..

~I -- 1~ I ~__,___

3 9080 02754 3435 AD 491 094

.INELASTIC LOBAR BUCKLING OF CYLINDRI CAL SHELLS

'UNDER EXTERNAL HYDROSTA TIC PRESSURE

Thomas E. Reynolds

Department of the Navy

August 1960

(\ji2
iu81971
I UENGI&EENRG LBRY

Distributed .. ,'to foster, serve

and promote the nation's
economic development
and technological
advancement.'

U.S. DEPARTMENT OF COMMERCE

This document has been approved for pubiic release and sale.
 Ob

C0

**

putC-48 two's. 3-3

 INELASTIC LOBAR BUCKLING OF CYLINDRICAL SHELLS
UNDER EXTERNAL HYDROSTATIC PRESSURE

by

Thomas E. Reynolds

August 1960 Report 1392

S-F013 03-02

1 "
__ -- I- - -;r _____ ___ - - ___ - - _ - . -1 - - a - ms - - . . - - I l- _.WA - L-L- IU . I L l-.
- - _1I-l_
- -I - ~-to-- -_I, -.

TABLE OF CONTENTS
Page
ABSTRACT................ ......... .
...................................................
r 1

INTRODUCTION ................... ......... ........... ....................... 1

ANALYSIS ...............................................................................
2
..........
Plastic Buckling Equations ... ... ................................................................ 9
Elastic Buckling Equation ...................................................
...................... .. ..........
5
Minimization of Expressions for Buckling Pressure................... ................... 6
Determination of Secant and Tangent Moduli............. ................................. 9
Inelastic Buckling ............... .......... ............. ...
.. ...................................
13

EVALUATION OF THEORY WITII EXPERIMENTAL DATA ON

STIFFENED CYLNDERS ...... ..... .. .......................... ........
................................. 15

......................... ........
CONCLUSIONS ...................................................................20

. ...............................
ACK NOWLEDGMENTS .....................................................................
20

APPENDIX - DERIVATION OF BUCKLING EQUATIONS ..................... 21

REFERENCES ................................................................................................... 25

LIST OF FIGURES
Figure I - Buckling-Mode Parameter 0Sas a Function of vMI/L ......................... 8
Figure 2 -Cmparison of Approximate and Exact Methods for
Determining Minimum Buckling Pressure ... 10
...................................
Figure 3 - Typical Stress-Strain Diagram ............................. ....... 11
Figure 4 - Graphical Determination of Buckling Pressure for Two
16
General Classes of Material .........................................................
Figure 5 - Compression Curves for Cylinder Materials.... ............ 18
Figure 6 - Comparison of Theoretical and Experimental Collapse Pressures......... 19

Figure 7 - Coordinate System and Cross Section of Stiffened Shell ..........................

22

LIST OF TABLES
Table 1 - 'roperties of Cylinders .. ................... 17

) Table 2 - Comparison of Theoretical and Experimental

Collapse Pressures.................. ..-. ................ 17

s..'
 NOTATION
A Arbitrary constant

AI Plasticity coefficients
Area of cross section of ring

Small dimensionless quantity

Faying width of frame

Bendng rigidity, £,A 3 /12 (1 - v 2)

Young's modulus
8, Secant modulus
E,
Tangent modulus

F (1 au to)

f Stress ratio, o/,

A Shell thickness
k n/R
L. Unsupported length of cylinder
t, a Integers
N (cosh O- cos 0)/(sinh 0 + sin 0)

N5, -N, N5* Forces per unit length

p Pressure

p. Elastic buckling pressure

Pe Inelastic buckling pressure

pp Plastic buckling pressure

a
S.
Radius of cylinder
8r
Circumferential coordinate
Shell displacements
 I~~
- ~1
-:-( ~YII*
btYLLI
IOYCI
~L4~ ~QI~WLI

* Axial coordinate

* (3/o#) (1 - E,/E,)
S.
0.0
0 (sinh 0 + sin 0)/(cosh 0 - cos 0)
*-
0
sinh - + sin -
2 2J
oh-
2
0 1- cos
2
y Shear strain V
eg, es Membrane strains

6 Strain intensity

9 (L//7IA) [3(1 - yi2)]

. m/L

v Poisson's ratio

VO Elastic value of Poisson's ratio

pR
Oa; x =2 Circumferential and axial stresses

ag Stress intensity

* /(Xt 2 + k 2 )

' Operator, C

V' (pV4)2
 ABSTRACT
A solution to Gerard's differential equations for plastic buckling of
cylindrical shells is found for the case of lobar buckling under hydrostatic
pressure. An approximate formula based on this solution is then obtained
for buckling in the inelastic region.
According to this formula, the buckling pressure is a function of the
cylinder geometry and the.secant and tangent moduli as determined from a
stress-strain intensity diagran for the shell material. Agreement with
experiments on ring.stiffened cylinders is found to be within 4 percent.

