:212ywfisg INIIH MIHW Iv.“ gm1_ g*?4*1*
3,311.‘ 33 - 3112-12?! 1:123 "333333 I n23 2231’ ham “3223332; , . r’ ~. _ 3293‘; .
a v ‘ 4‘:- 3 I I b n ‘.c ' ‘\ I . '$ :51}. _.&i4. ' . A L '
zufiiu -$fl-1%3 .1223” p 2 fifig-J,H‘dq§2:£§§ ;:I, 1 ,51,
:33; t3 “I 33b 34I333 “I3 I ,. I 533 33 3 3‘3" , 3_ ' ‘5 ‘3 ‘,
mmma w. TN W 3%P2 3 3HNNIMIMI -=‘fl- 5- . .
:‘ 123333;} 3 3W 3333133"‘3r1100331‘I523JI‘3‘: 'I I “13.9%“; ‘ $333331 ".6, ., ' ’ '.- = . .;.: 1 ,1. ’ -
11 313439;“ '3“ 3.33 33.33"“ :33 33: “m: ‘3331 .121: 13.1 333333 3"3'312I33 IIIIRIIWIIII L331: 33% .1 i4}. .- - __:1
t 1 II‘I 1“ 2:33
:, 2:1..3’ ‘3 k-"”*3a1 I31. " 2‘33“," :33”: 1325923331 2‘ ~313II’"II='I‘3I‘I3‘”3~ “JI- : , ,2; .1313 - ‘
$3 1M” 33'33‘I3i.,»3 3I1"II 31333334533332" ¥33333333LQ3J13£§RJ '~ 33333533333 33,3333 3333‘333 32-3; 333 3 “1‘ -1 313'4- ’ .
333:“: 3333 31“; )3“ 333 333;: 33331‘33‘3 33333331131'i‘1;;3133:3§f"3£1311333‘ 633—33333 ‘ I‘ 3r 333 3333333 gik 3 3333 3333' 3 313.33: 3 :33 ‘3
3'1’2'373 .31,- ~ 33' " 3' .1" 1‘? 3733‘ 3 '3 2' 3 - -
3'221317311‘ 11-11323 {IIIIHIIIIIII‘I‘RI 322233313:21:4111'21'333'32113313323353333 2.; “2‘2. 3‘33 ‘ t' » -f 1- ’ . 1 ' ~Z -'
6333132153711 ‘3‘? 3‘13 3333333 . MI HIM ' ‘3“259}; 3333333: 33 2:2;431331454?‘ 1.13‘535313‘ $233 33 '.“315': _ _1 - :31".
34321.33 2?: 2 Igi',,3:111§ ‘3‘! I! 2'15 232512;: ".{Wu I ,' ‘1 ‘ . ‘ .I, 1;: 1 3 ' ‘ '
:3314H'3 =I .33: 1 N ~3-[I I. I» {33’ 313"“ :4 t} ‘ L .'. '-
.13 III I, 2:", I .22 2223213: 212‘: 21"“ 1-2 332; 2 1 1222:2521 .. .. I
.14.; ‘13. ." g. ,' 3‘13I333’: "[3 '3’3“ ,I I.;g1\§3-= 5‘ {I} _ ‘ . ‘ .. ‘1- : 2J2" 33.3.
13.33% ;.I.fi1. I, . II'. 333‘ 135;; 21-33 .311; IE . -, 1 ,
I33 .,I 3213323323311, 21.533333“ ‘7‘! 331‘ 43332235“ 31,3733} ‘21,- ?2:42 .21.: -. . “-7
343.323" “3 '33' ’3' bk“ I .3 112 33; III3§52333”2§III§¥13}1223€I5532 4325;333" - ,7 33 1
123: I 222322m3h33.2 Iywkwww§.nfim p1: -- 35
361;“; :mI -, . . 31II}II[3.3:~' 3‘;€ 3333 333:333i ff ‘35 2.33333 25.. #3! 331133; :3; ‘- $.35: 333333 "34,311; ;-- . ' 1. 1-2
“:41.“ ' '3';er “ ;";; :3“ : I 33.3 'I‘ IIII‘I’I “23.3 33333313 “1 .. t :‘ g. -‘-‘ i; if“. ‘ . - 3H : ' 2&5:
323333333024.“ 3.313.323.3212; “.31.: ”3333 ’ 31‘3333m2 33I‘43'351333" 5.1.33 533233331 33 ' 1.533. 3 1.
3 3... I , . "I" g 3" [2333?1e? I. I. 1.
I" 33:3}i3’3i‘3,3333333 = v I}, 3333' 33"!“2' :3 333 3333 3331.321 ‘13}; 5,2,5“ .112211121- 3333333 333333 1“ Bits-“gw 3‘13”»
”43'; ~ I‘ ’ I3 x .3" “L" 3 l ‘1‘” ”45:535.;‘133/‘3. r; '-.!_' '_: ’ . a; “,
22211312311 3.33:1-223314;3.11-I 92.11.2213}! 31' 23:11.1,” “1110,1332! “I 1!, 2 III,- ., I 2.133 ,1 1,}
3“. 321222-12: ‘I‘ I" I ’4‘ ”2359.291 '3 3211,31: '31:.”713-2‘I iIgé: - I33 313 ' ' a??? «1-:
333331:3“.‘pu3333!:333333‘3333333’33333b3'3 £3. 3 »I{ :1! 'f ., r}: .:. £ . I. . ‘ 2
..,I 1:333‘ 3103‘ H 3.} f :{3 ,2, .3' 3‘3!
33(3l‘333333313 33323332I‘H,2”W313“32‘2'34333233”23‘33'231.13313”, ”33.331333233133333 ‘ 1 "33431214
[53111, '3 ’3333‘23 “I I": 33933.” ' {13333381313314 “‘31 {11 1,2; I133 . I§.§3}3‘ 3i; 31;;112331534.§:331§gfi§ ,
3 2- I'uuw2232 . : 3" ‘25222fi1
33"‘333; 3 33‘3“»: I ----- ’23333 33 .1 3453333 Hit-33
3 13 7 ' I 3:352; 333 'L'; z-;_§;‘_“
. 133,111;
I' II “-
33‘- . 35131 I=:‘1.I3' 4
gr. . 363" .
r" I 7' r. z“
333233222.
1. 3 4132323 '1
(,3 11W'3333fifii2,
{SEX-333 3113; 3'33331‘un 3.333343 3‘ 33'4ng 33‘: if:
3,3343 3%,!3 333;? w 33333393 ii; “in! “3333‘ 3:53;. {3d
3 22222221222 22””“2321 1]..
335‘“ 1332333”. i043 3'3333 333333 £43” éIj.
3 3 33 {3:3831‘33333‘33I’333 3’30 3 39:35:36 2433““!
‘WI 32' 3*
H I}, 3
. ~31-2 3 333333 33333 3
I), I I‘
.331 .4
,33'I.
t
13‘1é§,g;
. ,m
2 21332222
k 31 5 . ‘I ‘ :.
31:! .
. wpr
I 31,2213?” "35' 33331333 3
3-) "3312“ I21‘
. A _. #3333 ‘5‘ {(3'3;: E‘nsg
{1“I M:
F.
3-,
"ILI’ ‘ J‘; '33
33.3131"? 46.1w
"(‘1pr
2......
1:13.74
In“?! 3.33 ‘33354 ‘ 3;; ‘ -
33333 6;”; I333” d: 33"“‘I 5“?
I... 1112?: ‘3” 33!
3U‘3V%fl%33
3 I _..
‘.I I l' _-5_
233325-3311-
- I3:
[33133 .. :I‘MI
£333 o
.-I' II
3‘3311‘13‘1““ 3 "
1W3 :5
I
‘3 .3 \ . 3'3 E34
3 I ‘3-MI .1 . W2“! 112-13 .
$13131 ”1"“?an “i 3,3 @1331‘.” 1-3333334331‘ £63339 39:3 ’I 33 33333§3I333 3'333‘333'33 If: -_ 2be
33 23.332“ {I313 15333‘11.'”3‘ '3 ‘ I22333333 ”WM" "" .13 3‘3
. 1‘2, ‘ 23'3": ' L3333fi 3f?
'Ltaw
“‘3'
73:33:-
1.22:2“?3
'n'2;;.
.4“
I32
W2233VWMI
“$3
DM- 'mvi
' 113 3
'”"‘W
1' ‘. 'I‘31rl iII'. ;‘.‘I‘
I33 {I} h“ Q”
cyan -
,..._
1:.-
4o:
1.
....
1,3...
1::
13.;
..:..
H.
b"
..
“‘.,,
2-3...
:
..
1:.
a:
3” 33133. 33 '
33“”33‘333 3333333 3131333333324 2:
35¢ .-
.95»
v
1.
-‘:.~“~ “I“
. -NM;,:".; "1'.
. - . ,1 ..
-.. _
w
.2...
.__v
Cut-1-
-1111“
. I"! .3 r I 3 ‘
.1‘1i”""' III! 3332 31323 ,331I313IIE"‘.H.. 13%;“ 3333333” I113“ 3333 431:3 .431, .1, 3-H -"'
"1:2, '333" “l .1 , 2,33 ,' 3,333,133“? .‘ . ”333. {.1 3% “6"‘331133‘. 3| ”3‘ ’ ”I I‘M; I
1. 11232222 222211221221
3%! ,,3Mm32 .35WIIII IW3‘3R%$fi2%Ihflfifiij%mW
22- ;2%1 73m €22 qu$m3QIM$2$W%fi
1‘" ‘32" " "3521' 22' ‘33.; 1I 2 I II, 3 2213.13: 222232'33.:2-2222I'2223
2:212 .31 .3122 . I I‘ {131‘ .21 . I 3' 22321 1222 "2223:?
II .33 3331 - I- I2:
.23 3” " = ' 3’ ’ " ‘-
b
" . “2"”3' é”
V
3 " 33 I
. '3 I ‘ 3 ’} ‘3 3 .2 .
.{3- 3' ' 3333;3‘33'33333‘ hi. I}! 33
; . , ,a 3' H I.
‘33 l? 3. 2‘31? E33333‘f333353‘33‘3333 3
=3 =3: -2 WIMIPdap
2 I in 3.9 It}! 3112343333.! +1273!“
. 3‘ :III “3‘33”- 2.21.2- 33‘ 2‘31:
I‘ll’i‘ l ,1- ‘2“; 3-4 '3‘! 1f- ‘3' 35:3 I"; 1
2333 '2 II 3232312 = ‘
.3122- ' 3'} , If “"1217? ‘:
. . ‘ 132333343 MEI. "32113315333233;-
3 I ’9’3‘33’ - 33333‘ :I.‘3:333I3323¥3::!I "
:31}, 'IIIIII 332333 ‘ I}; ‘ 1.3;522213II221153f
3322.2 ~u122u 22 2222233
Tfifib‘ls
2 000
lllllllllllllllllllllll\llllHilllllllllll
3 1293 02074
This is to certify that the
dissertation entitled
Buckling and Post-Buckling Analysis of Neo-Hookean
Plates and its Correlation to a Direct
Energetic Stability Analysis
presented by
Sangwoo Kim
has been accepted towards fulfillment
of the requirements for
Ph . D . degree in Mechanics
Wlflw
M ljm> professor
Date August 27, 1999
MS U is an Affirmative Action/Equal Opportunity Institution 0-12771
PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.
DATE DUE DATE DUE DATE DUE
moo mim.po5-m4
BUCKLING AND POST-BUCKLING ANALYSIS OF NEO-HOOKEAN PLATES
AND ITS CORRELATION TO A DIRECT ENERGETIC STABILITY ANALYSIS
By
Sangwoo Kim
A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Department of Materials Science and Mechanics
1999
ABSTRACT
BUCKLING AND POST-BUCKLING ANALYSIS OF NEO-HOOKEAN PLATES
AND ITS CORRELATION TO A DIRECT ENERGETIC STABILITY ANALYSIS
By
Sangwoo Kim
The elastic stability of buckling and post-buckling deformations for
incompressible neo-Hookean rectangular plate subjected to a uni-axial thrust is
investigated. The buckling deformation is described by the small deformation superposed
on finite homogeneous deformations. Throughout the investigation the thickness of the
plate is not limited. The resulting nonlinear boundary value problem is analyzed by using
the perturbation expansion method in which an associated linear problem is solved at each
order.
Buckling onset is determined from the first order expansion and can occur in either
flexural or barreling mode shapes with any integer number of half wavelength in the
direction of thrust. The solutions from the higher order expansion correspond to post-
buckling deformations. The higher order problems inherit information from problems of
the previous order, both by the expansion procedure and by the application of
mathematical solvability conditions. The stability criterion for post-buckling deformations
is based on energy competition between the buckled deformation and unbuckled
homogeneous deformation in the vicinity of buckling onset. It is formally established that
the energetic favoribility correlates with the load following character of the buckled
solutions (progressive buckling vs. snap buckling).
Based on the expressions obtained by these procedures, it is found that the flexural
buckled deformation is energetically favored over the unbuckled homogeneous
deformation when mode number is small, otherwise the homogeneous deformation is
favored. The barreling buckled deformation is always energetically favored over the
homogeneous deformation. This contrasts with previous results of Sawyers and Rivlin
(1982), who obtain essentially opposite results for the elastic stability of homogeneous
deformation. The approach to evaluate the elastic stability by using the perturbation
method considered in this research gives more insights to understand the buckling
phenomena and is systematically applicable to higher order analysis.
Besides the main topic of stability evaluation, several approximate schemes for the
critical buckling load in neo-Hookean three-ply sandwich type plate were developed in
view of practical application. The schemes are based mostly on the Rayleigh quotients
approach and trial solutions. These schemes can be expanded to general multi-ply
composite plates and so reduce the effort to determine the critical buckling load.
To my parents
for their endless love, support
and teaching me the value of challenge and perseverance
iv
ACKNOWLEDGEMENTS
Thank God for giving me the opportunity to write this acknowledgment at last. He
always guided and strengthened me through His Words and taught me the values of
humility and integrity.
During the course of my graduate school experience, I have learned a lot about the
knowledge of applied mechanics as well as the life itself from many persons. The first
person I should thank was my respectful advisor Prof. Thomas Pence who accepted me,
taught me the detailed works of scientific research and advised me with valuable tips of
life. Your practical remarks such as “when you stuck, step back and consider a simple
example” and “follow the principles” will exist firmly in my memory. Also you showed
me the importance of honesty and integrity and I apologize to have you hard times due to
my insincerity. I appreciate your continuous encouragement to me when I am suffering.
My thanks also goes to my dear committee members - Prof. Dashin Liu, who made
me to think about the theoretical and experimental works, Prof. Ronald Averill, who
taught me the applied methods of mechanics and the fun of scientific jokes and Prof.
Baisheng Yan who granted me the confidence for my works and made me to continue.
They deserve my gratitude for their valuable supports and contributions to my work.
Very little of this work could have been accomplished without the help of my
family for their continuous support and encouragement. I am thankful to my father and
mother, Chang-ll Kim and Sung-Si] Baik for their enormous pray and priceless support. I
hope I will have a chance to recompense their sufferings. Also I thank to my parents-in-
law Hong-Soon Shin and Wol-Hwa Kim for their timeless pray and rigid assurance to my
work.
