Beruflich Dokumente
Kultur Dokumente
WILKINS, JR
C. W.
BERT AND
D. M.
EOLE
Although sandwich construction is being used more widely in shell configurations for
aerospace structures, only a limited number of the vibrational analyses referenced in a
recent bibliography [1] considered orthotropic facings, as exemplified by fiber-reinforced
composites. However, some did allow for the commonly used hexagonal-cell honeycomb
core by considering the simpler effect of an orthotropic core. The first sandwich-shell vibration analyses to consider orthotropic facings and core were done independently by Azar [2],
Vasitsyna [3] and Baker and Herrmann [4]. Azar treated axisymmetric free vibrations of
freely supported arbitrary open-ended shells of revolution, such as conical and paraboloidal
shell frusta but excluding cylinders. Vasitsyna analyzed free vibrations of circular cylinders,
while Baker and Herrmann considered the same case with the addition of a general state
of initial stress. Later, Bacon and Bert [5] extended Azar's work to include unsymmetric
modes.
Most of the above analyses used freely supported edges, while a few considered clamped
edges. None of them considered free edges. In fact, until very recently, the most difficult
case that had been analyzed for free edges was that of a homogeneous, isotropic, conical
shell. Hu [6] formulated his analysis to include both membrane and bending effects, with
solutions by Galerkin's method. However, the only numerical results which he published
for free-free boundary conditions included only membrane effects and they were not compared with any experimental results. A Rayleigh-Ritz analysis by Hu et al. [7] considered
the conical shell to be inextensional, i.e, the membrane strains were identically zero. For
a free-free shell, such an analysis yields only two meridional modes. Their results compared
t This research was submitted by the first author in partial fulfilment of the requirements for the degree
of Doctor of Philosophy in Engineering at the University of Oklahoma, 1969.
15
211
212
favorably for the two lowest meridional modes found in their previous experiments [8]
for low values of circumferential wave number, n.
Sewall's analysis, mentioned in a report by Mixson [9], was said to have been solved by
a Rayleigh-Ritz technique, and his results, for the lowest unsymmetric mode only, agreed
quite well with Mixson's experimental values for homogeneous, isotropic, conical shells.
A later Rayleigh-Ritz analysis by Naumann [10] compared favorably with various published
results for free-free isotropic conical shells.
The analysis of Krause [11], for the same type of shell, included both membrane and
bending effects, and used a modified Galerkin method in which it was not necessary to
satisfy the force and moment boundary conditions. However, Krause's results did not
appear to agree as well with the experimental values of [8] as did those obtained by the
much simpler analysis of [7].
The case of a sandwich conical frustum with orthotropic facings, perfectly rigid core and
free edgeswas analyzed by Bert et al. [12] using a simple inextensional theory and the RayleighRitz method. For the two lowest meridional modes, their analysis agreed well with their
experimental results for a free-free, orthotropic, sandwich shell.
The analysis reported herein was directed toward conical frusta, with the cylinder as a
special case, for symmetric and unsymmetric modes. Orthotropic facings and core and all
components of translational and rotatory inertia were included. The problem was formulated such that the boundary conditions need not be specified until the final step. Three
sets of end conditions were investigated: clamped-clamped, freely supported, and free-free.
2. FORMULATION
Expressions for the kinetic and potential energies of a symmetrical sandwich conical
shell with orthotropic facings and core were derived from basic principles, using Love's
first-approximation shell theory with the addition of transverse shear. Hamilton's principle
was employed to derive the differential equations of motion and the boundary conditions,
and then Galerkin's method was applied to the equations of motion.
All of the following assumptions are made in the analysis.
(i) The core is capable of resisting transverse shear, but not bending, extension, or inplane shear.
(ii) The facings resist extension, bending, and transverse and in-plane shear.
(iii) The facings are identical, so that the sandwich is of symmetrical construction.
(iv) Both the core and the facings are linearly elastic and can be orthotropic.
(v) The facings and core furnish both translational and rotatory inertia effects.
(vi) The shell thickness is small compared to the smallest radius of curvature of the
shell, so that zlr may be neglected when compared to unity.
(vii) In the expression for the radius, r, the term ZCOSlX is neglected.
(viii) All deflections are small, so that strain-displacement relations can be linearized.
