Professional Documents
Culture Documents
for loads lower than about 1.6 Ib, but for larger loads a signifi- would give better results, but it is much more complicated.
cant difference is observed. Possible reasons for this dis- Finally, for large loads, tangent rotation angle ft becomes
crepancy are summarized at the end of this section. large (see Fig. 7), and thus the assumption of small strain
In Figs. 6 and 7, theoretical values of e,e and fa,fid are is violated [see Eqs. (17b) and (37)].
shown as functions of indenter load P for the prestress ovo =
2510 psi. The values of e,e and pb,Pd are, at the most, only
References
weak functions of crr0. The curves for ovo = 850 and 1310
psi differ from these by less than 2% and hence are not 1
Bhatia, N. M. and Nachbar, W., "Finite Indentation of an
shown. The_upper bound value for the indenter load, corre- Elastic Membrane by a Spherical Indenter," International Journal
sponding to Pu = 0.6599 [Eq. (34b) ], is Pu = 2.021 Ib. This of Non-linear Mechanics (submitted for publication).
2
value of indenter load is found to be in good agreement with Jahsman, W. E., Field, F. A., and Holmes, A. M. C., "Finite
the two values of experimental rupture load in Fig. 5. Deformations in a Prestressed, Centrally Loaded Circular
The reasons for the small discrepancies between this Elastic Membrane," Proceedings of the 4th U.S. National Congress
theory and the experiments of Ref. 3 are now summarized. on Applied Mechanics, American Society of Mechanical Engi-
neers, 1962, pp. 585-594.
The first reason is that calculated values of <7r0 were assumed 3
Nachbar, W., "Finite Deformation of a Prestressed, Elastic
to remain constant during indenter loading for a fixed platen Membrane/' SUDAER 141, 1962, Dept. of Aeronautics and
load, whereas crr0 actually varies slightly with /3(r = a) and Astronautics, Stanford Univ.
4
consequently with P. Furthermore, o>o was probably af- Prager, W., An Introduction to Plasticity, Addison-Wesley,
fected by friction between the membrane and the outer Boston, Mass., 1959.
5
Downloaded by UNIVERSITY OF OKLAHOMA on April 22, 2013 | http://arc.aiaa.org | DOI: 10.2514/3.4673
supports when /3(r = a) is nonzero. The second reason is Koiter, W. T., "Stress-Strain Relations, Uniqueness, and
that the value of yield stress ay used in the theory is only Variational Theorems for Elastic-Plastic Materials with a
approximately the yield stress of Mylar. The third reason Singular Yield Surface," Quarterly Journal of Applied Mathe-
is the use of Tresca's yield criterion and the associated flow matics, Vol. 11, 1953, pp. 350-354.
6
Bhatia, N. M. and Nachbar, W., "Finite Indentation of
law of Tresca. Since directly under the indenter ar g^ ae = Elastic and Elastic-Plastic Membranes by a Spherical Indenter,"
ay, we are at a corner of the Tresca yield surface where there SUDAER 203, 1965, Dept, of Aeronautics and Astronautics,
is the largest discrepancy between the Tresca flow law and Stanford Univ.
7
the Mises flow law. It is possible that the Mises flow law Jahsman, W. E., private communication.
RAY W. CLOUGH|
University of California, Berkeley, Calif.
The stiffness matrix and three consistent load vectors are developed for a rectangular finite
element of a cylindrical shell. The present element is compared with others reported in the
open literature. Three problems are analyzed with the present element and the results are
compared to known solutions to establish their integrity.
