abridged Calculation for Maximum 4.5 Limitations on Application. 13
Stress . 5 4.5.1 Radial Load . 13
4. Cylindrical Shells 8 4.5.2 External Moment. 13
4.1 Sign Convention 8 5. Acknowledgments . 13
4.2 Parameters 9 6. References 14
4.2.1 Shell Parameter (r) . 9 Appendix .A-Basis for "Corrections" to
4.2.2 Attachment Parameter ({3). 9 Bijlaard's Curves 54
4.2.2.1 Round Attachment 9 Appendix B-Stress Concentration fac-
4.2.2.2 Square Attachment 9 tors for Stresses due to External Loads 6&
Local Stresses in Spherical and Cylinclrical Shells
due to External Loadings
by K. R. Wichman, A. G. Hopper, and J.l. Mershon
Stresses in Shell!J
shear stress on the x face in they di- direction with respect to the
rection, psi shell, psi
shear stress on they face in the x di-
rection, psi 2. General Equation
shear stress on the 1-1 fac(:!
In the analysis of stresses in thin shells, one pro-
shear stress on the 2-2 face
ceeds by considering the relation between internal
1.3 Nomenclature Applicable to Cylindrical Shells : · membrane forces, internal bending moments and
stress concentrations in accordance with the follow-
Vc concentrated shear load in the cir- ing:
cumferential direction, lb
VL concentrated shear load in the lon- 1N 6Mt
ut =KnT
--±K --
hT2
gitudinal direction, lb
Me external overturning moment in the Stress concentration factors should be considered in
circumferential·direction with re- the following situations:
spect to the shell, in. lb
ML. external overturning moment in the (a) the vessel is constructed of a brittle ma-
longitudinal direction with re- terial.
spect to the shell, in. lb (b) a fatigue evaluation is to be undertaken.
R,. mean radius of cylindrical shell, in. The designer may find the data on stress
l length of cylindrical shell, in. concentrations contained in Appendix B to
r0 outside radius of cylindrical at- be helpful.
.tachment, in. Much of the work contained in this Bulletin is de-
c1 half length of rectangular loading in voted to a synopsis of methods for obtaining mem-
circumferential direction, in. brane ft;>rces (N1) and bending moments (M 1)
c2 half length of rectangular loading in which have been developed by Professor P. P.
longitudinal direction, in. Bijlaard in his numerous papers written on this
T wall thickness of cylindrical shell, subject. This data has been obtained for a wide
m. range of cases by use of an electronic computer and
x coordinate in longitudinal direction is presented here in the form of nondiniensional
of shell curves.
y coordinate in circumferential direc-
tion of shell 3. Spherical Shells
cylindrical coordinate in circum-
3.1 Sign Convention
ferential direction of shell
a l/Rm For the most part, stresses will be considered in
attachment parameter the vessel wall at the attachment-to-shell juncture.
cdRm . Here a biaxial state of stress exists on the inside
c2!Rm and outside surfaces, so that one is concerned with
· Rm!T; shell parameter radial and tangential (principal) stresses as indi-
multiplication factors for· N"' and cated in Fig. 1. At this location, one can predict
Nx for rectangular surfaces given whether the stresses will be tensile ( +) or compres-
in Tables 7 and 8 sive (-) by considering the deflection of the shell
coefficients given in Tables 7 and 8 resulting from the various modes of loading.
bending moments in shell wall in
the circumferential and longi- p
tudinal direction with respect to
the shell
membrane forces in shell wall in the
circumferential and longitudinal
direction with respect to .the shell
normal stress in the circumferential
CASE I CASE II
direction with respect to the
shell, psi · To illustrate, consider Case I showing a direct
normal stress in the longitudinal di- radial inward load (P) transmitted to the shell by
rection with respect to the shell, the attachment. Here the load acts similar to a
pSI local external pressure load on the shell causing
shear stress on the x face in the cf> compressive membrane stresses. Also, local bend-
direction with respect to the ing occurs so that tensile bending stresses result on
sheli, psi the inside of the vessel at A and B while compres-
·shear stress on the 4> face in the x sive bending stresses result on the outside. ·
2 Strl'ssrs in ShPl!s
p
U = C1
0.875 Vlf::T
3.2.2 ATTACHMENT PARAMETERS. For spherical
shells, either round or square attachments may be
considered.
3.2.2.1 Rigid Attachments: In the case of a rigid
attachment, no attachment parameter is required
to use the nondimensional curves.
3.2.2.2 Nozzles: For a hollow cylindrical attach-
ment such as a nozzle, the following parameters
must be evaluated:
rm
T
t
T
p =
t
3.2.2.3 Hollow Square Attachment: If a hollow
square attachment such as a box beam is to be con-
sidered, the required parameters. may be approxi-
Fig. 1-Types of loading conditions at an attachment to a mated as follows:
spherical shell ·
rm
T
0 .875t
In Case II, the overturning moment may be con- T
sidered to act as a couple composed of equal and p
opposite .radial forces. Hence, tensile membrane
3.3 Calculation of Stresses
stresses result at A while compressive membrane
stresses result at B. As in Case I, local bending 3.3.1 STRESSES .RESULTING FROM RADIAL LOAD,
also occurs so that tensile bending stresses develop p
at A on the outside of the vessel and at B on the in-
side, while compressive bending stresses develop at Table 1-Sign Convention for Stresses Resulting from Radial
and Moment Load on a Spherical Shell
A on the inside and B on the outside.
In this manner, the signs (tensile +, compres-
sive -) of stresses resulting from various external
loading conditions may be predicted. It is to be
noted that these stresses are located in the vessel
wall at its juncture with the attachment.
Table 1 shows the signs of stresses resulting from 1 j
radial load and overturning moment. This table 'f 1
will facilitate the use of the nondimensional curves
(presented in the following procedure) and mini- ROUND ATTACHMENT SQUARE ATTACHMENT
mize concern for the signs of the calculated
stresses.
3.2 Parameters
The results of Bijlaard's work have been plotted
N, Ny
in terms of nondimensional geometric parameters T&-T-
by use of an electronic computer. Hence, the first
step in this procedure is to evaluate the applicable +
geometric parameters.
Bending
...
6M, Cu
Cl :;
3.2.1 SHELL PARAMETER (U). The shell param- 7
eter is given by the ratio of· the nozzle outside
Ou
OL ...
radius to the square root of the product of shell AL
.+
radius and thickness, thus:
Bending 8u
BL ;
6My
Cu
...
U _ __r_o__
- VRmT
T2
CL
Ou
Dl ..
NOTE: 1) Sign convention fot stresses: +tension,- compresJiion.
If a square attachment is to be considered, U may 2) If load P u~venes, oil signs in coh..nn P reverse.
3) If ovrMturning moment Mt tevMses, oil signs in column Mt rll!Yerse.
be approximated as follows: 4) For ravnd ottachment, overturning momenta M] and M2 may be combined
vectoriaffy, thu•: M !!.to/M{ ~ M.}
Stresses in Shells 3
3.J.l.l Radial Stresses ((Jr): (N,/T) by:
Step 1. Using the applicable values of U, 'r and
p*, read off the dimensionless membrane force
N,
T.
= [N,T VR:f] .[
M
M
T 2VRmT
J
(N, TIP) from the applicable curve which will be
found in one of the following figures: Figure SR-2 Step 4. By a procedure similar to that used in
Step 3, calculate the radial bending stress
or SP-1 to SP-10, inclusive.
Step 2. By the same procedure used in Step 1 (6Mx/T 2 ), thus:
above read off the value of dimensionless bending
mome~t (M, I P) from the applic'able curve. This
6MT =
T2
[MxVR:fl· [
M
6M
T 2VRmT
J
value will be found in the same figure used in
Step 1. · Step 5. Combine the radial mem,brane and
Step 3. Using the applicable values of P and T, bending stresses by use of the general stress equa-
calculate the radial membrane stress (N,/T) by: tion (para. 2) together with the proper choice of
sign (see Table 1); i.e.,
= K NT K 6M,
Step 4. By a procedure similar to that used in
(J
I " T ± h""T2
Step 3; calculate the radial bending stress (6M,/T 2 ) 3.3.2.2 Tangential Stress (cru): Follow the five
thus: steps outlined in 3.3.2.1, using the same figure to
6M,
T2
= (M'p ) . (6P)
T2
obtain (NvTVRmT/M) and (MvVR:fiM) used
to obtain (NTT!P) and (MT/P). It follows that:
Stt.r 5. Combine the radial membrane and
bending stresses by use of the general stress equa-
tion (Section 2) together with the proper choice of
sign (see Table 1); i.e.,
Stresses in Shells 5
Table 2-Computatlon Sheet for Local Stresses In Spherical Shells (Solid Attachment)
P Mr
1. Applied Loads *
lTy ·t::
and radial bending stresses given in Figs. SR-2 and DIAL LOAD (P).
SR-3, so that the following simplified procedure for Step 1. Calculate the value of the applicable
calculating maximum stresses has been developed. shell parameter (U) as given in para. 3.2.1.
3.6.1 MAXIMUM STRESS RESULTING FROM RA- Step 2. Enter Fig. SR-1 at the value of U found
6 Stresses in Shells
Table 3-Computation Sheet for local Stresses in Spherical Shells (Hollow Attachment)
, /Mr
I
1. Applied Loads •
2. Geometry
u ~·
~
fo =---
Vessel Thickness, T = 1n Stress Concentration Factors
Vessel Mean Radius, Rm= _ _ _ in. due to:
Nozzle Thickness, 1n s _ __
t membrane load, Kn HOLLOW ATTACHMENT
Nozzle Mean Radius, rm in. bending load, Kb ,. _ __
Nozzle Outside Radius, r0 _ _ _ in. NOTE: Enter all torce values in Rm
accordance with sign convention
SP-1 to 10 NxT p
::: ( N;T) •
T2 :::.
p Kn
Mx
p ::: Kb ( ~· )·
6P
TT
-
c·T~)
SM-1to10 NxT~ "::. M,
M,
Kn 0
T2~ =
6M 1
Mx~
-M-,-::: Kb
(Mx@
M, ) • T2jii;:T ::<\
NxTjRmT _
--M-,--- Kn (NxT~) Mz
• T•jRmT
6M2
• TZ'-/ RmT +
p
·rz-= -
- Kb (M;) •p X
61'
Kn (NyT~) 'rz~
M,
....