(ix) Lines which are straight and normal to the middle surface before deformation
remain straight during deformation, but do not necessarily remain normal to the
middle surface.
(x) The facing rotation, t/J~, is assumed to be identical in the inner and outer facings, in
view of hypothesis (vi). The same assumption also applies to '-Po.
(xi) The core is incompressible in the thickness direction.
(xii) All material damping, thermal, and initial-stress effects, as well as interactions
with surrounding fluid, are neglected.
213
r
Ro
_____ l. J _
Figure 1. Shell geometry.
(1)
+ l1W m =
'2
(2)
0,
",'2
(3)
27]5 cos e.:'v + 2Y/5 {w,o- 2"714C!fio.xx + {sin "'!fie.x)27]7 ?;2tj.Je,OO + 2(Y/5 + Y/14sin2 0(,2) !fie - (Y/8 +2Y/14) ?;!fi~.xe
2(7]7 + "714) sin ",,2!fi~,o - TJJJC!fio.xx + 'sin ex!fie.x) 'l}11 ?;2t/ie.oo + (27]5htcos2 a + 7]13 sin2e.:) ?;2!fio-
(4)
(5)
214
(6)
+,
+ 2'1]15 w,x -
(7)
The application of Hamilton's principle also provides the complete set of boundary
conditions at x = 0 and x = L, written as either
u=o,
or
(8)
either
v=o,
or
(9)
either
w=o,
or
(10)
either
or
M",e = (27]14 + 'lJ13)(fo.x - sin octf~,(j) + (7]13 + 2h27]12)(fe,x - sin a'fe + ~!fx.(J) = 0; (11)
either
or
M x = (27]6
O.
(12)
The circumferential dependence and the time dependence are removed from equations
(1-12) by assuming
u(x, (J, t) = U(x) cos nO sin wt,
tP!J(x,
tP;(x,
tPe(x,
(J,
t)
='
ne
t)
(13)
215
In order to use Galerkin's method to solve the equations of motion, modal shapes must
be assumed. This step is accomplished symbolically by letting
fi~
fie
2: A6mrP6",(E),
w= 2: A 3mrP3m(E),
m
(14)
III
fi.< =
L
A7tn rP7m( E),
m
where rP"II> rP2m, .., rP7111 are functions satisfying the appropriate end conditions. The use
of these dummy functions allows the problem to be completely formulated without regard
to boundary conditions. The specific steps in the application of Galerkin's method and the
entire computer code are given in detail in reference [13]. In this study, three boundary conditions are to be investigated : freely supported; clamped-clamped; free-free.
The freely-supported boundary condition is defined here as zero displacement in the
circumferential and normal directions and zero meridional stress resultant and moment at
each end of the shell. Thus,
(15)
The set of assumed modal functions for the freely-supported boundary condition is written
as
M,
0=
m=1
A1IIIR-v'o"cosm7TE,
fi~ =
Ml
V" =
m.=;1
fix =
M4
fiB =
m=1
m=1
W = L A 3msin m7TE,
m=1
2:
AS/IlR-v'o"COSln7TE,
m~1
M6
fie
M3
loiS
(16)
M7
2:
III =J
A 7mR- v'oxCOSln7TE.
A 4m sin m7TE,
0=
ni ce!
loll
fi~ =
V" = InL
A 2msin m7TE,
;;" 1
M3
m=1
fiB =
L A 4msin m7T,
m-l
2:
m=1
A smsin mne,
M6
fie =
W = L A l m sin m7T,
M4
loiS
m=1
A 6m sin m7T,
M7
fix =
11I =1
A7msin m7TE.
(17)
216
For the free-free boundary condition, the forces and moments must be zero at each end;
that is,
(18)
Unfortunately, no set of simple trigonometric series will satisfy the rather involved differential
equations represented by equation (18). From a physical argument, one comes to the conclusion that whatever series is used for the displacements and rotations, it must not be zero
at the ends. A "free end" implies, certainly, that the displacements and rotations cannot
be constrained.
The simplest form which is non-zero at the ends is an appropriate series of cosine terms.