Special recognition should be bestowed upon my close family members. I cannot
find how to thank to my only blood brother Prof. Junwoo Kim who rescued me from the
temptation to give up all my works by numerous emails and phone calls. His existence
have given me the strength to overcome every difficulties. And I owe everything to my
precious wife Kyoungsook Shin, who always showed the confidence in me and made me
comfortable in home. In addition, I deeply thank you for your patience. My first son David
made me to consider what is my problems and twin sons Samuel (Gom) and Daniel (Kan)
gave me not only sufferings but also the pleasures and reasons to live. The contributions
you all gave to my life are indescribable.
I also give my thanks to the staffs in Dept. of Materials Science and Mechanics for
creating warm and supportive atmosphere. Many friends and elders in Korean community
also deserve my thanks for their supports and advice to me and my family during our stay
in this town. Finally I want to give my appreciation to the people who maintain the nature
centers in Lansing area. In these places, I have relieved, thought, made a decision and
recovered my confidence. It is so fortunate for me to have these places in a close distance.
Now I am standing top of one small hill in the mountains. Soon I will climb up
another hill and may suffer the hard times. However I believe my experiences will help
much to overcome these obstacles. Oh God, bless me and my path, and strengthen me just
as you have given to me.
vi
TABLE OF CONTENTS
LIST OF TABLES ...................................................... ix
LIST OF FIGURES ...................................................... x
1. INTRODUCTION ..................................................... 1
1.1 Overview ....................................................... 1
1.2 Literature Review ................................................ 4
1.2.1 Elastic Stability ............................................ 5
1.2.2 Buckling Instability of Composite Plates ....................... 13
1.3 Thesis Organization .............................................. 14
2. PRELIMINARY WORKS ON NEO-HOOKEAN PLATE ..................... 18
2. 1 Introduction .................................................... 1 8
2.2 Problem Descriptions ............................................. 18
2.3 Bifurcation from Homogeneous Deformation .......................... 22
2.4 Energy Minimization of the Deformed Configuration ................... 26
2.5 Summary ...................................................... 27
3. BIFURCATION ANALYSIS BY PERTURBATION EXPANSION METHODS. . . 28
3. 1 Introduction .................................................... 28
3.2 Perturbation Expansion Methods .................................... 29
3.3 Analysis on Neo-Hookean Plate .................................... 31
3.3.1 Linear Differential Operator ................................. 34
3.3.2 Different Formulations ..................................... 35
3.4 Load Parameters ................................................ 37
3.5 Auxiliary Conditions Associated with Incompressibility Constraint ........ 39
3.6 Relation between Thrust and Load Parameter .......................... 41
3.7 Energy Formulation .............................................. 42
3.8 Summary ...................................................... 43
4. BUCKLING DEFORMATIONS FOR NEO-HOOKEAN PLATE ............... 44
4. 1 Introduction .................................................... 44
4.2 General Solution for the Governing Linear Differential Operator .......... 44
4.3 Nonhomogeneous Ordinary Differential Equation ...................... 49
4.3.1 Nontrivial Solutions to Homogeneous Problem .................. 49
4.3.2 Solutions to Nonhomogeneous Problem ........................ 53
4.3.3 Solvability Condition ....................................... 55
4.4 Buckling Onset (the First Order Solution k=1) ......................... 57
4.4.1 Buckling Initiation by F lexure and by Barreling .................. 58
4.4.2 Load Parameters Associated with Buckling Initiation ............. 60
4.4.3 Asymptotic Expressions for Load Parameters .................... 63
4.5 Post-buckling Deformation (Second Order Solution) .................... 70
4.6 Summary of Full Buckled Deformations .............................. 79
vii
4.7 Reduced Formulations Due to Symmetric Load Parameter ............... 80
4.7.1 Load Parameters .......................................... 80
4.7.2 Energy Equations .......................................... 82
4.7.3 Energy and Load Parameter .................................. 89
4.8 Summary ...................................................... 91
5. STABILITY EVALUATION FOR A NEO-HOOKEAN PLATE ............... 92
5. 1 Introduction .................................................... 92
5.2 Formal Determination of Stability Parameter .......................... 94
5.3 Numerical Determination of Stability Parameter ....................... 97
5.4 Asymptotic Study for Stability Parameter in F lexural Buckling at Low Mode . .
..................................................... 100
5.4.1 Critical Load in Classical Euler Buckling ...................... 101
5.4.2 Buckled Deformations with Asymptotic Equations .............. 104
5.4.3 Stability Parameters with Asymptotic Equations ................ 110
5.5 Discussion .................................................... 1 14
5.6 Summary ..................................................... 1 15
6. APPROXIMATE SCHEMES FOR BUCKLING LOAD OF MULTI-LAYERED
COMPOSITE PLATES .............................................. 117
6.1 Introduction ................................................... 1 17
6.2 Buckling Load of Multi-Layered Plates ............................. 118
6.3 Approximate Schemes to Determine Buckling Load ................... 123
6.3.1 Variational scheme ....................................... 125
6.3.2 Trial solutions ........................................... 126
6.4 Discussion .................................................... 129
6.5 Summary ..................................................... 131
7. CONCLUSIONS AND RECOMMENDATIONS ........................... 132
7.1 Conclusions of the Thesis ........................................ 132
7.2 Recommendations for the Future Work .............................. 134
APPENDIX A. STABILITY EVALUATION BY PERTURBATION EXPANSION
METHODS ........................................................ 137
A. 1 . Introduction ................................................... 137
A2. The Perturbation Expansion Method ................................ 139
A3. The Analysis Scheme by Sawyers and Rivlin ......................... 141
A4. Example 1: Elastica Problem ...................................... 141
AS. Example 2: Modified Elastica Problem .............................. 151
A6. Example 3: A Higher Order Problem Represented Neo-Hookean Buckling . . 154
A7. Discussions ................................................... 159
APPENDIX B. COEFFICIENTS FOR STABILITY PARAMETER .............. 163
BIBLIOGRAPHY ...................................................... 168
viii
LIST OF TABLES
Table 6.1 Summary of complete conditions and satisfaction for various schemes 131
Table A.1 Summary of the procedures and their corresponding equations for the
perturbation expansion. ..................................... 161
Table A2 Summary of the procedures and their corresponding equations for the
Sawyers-Rivlin method ...................................... 162
ix
Figure 2.1
Figure 2.2
Figure 4.1
Figure 4.2
Figure 4.3
Figure 4.4
Figure 4.5
Figure 5.1
Figure 5.2
Figure 5.3
LIST OF FIGURES
Description of the neo-Hookean rectangular plate under consideration.
The thrusts T are applied to the ends of surfaces at X1=ill and the plate
has a dimension of 211x212 x213. ............................... 19
The relation between the scaled thrust T5 and load parameters p and A.
The thrust is the compressive load so for this study, the load parameter is
restricted to 0
] .5, asymptotes is quite
close to exact expression ...................................... 69
Three-term asymptotic equation of load parameter p0 for barreling
deformation when 11 goes to infinity. When n>2.5, asymptotes is quite
close to exact expression. Otherwise, two expressions show totally
different values. ............................................ 70
Dimensionless stability parameter RS=R3/(ul,M4§23) for the flexural
deformation in (5.14) with v=1. At n=1.6283, the curve RS has a
discontinuity. Right before this, it is positive and after this, R5 is negative.
There are also sign changes in 11:0.6443 and n=1.305. Their details are
shown in Figure 5.2 and Figure 5.3. ............................ 98
The detailed curve of dimensionless stability parameter RS for the flexural
deformation in (5.14). At 11=nc=0.6443, the sign of RS changes from
positive to negative and at 11:] .305 the sign changes from negative to
positive. More detail near n=nc is shown in Figure 5.3. ............. 99
The detailed curve of dimensionless stability parameter R5 for the flexural
deformation in (5.14). At n=nc=0.6443, the sign of RS changes from
positive to negative. Hence when 112’2
) 213
211
Figure 2.1 Description of the neo-Hookean rectangular plate under consideration. The
thrusts Tare applied to the ends of surfaces at X1=i11 and the plate has a dimension
0f 2]] X212 X213.
The current configuration is defined by undeformed or reference coordinates so
that the deformation tensor is given by
x = x(X), (2.1)
where X(X,,X2,X3) and x(x,,x2,x3) are undeformed and deformed coordinates, respectively.
The deformation gradient tensor and Green’s deformation tensor are given as
F = (ax/6X), B = FFT. (2.2)
The condition of material incompressibility requires that the volume does not change afier
deformation so that
19
det F = 1. (2.3)
The Cauchy stress tensor for the incompressible isotropic hyperelastic material is then
given by
=_ 6W 5W) .. (6W) 2
where p is the hydrostatic pressure due to the incompressibility constraint, 1, and 12 are the
first and second invariants of B and W = W(I,,I;) is the strain energy density function of
hyperelastic material. The symmetric Cauchy stress tensor 1: described in a current
configuration is transformed to the Piola-Kirchoff stress tensor S in a reference
configuration which is given by
s = F41, (2.5)
after applying the incompressibility condition (2.3). The equilibrium equations in a
reference frame are then expressed as
DwsT=o. (am
The plate under consideration is subjected to a compressive load on each surfaces
X r = i1 1 . The boundary conditions for a frictionless thrust with an overall stretch ratio of
pare
S12 = SD = 0, on X] = ill, (2.7)
x1 = ipl, on X1 = i1,, (2.8)
where the case of compression requires 0 1 .
2.3 Bifurcation from Homogeneous Deformation
Motivated by boundary condition (2.11), we restrict attention to states of plane-
strain buckling taking place in the (X ,,X2)-plane. The buckling can be described as the
bifurcation from the solution of homogeneous deformation (2.14) so that the incremental
deformations of buckling are superposed on finite homogeneous deformations. The fully
finite deformation is then expressed as
X] = PXI +v,(X,,X2),
p“X2+ v2(X..X2). (2.18)
X3 = X3,
x2
and the pressure field is accordingly
22
TS ' w T s i
| 4 . I
8 _ I l
I l
. 2 .
comp essroh I .
6 r l , I COITIpI'CSSlOI'I
' o
4 . I l
l I
I '2 * I
2 r I i i
| _4 _ l
0 ' tension '
l
l '6 I
-2 1 I I
: tension -8 :
.4 l
| I
4 -10 1 -
0 1 2 3 0 1 2 7. 3
p
(a) (b)
Figure 2.2 The relation between the scaled thrust T s and load parameters p and A. The
thrust is the compressive load so for this study, the load parameter is restricted to
0
(X..X2>. (3.1)
9(X1rX2) = Zk_,3kp(k)(X1.X2)-
In vector notations with v=(v,,v2,q) and u‘k)=(u,("),u2("),p(")), the deviations v away from the
homogeneous deformation state are expressed as
v = sum + azum + e3u(3) + .... (3.2)
Here a is a measure of the amount of deformation away from the homogeneous solution
which is defined as
29
e = (v,u“))/(u('),u“)), (3.3)
in accordance with the orthogonality condition for the vector functions of each order 11")
and u“)
(u<1>,um) = o, j¢ 1. (3.4)
The brackets ( , ) denote the bilinear inner product on pairs of vector functions in a
domain 1'1 so that
__ 1
cm» — mliflmmdndxz. (3.5)
In order to acknowledge the evolution of the postbuckling path with the thrust, the
overall stretch ratio p is also expanded from the stretch ratio on buckling onset p0 as
P=%+2bfwr do
The stretch ratios p or 1. represent the load parameters as shown in Figure 2.2. and their
expansions are related by the expansion of equation (2.17) such that
lo = P62, M = 4135391. 12 = -pa3(202-3pa'pi),
(3.7)
13 = —2963(p3-3pa‘p1pz+295291),
where A=AO+SA1+82A2+83A3+0(84). For the simpler analysis, the expansions of p will be
used but for the interpretation of the thrust load, I. will be used afier converting by the
relations (3.7). If the odd terms in (3.6) vanish, pk=0, k=l,3,5,..., then the deformation
bifurcates symmetrically with respect to the trivial solution path. The symmetric buckling
which is common to the perfectly loading plate and the analysis much easier than the case
of asymmetric buckling.
30
3.3 Analysis on Neo-Hookean Plate
Entering the boundary value problem (2.28) with the expansions (3.1) and (3.6)
and collecting together common orders of a, give rise to the following general linearized
boundary value problem at each order 8":
“(“lfil + 1492) - Pfilpfi‘) = fl")(X,,X2),
mum+amrmwa=awana. inn.~ cw)
Pfilulii + Pouiii = fSk)(X1./X2),
ulk) = 0, “(um +p52ugf5) = 0, on 1“,,
“(149+ pazulf‘l) = gi")(X,,i12), on r (3.9)
“("53 -Pazui{‘l)—pop<") = g£*)(X,,i12), 2’
here k=1,2,.... The second in F, boundary conditions can be reduced to ugf), = 0 because
the first condition gives riff} = O. Primarily the above equation is a nonhomogeneous
second order partial differential equation except for order a (k=1). The expressions on the
right hand side in (3.8) and (3.9), f“) and g“), depend upon the lower order solutions
uU) = [ufl'h up, pm] and pj,j 933: G,u=0 on 1“,,qu = O on F2} and
let u e Hh(l‘I) be continuously differentiable. The subscript h as used here is to indicate
homogeneous boundary conditions. Notice also that Hh(I'I) is dependent on p0 because the
boundary operators G, and 62 depend on p0. In this Section 3.3, we disregard the order
superscript k in the equations for simplicity. Let F be the second order differential operator
defined in (3.14) and consider
(Fu,1—1) = 3717A ln[(Fu)Tfi]dX,dX,. (3.17)
Then integration by parts twice upon (3.17) with respect to the variables either X, or X2
yields
(F0. ii) = 4—llszl irziiMuL 1 " P62“; 2) - P6'Pll-‘1 + “(142,1 + P62”1,2)§21|rldX2
+ irllll(u1,2 + P62u2,1){‘1+{11(u2,2-P62u1,1)- Pupil-lzilrzdxi
—J'1‘2[{ M171, 1 - P132522) - P6113} “1 + “(172.1 + P6251,2)u2] ir,dX2 (3.18)
-irl[11(l-l1,2 + P132172, ,)u, + {P(l-42,2—P62{‘1. 1) — P0P} “2] lrde‘
+iiniu1{11(1-41,11+ 171,22)- PEIPJ} + “zillU-lz, 11 + l72,22) " P0132}
+P(P61{l1,1+ P0172,2)]dX1dX2]-
Applying the boundary condition requirements on u inherent in Hh(I‘I) to (3.18) shows
that the second of the two terms in the first integral vanishes, as does the entire second
integral, and the first of the two terms in the third integral. Hence
34
_ 1 , - _ _
(F0. 11) = miLliMMn’P6‘“2,2)—P61P}“1—P(u2,1 + P62u1,2)u2]|rldX2
_ ' -2' " _ —2" _ "
Ir,[p(u"2+p° “21)"1 + {11042.2 P0 ”1.1) Papiuzllrdei (3.19)
+ iinlu1{P(i-l1,11+51,22)‘P6‘P.1} +u2{P(l—12,11+l72,22)‘P0F,2}
+P(P6"-11,1+ Pol-‘2,2)]dX1dX2]-
This defines the adjoint operator to the triple consisting of the field operator F and the
boundary operators G, and G2. Let the associated adjoint operators be P" and G,°, 02°.
Then the integral over F, shows that G ,°=G ,, the integral over F2 shows that GZ°=G2 and
the integral over II shows that F‘=F. In other words, the linear differential operator F
restricted to Hh(I'I) is self-adjoint.