INTRODUCTION
Experimental studies of the buckling of stiffened cylindors under hydrostatic pressure
have shown that collapse of the shell plating.between frames is frequently preceded by yield.
ing of the shell material. This would indicate that inelastic shell buckling inay be an impor-
tant consideration in the strength design of pressure vessels, particularly when it is realized
that residual welding and rolling stresses often induce inelastic behavior at pressures well
below the design strength.
Inelastic buckling of cylindrical shells induced by external hydrostatic pressure can
take place in two basic modes: axisymmetric buckling, during which circumferential corruga.
tions develop along the axis, and asymmetric or lobar buckling, whereby inward and outward
lobes appear alternately around the circumference. Buckling of the first type has received
considerable attention, but it has usually been treated as a failure due to yielding rather than
a buckling phenomenon. Typically, an analysis is based on the concept of an ideal material
which, at a certain stress level, undergoes an abrupt transition to the perfectly plastic state.
The buckling pressure is determined not from stability considerations but from the state.of.
equilibrium stresses. Attempts to describe buckling of the. second type have been rather
limited and have usually depended on the intuitive use of a reduced modulus in place of
Young's modulus in the elastic buckling equation.
In recent years, however, advances in plasticity theory have made it possible to ap.
proach these problems more rigorously. Investigations by Bijlaard, 1 2
Ilyushin, 3 and
4
Stowell, among others, have conutributed greatly to the development of theory for the inelas.
tic buckling of plates and shells. More recently, Gerard s derived a general set of differential
equations for cylinders from which he obtained approximate solutions for torsional buckling
and axisymmetric buckling under axial compression for a strain-hardening material.
Lunchick 6 has since developed a similar, but more exact, theory for axisymmetric buckling

tRftereaces ,a Usted oa P0* 25.

 -~~~~~~- Lr ~ -- -~l- IL_ 1 -1-Y

of ring-stiffened cylinders under hydrostatic pressure. From recent tests*conducted at the

David Taylor Model Basin, this theory appears to be very reliable.
At the same time, it appeared that the work of Gerard might also provide'a worthwhile
approach to the problem of asymmetric or lobar buckling in ring-stiffened cylinders, since his
differential equations are sufficiently general to account for asymmetric deformations under
hydrostatic loading. Work was subsequently initiated to obtain a solution to Gerard's equa-
tions for the asymmetric problem. In this report the solution is derived, and an approximate
formula for inelastic buckling is obtained. Experimental data from ring-stiffened cylinders
are used to evaluate the formula.

ANALYSIS
PLASTIC BUCKLING EQUATIONS
In the Appendix of this r~i."t, the general differential equations of Gerards ror a fully
plastic cylinder are specialized for the case of hydrostatic pressure loading. The buckling
equations thereby obtained are:

ES I aU a2V 1 dwl 3 a2U 1 a's 1 a w

.- - +- + - - +- - + - - + - - = [1a]
2 2
2 2 az :CIs R C1 4 Jz 4 a *4 a.as

a2 1 d" 1 a2 1 iaS2
+-. 4 +- 4 -0 (lb]
, I aza
is 2
R da 4 aX2 4 faas

4
4EsA IE, (1 au a. E 1 a + ----
ao
+
a4o 3 a W
-+--+
a4w

3R Es 1 2 a a R E, 4 az 4
a 2
as 2
a 4
/ 4 r ax
Jar J

a2to N2 -
+ N, - ' N, + p-O0 tid
aZ2

where z and a are respectively the axial and circumferential coordinates,

are the axial, tangential, and radial displacements,
E, and E, are the secant and'tangent moduli,
EV~E
is the radius to the shell midsurface,
is the shell thickness,
is Poisson's ratio,
 is the bending rigidity - Eh 3 /1(1 - v 2),
N, and N, are forces per unit length in the axial and circumferential directions, and
is the hydrostatic pressure.
Since one function recurs in all three equations, let

Et1 du dae
- +a
-s
E2 az as R

With several differentiations, u and v can be eliminated from -Equations (lal and [lb] so that
one equation relating F and o results:

E 4F 8 F d'4F a4F 3 Et a'4

+ 8 - +4 - + 3
2 2 4
E, z az 8s a8 az R E, aZ:

With suitable differentiations, Equation flc) can be combined with Equation [3] so that a.
single eight-order equation in w is obtained:

EE 4(3 a4w 3 Os, as,

Df-
p~~
4 (ES ,
V + 1-E
E,
1V
'2
- -
Z
++-
_4_

aX2d,?I
+
-

4
r
*S
LE 4 ax
+ Os*aa2
-za

EA a4'W a 2W
+ -- +N V - 3( E, )O j
R2 aOZ Oz2 4 £2 a6

a2o 3 E dew
S
SN, - - ]+ [41
a2 4. Et a24a 2

where V4 indicates the operator + -

L 2
as?
A solution to this equation can be writton:

Wt - A sin ks sin Xz
where k -

*A %"o

1 is the length of the shell, and

a sad a are Integers.
 ---i~d
I~L-Z(SL
.IYLi; C Y 1 -~-_ - .-- ~~
_h
iP~-i-L
i~ ma- -~- --

This solution satisfies the conditions of simple support at the ends of the cylinder; i.e., that
a2to
-to and - vanish at z - 0 and a mL. These conditions are not unreasonable for stiffened
az2
cylinders since it is likely that the effective rotational restraint will be limited by the forma.
tion of plastic regions arising from high bending stresses near stiffehers or end supports.
By substituting the solution [5] into the differential Equation [41, the following
characteristic-value equation is obtained: .