Thus, for the free-free boundary conditions, the series are taken as
M1
0=
2:
AIm cos
mne,
m=O
v=
$~ =
M2
n1=O
A 2m c o s 1n1l' ,
M~
m=O
A sm cos m7T,
M6
$0=
1>1 3
m=O
A 6m R
.
(19)
COS 1n1l' ,
M7
m=O
M4
m=O
3. EVALUATION
The theory was first evaluated for the simplest case that can be considered, which is the
case of a homogeneous, isotropic cylinder.
For the freely-supported case, the experiments of Bray and Egle [14] and the analysis
of Egle and Soder [15] were used for comparison. Bray and Egle tested a steel cylinder
with a radius of 584 in. and length of 11907 in. with 0020-in. wall thickness. They found
the lowest natural frequency to be at 11 = 7 with a value of 380 Hz. The present analysis
predicts a lowest natural frequency of 3802 Hz at n = 7.
As a check case for the analysis, Egle and Soder used a steel cylinder for which Hu, Gormley
and Lindholm [16] have published experimental results. The dimensions were R o = 100 in. and
TABLE 1
2
5
6
7
12
14
m=l
Present Reference
[15]
analysis
63309
1597
1678
206'5
58106
7925
6335
159'9
1680
2060
5811
79.J.8
m=2
Present Reference
[15]
analysis
4833
3707
3255
5955
8025
4831
3705
3251
5949
801'8
m=3
Present Reference
[15]
analysis
9610
7244
581'4
632'9
825'4
9606
724'0
580'8
6323
8247
m=4
Present Reference
[15]
analysis
70H
868'3
7064
8676
217
L = 480 in . with wall thickness = 003 in. A comparison between the present analysis and
Egle and Soder's analysis is given in Table 1. The excellent agreement between the two
different analytical approaches supports Egle and Soder's contention that the experimental
shell was not actually freely supported, since the experimental frequencies were somewhat
higher than the analyses predict.
700
r---,.---,.---,--,.---,--,.----.,
m=3
60 0
m= 2
m=1
100
10
12
14
1tI=
2, () III = 3,
TABLE
m=l
m=2
m =3
0
2
3
4065
134-8
86'8
85'7
113'3
1537
2015
256'2
3173
3846
5359
13198
6249
3109
212-8
165-4
1606
1889
2365
2941
358'7
429'7
5894
1412-8
655-5
439 9
331'9
4
5
6
7
8
9
10
12
20
264'7
2356
2415
2757
3295
3954
469'3
636'3
15024
218
To non-dimensionalize, the frequency was multiplied by RoV p'(l- vB"i)/E;. The clampedclamped boundary condition was checked at another point for which the radius-to-thickness
ratio was 20 and the length-to-radius ratio was 2. Forsberg gave the dimensionless lowest
natural frequency as 032 at n = 3. For the present program, R o = 200 in., L = 400 in. and
t = 025 in. For n = 3, a frequency value of 0325 was found.
m=3
m=2
10
12
14
(6) m = 4, (0) m
TABLE
0
2
3
4
5
6
7
8
9
10
12
20
m=1
m=2
m=3
4065
177'2
1260
110-7
1267
163-5
212-3
2691
3327
4026
5599
13689
6618
340'1
2543
209'7
197'7
2148
2545
3100
376-1
450'2
617'7
1473-8
6884
474'7
376-9
314-7
2845
284-0
3097
3566
41%
4948
671-3
1575-9
For the free-free boundary condition, a cylinder tested by Watkins and Clary [18] was
used for comparison. The steel cylinder was 42 in. long with a 14-in. radius and a wall thickness of 0007 in. At n = 10, the lowest natural frequency was reported as 323 Hz. The present
analysis gave a value of 343 Hz. However, while the second lowest frequency at n = 10 was
reported as 328 Hz, the present analysis gave a value of 831 Hz.