3.3.2 Different Formulations
The nonhomogeneous boundary value problems (3.14) and (3.15) can be
formulated in different ways (see Reddy, 1986). The weak (variational) formulation is
stated as follows: find u=(u,,u2,p) e H(IT) such that
B(w, u) = l(w), for all w e {(w,, w2,r)|w,=0 on F,}, (3.20)
where B(w,u) is the bilinear form and l(w) is the linear form given by
B(w, u) = HHH[(“1,1W1,1+ “1,2W1,2 + “2.1%, 1 + “2,2W2,2)
+ P62(u1,2W2,1+ "2,1W1,2 - “2,2W1,1- “1,1W2,1)]dX1dX2 (3-21)
-i.in[P(P6'W1,1+ PoW2,2) + '(P61“1,1+ Po“2,2)]dX1dX2-
[(W) = —,i,in(w,f, + wzf2 + w3f3)dX,dX2 + in (g,w, + g2w2)|r2dX,. (3.22)
The solution space H(I‘I) indicates the boundary conditions that should be satisfied
35
H(IT) = {(u,,u2,p)|G,u = O on F,' qu= g on F2}. (3.23)
Note that B(w,u) is symmetric, i.e. B(w,u) = B(u,w). For sufficiently smooth functions,
the weak formulation (3.20) is equivalent to the direct formulation (3.14) and (3.15).
The energy functional on H(FI) corresponding to symmetric B(w,u) becomes
(Reddy, 1986)
E(u)
ll
%B(u, u) - l(u)
= gliniuin + “12,2 + ”in + ”£2 + 2P62(u1,2u2, 1 - “1,1“2,2)]dX1dX2
(3.24)
’“n [P(P6'u1,1+ P0112,2)+ (“L/1 + uzfz + “3/3)]dX1dX2
—IF1 (8W1 + g2u2)|r2dX,.
The energy (functional) formulation is to find u eH(I'I) which minimizes E(u). If B(u,u)
is positive for u eH(FI), then the weak and energy formulations are equivalent. In the
problem under study here, it is not clear under what circumstances B(u,u) is positive. The
first variation of E(u) gives (3.14) as its Euler equation, when the following boundary
conditions are specified:
u(u -—p'2u )—-p"p = 0 or u = 0
on F, 1,1 02 2.2 o 1 (3.25)
P(“2,1+P6 “1,2) = O 01' “2 = 0
F “(“1,2+P62u2,1) = 81 01' “I = 0 (3 26)
on .
2 P(“2,2‘P62u1,1)—P0P = 82 or “2 = 0
In the condition (3.25) and (3.26), the right sides correspond to essential boundary
conditions and the left sides to natural boundary conditions. Comparison of (3.25) and
(3.26) with (3.9), shows that the boundary condition u, = 0 on F, of the direct formulation
contributes the only essential boundary condition.
36
3.4 Load Parameters
The load parameters p represent buckling behavior on and after the buckling
initiation. The homogeneous problem (3.16) for the case of k=1 will only have nontrivial
solutions for certain special values p0 which define buckling initiation modes. At these
special values p0, the differential operator trio {F ,G,,Gz} is singular. The same differential
operator trio as the homogeneous problem appears in (3.14) for the case of k=2,3,..., and
the special values p0 are used here. These cases of k 2 2 will be a problem for solving
nonhomogeneous boundary value problem (3.14) and (3.15) for a singular operator trio
{F ,G ,,Gz}. For most right hand sides, solutions will not exist. But for certain special right
hand side of equation, the solution can exist - eventually this is explained by the Fredholm
Alternative Theorem for solvability of the nonhomogeneous equation as developed next
for this particular problem.
At a fixed value of po, let w=(w,, wz, r) be a nontrivial solution of the homogeneous
equation (3.16) and let u=(u,,u2,p) be a solution to the nonhomogeneous equations (3.14)
and (3.15) for given 1‘ and g. Then consider the expression
(Fu,w)— (u,Fw) = fin“{(Fu)Tw—uT(Fw)}dX,dX. (3.27)
Since Fw = 0 and F u = 1' in F1, the left side of the equation (3.27) is equivalent to (f, w).
After integration by parts twice and applying the boundary conditions G,u = 0 and G,w =
0 on F, and qu = g and sz = 0 on F2 to the right hand side of equation (3.27), it
becomes
1 1 l
(‘3 W) = m[,l(W181+ W282)|_2,2dX1~ (3-23)
37
Thus if the nonhomogeneous problem (3.14) and (3.15) is to have solutions, then it is
necessary that f, g, and g2 obey the solvability condition (3.28). In particular, since u“) is a
nontrivial solution of the homogeneous equation (3.16), any nonhomogeneous solution
(u,“",u2“",p(")), k=2,3,... to (3.14) and (3.15) for given fl") and g“) must satisfy
llnwumdx— l, gr>+ urgent dX. = o. (3.29)
In order for the nonhomogeneous problem of order 82 (k=2) to have a solution “(2),
the condition (3.29) must be satisfied with corresponding terms 1(2) and gm in (3.11).
Substituting 1‘” and gm into condition (3.29) gives
R1+ P1R2 = 0. (3-30)
where R, and R2 are constants defined as
R. = ”niui”(p“)ul,‘l),.—u£"(p"’ui,'l),,]dx
+ I in 2P"’)+p9>-p<*>us,'1>1dx
— zupaln (ui”u$,‘l - um usnnrzdx. - inpmusuirzdxr (3'32)
If R2 is not zero, then p, can be expressed as
p, = —R,/R2. (3.33)
Solutions “(2) to the linear equations of order two, will exist only if the equation (3.33) is
satisfied. Similarly for the nonhomogeneous problem of order 83 (k=3) with 1‘” and g”) in
(3.12), the solvability condition (3.29) for the existence of “(3), gives
R3+R4P1 +R5P12+R2P2 = 0, (3-34)
38
where R3, R4 and R5 are
R. = lintui‘Kp‘Z’ulfl +pu5?1>,,— ut"(p‘2’ui,'l +p<'>u1?1),,
+ 2p‘”(ui,'lu§,2i - uiiium) (335)
+ u£}I(p“’ui?l +p<2>u1,'r)-u1}1(pus?) +p<2>ur,'))1dx.
R4 = linl-p52(ui”p,‘12’-ui?lp“’) +(u£”p52’-p“’u$?l)]dx
(3.36)
— 211p53ir1(u1'>u1?1 — u1?1us'>)|,2dX.- lrlp<2>uil>lrzdxh
R5 = linlpa3(ui”pfi”-p“’ui,‘l)1dX+3upa4irl(ui”u$,‘l —u£”ui,'l)|r2dX1, (3.37)
so that
P2 = —{R3 + (P1R4 + PiR5)}/R2, (338)
if R2 is not zero. The higher order parameter p,, i=3,4,5,..., can be obtained in a similar
way. Note that solving for a specific pN requires full determination of um’s, i=1,..., N. In
particular, the conditions (3.33) and (3.38) must be satisfied for the existence of solutions
u“) and “(2). The freedom to choose the coefficients p ,, p2,... in (3.6) are used to meet the
solvability condition (3.28). Each term p,- in the expansion will give a key to the
postbuckling behavior of the system. Budiansky (1974) has discussed the mathematical
structure of general post-buckling problems through variational analysis and shows that
equation of the form (3.33) and (3.3 8) can be expected in the general case.
3.5 Auxiliary Conditions Associated with Incompressibility Constraint
A set of extra conditions from (38);, and (3.9), are derived for simpler calculations
of necessary formulations. First double integration on (3.8); gives
39
1351”” ude+ pollnugfwx = ”f/S’de, (3.39)
and on (3.9)l gives
”nuif‘ldx = Ilik’lridXz = 0, (3.40)
for k=1,2,3...., so that
””2455ng = p51 llnfgwx. (3.41)
Using the expressions for f3“) given in (3.10) to (3.13) yields
”n “iiidX1dX2 0,
“it “EidXflIXz = PEICDI’
(3.42)
”n “iiidX1dX2 = - P1P62¢1+ p51¢2,
ll” ult‘lXmdXz = P62(Pi96' - P2)(D1_ ptp52¢z + 95W.»
where
«D. = lln(ui,'lu£,'l -u131u53))dxldxs. (3.43)
«>2 = ”n[(ui,'lu£?l + uiilu£}I)-(ui,‘1ul?l + ulflulflfldXthz, (3.44)
<1). -- M [(434.31 + anus?) + mum)
n (3.45)
411131145?) + 111,21 "El + ”Pl “Find/Y1“:-
Here the results of each step in (3.42) was used for the calculations of subsequent step.
40
3.6 Relation between Thrust and Load Parameter
After bifurcation away from the homogeneous deformation solution (2.14) and
(2.15), the associated values of the thrust T is obtained by substituting the expansion (3.6)
into the expression (2.16) as follows:
T = -4u1213[(po - 953)+(t>1+ 3P6401)€ + (P2 + 3p6‘pz - 695591082
+ (P3 + 3P64P3 - 1213559192 + 10966pi)e31+ 0(34) (3.46)
5 TO + ST, + 2:sz + 83T3 + 0(84).
The first term T 0 represents the thrust at buckling onset and the other terms in (3.46) will
show the change in thrust load after buckling. If the terms except To have a positive value,
then the thrust must increase to get a larger buckling deformation after buckling initiates -
progressive buckling. For the opposite case, the thrust must decrease when the buckling
deformation grows so that there is a possibility of snap buckling in which the buckling
mode jumps to another mode. If the bifurcation growth is symmetric with respect to the
trivial solution, then p,=p3=...=0 so that the thrust (3.46) becomes
T = — 41213P[(Po — p53) + p.(1 + 53.):2] + 0(24)
0
= To + {:sz + 0(84).
(3.47)
For this symmetric case, if p2 <0, then T is an increasing function of s so that the buckling
is progressive. Otherwise, if p2 >0, then T is an decreasing function of a so that the snap
buckling is possible.
41
3.7 Energy Formulation
The energy equation (2.30) may be expanded accordingly. Substituting the
perturbation expansion (3.1) and (3.6) into the energy functional (2.30) and applying the
incompressibility (3.8)3, the energy functional for each order becomes
AE = 213(3E, + ezEz + 83E3 + 8454) + 0(85).
where E ,, E2, E3 and E, are given as follows:
E, = ”nMpa'ulfl + poni31)dX1dX2,
E. = ”nulp1(ui,‘l — 95214le (pour) + Pa'ui?3)1dXthz + 551..
E3 = ”filliiqulii + P63(P12 " POP2)ui,li}
+ 91041.21 - pazulil) + (9014,31 + pa'u£?l)ldX1dXz + #512.
E4 = ”null psuifl + p6“(- of + 2909192 - 939044535}
+ {92141.21 + 963(Pf - popz)ul?l} + 9.042) - 9521453)
+(pou1‘31 + pa'usfiiidXtdxz + gs... + #313-
Here we have introduced the notation:
at} = ”nWlpiulfh + ulfzulfb + “if’1u§{)1 + ugf)2u§{)2)dX,dX2.
(3.48)
(3.49)
(3.50)
(3.51)
(3.52)
(3.53)
Now direct application of the extra conditions (3.40) to (3.42) into the equations (3.49) to
(3.52) gives the following simplifications
42
(3.54)
(3.55)
133 = u((-291953)¢1+ 962432) + #312, (356)
E4 = Iii P63(3P12P6l " 2P2)+S.(l/C.dX2+L2)} (4.40)
+ p31C21iisde2 + 17,) + 52(‘ liczdxz + 472)}1.
B(Xz) =
The function f is specified in the right hand side of (4.11). The complete solution B
contains four as yet undetermined constant coefficients E, 1:, 117,, 117, that are available
to satisfy the four boundary conditions (4.17). Substituting the solution (4.40) into these
four boundary conditions (4.17), the four equations for these constant coefficients are
expressed in matrix form as:
C4X4J4xl = G4,,“ (4.41)
where
T
J = [L1 ,2 M1 M2] , (4.42)
54
92 F 2 ~ ~ - '
é, (1+ pmro + L3811+ 296“{M1*C2 + 41352}
G ___ of _ (1+964){L1C1-L5251}+29641MIC2-M552} (4,3)
g2 29641Li51+L2+C1i+962(1+064){M152+M3C2}
_2P64 'L151+L2C1} + 962(1 + P64)l—M152 +M2C2}d
and the coefficient matrix C is the same as that appearing in the homogeneous problem
(4.20). The fianctions with superscripts + or - denote that the functions are evaluated at
X2=12 or -12 respectively. Note that the matrix C is singular if and only if p,, satisfies either
(4.25) or (4.29). If p,, does not satisfy either of these equations, then (4.41) yields a mnque
vector expression for J. On the other hand, if C is singular, then (4.41) gives solutions if
and only if G is orthogonal to the null space of CT which generates the solvability
condition as we will show in the following section. For this singular case, if the solvability
condition is not satisfied, then there is no solution to the problem (4.16) and (4.17). If the
solvability condition is satisfied, then there exist infinite solutions which is shown in
(4.40). The constant vector J can be obtained by using the psudoinverse matrix of the
singular matrix C or by connecting the solvability condition to (4.41).
4.3.3 Solvability Condition
r
The existence of a solution J = [L 1 L2 M1 M2] for the linear algebraic equation
(4.41) when p,, satisfies either (4.25) or (4.29) requires satisfaction of an orthogonality
condition with the right hand side vector G and the null space of CT. If p,, obeys (4.25)
which corresponds to the symmetric case, then the null space of CT is given by
55
~ ~ ~ ~ T
NSS=Q3(1-p54) s, 4 514 C14 C14 . (4.44)
1+P6 1+Po 2130 2130
If p,, obeys (4.29), which corresponds to the antisymmetric case, then the null space of C T
is given by
~ ~ ~ T
C C s 5
NS = Q3(l—p‘4) I I I 1 . (4.45)
A 0 [1+0641+96429642P64]
Together the orthogonality of the null space and the matrix G can be written
67- NSSorA = 0, (4.46)
and so yields the following solvability conditions with the relations (4.25) and (4.29). For
the case associated with (4.25) and hence a symmetric nontrivial homogeneous solution,
[2
i~C CdX=1 4C1~+QS‘~ 447)
r2f(1—S 2) 2 —( —Po )[2p—6482 mgi] . (~
..[2
For the case associated with (4.29) and hence an antisymmetric nontrivial homogeneous
solution,
S] ~ QC] ~ ]|12
-l
£27151 - 552)dX2 = —(1 — Pfi4)[m82 + (1756—4381 (4-48)
Thus, if there exist nontrivial solutions 8,, to homogeneous versions of (4.16) and (4.17),
then there exist solutions to (4.16) and (4.17) with nonzero f, g, g2 if and only if these
functions satisfy either the condition (4.47) or (4.48), as appropriate
The solvability conditions (4.47) and (4.48) can be also obtained directly from the
ordinary differential equation problem (4.16) and (4.17) via similar procedure developed
56
in Section 3.4 to get (3.28) for the partial differential equation problem. For the problem
(4.16) and (4.17), exchanging u to 8, w to 8),, and f to f in the formulation (3.27) gives
(BB, Bh> — <13, B8,) = 3711,31. [(138)81. — B(BBMdXZ. (4.49)
2
Here F is the differential operator used in the left hand side of (4.16) so that i: B = f and
133,, = 0. Applying the integration by parts and boundary conditions in (4.17) yields
[2
~ 1 ~ ..