+ A PR(k + + 3 (£*-1) A6 + k (k2 + A2)2

R2 2 4 £p

+- -1 =0 [61

To simplify this equation the following substitutions are made:

jA2 1
W n 2L 2
P2+k z +

'E

The equation is then rearranged so that an expression for p,, the plastic buckling pressure,
is obtained:

D+C,[
2fDA1 1+ +3 1- ±+ 2

R# [1 - (1- [1 1 + 3

The smcuisopt fimcates

re th atss upr fe, the plastic egiom.
ELASTIC BUCKLING EQUATIONS
A similar procedure is followed in obtaining a solution for the elastic case. When
is set equal to unity, Equation [4) reduces to
ES

D 8
uEh a4. a2 = 0
D + - + ' (N + Ns ( 91
R2 a24 d) 2

which is the Donnell equation s 7 for the case of hydrostatic pressure loading. The bending
rigidity is now given by

EA3
[101
18(1- v)

where ve is the elastic value of Poisson's ratio. Substitution of solution [5] into Equation
(9] yields*

-I-+ -a
EA 12(1- 2) R12)
[11]

where f, is the stress ratio for the elastic region. This equation could have been obtained
directly from Equation [8] by using the elastic value for D.

*It is of some interest to compare this result with a similar equation obtained by Von Sanden and TlkeB in
their comprehensive study of stability problems in thin cylindrical shells. In considering the elastic buckling of
a ring-stiffened cylinder, they allow for the variability of the pre-buckling circumferential stress with the axial
coordinate. Their buckling pressure equation, in the tenninology of this report is:

2EA
1
L 12(- ) R A2
P. 1
i
+-
4"

PR
where (as) - Pt mdbay circumferential stress
2
PR
(eg) . 8 - - circumferential stress at a frame

esd , aMnd are determined from the theory of Von Sanden and Gunther.

If W does not very with x. then

ad the bucklInt ptessrtw equation reduces to Equation [11

MINIMIZATION OF EXPRESSIONS FOR BUCKLING PRESSURE
Equation [8] expresses p., the plastic buckling pressure, in terms of n, the number of
circumferential buckling waves, and m, the number of longitudinal half waves. The buckling
pressure can now be obtained in minimized form. It can be seen that pp is effectively
minimized with respect to n/m by setting

- 0 .[121]

The resulting equation defines the value of lp for which pp is a minimum:

44 r 1- 2, (1-,)+ 4e
CP 9 L3CC

t -4 (1-f, +f, -3-6402 (1-,) .

x 1 + 413C [
1-20 (1-f,) + 4

The corresponding equation for the minimized plastic buckling pressure is:

81 22 Efp
84 A f -- '-!
rl--&
1 + C4,2
4

3 --0 (I-f )+
3C3+O
?4
[14)

which is obtained through suitable combination of Equations [8] and [131. Since p, in
Equation ([4 is proportional to m 2 ,it is clear that for the minimum value of pa must be
equal to one in all cases where n is greater than 0.
The corresponding minimized equations can be obtained for the elastic case. After
minimizing p, with respect to 4b in Equation (11], one obtains the equivalents of Equations
[131 and [14] for the elastic case:

# 44 4 1 - 24 (1-fe)
1(V-2, fj
3-20, (-
 22 mE4
f A 2 ,17Fr 1 .
(1-, ) - 4(1-f)

These equations could also have been obtained directly fro,. Equations [13] and [14). Again
it is seen that m imust be equal to 1. For the case where f, is equal to % corresponding to the
prebuckling state of stress in an unstiffened tube, it can be shown that Equations [15] and
(16] are exactly those given by Windenburg and Trilling (Equations 210] and [21] of Reference
10) ri expressing the Von %lisesbuckling pressure in minimized forni.
Although Equations [141 and [161 are relatively simple in form, they contain the func-
tion 0 which is not readily determinable from Equations [13] and (15]. However, it is possi-
ble to obtain an approximate expression for 6 from a graphical representation of these
equations. Figure 1 shows plots of 0 versus '7W/L (with m equal to one) for the following
cases:
E, %0r
Elastic: -i 1 v e -0.3

1t f 1p

Plastic: 2 2 f

The value of : for E.'Es was chosen as a typical case for the plastic region. The two stress
ratios, ,' and 1, are extreme values which should bound all cases of practical interest. The
curves terminate at the line where 0 is equal to 1, since this is the case of axisyirunetric
(n - 0) buckling for which the minimized pressure expressions no longer have meaning. These
curves suggest that a simple linear relationship between cS and V/7JL might serve as an
adequate approximation for all cases. Use of such an approximation implies that the number
of circumferential lobes is independent of the material properties. After some investigation,
the equation