The only experimental work found that treats a sandwich cylinder or cone is the work
of Bert et al. [12]. Consequently, although their work was concerned only with the free-free
219
7 00
6 00
3,
5 01)
cQ)
=>
cr
40 0
:>
:::I
0
2
300
200
10 0 -
10
I~
14
(0) 111
111
= 3,
(.c.) 11l = 4,
TABLE
m =2
m=3
m=4
026
13-3
350
65'3
101-8
1433
1911
2457
3067
13057
7247
179'9
1302
1157
1338
1756
229'7
2905
3578
14407
7905
384'4
2606
2131
199'7
21704
260'5
3225
396'7
15597
8052
Jl
ot
2
3
4
5
6
7
8
9
20
m=5
4125
929'3
3397
901-8
366'7
913'1
4668
16890
956'8
Vex=
v;(J = 0,20,
= 10,
p' = 0265 X 10- 3 lb-sect/in",
5'07,
R o = 2245 in.,
K;=
L = 72'5 in .,
t = 00105 in.,
It = 015 in.,
E~ =
G."
Ee = 364 x
lb /irr',
G;x = Go. = G~(J = 10 X 106 Ib/in 2,
106
x;
= 32 X 104 Ib/in 2,
220
The results were as expected, with the frequencies for the freely-supported case being
higher than those for the free-free case, and the frequencies for the clamped-clamped case
being higher than those for the freely-supported case. These results are shown in Figure 2
and Table 2 for the freely-supported case, and in Figure 3and Table 3for the clamped-clamped
case.
r=
-1,0
~
o
~
m ol
~
"0
It>
.!::!
"0
...
+10
2:
- 1' 0
0 25
0 75
050
10
In =
For the free-free case, the analysis and experiments are compared in Figure 4, the analytical
frequencies are tabulated in Table 4, and the analytical mode shapes are shown in Figures 5
and 6. These results indicate that the inability to satisfy the boundary conditions for the
free-free case has resulted in good agreement only for the lowest natural frequency for
each value of circumferential wave number.
4. CLOSURE
It should be noted here'that there seems to be some inconsistency in the reporting of data
for the free-free case. For other boundary conditions, the meridional mode number, m, can
simply be thought of as indicating the number of half-waves in the deflected shape of a
generator of the shell. The modes can then conveniently be identified since the lowest frequency will always have m = 1, the second lowest, 111 = 2, etc. However, upon studying the
221
-10
#B
m=3
Cll
;0:;
Cll
"tl
-o
.~
+10
m=4
025
050
075
1-0
Figure 6. Modal shapes for a free-free sandwich cone with m = 3 and m = 4 and various values of
11.
free-free modal shapes of Figures 5 and 6, it is seen that the number of half waves (and the
number of nodes) varies with n for the lowest frequency, second lowest frequency, etc.
This means that a simple identifying number can no longer be used, but the modal shape
must be shown for each frequency. Consequently, in this work, the designation m = 1
denotes only a lowest natural frequency and tells nothing about the actual modal shape.
Similarly, m = 2 denotes the second lowest frequency, etc.
The evaluations of the present theory for the freely-supported and clamped-clamped
boundary conditions show generally good agreement with available experimental and
analytical data.
For die free-free boundary condition, however, the theory seems to give good agreement
for only the lowest natural frequency associated with each circumferential wave number.
REFERENCES
1. N. L. ROUST 1968 Hexcel Corporation, Dublin, Calif. Vibration and fatigue sandwich bibli-
ography.
2. J. J. AZAR 1965 University ofOklahoma, Norman, Okla., unpublished Ph.D, dissertation. Axisym-
222
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
APPENDIX I
DERIVATION OF THE EQUATIONS OF MOTION
FOR AN ORTHOTROPIC SANDWICH SHELL
The six strain components for the co-ordinate system shown in Figure I are
(AI)
where r = R o +- x sin ct +- zcos ct, and the subscripted u's are total displacements in the x,
8, and z directions.
In view of hypothesis (vi), the term zcos ctin the expression for r will be neglected as noted
in hypothesis (vii). In all that follows r is replaced by
r ,.., R o + x sin ct
(A2)
223
,=
(R a + x sin /X)-l.
(A3)
The displacements are now defined in terms of the middle-surface displacements and the
angles of rotation of normals to the middle surface in the meridional and circumferential
directions. For the core, these angles are denoted by !flx and tf'h while for the facings, they
are !fl; and !flIJ. The assumption is made that the core is incompressible in the thickness
direction.
For the core,
U~ = u(x, t) + ztfix, t),
e,
e,
(A4)
e;x = U, x + Ztfx.x,
eoe = '(v,e + sin /XU + cos /xw)
e~z =
0,
+ tfo.x].