(£31.) = —(ngh‘gIBh') - (450)
41,12 —1
Then substituting B, in (4.26) and (4.30) into (4.50) gives the same conditions as (4.47)
and (4.48) after rigorous calculations with the relations of (4.25) and (4.29). Hence the
condition (3.28) represents solvability at the partial differential equation level, the
condition (4.50) represents solvability at the ordinary differential equation level and
conditions (4.47) and (4.48) represent solvability at the linear algebra level. The three
conditions (3.28), (4.50) and either (4.47) or (4.48) for solvability are perfectly matched to
each other.
4.4 Buckling Onset (the First Order Solution k=1)
The first order solution u“)=(u,(”, uz“), pl”) of homogeneous boundary value
problem (3.16) represents the status of buckling initiation and was investigated by
Sawyers and Rivlin (1974). Since all the right hand side terms 1“) and g“) vanish, the
terms 1,, J,, K, in (4.7) and G, H, in (4.8) as well as the right hand sides of the equations
(4.11) _and (4.13) vanish. Therefore the boundary value problem for the case of k=1
becomes
57
Bf;1.””—912.n(1 + P6039?" + 914;..pr ’ = 0. (4-51)
(1
IL”
81.1.2“ 024541312 0.
F - 4.52
31.L"”-Q?;.(1+2064)B,‘.‘,2'—0, °“ 2 < >
In view of the relations (4.12), the other coefficients in (4.3) become
2
A112 = 5" 8112', C112= up—§(Bf.1"'-Qin8112)'- (4.53)
i;n aim
The first order solution of (4.51) and (4.52) can be derived from the solution of
homogeneous problem in (4.26) for the symmetric case, and the solutions of (4.30) for the
antisymmetric case by substituting 85],) into B), and adding subscript i to Q and 7].
4.4.1 Buckling Initiation by Flexure and by Barreling
Two different kinds of solution in view of symmetry represents two shapes of
deformation: flexural and barreling deformation. For the flexural deformation, the lateral
deflection uz“) is symmetric with respect to X2, so that B};‘,,)(X2) is an even function. It
follows essentially from (4.25) - (4.27) as
811209) = M1cosh(n.-..Xz)-s.,.cosh(n.,.paZX2)1, (454)
under the solvability condition of
(1 + p64)2tanh(nmpaz)—4pa‘5tanhn,-.. = 0- (455)
Here 1],”, which is given from (4.1) and (4.23) as 11,," = lez. The constant M denotes
the amount of buckling from homogeneous deformation so that M will be determined
according to the normalization convention. The value of M will be determined later in this
58
section. For the barreling case, the lateral deflection is antisymmetric so that B§,',,)(X2) is
odd function and expressed from (4.29) - (4.31) as
31,1206) = Mlsinh(Q.-;nX2) - S.;nsinh(91;nt>62X2)l, (456)
under the solvability condition of
(1+ pa“)2tanhn.-.n-4paétanh(n,~..paz) = 0. (4.57)
Nontrivial solutions will only exist, at fixed mode number n and initial geometries I, and 12
(which will be shown in 0,," and 11,"), for particular values of p,,. The conditions (4.55)
and (4.57) show the relations between p,, and 11,," which represent the load parameter
curve. The aspect ratio 5 for both types of deformation is
_(1+ 1» ) sinhzn...
2p0 Sinh“ + 962M”, + vsinh(1— p62)n,m’ (4-58)
i;n
where v=1 (v=-1) for flexural (barreling) deformation.
The value of the constant coefficient M in either (4.54) or (4.56) is determined
from normalization for which we define as
4171,,” [(11,113 (Lf:’)2+(8,‘1'_’)jdx,dx, = 1- (4.59)
This is in contrast to the normalization used by Sawyers and Rivlin (1982) who instead
require that u§‘)(l,, 0) = :1 for flexure and u5')(l,, 12) = i1 for barreling deformation.
Their results are simple and procedures are relatively easy by introducing above special
rules of normalization. But in this work, we will follow the definition of normalization,
Hum" = 1, as shown in (4.59). Applying the solutions (4.2) for k=1 to (4.59) with the
relations (4.53) and integrating over F, give
59
ii ( p6 B"")2+('B— "2+3 pg 1”"— 92 B“)' 2 (1X — 1 460
412 F2 (T3171 1;" [2 Dim i;n 1;") 2 _ ° ( ' )
Then substituting the solution expression for B,_.,,“) in the separate cases of flexural
deformation (4.54) and barreling deformation (4.56) and the aspect ratio 5 in (4.58) into
(4.60) and applying the solvability conditions (4.55) and (4.57) give the following
equation,
M2 2 4 nh(2n.,.) sinh(2952n.-.n))
_ + +0 2 2 2 _ —4 2 _
4—I§[{V(l 0' Po) (1 Po) T—Smn “711,115 90(1 Po ) (V 296271.35.
nh2( 052111..)
+s2 v l—cr2 +1+cr2 4.61
l ( ) ( )Si 295211,. ( )
sinhn,.,,(1- P62) SiflhT], n(1 + p02)
—2svl—0'22 ' +l+o'22 =1.
i( W 111511-952) ( p0) n1,"(l+p5)
where o=lz/l, and v=1 for flexural deformation, and v=-1 for barreling deformation. When
the geometry 1,, 12 and mode number i, n of given plate are supplied, the value of M for
flexural or barreling deformation is determined by (4.61).
4.4.2 Load Parameters Associated with Buckling Initiation
The load parameter values p,, at which the plate may initiate buckling, is dependent
upon the initial geometry through mode number, n”. Their relations for the case of
flexure and barreling deformations are given in (4.55) and (4.57) and are shown in Figure
4.1. Sawyers and Rivlin (1974, 1982) first reached these first order solutions in terms of
A=p0’2 and n by using the variational approach. It can be shown for each fixed value of i
and for each n (which then specifies a value for n,,.,,>0) that there exists exactly one
solution of which satisfies (4.55). Also for each fixed value of i and for each n, there
60
Po
0.9 b
Flexure mode
0.8 .
0.7 -
0.5 . p,,,
p—————————_—
0.5 ..
0.4 . Barreling mode
0.3 .
0.2 -
0.1
Figure 4.1 The load parameters for flexure and barreling modes. When 11 goes to
infinity, both modes converge to pm=0.543689 as shown in (4.65).
exists exactly one solution p}? which satisfies (4.57). Therefore there exist only two
possible solutions for each fixed Q”. The indices i=1 or 2 and n=1,2,3,... determine
special values for each solution
po = p57(i;n), p63(i;n), i = 1,2, n = 1,2,3, (4.62)
If p,, is not equal to one of the two special values, then no solution, other than the trivial
solution u")=0, exists for the homogeneous problem of k=1. That is buckling can not
initiate at loads other than those given by (4.62). On the other hand, if p,, is equal to one of
the special values given by (4.62), then the solution of the first order problem consists of
61
the single function from each of the infinite series in (4.2) that corresponds to the
particular i and n which satisfy (4.62).
Note that the two curves in Figure 4.1 are each monotonic and approach the
common asymptote p,,, as n—>oo in which they share the same load parameter. When
n-—>oo, both tanhnw and tanhpaznw have the same value since of is finite and hence the
conditions (4.55) and (4.57) lead to
(1 + 9.54? - 49.3" = 0. (4.63)
The solution except for trivial case of pw=1 in (4.63) is the real root of
95.6 - 312.54 - 13;} - 1 = 0. (4.64)
and becomes by using Cardano’s solution for cubic equation (see Qiu, et al., 1993)
poo = [1 +3JR+ iQ3+R2+3~/R_ ,lQ3+R2]-“2 = 0.543689... (4 65)
Q = —(4/3), R = 2.
The deformed shape of the plate at n—wo involves an infinite number of wrinkles. In view
of the loading mechanism, when the thrust T is increasing, the load parameter p,, is
decreasing starting from p0=1 according to (2.16) so that the buckling initiation modes
occur sequentially such as
Tf0 and n—mo. This gives four separate cases
corresponding to: (i) flexural deformations at low mode number ((4.55) as 1190); (ii)
barreling deformation at low mode number ((4.57) as 71-90); (iii) flexural deformation
corresponds to wrinkling ((4.55) as n—-)oo); (iv) barreling deformation corresponds to
wrinkling ((4.57) as '11—)00). Later in Section 5.4.2, the low mode number flexure
expansion (i) is used in an asymptotic stability analysis. The other cases (ii)-(iv) are given
here for completeness.
(i) p,, for flexural deformation when 11 goes to zero
63
When 11 goes to zero for the flexural deformation in (4.55), the parameter p,, goes
to one as shown in Figure 4.1, so that we assume the series polynomial expansion of p,, as
p,, = 1+k,n+k,n2+k3n3+... (4.71)
The hyperbolic tangent can be expanded in a series form when n has small value as
1n3+3n5—... (4.72)
tanhn=n—§ 1
Introducing (4.71) and (4.72) to the condition (4.55) gives the polynomial equation of n in
which each coefficient function vanishes simultaneously. The lowest order becomes 113
and its coefficient shows
k, = 0. (4.73)
Then the coefficient firnctions after substituting (4.73) become
l3§k,(1+ 3k,)n5 + 161.3(1 + 215,),6
3
+ 8(— $9., — 1:31., — 112k, + 16k; + 13—6k4 + 32k,k,)n7 (4.74)
4 19 1
+16(fik3 — -3—k,k3 — 211531., + 2k,k, + ik, + 2k,k, n8 + 0(119) = 0.
Because of small 11, each term in (4.74) vanishes simultaneously so that the first term
gives
k, = __. (4.75)
excluding the trivial case, k,=0. Subsequently the other coefficients show that
_1 19
k5 = 0, k6 = "Tg'g—O, k7 = O. (4.76)
64
Therefore the asymptotic equation for p,, when 1] goes to zero gives
1 1 19
which is shown in Figure 4.2 up to the fourth order accompanied with the exact values
which is computed by using (4.55). As we can see in (4.77), p,, is an even function of n.
The asymptotic equation (4.77) were also obtained by Sawyers and Rivlin (1982) in terms
oflask = p57- =1+§n2+gn4+0(n6).
Po
0.8 .
Exact expression (4.55)
0.6 .
Asymptotic expression
0-4 - up to the fourth order '
(4.77)
0.2 .
0 A 1 A A J
0 l n 2
Figure 4.2 Asymptotic equations of load parameter p,, for the fiexural deformation
when n has small value by using (4.77) up to the fourth order. For n<1, asymptotic
expression is quite close to exact expression.
65
(ii) p,, for barreling deformation when 11 goes to zero
For the case of barreling deformation shown in (4.57), the parameter p,, goes to
zero when 11 goes to zero. However npo'2 in hyperbolic tangent does not go to zero. A
consistent analysis of possible forms for the expansion of p,,(n) near n=0 shows that the
correct expansion form is
Po = km “2 + km + km” + 15,112 + km” + 0013). (478)
Then the expansion
11ng = k72—2k2kf3n “2 + (3k3-k,‘4 — 2k3k,‘3)n + (4.79)
so that
lim0(np52) = kfz. (4.80)
n->
Substituting (4.78) into the condition (4.57) and the Taylor series expansion of tanh(np0'2)
with respect to the value of (4.80) give an algebraic equation of n in which each
coefficient function vanishes. Then the first term becomes
k73(1 - 4k',’-tanhk,‘2)n‘3 = 0 (4.81)
which shows the numerical value of k,=0.500169... The following terms show that k,, k;,
and k4 equal zero and the fifth term shows
k,(1—6k?)
k5 = 24(- tanth;2 + 315301511153)
(4.82)
Using the value of k,, the equation (4.82) shows k5=-0.05234..., numerically. Hence the
asymptotic equation of p,, for barreling deformation when 11 goes to zero becomes
p0 ~ 050016911“2 - 0.05234115/2 + 0(03) (4.83)
66
which is shown in Figure 4.3 with exact equation.
0.6
Po
Exact expression (4.57)
0.5 -
0.4 .
Asymptotic expression (4.83)
0.3 1-
0.2 -
0.] I
Figure 4.3 Two-term asymptotic equation of load parameter p,, for barreling
deformation when n has small value by using (4.83). Again when n<1, the
asymptote is quite close to exact expression.
(iii) p,, for flexural deformation when 1] goes to infinity
When 1] goes to infinity, the parameter (3,, goes to a finite value pw so that p,, can be
written as
Po = 9.11 +800}. (4.84)
where 6 is remainder term that vanishes as 1] goes to infinity. The hyperbolic tangents are
expanded by using infinite series of exponential equation as
67
= __ -_— _ —2 —4 _ —6
tanhn 1+e-20 1 2e rI+2e '1 28 0+... (4.85)
tanh(n 952) = 1 — 262"“? + 2.540962 — 2e'6np52 +
where
npa2 = np;.2(1—25+382—483+...). (4.86)
The possible form of 6(1)) can be obtained from considering (4.85) with the conditions
(4.55) or (4.57) as
5 = k,e’7-‘1 + k,e‘4" + (4.87)
Substituting (4.84) and (4.85) into (4.55) for flexural deformation and equating each
coefficient function of each order of e to zero give algebraic equations which contain the
unknowns k,. The first term which is independent on 11 shows
(1 + p54)2-4p;1’ = 0 (4.88)
and the real solution to (4.88) give a limit value of pw=0.5437... excluding the case of
unloading, p0=1, by using Cardano’s rule as shown in (4.65). The next leading terms show
that
2
k1= p. ,
1-3p$+pw 489
k =(9-21p.2.+51>;1)Ic%--2p.2.(1+6k.) (' )
2 204917-91) ’
with the numerical values of p,,, the parameter p,, when 11 goes to infinity becomes
p0 ~ p,,,(l + 1.47395e72'1 + 2.98066e‘4" + ...), (4.90)
which is shown in Figure 4.4 with exact solution.
68
Po»
0.9 -
0.8 -
Exact expression (4.50)
/
0.7 -
0.6 . f
Asymptotic expression (4.90)
0.5
0 l 2 3 4 5 6 7 8 9 1110
Figure 4.4 Three-term asymptotic equations of load parameter p,, for flexural
deformation when n goes to infinity. When n>1.5, asymptotes is quite close to
exact expression.
(iv) p,, for barreling deformation when r] goes to infinity
For barreling deformation, the similar approaches can be used in the case of
flexural deformation when 11 goes to infinity. The expansions (4.84) and (4.87) are also
possible candidates for barreling. Then the results shows that
k _ (1 + PW
‘ _ 4(1- 393. + pt)’
(4.91)
_ (13-4295+1098)ki+(1+P$)(1+P$-8k1)
2 _ 4(1-3p?.+p:‘.) ’
which determines the asymptotic equation numerically as
p,, ~ pw(1—1.47391e-2'1 + 34,94543—40 + ...). (4.92)
69
The asymptotic equation (4.92) with exact equation are shown in Figure 4.5. and it
0.6 2 .
Po
Asymptotic expression (4.92)
0.55 . 1
1
Exact expression (4.57)
05 4 A n . . . . 1 1
0 1 2 3 4 5 6 7 8 9 10
1]
Figure 4.5 Three-term asymptotic equation of load parameter p,, for barreling
deformation when 11 goes to infinity. When n>2.5, asymptotes is quite close to
exact expression. Otherwise, two expressions show totally different values.
is shown that the asymptotic equations are well matched when 1] has larger values in both
Figure 4.4 and Figure 4.5.