are
(la)
Downloaded by UNIVERSITY OF OKLAHOMA on April 22, 2013 | http://arc.aiaa.org | DOI: 10.2514/3.4673
Rigid-Body Modes
It is easy to show that, for a rigid-body translation of com-
ponents 5X, 8y, 82 in the system of reference (xyz), the dis-
placement field must take the following form (see Fig. 1) :
0 0 "to
o o o oO O o o o o O O o o o ^ 6o o
! i
(10)
M,
Downloaded by UNIVERSITY OF OKLAHOMA on April 22, 2013 | http://arc.aiaa.org | DOI: 10.2514/3.4673
II o
Ss K
where Ki = Et/(l - ^ 2 ); #2 = pKi] Ks = |(1 -
Di = Et3/I2(l - M 2 ); D2 = fiDi] D3 = i(l - /*)Z)i. The
strains and forces are all taken in the middle plane of the
shell. In more compact form, Eq. (10) will be written
o o o o O e o o o o ^> « o o
(10a)
The strain energy in the element is
I I
U = ^ff{N}T-(e} d^df] (11)
? ^ ^ -o vo O O ,0
•^^co e o o O f M c o e o o ^ 3 O O O <N <M e or substituting for {A^} and {e} from Eqs. (9b) and (lOa):
I I I I I I
Denoting the integral in Eq. (12) by the symbol [k], the strain
I I
energy becomes
M
U = ^^[^{a,} (13)
ooooOcoooooOcoooooOcooo
I Finally, solving for {di} from Eq. (8) and substituting into
Eq. (13) leads to the stiffness matrix
1 1 [K] = [T-*F[k][T-i] (14)
The matrix [k] is given explicitly in Table 2.
A computer program was developed to test the rigid-body
properties of this stiffness matrix. The eigenvalues of the
stiffness matrices of several examples were obtained; in all
-OOO cases six eigenvalues were zero. In addition, knowing the
OcOCOCT)
I I II .«
exact form of the displacement functions for rigid-body modes
it was made possible to construct nodal displacement vectors
corresponding to each of the six rigid-body modes. Multi-
plying [K] by these vectors gives the associated nodal forces:
for all six rigid-body modes the forces were zero. On the
basis of these results, it was concluded that this element stiff-
A ness matrix adequately accounted for all rigid-body effects.
03
Consistent Loads
A finite element analysis requires that actual loads be
transformed into their consistent nodal load equivalents.
OooOooOooOooOoo^ooOooOc This analysis was carried out in the standard manner for
i I I I
this cylindrical element, and the results for a uniform gravity
load, a uniform pressure load, and a concentrated load are
OOooOrHOOOOOT-HOOOOOiHOOOOO given in Table 3.
0 k2
0 0 ks
0 0 0 0
0 0 0 0 fc4
0 0 0 0 0 0
0 k7 0 0 1 0 k8
0 0 0 0 0 0 0 0
0 0 0 00 00 Ofcio symmetrical
0 0 0 0 kn 0 0 0 0 kiz
0 0 0 0 0 0 -0 0 ku 0 fas
0 0 0 0 foe 0 0 0 0 kn 0 ku
ki9 0 0 0 0 0 0 0 0 0 0 0 ho
0 £21 0 0 0 0 £22 0 • 0 0 0 0 0 £23
£24 0 0 0 0 0 0 0 0 0 0 0 kzs 0 &26
0 £27 0 0 0 0 A;28 0 0 0 0 0 0 £29 0 &3o
0 0 0 0 0 0 0 0 A;32 0 £33 0 0 0 0 0 /bs4
0 0 0 0 /c35 0 0 0 0 fas 0 £37 0 0 0 0 0 /c38
0 0 0 0 0 0 0 0 &40 0 &4i 0 0 0 0 0 £42 0 &43
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
Downloaded by UNIVERSITY OF OKLAHOMA on April 22, 2013 | http://arc.aiaa.org | DOI: 10.2514/3.4673
n -iTTt.
a 7 jKigfe3 . Ksasb 7 r, , 7 T, , , Kia3b . Ksab3 . Diasb . Dsab3 , ^. , 7 _^ . Diab , Kitfb"1 .