- 1.23- (17
L

represented by the dotted line in Figure 1, waS chosen as a reasonably good. approximation.*
It is seen that this line falls roughly midway between the extremes of the curves p'esented.
Use of Equation [171 in conjunction with Equations (14] and [16] then provides approximate
expressions for p and pe. It is also helpful to make an additional approximation which

*Actually, the selection of the factor 1.23 was somewhat athbitrary since the buckling pressure is relatively
nsensitive to this parameter so long as it falls between LOa'nd 1.5. In following an equivalent procedure
for minimizing the Von Mises buckling pressure, Windenburg and Trilling1 0 obtained the value 1.265.
 A - -4. 1- _.- _--
- .. .~-- = Y-LI-jhrI Ii-=- I_~I ILYLi
IL)i

0.9

Ole I FzsS

0.7

0.6-
0.$

lo 0.5 /0,

0.5

0.2
0.3 ,

------- Exoct (Equtions 13 and 15)

----- Approximat (EquCtion 17)

0 O.a 0.2 0.3 0.4 0.5 0.6 0. 0.8 0.s

Figuro 1 - Buckling-Mode Pararnetr 0 as a Function of /RA/L

has the limiting value I corresponding to a it0 which defines axisymmetric backling.

simplifies Equation [14) for the plastic buckling pressure. That equation can be rearranged
so that

8, 2Er A) 2 2~
2 3::+
1+
- )
S4 (1-8)
where

2 3 (1-) +191

The subscript on 0p has been dropped, since the single function 0 is to be used in both the
elastic and plastic regions. It can be seen that a will take on its inaximuma value when 0 is
equal to 1. For the case E /E, = !, a is 0.091 when fp is ji, and 0 when fp is 1. Thus for
all cases where E,/E ~, O 0 < a 0.091. Since E,/'E will seldom be much smaller than !
whereas . will always be less than 1, the approximation that a can be neglected will intro.
duce only small errors. Equation [18] is thus reduced to

8r 2Etfp 2 (7, 2 1 +-

PT 9I1 L 3-2(2 1 [201

To examine the accuracy of the equations thus obtained, the results of the approximate equa.
tions are compared with those of the exact equations in Figure 2, where (s is plotted
as a function of '/7M/L, p being the theoretical buckling pressure. The solid curves represent
the exact Equations [141 and (16] for the plastic and elastic cases, respectively, with the
corresponding values of 6 determined from the exact curves of Figure 1. The corresponding
approximate results, indicated in Figure 2 by the dotted curves, are obtained from Equations
(20) and [161 using the approximate expression [171 for 0. It is seen that the approximate
method of calculation agrees quite closely with the exact method, even though Figure 1 shows
wide divergence between the approximate and exact values for 0, particularly in the upper
range of I /L.

DETERMINATION OF SECANT AND TANGENT MODULI

Before Equation [201 can be used, E, and E, must be related to the applied pressure.
The secant and tangent moduli

E, a

doi
d

are defined in the Appendix and are shown graphically in Figure 3. For hydrostatic pressure
Cl~k)i~L~'~Z~rU~~
L~YYI ~YI--Y

CIS I S J
'AAo zz.0

N S-- sn -- SC-- ° 0 ' o as

I.

0o 0 0 04 0. 0. 0.? 0. 0.)

Figuru 2 - Comparison of Approximate and Exact Methods

for Determining Minimum Buckling Pressure

loading, and with Pois;On's ratio equal to j, the stress and strain intensities are
S(o2 + F, ,)

2
T (12 + # 2 +

Thu cbaracteristic stress-strain curve of the shell material is first obtained from uni-
axial cohl)rOs s i o n te.ts. In this case o i and ej are identical with the axial stress and strain
(regardlou' of the value of Poisson's ratio). Hence E, and E -are readily determined from the
stress.strain curve. In practice, it is convenient to determine E by drawing tangents to the
curve.
Monion should he made of one difficulty which may be encountered in the interpreta.
tion o( tLo nxiP cae
tid col~rasln
o an ratr.
10 riareMenica
I
itressstrain data. Conventional strain-measuring wi~ th axal tres ad srai
equipment such as an automatic
recorling extonsometar, although
~pg~/~Ua n Lbfor
adequate
mio~ measuring
vllO
Hene yield strength,
o 6Posso's
andE~ae rrdiy may not be suffi. th
deermaedIr~
ciently gcourato for the dotermination
sues-elrlncure.
t i of Young's
c~venent modulus.
n prctie,
p dtetrin Unless
Erby aQwin
high-precision
tsgeos device
t th
 Figure 3 - Typical StressStrain Diagram I