4.5 Post-buckling Deformation (Second Order Solution)
The higher order solutions “(2), u‘”,... are the expansion in e as given by (3.1) to
account for the postbuckled deformation. This is determined by the nonhomogeneous
nature of the equations which effect f and g in (3.8) and (3.9). To determine the second
order solution bifurcated from one mode of the first order solution, we now choose either
70
fiexure or barreling and also fix i=1 or 2 and mode number n=1, 2, 3... in the first order
buckling solution u“) in (4.68) to (4.70). It will be convenient to rename i to j and n to m
as we wish to use j and m in what follows as the expansion indices for the chosen first
order mode. Thus for fixed j=1 or 2 and m=1,2,3,... as determined from the bifurcation
mode under consideration, we seek the coefficient functions BM”), A ”(2), Cm”), i=1,2,
n=1,2,3,..., in the second order case.
For the zeroth terms, the general equations (4.14) and (4.15) with the right hand
sides of the second order equation (3.11) become
II p8 ' I I I I I
11382) “—90%” = -u,—,—,—;(B,..B11.2 ) 5 90362) = OMB/($13}; ) ,
" 5 (4.93)
. Po . .
on F, 11832) —poC62) = #155519» 3,93 .
where
Brim = B};1ni"_'sz;InBJ(;iri° (4°94)
To within the rigid body motion given by a constant in 80(2), the solution for (4.93) is
I I I p8 I I
812’= 1258112810. C19 = {(311.2319) Jrq‘im 311.2] (4.95)
For the other coefficient functions, we will utilize the generalized solution for
nonhomogeneous equation, (4.43) with the solvability conditions (4.47) and (4.48).
Substituting {(2), gm in (3.1 1) into the equations (4.7) and (4.8) shows the following results
for the right hand sides of the equation (4.9) and (4.10):
(A) If the mode of u“) is j=1 and m, then only the expressions for i=1, n=m and for i=2,
=2m-1 of “(2) are non-vanishing:
71
Iizrii = —PIPBZQI;mCii}n Jig): = plCiirir'a
K132. == -p1(31!.’.'+p5201..A1!.2.). (496)
G132. = 2119195391....3111. H132. = 91(C112.+21195301;mr41!3.),
and
.... NIH [\N‘“b
112.3..-1 01,m(C1!.1.'31§2.-C1!,2.31!2.').
0an 915461354112.-C1§2.All.2.'),
(4.97)
Kiym-l = iQI;m(Aiirir’Bilrir _AilrirBiirir')a
l l
GSQm-l = ioleiirirCiirir’ Hiiim—r = iQI'JnAi‘IiIBiirir
(B) If the mode of u“) is j=2 and m, then only the expressions for i=2, n=m and for i=2,
n=2m of "(2) are non-vanishing:
[£2121 = _plp0202mC3irin J39: = plciirir'a
K113. = -p.(31!.2.'+p5202-,.A11.2.), (4.98)
0132. = 20910530253131, H132. = 91(C122.+ZMP5302..A1!,2.),
and
1 . .
119.. = —,92..(C1!.2. 8121-6111810. 1.
1 I I
J19. = 792.4610. A112. $121419. 1,
1 (4.99)
K59»: = —§QZ;M(A$irir'B£lrir-A$irirBii,rir')9
1 1
059m = —§QZ;InB£itiIC£irin H53»: = —§QZ.MA$LZIB£BV
For the modes which are not mentioned above, the right hand sides of the equation (4.9)
and (4.10) vanishes so that these modes have the same solutions as those for the
homogeneous case or u“). However, according to the orthogonality rule (3.4), the solution
72
of these modes must be eliminated. For the cases of n=m in (4.96) and (4.98), the
differential operator are the same as those of homogeneous equation so that in most cases,
the solution does not exist. But for the special arrangement of the nonhomogeneous terms
which satisfy the solvability condition, the solutions exist. Substituting (4.96) and (4.98)
into the right hand side of (4.11) and (4.13) with the relations (4.53) for Am,“ and C111,,"
gives
]=-4plp5'(B11."'-Q-2-3.1”)” (4.100)
1’" 1’"
and
g“, = ‘2P196'(B}-,in)"—sz;m 9643193)
2'. = 49195'(B,1.,.""-Q?BB-'-1'2)
Jv’" 1’"
(4.101)
Substituting from (4.100) and (4.101) into the solvability conditions (4.50) with Bh=B“)J-_.m
and integration by parts to the left hand side gives
2p1p5'[{(B,<,0" + 0,2,..prB<1.2)B<1.> 1|:
(4.102)
—2 1: {(31.202 + 9,2..(B1.‘,.2)214X2] = 0.
Considering the first order Fz-boundary condition (4.52),, the condition (4.102) becomes
p,=0 since the integral is positive definite. Hence the solution for the case of n=m exists
only if p,=0. If the first order load parameter p,=0, the right hand sides, (4.96) and (4.98)
vanish so the modes n=m have the same solutions as the homogeneous solutions. Due to
the orthogonality between the first order and higher order solutions, the solutions of
modes n=m must be eliminated. Therefore if the mode of j=1, m is chosen for u“), then
only the mode associated with i=2 and n=2m-1 in the second order solution expansion “(2)
73
is governed by an equation that has nonhomogeneous terms. That is, only Bfim, has
nonhomogeneous solutions so that 81%,} = 0 for all n and B5?) = 0 unless n=2m-1. On
the other hand, if j=2, m of u“) is chosen, then only the mode i=2 and n=2m, that is, 8533",
has nonhomogeneous solutions. Thus for fixed mode variables m=1,2,3,... and j=1,2, there
is exactly one nonzero B3,? which corresponds to only BS} governed by a
nonhomogeneous equation. The index i for this nonzero second order B3,) is always i=2
and n is given by n=2m-1 if j=1 and n=2m if j=2. Note that the mode number n of the
second order is different from the mode number m of the first order so that the differential
operators of the second order are different from those of the first order. Hence there exists
a unique nontrivial solution to the second order equation (4.16). After combining the two
cases in which the nontrivial solutions are possible to exist, the right hand sides become
1 l
1 Q—TWYIMSan—la 1 g—lfi—MKlmsnflm—l’
11?) = 31198 1 J1?) = 51198 ,
-mY2m5n2m9 -mx2;m8n;2m’ (4.103)
1 0’] 'msn‘Zm— 19
K?) = _ 2( ’ ’
in 2‘30 —O'2:m6n;2m2
and
'_1_'Bl Biirirsn'Zm-lr LBlunBiirir'an'Zm-l’
, 1 9....2"'~ ,1 or... '
G1,.) = 51198 1 H§,,2= 511123 1 (4.104)
'KTz—BimBiirirfinfim’ _EB'ZynBilyir'éan’
where the upper terms are for j=1 and the lower terms are for j=2 and A1,," and C”, have
been converted to Bj_.,,,. The notations used above are
74
(1., = B};1Iri"Bj(;lni‘-Bj(;lni'Bj(;iri'9 71;»: = B" BU)... iLMBJ(#i'i"
Jv’" 1;"! 14’"
= " BU)’ — 8,.,,,B<.1>"
Kj;m j;m jLM 1,"! 9
(4.105)
and SM is the Kronecker delta (6M = 1 ifp = q and SW, = 0 ifp i q).
For the equations (4.103) and (4.104), the mode numbers i and n in the left hand
side of (4.9) and (4.10) are changed accordingly. And from the definitions of mode
number in (4.1), Q," can be written in terms of Q”, as
02:2711-1 = 291m: Q2;2m = 292;».- (4-106)
so that the formulations for the second order can be considered as the generalized
formulations developed in Section 4.2 with 29m in place of Qw- Applying the relations
(4.106) and the right hand side terms in (4.103) and (4.104) into the boundary value
problem (4.11) and (4. 13) for both cases of j=1 and 2, have the following format
B12)” — 4912,,(1 + p54)B(2)" + 169,1, (5543(2)
' 1 .. ' ~ (4.107)
= mPO(Y/‘;m — 2Kj;m + iaJ‘M—ZQJZWGJW) 51(2)’
011 F2,
II 1 ' I ~
8(2) + 4szynp6430) = mp0(§aj;m + BjJflBJ(.lni) E gin,
B(z)m-4QJ'2;,,,(1 + 2p64)B(2)' (4.108)
1 . . r "
= wpo[Yj;m + iajmr " 2(1 + P04)sz;majzm " zflj;mBj(;iri :15 81(22):
where
3522m_ 2 19 for . = 1’
B<2> = ’ ' 55 = ( I, (4.109)
359m, _1’ for J = 2
75
According to the relations (4.12), the other coefficients become
2
Am 90 (3121233100, m)
2:2. 2
m
4.110
C121 —- 1153 B121"'—4Q.2.,,,B(21'wp0 —4Q2.55 +2 ( )
_ 40}, l ‘T(°‘Bm" m a)... 175)],
where
A2 , c2m_ , -___
A12) = 53”“ Co) = 5'3 ‘ J 1’ (4.111)
A13... C13... = 2-
Applying the first order solutions (4.54) and (4.56) with zero right hand side into (4.94)
and (4.105), then the right hand sides of (4.107) and (4.108) becomes
1‘” = -5500 2,..(p5 _15') 5,.,,,, (4.112)
and
812’ = l912591-2-...(7p5“+ 03,15,231!»
.III
2 ’ (4.113)
g1” = -255,a,2,,(2p,4 —1)(B}}.2')2
Substituting 3,1,1”; for flexure in (4.54) or for barreling in (4.56) reduces the equations
(4.112) and (4.113) to
im= ,3,,-va2s..9.-..p5(1-pr)21v(1-pa2)sinh{n,..<1+55%}
+(1+p5z)sinh{Q--,m(1—952)X2}].
(4.114)
gi2) = 4911421300130'1'1)Qj3yn[51nh(29j;n-X2)+5};m9625mh(291m902X2)
—Sj;m{ (1 + 962)Sinh{ 0);..(l 7‘ 962))(21 + V(1 - 962)Sinh{Q,-,m(1 - (262)21’2} } 1.
(4.115)
76
812’ = —90sz5(21>54 -1)Qj‘,m[cosh(2§2).mX2)+ 3,2;m964005h(201;m1362X2)
(4.116)
- V(1 + .9295“) - 28);..9521 cosh { 9,...(l + p52)X2} - vcosh { Q,...(1 - p52)X2} 1].
where v=1 for flexural mode and v=-1 for barreling mode. The general solution to the
differential equation (4.107) with the boundary conditions (4.108) has the expression
B121 = N, sinh(2p52(2,,,,X,) + N, sinh(20,-.,,,X,)
. . (4.117)
+1\1_w.smh{(p52 + 1)Q,,..X2} 211’5smh{(p52 ~1)Q),..X2}-
Here the first two terms are from the homogeneous part and the rest are from
nonhomogeneous part with constants N,’s. Introducing (4.117) to the differential equation
(4.112) and matching the coefficients give N3 and M, as
_ 3 (954-1)(p52+1)
N3 — mesjimflji’"p°(3p52+1)(p52+3)’
3 (p54— 1)(p52- 1)
N4 = -O)VWSj-;mflj;mp0(3p62_1)(p62—3),
(4.118)
and to the boundary conditions (4.113) give the values of other coefficients as
_ wngmpa
' _ 32A
(011129.. p0
= 1""
N2 16A
[4954D1905h(211)‘ (1 + p54)Dzsinh(2n)],
(4.119)
[p5zDzsinh(295zn) — (1 + PB4)D1°05h(290211)],
where
= 4p5fisinh(2952n)cosh(20) — (1 + p54)zcosh(295zn)sinh(20). (4.120)
and
01 = 211(952) + vd1(—95)—(7p5“ +1)[sinh(20)+51%mp5zsinh(2952n)]. (4.121)
77
D2 = d2(962)+vd2(-962)
_4(2964 -1)[(cosh2n -— v) + 512;»: 904(90511206211. _ v)], (4.122)
1— 2 2 ' ,
6W) = "31""(1 + ”if (2215,??? L“ + 7501521120 + 5..)11, (4.123)
512(E.) = 4.431149“ “112134“ 752)
(1 +3§)(3+e=,) +8§(1—2§2)]cosh(1+§)n. (4.124)
The denominator A represents the combination of the conditions (4.55) for fiexure and
(4.57) for barreling such as
A = 45153213132101: p54)2tanhp5zn -4p56tanhn}
.. .. (4.125)
+ C1C2{(1+ p54)2tanhn — 4955tanhp52011.
The solution (4.117) can be also obtained by modifying a generalized solution (4.43) with
replacing 0,," by 20m. The particular solution (4.40) contributes to the last two terms
including N3 and N4 in (4.117). Using the variational analysis, Sawyers and Rivlin (1982)
also found the same type of second order solution (4.117) which only differs in the
notations. Finally, the corresponding second order nonhomogeneous solutions
u‘2’=(u,(2), u212),p(21) become from (4.2) with (4.95) for the zeroth term and (4.110) and
(4.1 1 I) for the higher order as
up) = .A12)sin(ZQ,-,MX1)2
1
u)” '2'PoB(l)B(1)'+B<21cos(20,;mX1).
1;"! 1;"!
(4.126)
4
2‘” = 202191218) + 12)-18.818? “220828.220,
Jim
where Bfig has only homogeneous solution and 853,} has nonhomogeneous solutions
359m , or 8533", for j=1 or 2, respectively. The nature of coefficients B12) depends on the
78
choice of the first order solution - flexure or barreling mode and j =1 or j=2 and mode
number m.
4.6 Summary of Full Buckled Deformations
The fully finite buckled deformations v,(X,,X,), v,(X,,X,) with pressure p(X,,X,)
after perturbation expansion are defined in (3.1) and solutions of the first and second
orders are determined in the previous two sections. Now we will summarize the results as
a reminder of complex procedure. The buckled deformation with respect to X, axis is
given from (3.1), as
v2(X1.X2) = eu1'>(X.,X2)+ e2u1221X..X2)+ 0(22). (4.127)
More specific result is from (4.1), (4.69) and (4.126), as
V2(X1,X2) =
1 , . 2m—
831,2,(X,)srn( 211 7tX,)
+ 52{.p2_°B(g,),(X,)B(§,2,'(X,) + 81?;(X2)cos(2m, lit/1’1) } + 0023),
I
(4.128)
<
cB,l,2,(X,)cos(lflrtX,)
1
+ e2{9,-°Bs!.2.(X.)B12.2.'(X.) + 812.1(X2)cos(2,—’"-nX.) } + 00:2).
I
where the upper solution is for j=1 and the lower solution is for j=2. The coefficients 8,9,}
are given in (4.54) for flexural deformation and (4.56) for barreling deformation and the
coefficients 8,13,} are given in (4.117). The subscripts j=1,2 and m=1,2,3,... are indices of
the chosen onset bifurcation modes. Then deformation v,(X,,X,) along the X, axis and
pressure p(X,,X,) follow v, solution (4.128) in an appropriate order of e from (4.2), (4.53)
79
and (4.110). In conclusion, there exist four different types of solution which depend on the
deformation types such as flexure or barreling and chosen value of j=1 or 2. In each type,
there exist infinite modes of solution.