My~ 6M 1
-M-,-- 'l: 1\b (My::r ) 'rz~ 3
NyTyRmT::
Mz ~" (HyT~) Mz
• rz.(RmT
----
Myyrr;;;f -
Mz Kb (My~)
6M 2
• T2yR;l
CYLINDRICAL
SHELL
~~.~
normal to the surface of the vessel is zero, rr is the
maximuni stress intensity; i.e., s = <I.
3.6.2 MAXIMUM STRESS HESULTING FROM OVER- ON CYLINDERS: B IS MEASURED
FROM THE LONGITITUDINAL
TURNING MoMENTS, M1 AND M2. AXIS {B= 0")
OQI
TYPICAL ELEMENT
Step 1. Hesolve moments M1 and M2 vectorially
into a single moment (M) thus: Fig. 2-Types of loading conditions at an attachment to a
cylindrical shell
M = VMI 2 _+_M;;
NoTE: It is assumed that M1 and M 2 areorthog-
onally oriented. Also, M must not include tor- In using this formula, the sign conventions estab-
sional moment (MT) as shown in Fig. 1. lished in Table 1 should be used.
Step 2. Calculate the value of the applicable
shell parameter (U) as given in para. 3.2.1. 4. Cylindrical Shells
Step 3. Enter Fig. SH-1 at the value of U found 4.1 Sign Convention
in Step 2, and using the curve marked "external Stresses will be considered in the shell at the at·
moment M" read off the value of the nondi- tachment-to-shell juncture in both the circum-
VR::i'
inensional stress ( rr x T 2 I M). ferential and longitudinal directions as shown in
Step 4. Using the applicable value of moment Fig. 2. A knowledge of the shell deflections re-
(M), shell thickness (T), shell mean radius (Rm), sulting from various modes of loading permits one
and stress concentration factor (K 11 ) , calculate the to predict whether resulting stresses will be tensile
maximum combined sti·ess ( rr x) thus: ( +) or compressive (- ).
is zero, rr is the maximwn stress intensity, i.e., S = Consider Case I showing a direct radial inward
<I.
load, P. Here P acts similar to a local external
3.6.3 MAXIMUM STREss REsULTING FROM CoM- pressure on the shell causing compre~sive mem-
BINED LOAD (P) AND OVERTURNING MOMENT (M). brane stresses. Furthermore, local bending occurs
If load (P) and moment (M) are considered sep- so that tensile bending stresses result on the inside
arately as outlined in 3.6.1 and 3.6.2 above, it is of the vessel at C and D while compressive bending
possible to consider the combined loading condi- stresses result on the outside.
tion by superposing results of the two cases as In Cases II and III, the applied moments are
follows: considered to act as couples composed of equal and
opposite radial forces. Hence, tensile membrane
stresses result at Band D while compressive mem-
brane stresses result at A and C. As in Case I
8 Stresses in Shell8
Appendix A.
Table 4-Sign Convention for Stresses Resulting from Radial
and Moment Loading on a Cylindrical Shell 4.2 Parameters
" +
6M, C,
:;
.-..
'i:" "
0,
4.2.2.2 Square Attachment: For a square attach-
ment the parameter is evaluated by:
NOTE: 1) S•gn con.,.~rntion for stressr10: +lensoon,- comprrs••on, Ct c,
2) If lood at mom•i'lt direct tons tr't'ttu:, all s•gns
H'l opplocablr tolumn revers•,
(3 = !3t = !3z = -- = -·
Rm Rm
tensile bending stresses result at A and Con the in- 4.2.2.3 Rectangular Attachment Subject to Radial
side ~f the vessel, and B and D 'on the outside of the Load (P): For this case (3 is evaluated as follows:
vessel. Similarly, compressive bending stresses
Ct
result at A and Con the outside and Band Don the !3t =-
inside. Rm
In this manner Table 4 has been developed to
show the signs of stresses resulting from various If:: ? 1, (3
external loading conditions. These stresses are lo-
cated in the vessel wall at its juncture with the at-
tachment. Use of Table 4 permits one to use the
nondimensional curves presented in the following
procedure with a minimum of encumbency and If(!.: < 1, (3
(3~
concern for sign convention.
The numerous stress components can be readily
accounted for if a scheme similar to that shown in
Table 5 is adopted. In using this scheme it is to where K values are obtained from Table 6.
be noted that the Maximum Shear Theory has been
used to determine equivalent stress intensities.
Also it is to be noted that evaluation of stresses 4.2.2.4 Rectangular Attachment Subject to Circum-
resulting from internal pressure has been omitted. ferential Moment (Me).
Test work conducted by PVRC has shown that 4.2.2.4.1: When considering membrane forces
stresses attenuate rapidly at points removed from (N;): (3 = ~-r;;. Then multiply values of
the attachment-to-shell juncture, the maximum Nt!(Mc/Rm 2!3) so determined by Cc from Table 7
stress usually being located at the juncture. How- (see para. 4.3).
ever, in the general case of arbitrary loading, one 4.2.2.4.2: When considering bending moment
has no assurance that the absolute maximum (M 1): (3 = K,~(3 1 2(3 2 whereKcisgiveninTable7.
stress intensity will be located at one of the eight
points considered in the above discussion. The
maximum stress intensity could be located at some 4.2.2.5 Rectangular Attachment Subject to Longi-
intermediate point around the juncture under an tudinal Moment (M L).
arbitrary load, or under a longitudinal moment 4.2.2.5.1: When considering membrane forces
with the circumstances outlined in para. 4.4 and (N 1): (3 = -Y"{3 1 (3 2 2, Then multiply values of
Stresses in Shells 9
Table 5-Computation Sheet for Local Stresses in Cylindrical Shells
Stresses in Shells 11
and T, find the circumferential membrane stress Mr
(N"'/T) by:
Tot>, = Trot> = 27rro2T
N"'
T
N~ [ Mr
M,/R,. 2{3 . R,. 2{3T
=
J[ J If only shear stresses are being considered, it is to
be noted that the equivalent stress intensity is
Step 4. By a procedure sim'ilar to that used in twice the above calculated shear stress.
Step 3, find the circumferential bending stress In the case of rectangular attachments, torsional
(6M.!T 2 ). Thus: moment produces a complex stress field in the shell.
Acceptable methods of analyzing this situation are
6M"' [ M"'
T 2 = M,/Rm/3
J . [·.R,.T
6M, J
/3 2
not available at this time. If the designer has
reason for concern, the problem should· be resolved
by testing in accordance with estl\blished c.ode
Step 5. Combine the circumferential membrane procedures.
and bending stresses by use of the general stress
4.3.5 STRESSES RESULTING FROM SHEAR LOADS,
equation (Section 2) together with proper choice of
V, AND V L· Bijlaard has proposed • that shear
sign (Table 4); i.e.:
force ( V) can be assumed transmitted to the shell
q<l> = K !!_"' ± K ~M"' entirely by membrane shear force. Therefore
" T b T2 stresses in the shell at the attachment-to-shell
4.3.2.2 Longitudinal Stresses (u,): Follow the 5 juncture can be approximated as follows:
steps outlined in 4.3.2.1 except that [Nr! 4.3.5.1 Round Attachment
(M,/Rw. 2f3)] is obtained ul.'ing Fig. 4A; and
[M,j(M,jR,.B)], using Fig. 2A. It follows that: Trot> = v,T cos(} (max. at A and B)
J . [R,.M,{3Tl
1rro
N, [ . Nr
T = M,/R,. 2 {3 2 T"'' = VLTsin (} (max. at C and D)
1rro
6M,
T2 =
[ Mx
M,/R,./3
J. [R,.T
6M, J
{3 2 4.3.5.2 Rectangular Attachment
K Nr K 6M,
v,
Uz = " T ± b T2 4ciT
4.3.3 STRESSES RESULTING FROM LONGITUDINAL VL
Tot>z = -4 T
MOMENT, ML. c2
4.3.6 STRESSES RESULTING FROM ARBITRARY
4.3.3.1 Circumferential Stresses (uq,): Follow the LOADING. In the general case, all applied loads
5 steps outlined in 4.3.2.1 except that [N<~>/(M d and moments must be resolved (at the attach-
Rm 2f3)) is obtained using Fig. 3B; and [M"'/ ment-to-shell interface) in the three principal
(M d R ,1{3) ], using Fig. 1B or 1B-l. It follows directions; i.e., they must be resolved into compo-
that: nents P, V" VL, Mo ML and Mr.· If one then
proceeds in the manner previously outlined (e.g.,
para. 4.3.1.1), membrane, bending and shear
stresses can be evaluated at eight points in the shell
6M"'
75 =
[ M"'
M dR,./3
J . [R,./3T
6M J L
2
at its juncture with the attachment. These eight
points are shown in the sign convention chart,
4.3.3.2 Longitudinal Stresses (ur): Follow the 5 Table 4.
steps outlined in 4.3.2.1 except that [Nr/(M d
4.4 Nondimensional Curves for Cylindrical Shells
R"' 2{3)] is obtained using Fig. 4B; and !Mrl
(M d R ,.{3) ), using Fig. 2B or 2B-l. It follows The nondimensional curves which follow consti-
that: tute, in general, a replot of Bijlaard's data to a
semilog scale in order that certain portions of the
curves can be read with greater facility. Those
portions of the curves which are taken directly from
6M,
'f2 =
[ Mr
MdR,.{3
J. [R,.{3T
6M J L
2
Bijlaard's work are shown as solid curves; those ·
portions of the curves which have been modified on
the basis of recent experimental data, as discussed
4.3.4 STRESSES RESULTING FROM TORSIONAL in Appendix A, are shown as dotted curves.
MoMENT, Mr. In the case of a round attachment In the case of longitudinal moment loading and
(such as a pipe), torsional moment is assumed to axial loading (thrust), two sets of curves are shown
induce only shear stresses, so that shear stress ( T) for the bending components of stress-one set ap-
in the shell at the attachment-to-shell juncture is
given by: • See reference 14.