I I I

/ \f

is employed, it is best to obtain only the relative values E,/E and Er/E from the stress-strain
data and assulne a standard value for E.
Having deternminod Es/E and E,,/E as functions of ai, one must then apply them to the
hydrostatically loaded cylindrical shell. According to a fundamental hypothesis of plasticity
theory, the stress and strain intensities are uniquely defined. 4 Thus by expressing hydrostatic
pressure in terms of the stress intensity, a relationship between E, Et, and pressure will be
established. Since equilibrium requires that a x be equal to pR,/2h, only a s in Equation [20]
remains to be determined. As discussed in the Appendix, a, is actually a continuously vary.
ing function of z, whereas in this theory a, is treated as a constant. Thus a single value of
a. must be chosen, and it is taken to be the stress occurring midway between framhes. Since
12 in the shell, itI might be. regarded as a
this is, generally, the lmaximumn membrane stress
conservative choice, Hlowever, it should be noted that for a material exhibiting
E . .
a plateau-
type stress-strain curve, any other choice would probably overestimate the strength of the
is bet.eploedit
o otai s onl th reatie vaues£,/ an £/ frmth trscsri
shell.
In calculating o s it is particularly useful to make the further simplifying assumption
that as is proportional to the applied pressure in both the elastic and plastic regions. This
assumption is reasonable provided the deflections of the shell remain small compared with
its thickness. Then a. can be determined completely from the theory of Von Sanden and
Gunther9 with v equal to .:
I_ ______ _ __ ___ __ ____~___ __I _ _ I _____ __ __ _

Al i1
0.75 A- 23

1+
2 * A
where
Ssinh 0 + sinG
\ cosh O -Cos lo

P o,
sinb- + sin
2 2
W' 2241 [
(cosh -
2
- cos
2

p (2.25)%

Al is the cross-sectional area of the frame, and b is the faying width of the frame. The sub.
script p indicates that all functions are given for Poisson's ratio equal to ,.
It will be observed that the "beam-column" effect, dcmonstrated theoretically by
Salenjo and Pulos,tl is ignored in the assumption of proportional loading. This effect causes
a departure from proportional loading ia the elastic region. However, this departure is ordi.
narily small and, in view of the approximations already made, -to account for it would be an
unnecessary refinement. In those cases where the effect is large it can easily be included in
the value assumed for a.*
The stress ratio for Poisson's ratio equal to $j is then given by

p 0.75 A 1 25)

1-
1- P

*Additional departures from propoettonsl loading amw ehibited by cylindes whose generators are not initially
straight. This effect can be computed from a analysis by Lachick and Shot 1

12
Solving Equation [221 for the applied pressure, one obtains

20iAfp
p.* [S6]
Rft/ - fp + 1

A plot of p versus ai from this equation is a straight line for the case of proportional loading
but becomes a curve if the aforementioned nonlinear effects are included.

INELASTIC BUCKLING
Although Equations (16], [201, and [261 define the buckling pressure for the elastic
and the fully plastic regions, no solution is given for the inelastic region which lies between
thes two limiting cases. However, by employing an empirical correction factor wherein
Poisson's ratio is regarded as a variable, one can arrive at an expression which reduces to
the proper limiting values. Gorard and Wildhorn U have found that a can be accurately ex-
pressed as a function of E, in the inelastic region by the equation

v = -2 - --E 2 - v . [27]

which reduces to % when E,/E is zero and to ve when s,/E is one. Since Equation [20] is
for the fully plastic case where v has the value !., it could be written

3C6

)228]
1"

If ' is now replaced by v, a variable defined by Equation (27] and p is replaced by f, a

function of v, the equation for Pc, the buckling pressure in the inelastic region, is

1+ - -
Sf )2E,( )2 4 E
36(1- &) R L 3 - 20(1-f)

With qb given by Equation (171, p, reduces to pp when E,/E is zero and to p,, given by
Equation [161, when Es/E is one. From Equation (29] it can be seen that the inelastic
buckling pressure depends on both the tangeot and secant moduli.

13 *
II___I~ _~ _1_1 __

In determining the stress ratio f,os is again taken to be the stress midway between
frames as given by the theory of Von Sanden and Gunther, 9 but with v a variable defined by
Equation [271. It is found, however, that o, is practically insensitive to variations in v and
that it is sufficient to treat f as a constant which depends only on the geometry of the cylinder.
This is the sane as assuming that oa is proportional to the applied pressure. Since it has
been found in practice that variations in v are small, v, can be used for determininin the stress
ratio. In this way p, will still reduce to P, when E,'E is equal to one, and Equation [291
can be written

Pc P. 34 34 Es
PPC a I + E 301

where

(1'- )R 1 3-2 (-fe)

-l1.23 L17

0.5
Se. [al

1 A,+ 6A

and

sin e,
A ( sinh 0, +
cosh0, - os 0 ,

sink - sin

e- s cn - o.