4.7 Reduced Formulations Due to Symmetric Load Parameter
The first order solutions u“) obtained in (4.68) - (4.70) may affect the formulations
on the load parameter and the energy equations so that their equations will be reduced to
simpler forms. These simplified forms will make less efforts to investigate the stability
near buckling initiation.
4.7.1 Load Parameters
With the results on the first order solution in hand, the postbuckling formulations
given in solvability conditions can be much reduced. In particular, we now show that p, as
given by (3.33) must vanish. Applying integration by parts once with respect to X, to the
first two terms of R, in (3.31) yields
8. = ir2P(')(uI”“i,li"ui')ul,'i)|r|dX2
(4.129)
+ 11,3121'2018431 — u131u13))dx.dx..
Then applying the F,-boundary conditions in (3.9), to the first term, the numerator R, in
(3.33) becomes
12. = sllnpuxumum -u131u13))dX.dX.. (4.130)
Substituting from either option for the first order solution (4.68) into (4.130) and
performing the associated integration gives
80
R, = 0. (4.131)
Rearranging R, in (3.32) yields
R. = 41111552491201 — u12>p<0>dx + 11np(pa2u131— agndxr. (4.134)
I 2
Substituting the first order solutions (4.68) - (4.70) into (4.134) and integration with
respect to X, now yields
p8 I II II I
B. = -41211,72-B1"(Q%p5431"+31" )|,2— lrzuBs') )2+Q..(B1'> )2)dX.. (4.135)
The F, boundary term in (4.135) vanishes by virtue of the homogeneous form of the F,-
boundary condition (4.13),. Then
R. = 4011%11,1181”")2+£2.2(B12>')2)dX.. (4.136)
81
It is significant to note that
R, > 0, (4.137)
for nonconstant 8,1”. The results on R, in (4.131) and R, in (4.137) in conjunction with
(3.33) show that
p, = 0. (4.138)
This result simplifies the expression of p, in (3.38) to
p, = 772. (4.139)
The numerator R3 can be reduced via integration by parts with respect to X, on the first
two terms in (3.35) such as
R3 = irz[u(|>(p(2)u§_')+p‘11u§?3)- u§1)(p(2)u(]) +p(l)u(?))]|rldX,
41,, 029013142) — 4131u12))+p<2>(41}14531— :43) u13))1dx
(4.140)
+ 112500131421 - u131u121)dx
n
+ 11” 15944314412)—u131u12))+p<2>)|,2992+ 1, (u111u12>-43141201,de5 (4.143)
9. = 12012142413342” 4314420122.
1 (4.144)
- ..(u121u12>—u111u122+ 4431412221.Xm-
Substituting F, - boundary condition in (3.9),, all the first integrals in (4.142) to (4.144)
vanish so that
9. = -l,lu131u1'>|,zdx.,
(D2 = i,l(u1,‘lu12’-22131u12’)|,2dX1, (4.145)
9. = -l, —u131u12>+ 43149)), 420.
Similarly, after integration by part with respect to the variables for 12,01, the definition
E,J,’s in (3.53) are reduced to
E..- = 1201:1410 + us: 1221”)|,,dX2 + 1,011.4» + u155u22)|,de.
(4.146)
—lln 104:1. + 2415.291) + 04:21. + u1€2.)uy>ldx,
where i=1,2 and j=1 ,2,3,.... For i=1, the coefficients of u,” and 11,11) in the third integral are
substituted by the governing equations in (3.8),, for k=1 and due to the F,-boundary
condition in (3.9),, the first integral vanishes. Hence
E..- = 1, 043114» + u131u1/1)|,24X1--,l-,iln(p5‘p‘,'lu1”+ P0P(,liuy))dx- (4.147)
83
Integration by parts with respect to the variables of p“) in the second integral in (4.147)
and application of the boundary condition (3.9), yield
._ . . P .
=.,- = ((5213154) + 431119795949) 4X.
r2
(4.148)
1 . .
+ T1 iin(p5'u1,’1 + p5u2’2)p‘”dx-
The parentheses inside of the second integral in (4.148) is the left hand sides of
incompressibility conditions in (3.8), so that substituting the right hand sides of the
condition to (4.148) becomes for each j
3.. = 1,1041%!”u1,')usl>—%’p<'>u1”) 4X5 (4.149)
a. = l, (24,042+u131u122—39pmu122) 4X.
' ,1 ’2 (4.150)
+ 5,1 inp‘”(u1,'1u1.'l — u131u13))dx.
3.. = ((53,50).. u131u12>-%p<1>u12>) 426
(4.151)
+fillnp152dx (“52)
-,l,11n1u12>(p<3141}1—p<3)u131)+ 2422029124131 -p<,') 4111))dx.
Then, from the abbreviated equations (4.145) and (4.149) to (4.152), the terms in the
expansion of energy difference (3.55) to (3.57) have reduced formulation. Substituting
(4.145) and (4.149) into (3.55) yields
dX,. (4.153)
r2
B. = ,1, H013) - p52u1fl)-P,fp1"}u1”+(u111u1‘2- 5524311410]
The coefficient of u,(” in the first term in (4.153) vanish due to the F,-boundary condition
(3.9),, and after integration by part with respect to X, to the second term and applying the
F,-boundary condition (3.9),, E, becomes
E. = 51,013) +pa2u1}))u1'>|,2dX.. (4.154)
Substituting again the F,—boundary condition in (3.9), to the coefficient of u,‘” in (4.154)
yields
E, = 0. (4.155)
Similarly E3 in (3.56) with (4.138), (4.143) and (4.150) becomes
85
E3 = ”(P62¢2+Elz)
Po
= 1_-21__(1) 2) 1 —21 2)
1,1101.) 5.221)) ,p }u1+(u1,)+p.u1,1)u1] 4X. (4.155)
r2
+ ,1 11594511840 — 431439444X.
Applying the F, -boundary conditions in (3.9), to the first integral, E3 becomes
3: p u u —u u , ,. .
E 31”“ (0(1 3 1 ])dXdX (4157)
The integrand of E3 in (4.157) is the same as that of R, in (4.130) so that the procedure
leading to (4.131) also gives
E3 = 0. (4.158)
Based on the results (3.54), (4.155) and (4.158), the first nonzero term in the energy
difference is, at minimum, the fourth order term 13., so that
AB = 213E434+ 0(85). (4.159)
With (4. 138), the formulation of E4 in (3.57) becomes
1... .—
E4 = P(- 29133sz1 '1' 962(1’3 '1' 5:22 '1' :43)- (4-160)
Here (I), requires use of u“), 3,, requires use of u“) and “(2), and CD3 and 3,, requires use
of u“), “(2), and um.
We now establish, however, that E, can in fact be determined without first having
obtained “(3), Using (4.145), and (4.151) we may write
86
5,25, +2, = 1r. [-pa2(u12)u12>—u1}1u12>+ 440:4”)
dX,
+ u111u12>+ u1})u12>—%9pmu9>]
r2
(4.161)
+fillnp1pz+ u121u12> + u12)u12>+ 1,95%?) 4X.
I.2
+ 1 15940182112) + ut2)u1}1)- 0431542) + 2:121u13)))dx
1 (4.165)
+ , llnp<2>| dX (4 168)
2 Ian ’ r2 1 2 1‘, ’ O ’ r2 1. .
Again substituting the F, boundary condition in (3.9), for k=2 into the second integral
yields
11,506,251}, - u12>u131 4.,an- (4.159)
Hence R6 in (4.165) becomes
88
R. = 11,015,250, —u12>u131)|,de.
+ 11115940131421 + u12)u131)—(u111u1.2) + u121u13))1dx
(4.170)
+ i1 1np<2>1u131u91 — 4943092
51151212991213 91,8481) + 1420;015:131 —p<31 u13))1dx.
4.7.3 Energy and Load Parameter
In equation (3.47), the relation between the thrust T and the load parameter p, is
revealed for the case of symmetric bifurcation. Now we will investigate the relation
between the energy difference AE and the postbuckling behavior. The first non-zero term
of AE appears in E4 and the postbuckling behavior depends on the sign of p,. If p, is
negative, then the progressive buckling occurs in view of the relation between T and p,.
Right afier the buckling initiates, there exist an extension of trivial solution path and the
buckled paths. The actual deformation will follow the energy minimizer between these
solution paths — stable deformation.
Integration by parts to the last integral for R6 in (4.170) with respect to the
variables of p“) gives
_ % Jr2p(1)(u(2)u§1) — u§2)u(1))|r‘dX2 -%irlp‘”(u12’u1,'1 - u,2)u(1])|r2dX, (4 17,)
«1 11159013142) +u131u121—u131u12)—u13)u121)dX-
The first integral in (4.171) vanish after applying the boundary condition on F, in (3.9),.
Then R6 in (4.170) becomes
89
B. = ,1 Inp<1>uu10u121 + u12)u12))— 0:131:42) + u121u13).)1dx
, (4.172)
+ 11np<2>(u1})u131- u1})u13))dx.
Comparing R6 in (4.172) with the result on R, in (4.141) and using the relation for R3 in
(4.139), establish the following relation.
1 1
,R, = —-R,p,. (4.173)
R6= 4
Note also that R, in (4.134) and R7 in (4.166) are related by
R, = R,. (4.174)
NIH
In conclusion, the first non-zero term in the energy difference E4 in (4.164) with (4.173)
and (4.174) becomes
1
B, = Z8,5,. (4.175)
The relation (4.175) shows that E, is simply related to p,. Since in (4.136), R, is always
positive for the nontrivial solutions so that the sign of E4 depends on the sign of p,. If
p,<0, then E4<0 so that AE<0. In other words, the solutions corresponding to buckling
have lower energy than the trivial solution at the same load - the buckled path is stable.
Therefore when the progressive buckling occurs, the buckled path is stable in the vicinity
of the buckling onset. The other case is also clear. When the snap buckling is possible,
then the trivial solution path is stable in the vicinity of buckling onset. In view of the
above statement, if the values of p, are known, then the postbuckled behavior and the
stability of each path can be obtained.
90
The parameter p, is given by (4.139) involves R, and R3 so that the energy (4.175)
becomes
154 = --R3. (4.176)
The parameter R3 shown in (4.141) involves the first and second order solutions u“) and
“(2) determined in Section 4.6.
4.8 Summary
The solutions of the first and second order in the expanded linear type boundary
value problems are obtained by using the separation of variables and infinite series
method. Physically, the higher order solutions are explained as the bifurcation from one of
infinite modes of the first order solution or buckling initiation. With the first order
solution, it is shown that the first load parameter p, vanish so that the buckling behaviors
become symmetric with respect to the trivial or homogeneous deformation. According to
the symmetric nature, the formulations on the higher order load parameters and energy
formulations are much reduced. By substituting the first order solutions and pl=0 into the
energy formulations, it is revealed that the first nonzero terms come from the fourth order
energy equation for which only the first and second order solutions are necessary to solve.
91
CHAPTER 5
STABILITY EVALUATION FOR A NEO-HOOKEAN PLATE
5.1 Introduction
The energy difference (3.48) between the buckled and homogeneous deformations
determines the energy minimal, or stable, path after bifurcation occurs. The equations
(3.54), (4.155) and (4.158) show that the first nonzero term in the energy difference is the
fourth order, E4, so that the energy difference (3.48) becomes
AE = 213E424 + 0(85). (5.1)
Also as we have seen in (4.175), E, is linearly related to the second order load parameter
p, as E4=R,p,/4. Specifically, the sign of E4 is the same as the sign of p, since R,>0 in
(4.137) so that if p,<0, then E40, then the
configurations on the bifurcated path have less energy than those on the homogeneous
solution path. Hence R3>O gives that the nonhomogeneous deformation is more stable.
Conversely if R3<0, then the homogeneous deformation is more stable. In this Chapter, the
stability of the bifurcated nonhomogeneous deformations and homogeneous deformations
near the buckling initiation will be investigated by using the reduced formulations and the
buckling solutions.
Our comparison is to Sawyers and Rivlin (1982) who analyzed the stability of
those type of homogeneous deformations in a neo-Hookean rectangular plate by
comparing the energy of homogeneous deformation with that of the bifurcated
deformation. This was done in the vicinity of the bifurcation points. Their analysis,
however, is apparently not a direct energy comparison between the homogeneous and
nonhomogeneous bifurcated deformations, either flexure or barreling, at the same level of
loading condition as analyzed in this work. On the basis of their analysis, the following
conclusions were obtained: (1) From Figure 3 of Sawyers and Rivlin (1982), the
homogeneous deformation is more stable than flexural buckled deformation when n<0.32
and the flexural buckled deformation is more stable than homogeneous deformation when
n>0.32. This is based on their equation (8.4). (2) From Figure 4 of Sawyers and Rivlin
(1982), the homogeneous deformation is more stable than barreling buckled deformation
at all values of n. This is based on their equation (8.7). Note from (4.55) and Figure 4.1
that the value n=0.32 on the flexure branch corresponds to p=0.966. Also n—mo
corresponds to poo-90.5437. Thus the analysis of Sawyers and Rivlin predicts the
93
followings: For increasing compression ratio, and hence p decreasing from p=1, and near
bifurcation onset, (1) the homogeneous deformation is more stable than flexural buckled
deformation for p:l—>O.966, (2) the flexural buckled deformation is more stable than
homogeneous deformation for p:O.966—->O.5437, and (3) the homogeneous deformation is
more stable than barreling buckled deformation for p:0.5437—>O.
5.2 Formal Determination of Stability Parameter
We first examine p,, which is related with stability parameter R3. The denominator
R, of p, in (4.139) is always positive for the nontrivial solution in view of (4.137). Hence
the formulation (4.139) shows Sign(p,)=Sign(-R3). Direct substitution of the first order
solutions (4.68) - (4.70) and the second order solutions (4.126) into the simplified
equation of R3 in (4.141) give upon collecting values with X, and X,:
+
Q
,E
'9:
‘2':
8
+
N
h;
E
‘72
C5
r—‘fi
Cl)!