12 ))tresses in Shell.<
plying to the longitudinal axis, and the other apply- in Appendix A and to the original references to as-
ing to an area of maximum stress off the axes of certain the limitations of applicability for·the pro-
symmetry (longitudinal moment), o1· to the trans- cedure used.
verse axis (thrust). In the latter case, a portion of However, there are a few generalizations that
the original curves has been deleted in order to em- can safely be made regarding vesse] and attach-
phasize that the curves should not be used beyond ment geometry.
the limits indicated. This was done because the 4.5.1 ExTERNAL RADIAL LoAD. Stresses are af-
available data indicated that the "outer limits" of fected very little by the ratio of shell length to shell
the curves were appreciably unconservatiye, with radius (l!Rm). Therefore, no restriction is made
no feasible manner to "correct:• them (as expl!lined on the point of load application except in very ex-
in Appendix A). treme cases. The curves included in this report
In the case of longitudinal momen1., the exact lo- are for an l!R"' ratio of S, which is sufficient for
cation of the maxinmm stress l.~atmol be defined most practical applications. On the basis of data
with certainty, but. Fig. A -14 will provide an esti- pJ;esented in Bibliographical Reference 2, results
mate of its location (considering that the location based on an l/R,. ratio of 8 will be slightly con~
of maximum stress under internal pressure and servative for lesser values of l/R .. ratio and un-
longitudinal moment was e..~ntially the same on conservative for greater values of l/Rm ratio.
liT model "C-1," as shown on Figs. A-2 and A-3). However, the error involved does not exceed ap-
It should aJso be noted that., to the best of our proximately 10% of all l/R,. values greater than 3,
knowledge, t.he curve::; for "maximum stresses off which should be sufficiently accurate for most cal-
the axes of symmetry" (l<'igs. lB-1 and 2B-1) would culations. Since for lesser values of lj R ,,., the re-
apply only to the case of a round, flexible nozzle sults are conservative, no restriction will ordinarily
connection; it is conceivable that a similar effect be necessary on l!Rm ratio or the point of load
might apply. to a rigid square or rectangular at- application. For extreme cases or for "off center"
tachment, for which the shell at the outer edges of loading, one may make corrections by use of the
the attachment might take a greater part of the curves presented on page 8 of Bibliographical Ref-
load than that portion of the shell adjacent to the erence 2, if desired.
longitudinal centerline. However, we know of no Results are not considered applicable in cases
dired evidence to support su~h an assumption. where the length of the cylinder (l) is less than its
4.4.1 LI&'T OF NONDIMENSIONAL CURVES FOR radius (Rm). This applies either to the case of an
CYLINDrtiCAL SHJi:LLS open ended cylinder or closed ended cylinder where
Fig. IA Moment M.;(Mc!Rff,{1) due to the stiffness is appreciably modified from the case
Me considered.
Fig. 2A Moment Mz/(MJRmti) due to 4.5.2 EXTERNAL MoMENT. Results are appli-
M< cable in the case of longitudinally off center attach-
Fig. 3A Membrane force N,;/(Mc/R,._:tl) ments (a more usual case) provided that the at-
due to .~fc tachment is located at least half the shell radius
Fig. 4A Membrane force Nz/(Mc/Rm 2 f3) ( 1I~ R "') from th~ end of the cylinder.
due to Me. 4.5.3 ATTACHMENT STRESSES. The foregoing
Fig. lB or lB-1 Moment. Mj( M 1,jR,,{J) due to proced1.1re provides one with a tool to find stresses
ML in the shell, but not in the attachment. Under cer-
Fig. 2B or 2B-1 Moment M.J(MJR ..{J) due to tain conditions, stresses may be higher in the at-
!Iff... tachment than they are in the vesseL For ex:~
Fig. 3B Membrane force N./(M·dR,/{3) ample, in the case of a nozzle, it is likely that the
due to ML stresses will be higher in the nozzle wall than they
Fig. 4B Membrane force NJ(MdR,.. 1{3) are in the vessel wall if the nozzle opening is unre-
. due toM 1... inforc:ed or if the reinforcement is placed on the ves·
Fig. lC Moment M ¢/P due toP sel wHll and not on the nozzle.
Fig. lC-1 Moment M xl P due toP
Fig. 2C Moment M xl P due to P
5. Acknowledgment
Fig. 2C-l Moment M ¢1 P due toP
Fig. 3C Membrane force N IPIP/Rm and The authors wish to acknowledge the significant
N xl PI R m due to P contributions made by J. B. Mahoney of Applied
.Fig. 4C Membrane force N XI pI R m and Technology Associates Inc. and M. G. Dhawan of
N '~>IPIR,. toP the Bureau of Ships during the preparation of this
4.5 Limitations on Application . <
paper. In addition, the comments received during
Where relatively large attachments· are consid- the review of this document by the members of
ered, or when situations are encountered that the PVRC Subco~mittee on Reinforced Openings
deviate considerably from the idealized cases pre- and External Loadings are deeply appreciated.
sented herein, the designer should refer to para. A.3
· Stresses in Shells 13
18.· Dally, J. W., "An Experimental Investigation oflheStr""""" Produced
6. References in Spherical Vesseh by External Loads TrantJferred by a Nozzle," Ibid .• No.
84, (Jan. 1963).
1. Bijlaard, P. P., ..Stres..'"ieS from Local Luadings in Cylindrical PreBBure 19. Kaups, ~raavi, •·Effect of External Moments ·on 190-foot Diame~r
Vessels," Trans. A.S.M.E., 71,805-816 (1955). Horton•phere," The Water Tower, XLVI (1), (Sept. 1959).
2. Bijlaard, l'. P., ..Sb·esses from Radial Loads in Cylindrical Pressure · 20. Lee".de, F. A .. and Penny, lL K .. uSt.ra3S Concentration Factors for
Vessels," Welding Jnl., 33 (12), Research Supplement, 615-s to 623-s (1954). the St~esses at Nozzle Intersections in Pressure Vesseh;' Welding Research
3. Bijlaard, P. P .. "Stresses from Radial Loads and External Momenta in Counctl Bulletin No. 90, 19 ·26 (Sept. 1963).
Cylindrical Pressure Vessel," Ibid., 34 ( 12). Resc>rch Supplement, 608-s to
21. Harden~>er~h, D. E .. Zamrik, S. K., an1 Edmon•~n. A. J., "Experi-
617-s (191\5). mental Investigallon of Stresses in Nozzlt:S in Cylindrical Pre8:bure Veseela"
4. Hijhard, P. P., '"Comput3.tion of the Stresses from Local Loads \n Ibid., 89, (July 1963). '
Spherical Pressure Vessels or Pressule Ve3Sel Heads," Welding Research
Council Bulletin No. 34, !March 1957). 22. Hardenhergh, D. E .. and Zamri 1<, S. K., "Effects of External L<ladings
5. Bijlaard, P. 1'., "Local Stresses in Spherical Shelb from· Radi-11 or on Large Outlets in a Cylindrical Pressure Vessel," No. 96 U-23 (May
1964). •
Moment Loadings," Welding Jnl., 36 (5), Reaearch Supplemenl, 240-s to
243-s (1957). 23. Riley, W. F .. "Experiment>\ Determinotion ofStreso Distributions in
6. Bijlaard, I'. P., "Stresses in a Sphetical Ve3Sel from Radi.! Loads Tnin Wall'Od Cylindrical and S;>herical Pressura VelSel• with Circular Noz-
Acting on a Pipe," Welding Research Council [Jull•tin No. 49, 1-·30 (April zles," IITRI Final Report, Project no. M6053 March 1&, 1965. (To be pub-
1959). . lished in Welding Res.arch Cou~cil Bull!tin No. 108, Sept>mher 1965).
7. Bijlaard, P. P., "Stresses in a Spheric>! Ve1sel from External Moments 24. Leven, M. M., "Photoehotic Determination of St........, Due to the
Acting on a Pipe," Ibid., No. 49, 31··-62 (Ap.ritl959). Bending of Thin Cylindric•! Nozzles in Thin Sphmcal Veseeb," Westing-
8. Bijlaard, P. P., "Influence of a Reinforcing Pad on the Stre3Ses in a house Re3earch Labs. Repcrl63-917-514-R2, April 15, 1963.
Spherical Vessel Under Local Loading," Ibid., No. 49, 63-73 (April 1959). 25. Naghdi, A. K., and Erilll(en, A. C., "Stre..s Anslyois of a Circular
9. Bijlaard, P. P., "Stre'""" in Spherical Ve.,eh from L.:cal Loads Trans- Cylindric>! Shell with Circular Cutout," General Technology Corp. Report
ferroo hy a Pipe," Ibid., No. 50. 1-·9, (May 1959). Nc. 3-2, Jan. 1963.