The subscript e designates functions based on the elastic value of Poisson's ratio.
 The inelastic buckling pressure P. can now be determined as follows: from the charac-
teristic stress-strain curve of the shell material, E, /E and Et/Z are defined in terms of ai,
the stress intensity. Hence Pc can be plotted as a function of ri using Equation [301.
Similarly, the applied pressure p can be plotted against oa from Equation [26. Since Equation
[30] is valid only when Pc and p are equal, the buckling pressure is obtained from the
intersection of the two plots.
Figure 4 illustrates the two general types of material encountered in practice. The
first is the strain-hardening type (Figure 4a) vhich exhibits a continuous stress-strain curve.
Plots of Equations ([6] and [30] are shown in Figure 4b, where the buckling pressure is
defined by the intersection of the two curves. Figure 4c illustrates the case of an elastic-
perfectly-plastic material. Since the buckling pressure abruptly drops to zero when the plastic
region is reached, collapse is simply defined at the elastic limit of the material, as shown in
Figure 4d.

EVALUATION OF THEORY WITH EXPERIMENTAL DATA

ON STIFFENED CYLINDERS

To evaluate the theory, results were examined from previous tests of seven cylinders
in which asymmetric (lobar) shell failures were observed. The properties of these cylinders,
all of which had external stiffeners, are listed in Table 1. Two were machined from seamless
"teel tubing and had rectangular stiffeners. The other five had T-stiffeners and were rolled
and welded from U.S. Steel Carilloy steel pla:e. As indicated by the symbol (c) in Table 1,
in some cases specimens were taken from the collapsed cylinder, whereas in others (unmarked)
they were taken prior to fabrication. Data from uniaxial compression tests of these specimens
were used for the measurement of yield strength and for the determination of the secant and
tangent moduli. Since all data were obtained with an automatic recording extensometer, which
was not sufficiently precise for the absolute determination of Young's modulus, only the rela.-
tive values , /E and E,/E were determined from these data. For all cylinders a standard
value of 30 106 psi was used as Young's :.odulus E. Plots of oaversus Cifor the seven
x
cylinders are shown in Figure 5.
In Table 2 the experimental collapse pressures are compared with the inelastic buck-
ling pressures calculated from the theory of this report (Equation [30]). This information is
also presented graphically in Figure 6, where the ratio of theoretical pressure to the elastic
buckling pressure Pe (Equation (16]) is plotted against the ratio of experimental pressure to
Po*
Table 2 also lists failure pressures predicted by several other criteria. Elastic buck-
ling pressures were calculated from Equation (161 of this report, from the theory o(Von Sanden
and T5lke,s and from the theory of Von Mises using the approximate EMhB formula (101. to
Shell failure pressures are given for Von Sanden and Gunther formula [92a] 9 (based on simple
yielding of the exterior shell fiber. at midbay) and for the llencky-Von Mises criterion applied
 V I 1/ '
4E..
Strain - Hordening

Qauofion 30 .

PP

Elostic - Perfectly Plastic

1lur. 4e Flg=* 44

Figure 4 - Graphical Determins.zaon of Buckling Pressure for 7wo General Classes of Material
18

I
 TABLE I

Properties of Cylinders

Unsupported
Radiuss Shell LenjUl Frame Frame Yield
Cylinder Thickness- of Shell Area Faying Width Strength
A L A 6b '
in. . iin. in. sq in. in. psi
T- 2 38.87 0.264 .7.24 1.885 0.260 88,000
(c)
T- 3 38.87 0.260 8.74 1.625 0.260 108,000
Welded (C)
wrth T-6 26.87 0.256 7.24 L170 0.760 115,000
T.Franes (e)
T- 2A 38.87 0.254 7.24 0.196 0.260 103,000
T- lA 26.87 0.263 8.74 0.683 0.260 84,000

Machined
wit; U 12 4.081 0.0446 ' 1.06 0.0248 0.079 61,000
Rectn ular U-22 4.071 0.0356 0.834 0.0185 0.068 70,500
Frames

*All yield strengths are defined at offset strain of 0.002

(c) Spetjr.ens of shell material taken from collapsed cylander.
Specimens for altother cylshders were obtained prior to.fabricatie.

TABLE 2

Comparison of Theoretical and Experimental Collapse Pressures

Welded 'fjchined
Cylinder I4umber T-2 T-3 T-6 T-2A T-7A U-12 1U-22
Experinmental Collapse 553 1005 680 70 95 735
Pressure 670 553 1005 680 770 975 735

Inelatic
8ucklinS Equation [301 (PC) 696 563 1016 705 748 938 734

Von Sanden and T6Ike 8 930 631 1258 773 1032 2014 1054
Elastic Equation [1] (p,) 906 626 1259 756 1010 1907 1002
Duckling Van Mises l o
(EMB FormUla [101) 786 585 1180 705 995 1786 963

Hencky-Von Mises
Shell midbay, midplane 903 953 1429 912 978 1081 777
Yield Von Sanden and Gunther
Formula [92A19 695 742 1121 733 799 788 622
U,

0I-
:
0

E
E
00

0 2000 4000 0 2000 4000 0 2000 4000- 6000 8000

Strain Intensity, i in Pi/in./in.
(All Strains Corrected on Basis of E 30 x 10 ps)

120

100

so

60

40

,/ T-7A . U-t2

4000 0 2000 4000 0 2000 0

Strain Intensity, in P/in./in.
(All Strains Corrected on Basis of E 30 x 1t psi)

Figure 5 - Compresstion Curves fot Cylinder Materials

(0) MI teald tmake from sepsed cylltndw
Axis of load colreeponds to alrmferentid drectio aIn
cyIadrW

18
 09

08

0.6

0.5

0.