Where: fa = ——— + ——— ; kz = K\ab\ ks = Ksab; fa = ——— + ——— + ——— + ——; fa = Kzab; k& ~ Kiab + ———; fao = ————— +
12 12 12 12 12r2 3r2 r* 200704r2
+
448 800 448 1600 ' "= 960r ~ 24r ~ 40r ~~ 6r ' 12
~~ 35840r2 80 40 112 20 ' "~
Or2 +
35840r2 80
80 +
80
80 +
80 ' ' ^15"= 5376r2
80 5376r2+ ~4 4 +
20' ' " =
20 " 80r
80r~ 2r ' *" ~" 5376r2 + 4~ +
20 ' ^18 =
z3 - _ J.\.2U,"U" , _ J.\-1U"U"
3Z>ig 6; fag
\D\a b', kig = —-——: kzo = —-—— + —-— + ——— + ——-—— + —-—— ; ^21
fai = ———:
———; kzz = —-—— — —-— — —-—: kn = —-—— + —-— +
960r2'
960r2 35840r2 ^ 112
112 40 80
80 20 ' ' 21 ~ 144r
20 144r ' 22 ~ 144r
144r 144r " 6r 6r 6400r2 20
3
+^
18 +
20 +
9 ' ^ =
144r ' k* =
6400r2 + 20 +
12 3~' 26 26~ ~ 960r2
3~' 3 3 ' 2?
12r ' 28
12r
29
29 33 10
10 33
r ' ~ 960r2 3 3 ' ° ~ 80r2 '' 35840r2
35840r2 80
80 80 80 ' 6400r2 4 4 '
3
•———
—— + + ———
—-—— -f*
4- ———
—^- ;: A;36 =
fas = ————
——— — ————•
—-— —— ————;
—^—: fcse
fcse =
= ———
—-—— -K + ———
—-—— ++ ———
—-— + ————; k& = ———— 4- Z)2g &; fc3s = ————
5376r2
!7«^2 ^ 4 A 20
9H 144r
1AA» 6r
R* 3r
%*• 6400r2
Rdnn-r2 12
19. 20
9O 3 960r2 960r2
4 3 43 3 44
' ° "" 6400r2 4~~' " ~ 960r2 ° ' ^ ~ 960r2 ' ~ 144r2 ° ' ~ 80r ' ^ ~ 5376r2 20
)5
J- D 7)3- fr = ^ia^7 4. qn 7,3. fr = ^2Q^3. i. _ -^ia^3 _ 2Z>ig6 _ Kiazb5 . DzabJ D\a*b^ , __ Kia3bs , _ , _ .K:
20 5 47 10 5 48
~" 448r2 "^ ~ 12r ? 49 ~ 12r T"5 6° ~ 960r2 "^ ~3~ 3 ; " ~ 144r2 "*" 2
° 5 *62 ~ ~8i
w (ft)
center
2.0
1.0
-0.002
2 0 ° 40°*
Downloaded by UNIVERSITY OF OKLAHOMA on April 22, 2013 | http://arc.aiaa.org | DOI: 10.2514/3.4673
50.
-O.OOOI
•THEORETICAL SOLUTION
u(ft)
end
.010
.005
;005
20° 40
.02 .03 .04 05 -
N
N-number of elements
Fig. 4 Displacement w along AB and u along DC for Fig. 6 Effect of rigid-body modes on the convergence of
problem 2. the solution for problem 2.