10. l!ijlaard, P. P., "Additional. Ddts on Stres>es in Cylindric•! Sheils 26. Koh, S. L., Thiel, C. C., and Erin.:en, A. C., "Comput>tions for Streos
Under Local Loading," Ibid., No. 50, 10 ·50 (May 1959). an-i Stre38 Concentraticn in a Circular Cylindrical Shell with Circular Cut-
11. Kempner, J., Sheag, J., and Pohle, F. V., "Tables and Curves for De- out," General Technology Corp. lb>port No. 3-3, April 1963.
formations and Stresses in Circuhr Cylindrical Shells Under Localired 27. Eringen, A. C., Nagbdi, A. K., and Thiel, C. C., "State of Stress in a
Loadings," Jnl. A ern. Sci., 24, 119 ·129 (191>7). Circ:.dar Cylinrlrical Shell with a Circular Hole," Welding Research Council
12. Reissner, E., "Stress and Small DisJ>hcements of Shallow Spherical Bulletin No. 102, (Jail. 1965).
Shells," I., J. Math. Phys., 21i, 80 ·85 11946). 28. I.an~ter, B. F., "PVRC Interpretive Report of Pressure Ve....I Re-
13. Foster. K., "The Hillside Problem: Stresses in a Shall£>w Spherical se~rch ·Section, 1., Design Considerations; Section 1.6, Extem.al Loadins."
Shell Due to External Loads on a Non-Hadial Ri~id Cylindrical In....rt," lbd., No. 95, ZS-33 (April 1964). · ·
Ph.D. Thesis, Cornell University (1959). 29. Pete.-.on, H. E., "Stre"" Concentration Design Factors," John Wiley
14. Bijhsrd, P. P., "On the Effect of Tangenti~l Loads on Cylindrical and Son•. Inc., New York, 1953.
and Spherical Shells." Unpublished, available in the files of PVRC Weldin~ 30. Heywood, R. 1:!., "Designing by Photoelaoticity," Chaptrutn and Hall
Research Council. ' ~ London, 1952. '
15. ?-:.:.:,ny, R. K., ..Stress Concentrations at the Junction o( a Spherical 31. Van Dyke, P., "Stresses About a Circular Hole in a Cylindrical Shell"
Pressure Vessel and Cylindrical Duct c:1used hy Certain Axisymmetric Harvard Univ. Technical Report No. 21 under Contract Nonr-1866(02,),
Loading," Proceedings of a Syrnoosium, Hoyal College, Ghssow. May 17- Sept. _1964.
20, 1960, Buttcrworths, 88 Kin~sway, London W. C. 2. ~2. Owaltn~, R.C., ET AL '11w>oretical aad Experimenlal Stress Analysis of OlltNL
16. Tentative Structural Design Basis for Reactor Pressure Vessels and Thin-Shell Cylioder-14>-Cylinder Model No. 3".11teport No. ORNL-5020, June 1'75.
Directly Associate<! Components (Pressurized Water Cooled Syst.ems) 33. Masl!on, J.L., "PVRC Reoea.rch oo Reiaforcementof()peain&< in ~'rasure Vessels"
December 1958 revision. ' WRC Bulletin No. n (Tables B-7a and B-7b, poge 48), May 1962. •
17: Bi)la&rd, P. P., and Cranch, E. T., "Interpretive Commentsry on the 34. Corum, J.M.,et.U. "Theoretical and ExperimentaiSitcSsAaalysisofOilNL Tbin-SheU
Apphcat10n of Theory to Expertmental Results for Stresses and Deflections Cylinder-10-Cylioder ~No. I", Report No. ORNL-4SS3, Oct. \972.
Due to Local Loads on Cylindrical Shells," Welding Research Council Bulle-
tin No. 60, 1·-2 ·(May 1960).
14
Fig. SR-1--Maximum stress due to external loading on a spherical shell (rigid plug)
Hi
Fig. SR-2-Stresses in spherical shell due to a radial load P on a nozzle connection (rigid plug)
Strrssrs in Shr!fs
0 0.5 1.0 1.5 2.0 2.2
Fig. SR-3-Stresses in spherical shell due to overturning moment M on nozzle connection (rigid plug)
8iressrs in 8hr//s 17
0 0.5 1.0 1.5 2.0 2.2
18
0 0.5 1.0 I .5 2.2
Stresses in Shells 19
0,5 LO L5 2,0 2.2
20 8t1'essc~:; £n Shells
Fig. SP-4-Stresses in spherical shell due to radial load P on a nozzle connection
StreS!iCS in Shctl8 21
0.5 1.0 1.5 2.0 2.2
22 Stresses in Shells
Fig. SP-6-Stresses in spherical shell due to radial load P on a nozzle connection
23
Fig. SP-7--·Slresses in spherical shell due to radial load P on a nozzle connection
24
ro
u
YRmT
-;0001
0 0.5 1.0 1.5 2.0 22
Fig. SP-1!---Stresses in spherical shell due to redial load P on il nozzle connection
StretJIJI!Il in Shells 25
0.5 1.0 1.5 2.0 2.2
Fig. SP-9-Stresses in spherical shell due to radial load P on a nozzle connection
26 8trrssrs m 8hr/1s
0 0.5 1.0 1.5 2.0 2.2
Stresses in Shells 27
0 0.5 1.0 1.5 2.0 2.2
29
Fig. SM·3-Stresses in spherical shell due to overturning moment M on nozzle connection
30 Stressel) in Shells
0 0.5 1.0 1.15 2.0 2.2
Fig. SM-4-Stresses in spherical shell due to overturning moment M on nozzle connection
82
- - -~__::. ______ ----r--------
• • •- . . .
~ • T
'
••
' : •
•
, __ ,
' .
. '
. ; .
•
::
8irt'88!'8 in 8/l.r-/.fs
0 0.5 1.0' 1.5 2.0 2.2
36
0 0.5 1.0 1.5 2.0 2.2
38 Strrsses in Shrlls
Fig. 2A-Moment M,/(M,/R ...i3)~due to an external circumferential moment M, on a circular cylinder
Stresses in Shells
0 005 0.10 015 020 0.25 0.30 0.35 0.40 0.45 0.50
Fig. 3A-Membrane force Nq,/(M,/Rm'/3) due to an external circumferential moment M, on a circular cylinder
:ilrc~;ses in Shell,~
Fig. 18-Moment Mq~/(ML/Rm/3) due to an external longitudinal moment ML on a circular cylinder (Stress on the longitudinal
plane of symmetry).
42 Stresses in Shells
Fig- 18-1-Moment M<~>/(Mr/Rm/'l) due to an external longitudinal moment M .. on a circular cylinder
Slri'N.~cs £n Shrlls
Fig. 28-Moment M,j(Mr./Rmll) due to an external longitudinal moment Mr. on a circular cylinder (Stress on longitudinal
plane of symmetry)
44 8/!'I'Ssr·s in 8hl'l/s
0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.'15 0.50
Fig. 28-1-Moment M,/(ML/Rm/'l) due to an external longitudinal moment M1. on a circular cylinder
8/n'ssr's in 8h.l'/l.~
Fig. 38-Membrane force Nq,/(Mr./Rm 2fl) due to an external longitudinal moment ML on a circular cylinder
46 8tn:sses in Shell.~
0 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50
Fig. 4B-Membrane force N,j{ML/Rn8l) due to an external longitudinal moment ML on a circular cylinder
Stresses in Shells 47
0 0.05 OJO OJ5 0,20 0,25 0.30 0:35 0.40 0.45 0.50
Fig. lC-Bending moment Mq,/P due to an external radial load P on a circular cylinder (transverse axis)
48 Stresses in Shells
'0 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50
Fig. lC-1-Bending moment M'X/P due to an external radial load P on a circular cylinder (longitudinal axis)
49
0 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50
Fig. ZC-Bending moment Mx/P due to an external radial load P on a circular cylinder (transverse axis)
50 Stresses in Shells
Fig. 2C-1-Bending moment M•/P due to an external radial load P on a circular cylinder (longitudinal axis)
!il
0 0.05 0.10 0.15 0. 0 0.25 0.30 0. 5 0.40 0.45
Fig. 3C-Membrane force N4>/(P/Rm) due to an external radial load P on a circular cylind.er (transverse axis)
-Memorane force N~<l(P/R,..) due to an external radial load P on a circular cylinder (longitudinal axis)
;)2 8/l'esses in 8hPI!s
Fig. 4C-Membrane force N:.<j(P/R,,) due to an external radial load P on a circular cylind?r (transverse axis) ·
-Membrane .force N¢/(P /R,,) due to an external radial load Fl M a circular cylinder (longit.ucH.nal &xi a)
APPENDIX A~BASIS FOR "CORRECTIONS" TO BIJLAARD'S CURVES
A.l Introduction
mately 37, 80 and 92, which parameters are all well
During the past year or more, data have become within Bijlaard's limits, as is indicated on Fig. A-1.
available indicating, or emphasizing, certain limi- Subsequently, Westinghouse Research Labora-
tations in Bijlaard's work for external loadings on tories tested four photoelastic models for the
nozzle connections. In some cases, the discrep- Bureau of Ships, under moment loading only, as
ancies involved were quite large and seemingly reported in Reference 24; these models had a
indicated a necessity for providing interim modifi- Dm!T ratio of 51.0, one with a dt/D; ratio of 0.13,
cations to a portion of Bijlaard.'s data pending de- two with ad;/ D; ratio of 0.27, and one with ad;/ D1
velopment of a more adequate analytical treatment ratio of 0.50. Similarly, IIT Research Institute
of the problem. The following summary is pre- has tested one steel model for PVRC, having a
sented to document the nature of the discrepancies Dm!T ratio of 236 and a d;jD; ratio of 0.50, the
and to explain the manner in which the curves preliminary results of which are reported in Refer-
based on Bijlaard's data have been modified ence 23. As indicated on Fig. A-1, these models
herein. provide one point well within Bijlaard's limits, two
approaching those limits and two well outside those
A.2 Spherical Shells limits.