0.3 0.4 0.5 0.6 0.7 08 0.9 1.0

xi.erimentoI Pressure / e

Figure 6 - Comparison of Theoretical and Ev.periniental Collapse Pressures

Pc = Inelastic Duck!ng Pressure (Equation 130)
pe WElastic fuckling Pressure (Equation [161)

to the membrane stress at midhay, wherein failure is assumed to occur when the btress intern
sity, al, computed from the theory of Von Sanden and Gunther, reaches the yield strength of
the shell material.
From Table 2 and Figure 6, it can be seen that Equation [301 is in close agreement
(within 4 percent) for all cylinders reported. In view of the many approximations contained
in the theory, this agreement is considered surprisingly good. It is also noted that for all
welded cylinders except T-TA the.press.ures according to Equation [301 are higher than those
observed. This is to be expected, since such cylinders are weakened by residual stresses
and geometrical imperfections introduced during fabrication, none of which are accounted for
in the theory. However, good correlation in these cases indicates that such weal. ning
_ _ ~_ _ _~ _.P _L-~- I~UII
-- ~LII LIYII~- - ~-- ll~
--L1I- --

offects may not be as severe as had previously been vuspected. From the uniformity of the
results there also appears to be no significant difference for the limited available test results
between those cases where test specimens were taken before fabrication and those where they
were taken from the collapsed cylinder.

CONCLUSIONS

S1. The theory of this report, on the basis of the limited test data available, appears to
predict the inelastic (lobar) buckling pressure with good accuracy.
2. Final evaluation of the theory must await additional experimental data, which should
include tests of cylinders with internal frames.

ACKNOWLEDGMENTS

This work was initiated at the suggestion of Mr. John G. Pulos and has proceeded
under his general guidance. The author is greatly indebted to Dr. Myron E. Lunchick, who
provided valuable suggestions and advice. Thanks are also due Mr. John E. Buhl, who
supplied the experimental data, and to Mr. Abner R. Willner, from whom the stress-strain
measurements were obtained.
 APPENDIX
DERIVATION OF BUCKLING EQUATIONS

In deriving general plast' "ty equations for cylinders, Gerard s defines stress and strain
intensities for Poisson's ratio equal to ' according to the octahedral shear law:

O, . (o of o, +32
+ %
[331

The secant and tangent moduli are then defined as

d,. [34]
d'ac"
-

With the assumption that Poisson's ratio v is eqtal to ! in both the Olastic and plastic
region, the stress-strain relations become

s --
E as 2
w 1f Q -- ox
[35)

y M

The subscript s replaces y in Gerard's notation and refers to the tangential direction, as
shown in Figure 7.
In treating the general case, Gerard considers a cylinder subjected to external loads
Nx, N,, and N., per unit width and externil pressure p.. The general differential equations
of equilibrium are:

2
A
Oa A 13 a u A3 a'u A 13 a 2s. A 3 A' 02
a7z 2 4 c* 4 aR
O 2 4 czdc

A23 a2. A13 det A23 de

+~ - 0
4 4a' 2R as 4R a.
0

21
__ I ____

Figure 7 - Coordinate System and

O Section ot Stftened Shell
Cau

I ~t
)-----t ------

A2 3 a2 s 43 j2p A 13 a 2 u A2 A31
A2
a. 2 2 aza. 4 aa2 4 az 2 +''I' -
.[3s]
Az2 3 2
A2 ad A 2 3 dt
4 +- - - WO
4 as2 R a. 41R az

d'2 a4W
D A- - 1113 - (A12 + A ) A23 - + A2-
a:ZG. 3
3
az 3 da da st . Cs4

4,A (A au A2 3 adu A2,d

+ - -
12

a.
4 - a4. - +A2e
3R .2 dr 4 &a .
oa2' a2 ' a2to .
+ Nx - + 2Ns,- + N,.- +p 0
aX . a.a da2

In this analysis, long cylinders which buckle in the oval (n - 2) mode will be excluded from
consideration, -since the differential equations are deriied with certain approximations which
are valid only when a 2>>L
4q
I

I
 The-ba.di -- rr dity for v equal to % is

0- 9
[81']

Gerard defines the plasticity coefficients as follows:

ao 2
A 1 ------

A2 = 4

AS i -a1r2
[381

Az= Al

A 3 1 - A13 =aQ T

A32 A23 -

where
a=--- 1- [391
o El
in the case of uniform hydrostatic pressure, it is readily seen that

NX,

(40]
N pR
A 2A

On the other hand, o, is not so easily disposed of. La deriving his equations, Gerard con-
* siders N, as independent of z, -an assumption which is correct in the two cases (axial
compression and torsion) for which he obtained'solutions, but one which is clearly not
correct in the case of hydrostatic pressure if the ends of the cylinder are restrained.
However, without such an assumption it would be difEcult, if not impossible, to solve the
problem by moans of differential equations. Furthermorce, in order to simplify the plasticity
coefficients, the additional assumption is made that
N, pH
* *
,A
*- AI"
[41]
 whereby the stress intensity becomes

1421
and the resulting plasticity coefficients are

A Et
-4 4E.