~N
Ql
LL 1 \
+
Q
55
65
OR
13
A
'-.J
L/
(5.3)
+02) Aume _ _ +A(”B(‘>' _ _
Casi C3Ci
+-C62’ Aurgm _ +A(1)B(')' _
2 5.2 cf
Here the following trigonometric notations have been used,
C, = cosQXl, S, = sinflXl,
' sinZQXl,
_ (5.4)
C3 = cos2QX1, S3
94
and upper (lower) terms are for j=1 (j=2). Note that the subscripts j and m of the
coefficients A“), 3"", Cl") and (2 which denote the chosen mode of the first order solution,
are ignored for simpler expressions. Integration (5.3) with respect to X 1 gives
R3 = Q]l J12 [-coC(')(A(2>’B(”+A“)B(2)'+2A“)'B(2)+2A‘2)B“)')
42 (5.5)
—o)C‘2)(A(”B“)' _Am'Bm) + C(1)A(11362)' + C52>(A<‘>B“))']dX,.
where m=1 or -1 for j=1 or 2. According to the relations A“) and CI") to 8"" shown in
(4.53) and (4.110), the equation (5 .5) becomes the function of only B“) as
2 p4
R3 = 1491i]! [,2'3'5'2(2(0961QA + QB)dX2’ (5-6)
where
_25' { 3(2)"B(1) + 4(8“)'B(2))'} + a(B(2)"—4QZB(2))',
-a{a" - 40201 + 2([3"B(‘) — B'B(”')} + 213'(a'B(') + 2018“") (5.7)
+ 3(3(1)3(1)')1{2p'3(1)'+ szflgamury},
QA
Q3
Here B and a are defined in (4.94) and (4.105). Note from the new notations that Q, is
function of B“) and 3(2) and that Q, is function of only 8‘”. Note also that the second order
solution 8‘” is more complex than B“). One approach to evaluating (5.6) is directly
substituting B“) in (4.54) for flexure or (4.56) for barreling and 3‘” in (4.117) into (5.6)
and integration over -l,0 and corresponding values of load parameter p0 according to the
relation (4.55) or (4.57) into s in (4.58) and A in (4.120), and then substituting obtained
values into the stability parameter R3 in (5.13). For the purpose of numerical setting, we
introduce the dimensionless stability parameter Rs = R3/ (ullM4Q3). Then
Rs = 2{nn2,01+2,‘.1.(1'11.k1+ ”é; kl) 41111121111}. (5.14)
The Figure 5.1 and Figure 5.4 show the relations between RS and n for flexure and
barreling modes, respectively. For the flexural deformation, Figure 5.1 and its detailed
Figure 5.2 and Figure 5.3 show that Rs is positive when 01.6283. According to
97
the relation (5.2), the signs of R and R5 are different from the sign of AE. Therefore when
n1.6283, the unbuckled state is energetically favored and hence stable. Note that when 11
goes to zero, the undeformed geometry looks like a rod subjected to thrust at its ends
which is similar to the conventional elastica problem. At n=1.6283 which corresponds to
7&0 = po'2 = 3, there exists discontinuity. This comes from the fact that N4 becomes infinity
at this value according to (4.118). For the case of barreling, Figure 5.4 shows that RS is
RS . .
I
400 . l
I
I
200 . l
I
I
I I I
0 I I I
. . I . I . . -
Posmve I Negative I Posrtive : Negative
I I I
'200 '- I I I a
| | I
I l l
-400 - I I I -1
I | l
1 . 4 l k 4 . J . n r .
o 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 TI 2
Figure 5.1 Dimensionless stability parameter RS=R3/(ul,M“Q3) for the flexural
deformation in (5.14) with v=1. At n=1.6283, the curve Rs has a discontinuity.
Right before this, it is positive and afier this, RS is negative. There are also sign
changes in 11:0.6443 and n=1.305. Their details are shown in Figure 5.2 and
Figure 5.3.
98
_n
d
1
0.8 -
0.6 -
0.4 .
0.2 ~ '
-0.2 ,
-0.4 .
-0.6 .
1.4 1.6 1.8 2
11
Figure 5.2 The detailed curve of dimensionless stability parameter R5 for the flexural
deformation in (5.14). At n=rIc=0.6443, the sign of Rs changes from positive to
negative and at n=l .305 the sign changes from negative to positive. More detail
near n=m, is shown in Figure 5.3.
positive in a whole range of n so that the buckled state is always stable. Hence all buckled
barreling deformations has larger energy than homogeneous deformation.
Compared with results by Sawyers and Rivlin (1982), they conclude in Figure 3
and 4 of their paper that the homogeneous state at which bifurcation occurs is stable when
nj vanish in the process. The
undetermined coefficients am, a“, a,0,.. ., ay- are calculated by substituting the proposed
solution (5.25) into (5.23) and (5.24) with (4.77) and equating the coefficients of various
products of C; and 11 to zero. In fact, we immediately find that aI-0=0, for i=1,2,3,... by
considering the 0(11‘") term in (5.23). From (5.24), after substituting (5.25), the 0(1) term
gives
1
all = —§. (5.26)
Substituting (5.26) into the expanded equations of (5.23), the 0(1) terms give
1
022 = —§. (5.27)
The same result is separately obtained by considering the 0(r1) term in (5.24),. The other
coefficients can be obtained similarly. Namely the 0(Q2112) term in (5.23) gives
a33 = -1_;—44 and the 0(112) term in (5.24)I gives a,l = %. Then the 0(112) term in (5.23)
after substituting known coefficients give a,, = ——1-. The same result for 03, is obtained
18
simultaneously considering the 0(113) term in (5.24),. Finally the 001‘) term in (5.24)]
gives a3l = 415-. In summary, the first order solution B(11,§) becomes
2 4 6
B(n.€)= C{1—-2-n2+(3 (9— -§§-4)n‘+(4-% —%—I-I--I—%—4)n6+0(n8)}. (5.28)
105
The constant coefficient C will be obtained from the normalization process explained in
Section 4.4.1. The reduced normalization equation (4.60) can be rewritten by using new
variable (5.22) as
2
1 6263 d 2 d 1 6213 _
m -1I—n'2—(21g) + Bz+pg{E[?a?—BII Idg — 1, (5.29)
where o=l,/l 1- Substituting the solution (5.28) into (5.29) with p0 in (4.77) gives
C = [21211 + 0012)}. (5.30)
The 0(112) correction to (5.30) will put terms of order 112", k=2,3,4,... in (5.28), but we have
previously shown that aI0=0, for all i. Hence we conclude that C = J21, so that
= C2 2 (C2 C 4 (4? C“ C6) 6 3
B(‘LQ filz{I-7Tl + 3‘? 11 + E-Tg-m 11 +001) - (5-31)
The complete solution (5.28) with (5.30) may be compared with the direct small 11
expansion of analytic solution (4.54). Here one finds that
= 22(16Q4Il6 263996 8
B(TlaC) M{§Tl + 4—5—‘3- '1 + ‘9—43—75-‘1- 11 +001) , (5-32)
and the coefficient M in (4.61) for small 11 expanded flexural deformation is found as
_ 3 _2 4 244 2 4
Note that M is required to be positive by virtue of (4.61). Therefore this expansion is valid
only for 11 « 1.37. Thus both solutions (5.31) and (5.32) with (5.33) are the same in the
view of small 11. The same expression for solution (5.31) can be also found in (7.7) of
106
Sawyers and Rivlin (1982) except for the constant coefficient C = J21, which is caused
by adopting the different normalization condition.
For the second order solution B(l,§) = BI2’(X,), the boundary value problem
(4.107) and (4.108) is rewritten here in terms of B(11,§) and B(n, Q) as:
1 4141? “+0604”? 4- _ 30) (064-1) d( c123 dBd
4'44?— TEEHM B ‘ 5690—17—12 IRE-2742" (534)
And onq = i1,
1d21‘3 _4_ _ 1(1) _4 dB
TEE-MPG B — 212—90090 +1)Bd—§
(5.35)
31:11 4(1+2pa“ 0)d3_ _9 (2pa4-1)d_
1134K3 11 dc lzp"
Then substituting small 11 expansion of the first order solution (5.31) into the right hand
sides of (5.34) and (5.35) gives
1d473 (1+064 )dzB
fid—Cr 4—nz°_d__§2+16p543 = —2ml,§{8114+(%+ +-§2)11 6}+0(118), (5.36)
andonq = :1,
d2? _
lid—2+ 40643 = -2wlz(4n2+gn‘+§06) + 0(113),
n C 3 15 (5 37)
1d31'3 (1+2p54 )dB I 3 5 5 7) 9 '
1-1—33E— T3: = -2(Dl, 21’] +411 +3311 +00] ).
After considering the solution (4.117), the expression for 73(11, Q) in (5.36) and (5.37)
must be an even function in T] and an odd function in Q. This motivates
E(Tl, C) = —2wl,§{b”112 +(b21+ b,,§2)114 +(b31'I’ b32C2 “I b33C‘)n6
(5.38)
+(b41 ‘I' b42§2 + b43C4 ‘I' bug‘s)“8 + 00110) l-
107
Substitution of (5.38) into (5.36) and (5.37) and equating the coefficients of left and right
hand sides in terms of the product Q and 11, give the undetermined coefficients b,,—’s. The
details to obtain bIj’s are as follows: 0(112) term in (5.37)I and 0(11) term in (5.3 7), give b”
and b,,, 0(Q112) term in (5.36) gives b,,, 0(Q211‘I) term in (5.36) gives b411, 0(114) term in
(5.37), and 0(113) term in (5.37), give b,I and b,,, 0(Q114) term in (5.36) gives b,,,, and
0(116) term in (5.37), and 0(115) term in (5.37), give b3, and b,,,. In summary, the
asymptotic second order solution 73(11, Q) for flexural deformation when 11 is small,
becomes
3014) = -21912C l112- Z-ICZ 114+ g-ZCHIC‘ 116
4 6 2 72 9 6
(5.39)
2
110741-33? + 5%“ + 4:5?)118 + 0(11 10) I
The coefficient b,,, is undetermined yet but it is not necessary for future calculation. As we
expected, since the B matches the second order solution of Sawyers and Rivlin (1982), the
asymptotic second order solution (5.39) is the same as the series equation (7.14) of theirs
except for the sign which is due to the difference of definition of (0 and the normalized
coefficient. The asymptotic solution (5.39) can be compared with the small 11 expansion of
analytic solution (4.117) for verification of its accuracy. Now we find from (4.118) and
(4.119) that
o 9 69 6537
= _ ___— 2 __ 4 6
N3 ”C2112I32 160TI +22400'I +0“ I}:
(5.40)
3
1
3 1 2107 4 6
- mCZQI—§+§11 “274-611 +0(11)},
108
and
N _mczoI 21 107 103
I-
_ __ __ 2114 6
112 64+640'I+40'I +0“ )I
_ 29 3 169 2_447116
N2“°°C1?I671+a—o" 2'37)" “10‘ I}
(5.41)
where C = J21, Here small 11 expansions of coefficient M in (5.33) and the notation A
in (4.120) as
have been used. Then the second order solution B(n, Q) in (4.117) becomes
7301, o = 22.411 + §0.594, the homogeneous deformation has less energy.
These stability conclusions are similar to those obtained by the numerical
procedures in Section 5.3, except that the asymptotic analysis predicts a critical 11=0.594,
while the numerical analysis gave a critical 11=O.6443. Note that the curve generated by
112
(5.53) can not be directly compared with the numerical curve of Figure 5.3 because of the
presence of the normalizing M in (5.13). Performing a similar normalization on (5.53)
using asymptotic coefficient M in (5.3 3) makes dimensionless stability parameter RS as
R301) _ 16 9I 7
= —_ _ _ __ 2 4
15,01) 01.444423 8111 2 511 +0011}. (555)
where 11 = Q], was used. The transition value here is 11=l.l95 but this value is not
important since the expansion for M in (5.33) is valid only for 11 « 1.37. The comparison
of RS in numerical results (5.14) and asymptotic results (5.55) is shown in Figure 5.6.
0.15 . . . . . 4
Rs .
Asymptotic result (5.48)
0.1 . \ I
Numerical result (5.14)
0.05 t \ .
o
I
l |
-005 . I l
| I
I |
-0.1 . . I 4
_ =1.19
-0.15 _ 11—0.6443 '1 I
-0.2 A A A . A A
0 0.2 0.4 0.6 0.8 1 1.2
Figure 5.6 Comparison of dimensionless stability parameters Rs of numerical result
(5.14) and asymptotic result (5.55) for flexural deformation. The asymptotic result
is valid only for very small 11.
113
When 11 is near zero, the signs and slopes are close so that the results are consistent with
the numerical procedures.
5.5 Discussion
The sign of p, is opposite to the sign of R3 according to (4.173) and the sign of E4
is the same as that of p, according to (4.175). E, is the leading term in the energy
difference AE. Positive (negative) values of p, denotes that the load must decrease
(increase) after the buckling onset value p0 in order to follow the bifurcated branch of
buckling solutions. In other words, by converting p, to 71, with 7L, = —2 p,ij3 , the load
must decrease if K,<0.
For flexural deformation the numerical results show if 11<0.6443 then Rs>0 which
means the bifirrcated path involves less energy than the trivial solution so that the
homogeneous deformation near the bifurcation initiation is unstable. Otherwise when
11>0.6443, the homogeneous deformation near the bifurcation initiation is stable. The
numerical results also suggest additional stability transitions near 11=1.305 and 11=1.6283
(Figure 5.1). However, these results are highly sensitive to the numerical evaluation
procedure and so are rather suspect since they involve R5—>oo. Further it is not obvious
how to treat these by asymptotic or perturbation procedures. These Rs—mo transitions that
are only detected numerically will be dismissed from further discussion. In contrast, the
stability for small 11 on the flexural branch is consistent with the asymptotic analysis near
11=0 (although the value of 11 for stability transition found by the asymptotic procedure is
different).
The numerical results also show for barreling deformation that Rs>0 for all 11 so
that the homogeneous deformation is unstable compared to the barreling deformation.
114
These trends are in fact opposite to those found by Sawyers and Rivlin (1974) in the view
of stability evaluation. They also find that 11=0.32 gives the transition in stability on the
flexural branch.
5.6 Summary
The stability of post-buckled deformation near buckling onset was evaluated by
using energy minimization scheme. Extensive use of symbolic algebra procedures enabled
certain simplifications, but the problem still remained very complex. Accordingly, a
combination of asymptotic and numerical procedures were employed to attempt to
determine stability transitions. The physical buckling behaviors are as follows. As the
thrust load increases from the original zero value, a family of infinitesimal flexural
deformation competes with the homogeneous deformation until the thrust reaches the
value associated with p0=0.5437 (TS=5.6786 in (2.16)). This thrust is known as the
wrinkling load. Then as the thrust exceeds the wrinkling load, the infinitesimal flexural
deformation family ceases to exist and is replaced by an infinitesimal barreling
deformation, which again competes with the homogeneous deformation family.
The energy analysis shows that the infinitesimal flexural deformation family is
energetically favored over the homogeneous deformation family at small loads (implying
small mode number), but that the homogeneous deformation family is energetically
favored at large loads (again dismissing Rs—wo transitions). Thus there is a transition load
value, and a corresponding transition mode value 11=0.6443 (See Equation (5.14)) for this
exchange in stability. In contrast, the infinitesimal barreling deformation family is always
found to be energetically favored over the homogeneous deformation family.
115
Asymptotic analysis, valid only for small 11, was employed to attempt to verify the
behaviors of stability pattern for the flexural deformation. The results agree with those of
the numerical approach. A leading order value for transition of stability gives 11=1.195
(See Equation (5.55)). This precise value (11=1.195) is not of importance since the
asymptotic analysis is only valid for 11 near zero.
Clearly there remain significant questions with respect to this work, especially
with regard to precise numerical transition value. It must also be admitted, since the
stability interpretation is dependent on the (+/-) sign of very complicated expressions
(Equation (5.13)), that additional efforts are necessitated for confirming those results.
This, however, should not obscure the fundamental basis provided by this work. Notably
the consistent perturbation analysis provides strict order expansions both of the energy
competition between homogeneous and bifurcated solutions (Equation (4.159)) and of the
relation to the bifurcated path near buckling onset (Equations (4.2), (4.69), (4.126),
(4.127)). Here the essential and consistent coupling between these is provided by (4.141)
and (4.176).