Bijlaard's work on spherical shells was based on Also recently, the work of Penny-Leckie 20 be-
shallow shell theory, and the limitations which he came available, which is based on "not-shallow
placed on the theory were essentially as indicated shell" theory and which might offer hope of a more
in Fig. A-1, from which it will be observed that the adequate treatment at the larger diameter ratios.
limiting dt/D; ratio is about 1 / 3 for "thin" shells, A summary of the parameters for the above
and somewhat less in thicker shells (D,JT ratios of mentioned models is contained in Table A-1, and a
20-55). The experimental work at Cornell Uni- summary of the calculated and measured stresses
versity which was performed to verify the theory, in Table A-2. Reference 20 provides curves only
as reported in Reference 18, was for a d ;/ D; ratio of for the maximum of the two stresses, and states
approximately 0.10 and Dm/T ratios of approxi- that " ... for small values of tjT, the hoop stress
10 100 1000
Dm/T RATIO
Fig. A-1-Relation of spherical test models to Bijlaard's "applicability limits"
54 8trrsses in 8hrlls
Table A-1-Parameters for Model Vessels Tested with External Loads on Nozzles
u•
( = 1.82
(3"
( = 0.875 Fillet
ro radius, r,,,~
• Bijlaard's parameters.
Table A-2-Comparison of Calculated and Measured Stresses in Spherical Models under External Nozzle Loadings
~-Calculated stresses, • k s i - - ~---Measured stresses,• ksi--~
~--Bijlaard--~ Adjusted
ffx Penny
Moment loading
WN-50A 2.72 2.13 3.64 4.81 6.05 5 .03/J
WN-50B 3.18 2.51 3.37 4.59 5.73 4. 78°
WN-50C 2.40 0.554 2.52 3.83 4.45 3. 97'
WN-50D 2.11 0.447 2.09 2.73 3.15 2.81'
S-1 14.5 9.95 14.4 11.8d
Direct axial load (6000# pull)
S-1 23.4 16. 56' 20. 66' 17.0d
18.31 23.5 1 19.3d
NoTES:
• Stresses due to moment loading are reported as a ratio of the stress in question to the· calculated bending stress in the
nozzle, as was reported by Westinghouse for the photoelastic models. These "base" nozzle stresses are as follows: WN-50A-
0.00398Mb; WN-50B-0.0135Mb; WN-50C-0.0274Mb; WN-50D-0.118Mb; S-1-0.0903Mb, where Mb is the applied
moment.
b Based on local SCF of 1.20.
'Based on local SCF of 1.12.
J Based on local SCF of 1.22.
' Average of eight separate measurements around nozzle.
I Maximum of eight separate measurements around nozzle.
( fi ") is higher than the meridional stress (ax). calculated values for fix are greatly below the meas-
With increasing tjT, fiv becomes smaller and ax ured values. As was explained in Reference 28,
larger, until fix begins to dominate. This change- the primary reason for this difference is that both
over takes over when tjT is approximately Bijlaard and Penny-Leckie provide only for the
0. 75 .... " On this basis, the stresses from Penny- calculation of the stresses in the shell; however, for
Leckie are presumably for fi 11 for the four photo- all four of these models, the maximum stress
elastic models (WN-50 series), and ax for the steel proved to be across the base of the nozzle, charac-
model ("S-1 "). terized by a high bending stress in the axial (fix)
In the case of the photoelastic models (moment direction.
loading), it will be noted that the calculated In the case of the steel model (S-1), the calcu-
stresses from Bijlaard and Penny-Leckie agree al- lated stresses from Penny-Leckie agree almost ex-
most exactly for the smallest d j D; ratio (WN- actly with the maximum measured stresses under
50D), Bijlaard is about 5% lower than Penny- both moment and axial loading, and would be
Leckie for models WN-50B and WN-50C, and 25% about 25-30% conservative in relation to average
lower than Penny-Leckie for model WN-50A, indi- measurements adjusted for local stress intensifica-
cating a progressive deterioration of shallow shell tion. It should be emphasized, however, that
theory. However, it will be noted that for these present figures are preliminary and "average" fig-
four models, all of the calculated values for a 11 are ures may not be the most1 valid representation of
somewhat below the measured values, and all of the the data.
Stresses in 8hells
In normal design procedure, it must be assumed complete stress data based on this theory, including
that the maximum pressure and maximum external the stress distribution in both the shell and nozzle.
loading stresses occur at the same point. In the
case of Model "S-1," this was true insofar as can A.3 Cylindrical Shells
be determined from the preliminary data.. In the About two years ago, PVRC undertook testing of
case of the photoelastic models, a duplicate of a series of simple, fabricated tee type models consist-.
Model WN -50B was tested under internal pres- ing of two models ·at a D;/T ratio of 18.0, with
sure; for this model, the points of maximum stress d;/J) 1 ratios of 0.63 and 1.00, and two models at a
did not quite coincide. Assuming that the point D 1/T ratio of 230, with d;/D 1 ratios of 0.50 and
of maximum pressure stress will be controlling, 1.00. The primary purpose of these models was
the external (moment) loading stress at that point to provide external loading data at the larger
was perhaps 10-15% less than its maximum. On diameter ratios, in the hope of being able to extrap-
the basis of present evidence, it should not be as- olate Bijlaard's curves for cylindrical shells on up
sumed that there is any large conservatism in to a djD 1 ratio of 1.0.
considering the points of maximum stress as coin- The results of Lhis work have recently become
ciding (in the case of the larger diameter ratios in available, the data on the two "thick" shell models
spherical shells at least). being reported in Reference 22 and the preliminary
Quoting Reference 28, the status of the theoreti- data on the smaller of the two thin shell models in
cal work on spherical shells can be summarized as Reference 23. The results from the thick shell
follows: models indicate discrepancies in Bijlaard's ·"ex-
"The theoretical solutions for the stresses and tended range" data (page 12 of Reference 10) of a
deflections in (spherical) pressure vessels produced magnitude consistent with that which would be ex-
by externally applied forces and moments have pected from shallow shell theory (as indicated by
been developed to the point where they can be of the work on spherical shells). However in the case
considerable value to the designer if used with dis- of thin shell model, the results indicate that some
cretion. The discretion which must be used con- of this extended range data is greatly in error; in
sists of cognizance of the following limitations: addition, it appears probable that some of the
(1) When the loads are applied through rela- original curves 2• 3 are significantly in error in the
tively thin walled nozzles, the rigid-insert approxi- · very thin shell region (say, for values of {3 greater
matwn•supresseslJfe stresses circumferential to th:e than 0.15 and 0.10 at values of 'Y = 100 and 3oo;
nozzle. Whereas this approximation also exag- respectively). It appears that the basic reason for
gerates the meridional stresses, there is no reason this discrepancy is that, in thin shells, the longi-
to believe that the calculated meridional stress is a tudinal axis is relatively :flexible and free to deform
good approximation of the actual circumferential in relation to the transverse axis, causing the trans-
stress. verse axis to cauy a disproportionate share of the
(2) When the loads are applied through re.la- load. This effect was not fully provided for in
tively thin-walled nozzles, the highest stress may Bijlaard's treatment of the p~oblem, which treated
occur in the nozzle and a solution which gives only the nozzle as an "equivalent" square attachment.
shell stresses (as do Bijlaard's and Penny-Leckie's Actually, from superficial examination, some of the
solutions in their present form) may seriously un- test results appear so improbable as to create
derestimate the peak stress. suspicion of major deficiencies in the test model.
(3) None of the theories are capable of con- However, upon detailed comparison with available
sidering the geometry of the junction in detaiL internal pressure data, there is very good reason to
Therefore, the concentrating effect of a sharp cor- . b©_lieve that the ro,sults are essentially correct.
. ner must be estimated separately. Also, the addi-
tion of even a small fillet or weld bead can signifi- A.3.1 "Thick-Walled" Model Data
cantly affect the stiffness of the junction and result A summary of the parameters for the models in
in discrepancies between the actual and calculated question (Penn State Models "R" and "S" and
stress .... " IITRI Model "C-1") is contained in Table A-1,
On the basis of the foregoing, no changes in together with similar data (subsequently used for
Bijlaard's curves for spherical shells are considered comparison purposes) for Penn State Model "L,"
necessary, but particular attention should be paid reported in Reference 21.
to these limitations, and to the limitations which A summary and comparison of calculated and
Bijlaard placed on his own work (as summarized measured data for the three "thick-walled" models
on Fig. A-1). Since Penny-Leckie's theory ap- is contained in Table A-3. These data indicate
pears to give the same results as Bijlaard's at small that for Model "L," which presumably is at about
diameter ratios but does not have the limitations the upper limit of shallow·shell theory, the calcu--
of shallow shell theory, and in addition covers both lated stresses under moment loading range from 10
:flush and protruding (balanced) nozzles, the PVRC to 50% conservative; for Model "R," which in-
subcommittee hopes to provide, in the future, more ~olves an extrapolation of Bijlaard's curves, the
56 .
Stresses in Shells
Table A-3-Comparison of Calculated and Measured Stresses in Thick Walled Cylindrical Vessels with External loads on Nozzles
-·---C:aiculated stresses, ksi---.~ Maximum measured stresses, ksi
Membrane Bending Total Gross Adjusted
Model "L"
Longitudinal moment -u"' 10.45 17.1 27.55 20.5 20.5
(250,000 in.-lb) -u, 3.36 26.9 30.26 35.4 27.2d
Transverse moment -u"' 3.41 45.4 48.8 44.7 34.4d
(250,000 in.-lb) --~ux 6.37 24.7 31.1 21.2 21.2
Model "R"
Longitudinal moment -u"' 7.58 5.37 12.95 14.6 14.6
(500,000 in.-lb) -u, 3.60 8.03 11.63 23.7 16.9b
Transverse moment -u"' 4.16 31.5 35.66 57.1 40.7b
(500,000 in.-lb) -u, 10.0 15.0 25.0 26.9 26.9
Radial load (30,000# pull)
Longitudinal axis --o-q, 3.55 3.24 6. 79 1.4 1.4
-a, 2.08 5.78 7.86 2.2 1.57·'
Transverse axis --u:,. 2.08 5.78 7.86 14.4 10.3"
-u, 3.55 3.24 6.79 8 2 8.2
Model "S"
Longitudinal moment -u"' 4.8 4.8
(500,000 in.-lb) --ax 9.7 6.9''
Transverse moment -u"' 9.8/6)0 . 8.15<
(500,000 in.-lb) -u, 3 . 3 ( ( aXIS 3.3
NoTEs:
• Based on S.C.F. of 1.30.
b Based on S.C.F. of 1.40.