A2 f A2 1 A= 2 -
[431

A s s W A 3 1 - A2M A32 - 0

Although it would logically follow that N s in Equation (361 should be replaced by pR, this
substitution is not necessary for the solution of the differential equations. Instead, by -
retaining N, as an arbitrary factor, a more general solution can be obtained. The load N s
can later be determined from appropriate theory, as described in the body of this report. The
resulting differential equations are:

-- + -- +-- 1 4- a2-v +-1 -+-

2 1 - WO (lb)
0
2E,\ 2 az2 aaa R da 4

Eg 1a 2u
s a2 1 at 1
. +2- 1++ 2
[Ib)
a.- deas Ra8 4' a 4 a dsd
I

4r Es )(2 a. toW
+- + - +RD 4 aZ4 a. 2 aS2
t
£.
1i
a2W
+-+- a4l 4 " + Ns -
a4+- s a
4 t
a4a at2 4,d
r
+ N, -- +p O [lc]
de5

0
 REFERENCES
1. Bijaard, P.P., "Theory and Tests on'the Plastic Stability of Plates and Shells,"
Journal Aeronautical Science, Vol. 16, No. 9 (Sep 1949).
2. Bijlaard, P.P,, "On the Plastic Stability of Thin Plates and Shells," Verb. Koninklijke
Nederlandsche Akademie van Wetenschappen, Vol. L, No. 7 (Sop 1947).
3. Ilyushin, A.A., "The Elasto-plastic Stability of Plates," National Advisory Ceonittee
for Aeronautics, Technical Memorandum 11SS (Dec 194T).
4. Stowell, E.Z., "A Unified Theory of Plastic Buckling of Columns and P' ,* s,"
National Advisory Committee for Aeronautics Report 898 (1948).
5. Gerard, G., "Compressive and Torsional Buckling of Thin-wall Cylinders in Yield
Region," National Advisory Committee for Aeronautics Technical Not' 3720 (Aug 1956).
6. Lunchrk, M.E., "Plastic Axisymmetric Buckling of Ring-Stiffened Cylindrical Shells
Fabricated from Strain-lardening Materials and Subjected to External klydrostatic Pressure,"
David Taylor Model Basin Report to be published.
T. Donnell, L.I., "Stability of thin-walled Tubes under Torsion," National Advisory
Committee for Aeronautics Technical Report 479 (1933).
8. Von Sanden, K. and Thlke, F., "On Stability Problems in Thin Cylindrical Shells"
(iiber Stabilititsprobleme Dinner, Kreiszylindrischer Schalen), Ingenieur.Archiv, Vol. 3,
pp. 24-66 (1332), David Taylor Model Basin Translation 33 (Dec 1949).
9. Von Sanden, K. and Gunther, K., "The Strength of Cylindrical Shells, Stiffened by '
Frames and Bulkheads, under Uniform External Pressure on All Sides," (t'ber das
Festigkeitsproblem. guerversteifter 1Hohlzylinder unter allseitig gleichmissigem Aussendruck),
Werft-Reederei-flafen, Vol. 1 (1920), Nos. 8, 9, and 10, and Vol. 2 (1 21), No. 17. See
David Taylor Model Basin Translation 38 (Mar 1952).
10. Windenburg, D.F. and Trilling, C., "Collapse by Instability of Thin Cylindrical Shells
under External Pressure," Transactions, American Society of Mechanical Engineers, Vol. 56,
No. 11 (1934). Also Experimental Model Basin Report 385 (Jul 1934).
11. Salerno, V.L. and Puios, J.G., "Stress Distribution in a Circular Cylindrical Shell
under Hydrostatic Pressure Supported by Equally Spaced Circular Ring Frames," Polytechnic
Institute of Brooklyn, Department of Aeronautical Engineering and Applied Mechanics,
Report 171-A (Jun 1951).
12. Lunchick, M.E. and Short, R.D., Jr., "Behavior of Cylinders with Initial Shell Deflec-
tion," Journal of Applied Mechanics, Vol. 24, No. 4, pp. 559-564 (Dec 1957).
13. Gerard. G. and Wildhorn, S., "A Study Of Poisson's Ratio in the Yield Region,"
National Advisory Committee for Aeronautics Technical Note 2581 (Jun 1952).
~z - -1.-- ---- -- __ ,,__ __ -- ------ --

MAY 28 1874

MAY 1 5 1980
MAY 5 1981
OCTJAN
7 198

19871