1062 G. CANTIN AND R. W. CLOUGH AIAA JOURNAL
Bogner et al. (48 X 48) Cantin and Clough (24 X 24) (24 X 24) without rigid-body modes
Mesh No. of Eq. Displacement, in. Mesh No. of Eq. Disp., in. Mesh No. of Eq. Disp., in
1 X1 48 -0.0025 1 X3 48 -0.0297 1X 3 48 -0.00099
1 X2 72 -0.0802 1X5 72 -0.0769 1X 5 72 -0.00213
2 X 2 108 -0.0808 2X5 108 -0.0780 2 X 5 108 -0.00266
1 X3 96 -0.1026 1 X7 96 -0.0987 1X 7 96 -0.00352
2 X 3 144 -0.1036 2X7 144 -0.1002 2 X 7 144 -0.00420
1 X4 120 -0.1087 1X9 120 -0.1057 1X 9 120 -0.00513
2 X 4 180 -0.1098 2X9 180 -0.1073 2X 9 180 -0.00590
3 X 49 1200 -0.1128 3X 49 1200 -0.05583
Downloaded by UNIVERSITY OF OKLAHOMA on April 22, 2013 | http://arc.aiaa.org | DOI: 10.2514/3.4673
the element containing rigid-body modes are compared to significant discrepancy in the physical behavior of the system
those obtained with the same element without rigid-body unless a very fine element mesh is used. However, the de-
modes; the importance of these rigid-body terms is evident. velopment of an effective stiffness matrix for the curved ele-
A comparison with the (48 X 48) element of Bogner, Fox, ment is greatly complicated by the problem of adequately
and Schmit shows that the present (24 X 24) element gives accounting for rigid-body displacement components. The
results that are slightly lower for the pinched cylinder prob- procedure which is presented herein for including such rigid-
lem for the same computational effort. body terms in the displacement functions is both simple and
effective, and makes the curved finite element a practical
u(ft) ———— THEORETICAL SOLUTI ON
computational device.
.0127 / Results obtained in the analysis of cylindrical shells with
edge fr <5 -————0 this element show that the solutions are well behaved; coarse
^ mesh solutions provide good approximations to the exact
.OIO -
0 GC (24x24) CURVED RECTAN( LE
results, and monotonic convergence toward the exact results
O AJC (27x27) FLAT TRIANGLE
is obtained with mesh refinement. Comparisons of two sets
0 CPJ (15x15) FLAT TRIANGLE
of results obtained with this element, including and excluding
.005 - the rigid-body displacement terms, demonstrate that the
inclusion of these terms leads to significantly better answers
for coarse mesh systems. Similarly, comparisons of curved-
and flat-plate elements tend to indicate a useful improvement
resulting from the curvature. On the basis of these studies,
1000 2000 Neq it may be concluded that this cylindrically curved element
Neq - number of equations provides a practical tool for the analysis of arbitrary cylindri-
cal-shell systems.
Fig. 7 Effect of the curvature of the element on the
solution for problem 2.
References
Conclusions 1
Gallagher, R. H., 'The Development and Evaluation of
The advantage offered by a cylindrically curved element Matrix Methods for Thin Shells Structural Analysis," Ph.D.
for the analysis of an arbitrary cylindrical-shell system is thesis, 1966, State University of New York, Buffalo, N.Y.
2
obvious; a flat-plate element idealization must lead to a Bogner, F. K, Fox, R. L., and Schmit, L. A., "A Cylindrical
Shell Discrete Element," AIAA Journal, Vol. 5, No. 4, April
1967, pp. 745-750.
3
Johnson, C. P., "A Finite Element Approximation for the
Analysis of Thin Shells," Ph.D. thesis, 1967, University of Cali-
fornia, Berkeley, Calif.
4
Novozhilov, V. V., The Theory of Thin Shells, Noordhoff,
Fig. 8 Pinched cylinder: Groninger, The Netherlands, pp. 19-38.
5
P = 100 Ib, r = 4.953 in.; Scordelis, A. C., "Computer Program for the Analysis of
L = 10.35 in.; E = 10.5 X 106 Multiple Cylindrical Shells," computer program, Dept. of Civil
psi; t = 0.094 in.; /* = 0.3125. Engineering, University of California, Berkeley, Calif.
6
Carr, A. J., "A Refined Finite Element Analysis of Thin Shell
Structures Including Dynamic Loading," Ph.D. thesis, 1967,
University of California, Berkeley, Calif.
7
Timoshenko, S. and Woinowsky-Krieger, S., Theory of Plates
and Shells, McGraw-Hill, New York, 1959, pp. 501-506.