< Based on S.C.F. of 1.20.
4 Based on S.C.F. of 1.15.
calculated stresses under moment loading range axis of the vessel, with the maximum value being
from about 7 to 45% unconservative, with all four better than twice that directly on the longitudinal
stress quantities being 38 to 55% less conservative axis; a very similar effect was noted under internal
than was the case for Model "L." This effect is pressure. Whereas we had forewarning of this
believed attributable to limitations analogous to possibility under internal pressure, the effect
those of shallow shell theory, and is of a magnitude was quite unexpected under longitudinal moment.
not inconsistent with the effect noted in spherical Because this result seems somewhat irrational, a
shells. special effort was made to evaluate the probable
For the case of radial load, d!lta were not ob- validity of the results, summarized as follows:
tained on Model ''L." For Model "R," the maxi- (a) The vessel was not a machined model and
mum measured stresses on the longitudinal axis ~as slightly out-of; round. Such out-of-roundness
(both u"' and ur) are less than one-fourth the cal-· may have some effect on the measu,·ed pressure
culated values. On the transverse axis, the meas- stresses, but we do not believe that it would
ured longitudinal stress, u x (longitudinal with re- significantly affect the stres.<;es due to external
spect to the shell but circumferential with respect loading. The fillet and the area adjacent to the
to the nozzle) was of the same order of magnitude nozzle-shell juncture were checked with templates,
as the calculated stress, although there is perhaps and it is believed that any deviation in thickness or
an indication that the membrane portion of the local contour is minor and does not constitute an
calculated stress is ''low" and the bending portion explanation for the effects noted.
"high." In the case of the circumferential stress, (b) Ip. testing the model, only one quadrant
u"' (circumferential with respect to the shell), the was completely instrumented.* However, for
calculated stress is significantly lower than the the axes of symmetry, supplementary instrumen-
measured stress, but in this case the maximum tation was installed 180 o opposite the primary in-
measured stress was across the base of the nozzle. strumentation. Alw, for the external loading tests,
The next highest reading was in the shell and would the loads were applied in both the "positive" and
seem to be quite consistent with the calculated "negative" directions, giving a total of four read-
value. ings for each nominal location on the axes of sym-
A.3.2 "Thin-Walled" Model Data
metry and two readings for locations off the axes of
symmetry. In all cases, the critically stressed re-
A summary and comparison of the calculated gion was found to be directly in the fillet at the
and measured data for IITRI Model "C-1" is nozzle-shell juncture. A plot of the stresses along
contained in Table A-4. From this comparison, it this fillet, under the four loading conditions used,
will be noted that -in some instances there are very is shown in Figs. A-2 to A-5 inclusive, based on
large discrepancies between, the calculated and arithmetic averages of the available data. The
measured stress values'. Further, it will be ob- total scatter in the data for a given location slightly
served that under a longitudinal moment loading,
the maximum stress occurs well off the longitudinal '*'See note a.t the enrl of Secliou A.3.2.
Stresses in ~hells 57
Table A-4--Comparlson of Calculated and Measured Stresses in Thin Walled Cylindrical Vessel (IITRI Model C-1) with External
Loads on Nozzles
----Calculated stresses, ksi--~ ~Measured stresses, ksi~
Membrane Bending Total Gross Adjusted
Longitudinal moment (18,000 in.-lb) .
Longitudinal axis-u,. 5.74 0.72 6.46 11.8 11.8
Longitudinal axis-u, 4.02 0.93 4.95 12.6 10.3•
Maximum 6-cr. 15.9 15.9
Maximum 6-u, 29.5 24.15•
Transverse moment (3000 in.-lb)
Transverse axis-.-,. · 0.62 6.54 7.16 23.85 19.5•
Tranaverse axis--.-, 3.97 2.30 6.27 10.5 10.5
Radial load (1000# pull)
Longitudinal axiP-u,. 4.16 1.44 5.60 2.7 2.7
Longitudinal axis-.-, 1.12 3.66 4.78 3.0 2.46•
Transverse axis-.-,. 1.12 3.66 4.78 26.1 21.4a
Transverse axis-u, 4.16 1.44 5.60 1~.3 15.3
NoTES:
• Based on local S.C.F. of 1.22.
b Maximum stresses were located 60 -70° off the longitudinal axis. These stresses are derived from strain gage measurements
oriented radially and circumferentially 'with respect to the nozzle (or the hole in the shell). The maximum principal stress at
this location may be somewhat higher (but by definition cannot be lower) than indicated by these measurements. "• desig-
nates the stress normal to a plane at this section ( = circumferential with respect to the nozzle) and the stress in the plane a.,
of such a section (axial with respect to the nozzle). Therefore, on the longitudinal axis, a. = u., and"' = .-,; on the trans-
verse axis, u. = u, and a, = u.,.
exceeded 10% for th~ worst case (except in low the nozzle; see NoTE (2) in Table A-4] in the region
stressed areas where considerable scatter-· -in per 45-70° off the longitudinal axis under internal
cent--is normal); in most cases, the scatter was pressure and longitudinal moment (Figs. A-2 and
less than 10%. From the plot of the data on Figs. A-3, respectively). The readings off the axes of
A-2 to A-5 inclusive, it will be noted that the symmetry were obtained with two-~lement ro-
consistency of the data is good except for the case settes, readings from the third element having been
of the stress, e1,. [circumferential with respect to discarded because of excessive scatter; the maxi-
! i i: I IIIII i i !'I
i i' I ! ...
j ; II
I
1: i! !! I L Ill:
I : ! ! ;; i ! ' ' 1
;
!!' iI i ii !i
;:li:: !i:llil\i I I : i i !l I lii\1111\!il il
I j
'
Ii
1
i '
li
i
;
I
I t
;: i '
;. •1ti
'!:
"
·1 1
i i i
'
lii
\ ' '!
... _L .. ..
yL f'-t.. •
. I!
u
Ll I! : ;. ll- i
l!i-1
~
·lo·
: '
w·. ·. ·. ·. [i·
Fig. A-2-Measured stresses in fillet of IITRI model C-1 tested Fig. A-3-Measured stresses in outside fillet of IITRI cylind~ical
. under internal pressure (30 psi pressure) shell model C-1 under longitudinal moment loading of 18,000 m-lb
58 Stresses in Shells
.!t ~H:: lit ~nn nn.
ilillh IIil II
++~++1-J-.1+14-++q'
!1.'
·:.
' \ iI \ \'!II '\:; l
\ i i :\
! .. ' ; j i I' I
\ i 1:
II1
)!
!I
u·· . II'
i:11 1\ · :u. r:!I'\, ! t; I !1 H-H~H 'LH-H
I+ lit !II
Fig. A-4-Measured stresses in outside fillet of IITRI cylindrical Fig. A-5-Measured stresses in outside fillet of IITRI cylindrical
shell model C-1 under transverse moment loading of 3000 in-lb shell model C-1 under a radial load (pull) of 1000 lb
mum principal stress at each location may be Q increases, the maximum membrane stress cir-
higher than indicated by present readings, but cumferential to the hole and the maximum stress
cannot be lower (by definition, assuming the· basic on both surfaces all shift off the longitudinal axis.
validity of the data). (NoTE: The basic reason This "shift" develops first on the outside surface,
for the scatter in the third element is believed followed by a shift in the membrane stress and
attributable to difficulty in accurate orientation of finally by a shift in the inside surface stress. In
the 1I 32 in. gages; these readings will be checked each case, as the value of Q increases, the maxi-
using 1 I 16 in. pre-assembled, three-element ro- mum stress increases progressively in magnitude
settes). and also moves progressively farther away from
(c) As a part of the PVRC reinforced openings the longitudinal axis. This is illustrated in typical
program, an effort is being made under the direc- fashion for the membrane stress, as shown on Fig.
tion of Dr. A. C. Eringen to provide an analytical A-6. Stress profiles for all three stresses are shown
solution for the cylinder-to-cylinder intersection on Fig, A-7 for a value of Q closely approximating
problem under internal pressure. The basic theory that of the IITRI Model C-1. It should be empha-
and the first numerical results frbm this work are sized that these present results are for a hole with a
contained in References 25, 26 and 27. Dr. membrane closure (tiT = 0), and that the work is
Eringen has shown that the stresses in such an based on shallow shell theory, which would pre-
opening can be related to a single parameter, Q = sumably limit its validity to diameter ratios in the
(dmiDm)VDmiT. Although present numerical order of 1 /a. Nevertheless, the results may give
results cover only the case of an opening with a qualitative trends for larger diameter ratios and
membrane closure (tiT = 0), these results show they definitely indicate that, for the cases studied
that as the parameter Q increases, the maximum (tiT = 0), an instability or bulge of increasing
membrane stress and the maximum surface stresses severity develops as the value of Q increases.
both shift off the longitudinal axis. Although Although it is unfortunate that similar data are
the numerical results presented in these reports not yet available for finite tjT ratios, it seems ob~
cover values of Q only up to a value of Q = 2.8, vious that attachment of a nozzle will tend to re~
Van Dyke 31 subsequently extended the range of strain this localized deflection and rotation at the
calculated data up to a value of n = ,...__,8.0. The edge of the opening. In such case, there is every
results from this work show that, as the parameter reason to believe that the high circumferential
Stresses in Shells
_,,
_· l1 I! .. IIIJ ·- --
aI
!I 11 1 I ! I!I
I I !\
II 1+11 l!illil I '
It Ii['IH,. ill( 1:,\,·
1;i·r ! 111 I_'
\i
lj- ·lit~if . jl}1
•J
I
LJ -I
-- i
!
R"-~ -:Jll I ~ -l- . ~li [!-·- --
---
-·
'
t - \H l,i " --
..