(AS)
uh =
u~, u~ =
(A6)
e~x, e~x =
e~o, eho
e~z, e~"
e~o, e~o
0,
(A7)
Core strainenergy
Due to hypothesis (i), the core strain energy is the energy due to transverse shear strain
only, so that
V" = ~
or
(A8)
z
I,
VC =
f JJ [GzxCe~x)2 +
x
-I.
G ol eez)2]dz'-l dO dx.
(A9)
224
Squaring e;x and e9z from equations (AS), and integrating over z, gives
,-1 dOdx.
(AIO)
It is noted that the term in braces in equation (AIO)is hereafter replaced by 1, sinceh 2 ,2
<l
1.
:2I
f Jf
(0'
0
xx
0
+ o Ixx eIxx + 0'000 eoo0 + aooI eoo
i
+
exx
x 0 ::
(All)
0'00
u"x =
G~xe::x,
(AI 2)
where E; = E~/(1 - Vex v;o) and Eo = e/(1 - vox v;o). The superscripts a and i have been
omitted in equations (AI2) since they apply both to the outer and inner facings.
Substitution of equations (AI2) into equation (All) results in
JJJ{E~[(e~x)2 + (e~x)2] +
vf = ~
B9[(eeo)2 +
(e~o)2] +
+(B~ Vex
(AI3)
The strains are now substituted from equations (A7) and the integration over z is performed, with the inner facing terms integrated from -h - 2t to -h, and the outer facing
terms integrated from h to h + 2t. The following integrals are necessary:
,.,.21
J
-h
dz= +
11+21
-h
J zdz=- J
11
.r
zdz=+2t(h+t);
-h-2t
,.+2t
11
dz=+2t;
-1I-2t
-h
Z2 dz =
-11-21
Z2 dz =
(AI4)
225
yr =
'2
2,2
'2
'2
- 2' sin aVV,x - 2"1 sin IX!f;U t/Je,x] + 4ht2G;o[ if;x.e !f;;,u sin rt-tftx,e!J1o + 'ifx.e tft!J.x sin at/J;.e!f;o +
2
+ 'iKe tfte,x + '2 sin rt-tfte t/Jo - 'sin extfto !f;e,x - 'sin atft!J !f;e.:< + !f;o,:< tftL] + (8/3) t 3 G;e['2 tft;;o +
+ '2 sin2 at/Jfl + if;o~x - 2,2 sin rxifJ;.e tfto +
+ n!f;;.e if;L" - sin rt-tfto tfte.x] + 2tKeGoz['2 w,~ +
+ cosZav 2 + cos? eyN 'f~ +!J1? cos etVW,u +
+ nif;o w,O - cos avt/Je + 2ht'Z cos! rt-tftu !f;o] +
+ 2tK; G;x[W,~ + 2w,xif;; + t/J;2]} d8dx.
- ,2
'2
2''2
2'
'2
2'2
,-I
(A1S)
Total strainenergy
The total strain energy is formed by adding equation (AlO) and equation (A1S) and
collecting like terms.
y =
'l]2
{J
'2
'2
2,2
226
+ sin cx.r/Jo.o r/J~ + sin 2 rxt/lx t/l~] + 7)d~2 u,~ + '2 11 2 t/l~,o + '2 sin 2 rxv 2 +
+ ~2 h 2 sin21Xr/J~ + v,i + h2 r/JL - 2'2 sin cx.U,f) v+
+ 2~u,e v,x- 2~21t2 sin 1X00x,fi 0/8 + 2'h2t/lx.fJ r/Je.x - 2' sin cx.vv,x - 2~h2 sin cx.r/Jo !fro,xl + 7J13 [~2 !frx.fi t/l~.f) - ~2 sin rx!frx, 0 t/lo + 'iflx.fi !frL" - {? sin rx!fr;.o iflo +
+ N;,e t/le.x + sin 2 cx.!fro!fro- 'sin 1X1hJ!frL2
- ~ sin cx.!fre iflo.x + !fro.x o/e.x] + T]I4(~2 o/~,fi + '2 sin? et.!fre +
+ r/J&~x - 2~2 sin rt.r/J;,fi iflo + 2 'iflL 0 t/le.x - 2' sin Ci.1f& !frL] +
+ 7)dr/Ji + 2iflx w,x + w,iJ + 7)16[,2 w,~ + '2 COS2cx.V2 + ifl~ +
+ 2'w,0!fre - 2~2 cos rxW,fi V - 2' cos rxvr/Jo]} '-1 dO dx.