·,. II
.. ·- -
I
I
.!
Illl·
. !' ~'r
JT=-1
i'li
! ,,I '
I'
Ii ~J'il
i' '
i!lll
1' , .. ,
stress (in relation to the nozzle), would be partially excellent agreement and there is nothing in the data
replaced by an axial stress which should "peak" at which would indicate any serious deficiency in the
essentially the same point. In this respect, then, model.
it should be noted that the form of the curve for the In the case of radial load on the nozzle, prior
axial stress, 0' 1, shown on Fig. A-2, is quite consist- tests on Penn State Models "D," "E" and "R" 21 • 22
ent with the form of the curves on Figs. A-6 and indicated stresses on the transverse axis 3-5 times
A-7 [although the peak of the curve for the model those on the longitudinal axis. In the case of
(Fig. A-2) is farther from the longitudinal axis than Model "C-1," the ratio is 5.7 for ()'~' and 8.7 for
indicated by the calculated data]. In spite of the () x· The qualitative effect is therefore the same,
latter difference, we feel that the calculated data but the difference is seemingly accentuated in the
provides good qualitative evidence of the validity thin shell model.
of the experimental data under internal pressure In the case of moment loading, results from Penn
loading. State Model "R" gave maximum stresses under a
(d) For the external loading conditions, judg- transverse moment approximately 2-2.5 times
ment of the validity of the data must rest in consid- those due to an equal longitudinal moment.
erable measure on the internal consistency of the Calculations based on Bijlaard's curves predicted a
data itself. In this regard, a detailed comparison similar difference, although the absolute values of
of the stresses in the fillet on the axes of symmetry the calculated stresses were somewhat lower than
for the three external loading conditions is con- the measured ones in both cases. For equivalent
tained in Table A-5 for those cases where the stress moments on Model "C-1," the maximum stresses
was large enough to be significant (greater than 2.0 due to a transverse moment are 4-5 times as great
ksi). In each case, the value listed is an average as those due to a longitudinal moment, with the
of readings obtained in the positive and nega- maximum stress. being located 60-70° off the longi-
tive loading directions. For all such cases, agree- tudinal axis in the latter case; for the stress directly
ment between one axis of symmetry and its coun- on the long·itudinal axis, the ratios are 12.1 and 5.0
terpart 180 o opposite is within the range of 3 to for 0''~' and O'x, respectively. The comparative ef-
12%. Considering variations in fillet radius and fects noted in the Penn State and liT models are
difficulty in exact placement of the gages, this is therefore qualitatively similar, with the added fac-
tor of an apparent "instability" or local bulging in for 'Y = 300 (i.e., for values of Q greater than say,
the thin shell model. In the latter connection, it 2.0 - 2.5).
should be noted that the stress pattern for the axial
stress, u z. under longitudinal moment, as shown A.3.3 Modification of Curves
on Fig.- A-3, is remarkably similar to that under
internal pressure, Fig. A-2 (which, as has already Since the experimental data indicates that the
extended range of Bijlaard's data may be in error
been shown, would seem to be qualitatively con-
by factors of as much as 5.0, it seemed necessary to
sistent with calculated data).
On the basis of the foregoing, there seems to be
provide interim "corrections" to Bijlaard's curves
no reason to question the qualitative validity of the until such time as better analytical methods are de-
data.* The reason for the high stresses on or adja- veloped to compute these stresses. Unfortu-
cent to the transverse axis appears to be that, in a nately, we have little basis beyond the experi-
mental data on which to make these "corrections,"
thin shell, the longitudinal axis is relatively flexible
and free to deform, and that the loads are thereby and time may prove that their only virtue is that
transferred to (or toward) the transverse axis which they are in the "safe direction." Because of this
uncertainty, it seems necessary to document the
is less free to deform. On the basis of the data
available, it would seem that a large part of Bij- exact manner in which the curves have been
modified, as outlined in the following paragraphs.
laard's "extended range" data may be unconserva-
tive; further it would appear that the very thin A.3.3.1 REDUCTION OF DATA. In proposing any
shelfregion of his original data 2• 3 may be signifi- modification to Bijlaard's curves, the first problem
cantly low. That is, values of 'Y = 300 and (3 = to be faced is that of placing the experimental and
0.25 would correspond to a value of Q = [(dm!Dm) calculated data on an equivalent basis, which we
V'Dm/T] of approximately 6.95. However, the have endeavored to do by "correcting" the experi-
liT model, with a value of Q only slightly higher mental data for local stress intensification. The
than this, shows a very well developed condition of next step is to try to break the experimental data
instability . under longitudinal moment and a down into membrane and bending components, in
marked shift of the load to the transverse axis un- order to determine the nature ofthe specific modifi-
der the other two loading conditions. On the basis cations required. For both the Penn State and
of the data reviewed above, it is considered that the IITRI models, stress distributions were obtained
original data is open to question at d;/D 1 ratios in both the nozzle and shell on the axes of sym-
greater than 0.15 for 'Y = 100 and greater than 0.10 metry. In the case of the liT model, all such data
can be broken down into its membrane and bending
*NoTE: Subsequently, this high stressed zone was instrumented in the
components except for the. reading directly in the
remaining three quadrants on the mode}, two with I/3 2 in. two-element fillet (the closest "valid" points being 1 I 4 in. from
rosettes, and the other with 1I u in. three-element rosettes. The results are
summarized as follows (each value being an average of readings in the poSitive the fillet, on both the nozzle and shell). Such data
and negative loading directions):
seems to indicate that the membrane components
Tangential Normal of stress as calculated from Bijlaard are relatively
Gage stress, stress, accurate, but that the bending components are
Position length, in. ut, ksi CTn, ksi
60' (original), 11a2 29.5 15.9 sometimes greatly in error. Also, in a number of·
60' (retest) 1 /a: 30.6 17.1
cases, the bending stress is large in relation to the
120' l/32 31.8 18.6
300' \h't. 27.8 15.3 membrane stress; for such cases, large percentage
240' 'h• 24.5 14.1
increases in membrane stress would be quite in-
The t / ts in. three-element rosette confirmed that the stre5Ses measured by effectual in correcting the over-all total. For these
the two-element roBE'ttes were essentially the principal stresses. Although
there i£r some scatter in the data, and the readings obtained with the 1/u in. two reasons, it was decided that the major correc-
rosette were 15-20% lower than the average of those obtained with 1 /32 in.
rosettes, it is apparent that a high stressed zone exists at this location in all
tions should be made to the bending stress curves,
four qt>adrants (materially higher than on the longHudinal mds, itself). although relatively minor changes have been
These results would seem to remove any question concerning an isolated local
deficiency in the model and concerning the qualitative validity of the data. made to the membrane curves in a couple of cases.
Stresses in Shells
1.0 , - - - - - , - - - - - , - - - - , - - -
..
I ,I ! H !
il. -
I I
I I I IIi II
!·: !I ·+' LLfl!
It
j j.
r i,., flil
I .
i
OJO
'. 5
'•15 1
.,~
:I '!:_E
:,;
'-100
0.01
1' •300
fJ-
Fig. A-S-Moment M 1,/(M/R,f3) due to a circumferential
moment M Fig. A-9-Membrane force N¢/(M/R,, 2(j) due to circumferential
moment M
,_
( ' •) S/.rcssrs in Shf'l/s
1.0
0.10
>•5
-r~i
;>; cr
.,. :!5
t
except that no correction of the membrane stress inch away in' both the nozzle and shell gave mem-
was considered warranted. The corrections to the brane stresses approximately 30 higher than the
bending components of the stress are shown on Fig. calculated value; also, simple inspection of the
A-10, amounting to roughly 10 for Model "R" original curves would indicate a possible too-rapid
and a factor of 2.72 for Model "C-1." convergence in the thin-shell region. Under this
circumstance, the membrane curves were adjusted
A.3.3.3 LoNGITUDINAL (IN-PLANE) MoMENT.
upward a commensurate amount, as indicated on
Consideration of corrections required to the c11rves
Fig. A-12. The remaindet· of the required correc-
for longitudinal moment is complicated by the fact
tion was applied to the bending component,
Lhal for the thin shell model, the maximum stresses
which was adjusted upward by a factor of 5.2, as
were off the longitudinal axis. Under the circum-
stances, it was decided to provide two sets of indicated on Fig. A-11.
curves, one applying to the longitudinal axis and (b) Longitudinal Shell Stress, cr"' on the Longitudinal
the other covering the maximum stresses. Ac- Axis: Simila1· corrections were required for the
tually, it would appear that only the maximum longitudinal shell stress, except that no correction
stresses are of interest, since the available data of the membrane curves were considered war-
(Figs. A-2 and A-3) indicate that the stresses due to ranted. The increase in the bending component
internal pressure and longitudinal moment peak of the stress was approximately 66% in the case of
at the same location and must be considered Mode] "H," and a factor of 6. 75 in the case of
additive (at least in the case of the axial stress, Model "C-1," as indicated in Fig. A-1:3.
u,, which is the critical stress). A detailed descrip- (c) .,o/laximum Stresses: The only basis for e-sti-
tion of the corrections follows: mating the progressive divergence of the maximum
(a) Circumferential Shell Stress, u"', on the Longi- stresses from the longitudinal axis is to assume that
tudinal Axis: In the case of Model "H," the re- the effect is similar to that indicated by the calcu-
quired correction was relatively modest, and was lated data under internal pressure (perhaps ad-
applied only to the bending component. This cor- justed to be consistent with Model C-1), which is
rection amounted to about 18 as shown on indicated graphically on Fig. A-14. As a corollary,
Fig. A-11. it was also assumed that the maximum stresses due
In the case of Model "C-1," measured data 1/ 4 to internal pressure and longitudinal moment have
Stresses in Shells
.·~··
.. 1-.