(A16)
'2
T=
f f{/n(u,i + v,; + w,~) + J(if1i.t + ifiL)+ 2JI[(if1;.1)2 + CtPL)2]} '-1 d8 dx, (A17)
x
in which
Iii = 2(ph
+ 2p' t),
= (2/3) ph3,
J' = 2p' t[(t 2/3) + (11 + t)2].
J
(A1S)
Hamilton's principle requires that the first variation of the time-integrated difference
between the potential and kinetic energies be zero:
J(V-T)dt=O,
t1
(A19)
11
Performing this operation after combining equation (AI6) and equation (AI?) gives the
equations of motion and the boundary conditions, which are shown as equations (1)-(12).
APPENDIX II
NOMENCLATURE
E~, E~
E~, E~
== E~j(l
A tm
Vox v~o),
J
J'
K"" K o
K;,
K~
L
Me
M"
Me
m
In
n
Q", Q e
R
Ro
Ro
r
T
Vex)
V
u
V
V e , yf
V
v
W
W
w
x
z
ex
~
mass moment of inertia of core about core middle surface per unit surface area
(lb-sec-/in.)
mass moment of inertia of one facing about core middle surface per unit surface
area (lb-see'jin.)
core transverse shear coefficient in z-x and 8-z planes, respectively (dimensionless)
facing transverse shear coefficient in z-x and 8-z planes, respectively (dimensionless)
shell slant length (in.)
upper summation limit in assumed mode series
bending moment (in-lb/in.)
twisting moment (in-lb/in.)
meridional mode number; summation index
composite shell mass per unit surface area (Ib-sec-jlrr')
number of circumferential full-waves
transverse shear stress resultant in z-x and f)-z planes, respectively (lb/in.)
== Ro + E sin et (dimensionless)
radius of the middle surface at the small end of the shell (in.)
== Ro/L (dimensionless)
== Ro + x sin et + z cos /l( (in.) (see assumptions)
half-thickness of one facing (in.); time (sec)
total kinetic energy(in-lb)
normal mode form of u (in.)
U/L (dimensionless)
middle surface displacement in x-direction (in.)
strain energy (in.-lb); normal mode form of (in.); volume (in")
strain energy of core and facing, respectively (in.-lb)
Y/L (dimensionless)
middle-surface displacement in 8-direction (in.)
normal mode form of w (in.)
= W/L (dimensionless)
middle surface displacement in z-direction (in.)
meridional co-ordinate (see Figure 1)
thickness co-ordinate (see Figure 1)
cone semi-vertex angle (see Figure 1)
s=x/L (dimensionless)
== (R o + x sin OC)-I (in-I)
'Y]I ==2tE;
'Y]2 =2tE~
'Y]3
2t(E~ vo" + Eo v~e)
'Y]4 ==2tK;a;,.
'Y]s ==2tKiGo
'Y]6 ==
4t 2
31)1
41 2
'Y]7 ==T'Y]2
41 2
n == T'Y]3
16
227
228
=o2ht1]1
1]10
htn,
1]11
=0
2ht1]2
1]12
=o2tG~e
1]13
=0
1]14
=0
1]15
ss.o.;
2ht1] 12
4t 2
1]16
,
).lex,
f)
).Ixe
p
p'
ull
1>/J
!fix, !fi~
s: f~
!fie, !fie
fe, fe
w
31]12
=ohKeGez
angular circumferential co-ordinate
major and minor Poisson's ratios, respectively (dimensionless)
density of core material (lb-sec'/in")
density of facing material (lb-sec-jin")
stress (lb/in 2)
assumed mode functions
angle of rotation in the meridional direction of the normal to the middle surface for
the core and facing, respectively (radians)
normal mode form of !fix and !fi~, respectively (radians)
angle of rotation in the circumferential direction of the normal to the middle surface
for the core and facings, respectively (radians)
normal mode form of !fie and !fie' respectively (radians)
circular frequency (rad/sec)