'I'
I
,, t
I! I
64 Stresses in Shells
p__._
Fig. A-15-Moment Mn/(M/Rmfl) due to a longitudinal moment M Fig. A-16-Moment Mtf(M/Rmfl) due to a longitudinal moment M
this procedure does not distinguish between the available for the edges of the attachment are those
values at the edge of the attachment on the obtainable from a cross plot of the curves presented
longitudinal axis of the shell vs. the transverse axis in Reference 10; further, the latter data were for a
of the shell. A summary of the experimental re- value of a = 4 rather than 8, and were limited to
sults in comparison with the calculated (as taken values of {3 no greater than 0.25. Under these cir-
from Table 3 of Reference 17, for Attachment 2) is cumstances, the comparisons of Tables A-3 and -4
shown in Table A-6. From this comparison, it will were made on the basis that Bijlaard's calculated
be noted that the agreement between theory and stresses, for the center of the attachment, apply
experiment was quite good on the transverse axis, at the edge of the attachment on both the longi-
but that the theoretical results were conservative tudinal and transverse axes (even though the avail-
by a factor of, say 1.5-2.0, as applied to the stresses able evidence for a model well within the presumed
on the longitudinal axis. Prof. Cranch therefore limits of Bijlaard's theory indicated that the
suggested that, in the case of the· circumferential calculated stresses might be appreciably conserva-
stress, (]""' on the longitudinal axis, no "shift" in the tive as applied to the longitudinal axis).
stress from the center of the attachment to its edge The test results on the longitudinal axis of both·
is necessary. However, the only calculated data Models "R" and "C-1" indicate that Bijlaard's
Table A-6-Summary of Calculated and Measured Stresses for Cornell Attachment No. 2a under Radial Load (Pull) of 17,700 Lb
(WRC Bull. No. 60, Table 3)
Calculated stresses, ksi Measuredb stresses, ksi
Membrane Bending Total Membrane Bending Total
Transverse axis
5.74 26.4 32.14 3.71 27.6 31.3
""'
<Tx 6.7 17.9 24.6 5.3 16.8 22.1
Longitudinal axis
5.74 26.4 32.14 4.95 10.4 15.35
""'
<Tx 6.7 17 .9<- 24.6 3.18 13.6 16.78
NoTES:
"Model parameters: Dm/T = 78; dt/D 1 = 0.126; t/T = 0.448; (dm/Dm)vDm/T = 1.14; -y = 39.0; fJ = 0.119.
b Tabulated stresses at edge of attachment were obtained by extrapolation of measured values from strain gages located
ll/ 16 in. away (outside the edge of weld fillet).
Stresses in Shells 65
:l
! "\..V. ! l! ; i j i \. ! ! .
•••
I_ )\'1-:•: '' ,,
curves for axial load are appreciably conservative curves is considered feasible, and it can only be
as applied to the stresses on the longitudinal axis, warned that for large values of (dm!Dm)VDm/T
as was the case for Cornell Attachment 2. How- say, above 2.0___:_the curves are probably signifi-
ever, as applied to the stresses on the transverse cantly in error (unconservative).
axis, they are slightly inadequate for thick shells Under these circumstances, two sets of curve::,
(Model "R"), and greatly inadequate for thin are shown: (1) Bijlaard's>~original curves, which are
shells (Model "C-1"); in the latter case, the calcu- considered adequate (or more than adequate) for
lated values were low by a factor of 2.5-3.0 for the stresses on the longitudinal axis, and (2) a sec-
a"x( = un) and about 4.5 for u,.,( = u 1). Further- ond set of curves for application to the transverse
more, a plot of the test results would seem to indi- axis, which have been limited to "small" values of
cate a compression of the curves into a very narrow (dm/ Dm)VDm/T, as indicated in principle by Figs.
band, or (more likely) a "cross-over" of the curves. A-17 and A-18.
Under this circumstance, no "correction" to the
NoTF::
* Bijlaard's treatment of radial loading provided stress resultants at the edges of a rectangular loading surface.
However, experimental data indicated that some of these values might not be adequately conservative. Therefore,
in the interests of conservatism, he then recommended that the calculated stresses for the center of the loading
surface be applied at its edges, both on the longitudinal and transverse axes. However, as noted above, this
procedure does not allow for possible differences in the magnitude of the stresses on the two axes; also, it does not
make any distinction in terms of possible differences in orientation of the maximum stresses on the two axes. In the
latter respect, if one considers the case of a nozzle attached to a flat plate or a "small" nozzle on a cylinder, it should
be apparent that the cr stress on the longitudinal axis is the equivalent of the cr ( = cr<~>) stress on the transverse axis,
both being radial with ~espect to the nozzle. As such, it can be anticipated that these two stresses will be most affected
by the discontinuity between the nozzle and shell (or plate) and will have relatively high bending stresses as
compared to the stresses oriented 90" thereto (circumferential with respect to the nozzle). Ofsix experimental models
currently available, this is true in every case. From Table A-6, it will be noted that the calculated stresses arc
qualitatively consistent with the measured stresses on the transverse axis, but not consistent with those on the
longitudinal axis (neither with respect to the bending stress nor even the membrance stress). This was also true for
two other models having comparable diameter ratios, viz., ORNL-3 (Ret: 32) and Franklin Inst. model "EF" (Ref.
33). In· these cases, the matter seems relatively unimportant, since the calculated stresses are appreciabl~
66 Stresses in Shells
conservative as applied to the longitudinal axis. Other recent data at larger diameter ratios show this same general
inconsistency, but in addition have disclosed rases where the calculated stresses are inadequate, such as for model
ORNL.J (Ref. 34), which had stresses as follows under a radial load (pull) of 300 lbs.:
It will be observed that if the "labels" of the calculated stresses were reversed, the qualitative consistency would be
much improved, and the two cases of"low" calculated stress would then be adequate. This was also found true for
the two other models of relatively large diameter ratio. In light of this, for this March 1979 Revision, the titles on
the curves for radial load (Figs. I C-1, 2C-I, etc.) have been revised to reverse the orientation of the stresses for the
longitudinal axis. Whereas this will make the curves (calculated stresses) adequate or more than adequate for all
presently known cases of shell stress on the longitudinal axis, it does not alter the possibility of inadequacy for very
thin shells and flexiple nozzles, or the fact that stresses in the nozzle can sometimes be considerably higher than in
the shell, particularly when there is little or no reinforcement in the nozzle wall.
Stresses in Shells
67
APPENDIX B-STRESS CONCENTRATION FACTORS FOR STRESSES DUE TO
EXTERNAL LOADS
B.l Introduction and Terminology radius between the two sections of the bar, as illus-
Presently available analytical methods for trated in Fig. R-1. Heywood's curves conform to
stresses in nozzles, shells, etc., do not take into ac- the following formulas:
count the localized stresses in fillets and transitions. Tension case
The following data may prove useful itl performing
a fatigue analysis where such effects must be con-
sidered. In presenting these data; "the following
terminology will be used:
H thickness of thicker section of stepped
bar, in.
h thickness of thinner section of stepped Bending case
~ ~ ~~
HJl2(5.37~- ~]O·::I
bar, in.
r fillet radius between two sections of bar, 1
or between nozzle and shell, in. X
d1
dn
inside diameter of nozzle, in.
outside diameter of nozzle, in. (see Fig.
4.8) r
B-4)
For the case of a plate of "infinite" width (H very
dv outside diameter of attached pipe, in.
large in relation to h), these formulas would re-
(see Fig. B-4)
duce to:
thickness of pipe, in.
t1 thickness of reinforced section of nozzle,
in. (see Fig. B-4) Tension Kr = 1
1
+ --;. ]0. 65 (3)
T thickness of vessel wall, in. [ 5.6 h
o angle of taper between two sections of
bar or nozzle, degrees (see Figs. B-3 Bending Kn = 1 + - ·1 - ]0.85 (4)
and B-4).
Kr stress concentration factor at fillet of a [ 10. 74 ~
stepped bar in tension (as related to
the stress in the thinner member) Upon comparing Heywood's computed curves with
K8 ditto, for a stepped bar in bending (as the corresponding curves from Peterson, 29 it will
related to the surface stress of the be found that the tension curves are quite con-
thinner member) sistent with Peterson's data, but that the bending
Kn stress concentration factor applied to the curves are seemingly somewhat "low" for small
membrane portion of the stress due to values of r/h. The following alternate formula
external nozzle loadings ("tension" provides a curve which is more consistent with
curve on Fig. B-2; see para. B.3.1) Peterson's data (and somewhat conseevative in
Kh stress concentration factor applied to relation to Heywood's data):
[_]--;:JO '80
the bending portion of the stress due
to external nozzle loadings ("bend- K 11 = 1 + (5)
ing" curve on Fig. B-2; see para.
B.3.1) 9.4 h
] {0 stress concentration factor at fillet of
Curves based on formulas (3) and (5) are shown on
stepped bar for case of 0 = 0 o
Fig. B-2, taking Has infinite and taking has equal
/( 1 stress concentration factor for inclined
to 2T when applied to the vessel shell and equal t.o
shoulder
8.2 Stress Concentration Factors
Peterson 29 and Heywood 30 provide a consider-
able amount of data covering the stress concentra-
tion factors for various design problems including
the cases of two-dimensional, stepped bars in both
tension and bending. In both cases, curves are
provided giving stress concentration factors in
terms of the H jh * ratio of the bar and the fillet
-----·------------·----------·- - - - - - -
• 'l'crminnlogy changt~d to nvnicl conflict. with terminology in this text. Fig. B-1-Stepped bar
·}· l-oR)
*Since less data are avaHn.hle for stre~"' concentration fact.orR in Hhafls
t.hnn in bars or plates, Heywood (Heference 30, page UJfi) recommended that
two~dimensional plate data be used. In general, it is believed that such data
nre ~lightly conservative in re1alion to three-dimensional data. Fig. B-4-Nozzle configuration
69
811'1'881'8 in Shells
h
T
T T
70 Stresses in .Shell,s
INTENTIONALLY LEFT BLANK