116
CHAPTER 6
APPROXIMATE SCHEMES FOR BUCKLIN G LOAD OF MULTI-LAYERED
COMPOSITE PLATES
6.1 Introduction
The previous Chapters were concerned with the buckling and post-buckling
behavior for a noncomposite single ply plate and the stability of the various competing
solutions with respect to each other. However in this Chapter, we will investigate
somewhat practical topic - the critical buckling load on a composite plate. In a structure,
the critical buckling load plays an important role since it gives the lowest load to resist
against the compressed load. We had the critical buckling load on a single ply plate by
solving rather simple equations in (4.55) and (4.57) and showed the result in Figure 4.1 as
the relation between the load parameter p0 and the mode ntunber 11. The curves of
buckling onset which give the load at the buckling onset for specified geometry and mode
number, are monotonic with mode number (increasing for the flexural deformation and
decreasing for barreling deformation). Hence the critical (lowest) buckling load is always
mode-1 flexural deformation. But for a multi-layered plate such as the three-dimensional
geometry of Figure 6.1, this behavior may be seriously altered. Pence and Song (1991)
and Qiu et. al. (1994) showed that in symmetric three-ply plate composed of two different
types of neo-Hookean material, there exist another family of buckling paths and they are
not always monotonic. This means the mode of the lowest critical buckling load is not
always mode-1 flexural deformation.
As the number of layers in a composite plate increases, the direct algebraic
analysis of the bifurcation conditions becomes increasingly complicated since it involves
117
I A( /
13
<— 211 >/
Figure 6.1 Geometry of the symmetric three-ply composite layer. The buckled
configurations involve deformations in the (X,,X,)-plane.
seeking roots to a determinant equation for which the matrix dimension grows with the
numbers of plies. Thus it is useful to seek alternate methods for determining buckling
onset load in the manner of simpler approaches and closer to the exact values. The
purposes of this final chapter is to present some developed observations on these issues.
6.2 Buckling Load of Multi-Layered Plates
The prediction for buckling onset load may change if the material is composed of
multi-layered composite plate. In this Chapter, we will consider the general N-ply plate
stacked along the X, direction. The undeformed configuration of whole plate occupies the
region 21,x21,><213 and all plies are neo-Hookean materials. The shear modulus of each ply
is either III or 11" and alternate by ply. Perfect bonding is assumed across the ply interfaces.
Again our attention is restricted to plane strain deformation where buckling takes place in
the (X 1,X,)-plane as shown in (2.18). Then the mathematical formulation of composite
plate problem is similar to those of single ply plate studied in Chapter 2. The differences
118
are (i) the shear modulus p of single ply plate alternates between Ill and u", and (ii) the
assumption of perfect bonding yields the following interface conditions on the traction
and displacement.
XXX?) = X.(Xi )
+ on interfaces i = 1,2,3. (6.1)
521(le = 521(X2 )
Here the conditions for i=3 are automatically satisfied in a plane strain setting.
Let A” be the sum of original areas normal to X 1 direction of plies whose shear
modulus is p’ (i=I,II) so that AI +AII = 41,13. Then the total thrust on X, = ill for
homogeneous deformation can be modified from (2.16) to
T = -(P - 9’3)(H1AI+ WA”)- (62)
The buckling onset can be analyzed by the incremental deformation superposed onto
homogeneous deformation such as (2.18). The complete boundary value problem of a
composite plate then consists of that of noncomposite plate (2.28) in which It changes to p"
and the interface continuity conditions
[Pj(V1,2 + 13-2121) + qu,1IX2+ = IHJIVLz + 13.2121)+ 9V2, 11x,
[11102.2 ' P72V1, 1) -‘I(V1,1 'I‘ NIX; = [111.(v,,, — p-2v1, 1) —9(V1,1 + 9)]X5 (6 3)
IVIIX3 = [VIIXi
IV2IX; = [VZIXE
on interface. Here 11! is the shear modulus of top ply and pi is of bottom ply on that
interface. Since the buckling onset occurs at the first order (k=1) in perturbation expansion
of deformation, we will consider only the homogeneous boundary value problem modified
from (3.16) with appropriate interface conditions (6.1) such that,
119
Flu = 0 in IT,
GIu = 0 on I], (6.4)
0
on F,,
and
[61'2“]3 = [szulXi
on interfaces (6.5)
IG3“IX2+ = [63“],i
where superscript j=I,lI, denotes the differential operator of ply j in a composite plate. For
simplicity in the expression of the first order equation, the superscript I” will be
suppressed here and after. Here 0, is the constant matrix
4 = 1:, 1 31
The difference in potential energy E in (2.30) for noncomposite plate can be used for
composite plate as the sum of the energy of each ply. For buckling onset, the energy
equation have the value up to the second order so that
0')
E2 = 213IIIIE5'12902(“1,2“2,1 ‘ “1,1“2,2) + ”1,1 + “22.2 + ui, + “i, 1 }dXIdX, (6+7)
where the integration on X, is sum of the integrations of each ply.
Following the similar analysis to noncomposite case, this composite plate problem
is reduced to one homogeneous ordinary differential equation, 4 boundary conditions and
4(N-1) interface conditions. Introducing the proper general solutions which is similar to
(4.18) with discrete constants L1“), L5"), MI"), My"), m=1,2,...,N of differential equation
(6.4), to the boundary and interface conditions (6.4),, and (6.5) form a homogeneous
120
4Nx4N matrix equation. For the buckling onset load (nontrivial solution), the determinant
of this matrix must vanish.
In particular case of symmetric plate, the dimension of matrix can be reduced to
two 2Nx2N - one for symmetric mode and the other for antisymmetric mode as explained
in Chapter 4. For the simple example of symmetric plate, the three-ply composite plate
(N=3) was considered by Pence and Song (1991, 1993). Here the central ply (material 11)
has the thickness 2R (<21,) and the shear modulus Ir" so that the shear moduli of outer
plies (material I) are 11'. This problem then simplifies to 12 homogeneous linear equations
for the 12 constants L1“), L5”), MI"), My") , m=1,2,3. The vanishing of the determinant of
12x12 coefficient matrix gives the loads for buckling onset. Due to the aspect of
symmetry, this problem can be considered by two 6x6 matrix equation - symmetric
(flexure) deformation and antisymmetric (barreling) deformation along the X, direction
with four dimensionless parameters as
2.0 = p52, 11 = (21,, B = pII/IrI, or = R/l,. (6.8)
When [3:] or a=0 or CF], this problem reduces to the noncomposite case which is
analyzed previously by Sawyers and Rivlin (1974,1982).
The numerical computation for three-ply plate under various parameter sets shows
in Figure 6.2 that (i) the buckling onset load for composite plate does not guarantee its
monotonicity, i.e., the critical load is either mode-1 flexural deformation or wrinkling
deformation in which the mode number is infinity (Pence and Song, 1991 ), (ii) there exists
additional solutions for each original solution of flexure and barreling - we categorize
these into a new family and the original family of solutions, respectively, and (iii) the
wrinkling load of original family converges to that of noncomposite plate which is
121
A 15 1 fl I
0 \I ' \ (I3,a)=
\ new barreling (2.0,0.5)
' I \\//i”
11 2 new flexure _
Aw 3 ,
flexure
l .1
l l 1 l
o 2 4 6 a 10
I I I =
71'0 194 _I (Baa)
(05,05)
17 —
15 _ new barreling ~
13 4 _
1, I. new flexure l a
u
9 — _
11
It
7 _ 2
5 a barreling 4
kw 3 1— _.
1 flexure _
1 l l
o 2 4 6 a 10
Figure 6.2 The buckling onset prediction curve for a symmetric tri-layer with different
values of B=uII/I.rI. The shear moduli of shaded plies are twice as large as those in
the unshaded plies. In both cases, the volume fraction of central ply, a, is 1/2.
Nonmonotonic behaviors, new curves at large 3. values, and asymptotes of all
curves are shown (Qiu, et a1, 1993)
122
constant for the material parameters and the wrinkling load of the new family converges to
a higher value than that of original family depending on the stiffness ratio 13 (Qiu, et al.,
1993).
An asymmetric two-ply plate (N=2) is discussed in Qiu, et al. (1993). Due to the
lack of symmetric property in shape, the buckling deformations become a mixed mode of
flexure and barreling. Solving the resulting 8x8 determinant shows that there are three
solutions. The lower two curves are original family and the higher one is new family. The
wrinkling load of original family have the same values of noncomposite plate and the
wrinkling load of new family depends on the shear modulus ratio.
6.3 Approximate Schemes to Determine Buckling Load
To construct buckling onset prediction curves of buckling load vs. mode parameter
at fixed values of appropriate volume fraction and stiffiress ratio, will be a key to
determine the critical buckling load and eventually require complicated numerical
procedures. For the general N-ply sandwich plate, this problem is reduced to solving a
4Nx4N determinant equation. An exact analytic solution satisfies the complete conditions:_
the nonlinear constitutive equation (CE) in (2.13) for noncomposite plate, equilibrium
equation (BE) in (2.6), boundary conditions of a free surface condition (FSC) in (2.9) and
conditions of interface displacement continuity (IDC) in (6.1), and traction continuity
(ITC) in (6.1),. The simultaneous satisfaction of all these conditions gives much difficulty
when the plate consists of large number of ply stacking. This difficulty is stems from the
fact that standard numerical procedures to find the roots of the necessary deterrrrinant are
subject to various numerical errors and numerical instabilities. The possible
approximation schemes may involve procedures (specifically trial fimctions) that do not
123
satisfy certain conditions mentioned above. Satisfaction of all these conditions gives an
exact solution and thus an exact prediction of the buckling load. Therefore the goal of the
research described in this Chapter is to construct useful approximation schemes which by
sacrificing some of the conditions, give a simpler mathematical formulation. The effect on
accuracy of these sacrifices will then be examined.
The simplest approximate scheme is based on equivalent modulus where the
composite structure is treated as a homogeneous media with volume averaged stiffness
modulus. For example as described in Figure 6.3, the three-ply plate which the central ply
1
IA11 1th
11“ I211 212 -> =“I(l’°‘)+“"a 212
I
I‘ v
<—fll—> <——211———>
Figure 6.3 In the equivalent modulus scheme, the composite plate is treated as a single
layer of volume averaged stiffness, ucq.
has It" and top and bottom plies have 11' considered previously, is treated as a single ply
which has the equivalent stiffness modulus ucq = IrI(1-01) + uIIa where a is the volume
fraction explained in (6.8). Note that this scheme does not satisfy constitutive equation
pointvvise, but does so in a volume averaged sense. However the buckling onset prediction
curves of single ply plate does not show the dependency on system parameters so that any
combinations of plies have the same buckling onset prediction curves as the noncomposite
124
plate. Therefore the critical buckling load for equivalent modulus scheme is always mode-
1 flexure.
6.3.1 Variational scheme
Most approximation methods used in structual analysis are based on variational
mechanics in which the approximate solutions satisfy the weak (variational) form or
minimizes the energy functional (Reddy, 1986). The buckling load in a structual problem
or the natural frequency in a vibration problem can be determined approximately by so
called Rayleigh quotient obtained from the variational method.
The critical buckling load for the composite plate considered here can be
constructed from the boundary value problem of composite plate (6.4) and (6.5). The inner
product
(u,Ffu) = 0 (6.9)
followed by the integration by parts once and applying boundary conditions (6.4),, and
continuity conditions (6.5) gives the critical buckling load as
(I)
IIin(ui,1+ “3,2 + “i,2 + “inldxrdxz
(6.10)
x0 = on _
.I.In “(1)011, 1u2,2 " “1.2“2, 1)dX1dX2
where kozpo‘z is the load parameter and the integration on X, is sum of the integrations of
each ply. This quotient can be also obtained by energy formulation based on the fact that
deformed configurations with less strain energy than the unbuckled homogeneous
configuration only become available once the buckling load is attained. The difference in
energy from the homogeneous state to buckled state of a composite plate in (6.7) must be
125
negative when the buckling takes place. At the buckling initiation, E, becomes zero.
Equating E,=0 gives the same quotient for the buckling load A, = p52 as (6.10). If we
have the exact solutions a, and u, of buckling onset, then the quotient (6.10) will give the
exact buckling onset load. However the procedure to determine the exact buckling
solutions u, and u, of the general ply composite plate is not that easy. So the approximate
(trial) solutions which will satisfy part of required complete conditions must be
considered. These will give a closer prediction to the exact buckling load.
6.3.2 Trial solutions
Recall that the exact solutions satisfy the requirements of complete conditions: CE,
EE, FSC, ITC, IDC. A simple approximation for the composite plate is to use exact
solutions of noncomposite plate (4.68) with (4.26) for flexure and (4.56) for barreling.
One approach is that the composite material can be considered as the combination of
corresponding single plies (combined single ply solution). For example as shown in
Figure 6.4, the geometry of mode-m flexural buckling of composite plate is similar to
mode-m flexure in each ply, while mode-m barreling of composite material is similar to
mode-m barreling in the central layer with mode-m flexure in the outer layers. Note that in
this approach, the length of X, as well as 11 in each single ply are scaled to those of single
ply and the X, coordinates in each ply are transformed to the origin. This approximation
does not satisfy the interface displacement (6.5),, although the displacements are close.
For a symmetric three-ply plate with (B,a)=(0.l,0.5), the buckling onset curves in
Figure 6.5, are generated by the quotient with this combined single ply solutions. The
exact curves and the curves by equivalent modulus scheme are also shown. For flexure,
the variational scheme with combined single ply solutions gives better results than the
126
=1
_/_ _/— flexure
f f I131:)lrure
-> .
I _/ f. 321....
:12'R f f
=1 flexure
2R , f fl:)l(ure
ter f f _
—/— /— garielin
2’! x I" ,
N k
x XI rfrl=l
n=1 barreling
Figure 6.4 In the direct energy scheme, the overall deformation is approximated by the
single layer deformations as shown. The deformation of the central layer
distinguishes overall flexure from overall barreling.
equivalent scheme. This is because the single deformed shapes are well matched to the
three-ply plate for flexure. However they are not well matched for the barreling case.
The other approach is that the single ply solutions can be applied directly to all the
plies (direct single ply solution) since the final displacement of deformed shape of
composite plate is similar to that of single ply plate. For example of three-ply case, the
mode-m flexural buckling of composite plate may use the solutions of single ply mode-m
127
equivalent modulus
I
combined single ply solution
\ exact
o l l l
o 2 4 6 8 10 TI
A0 5 1 1 1 1
4 2 / combined single ply solution _
equivalent modulus
3 l— A
‘/ exact
2 A 4
l P _1
o l l l l
o 2 4 6 a 10 TI
Figure 6.5 The onset prediction curves as given by the equivalence scheme, exact
scheme and the variational scheme with combined single ply solution for flexure
(top) and barreling (bottom) for tri-layer with (B,a)=(0.1,0.5).
128
flexure. This approximate solution satisfies the boundary and interface conditions, but not
the constitutive equation.
For a three-ply plate with fixed pairs of (B,a)=(0.5,0.5) shown in Figure 6.6, the
buckling onset curves are close to the exact solutions for both flexural and barreling
modes. They form an upper bound as expected from the fact that the Rayleigh quotient
gives an upper bound.
6.4 Discussion
For an analysis for determining buckling onset load of a composite plate, we
examined some approximation schemes in which the approximate trial solutions satisfy
some parts of the complete conditions. Since general multi-ply plate is combination of
even or odd stacking, the approximate schemes developed previously can be applied to a
different ply stacking (like two-ply plate) in order to determine whether these schemes can
predict the buckling onset curves well enough.
Also we can consider other trial solutions based on combined single ply solution in
the variational scheme so as to satisfy the interface boundary conditions (IDC) in (6.1) 1-
One alternative is obtained by multiplying a suitable mollifier function of X, to a single
layer solution (4.26) for flexure or (4.56) for barreling. The example of mollifier function
is a simple polynomial forms with undetermined coefficients and play a role as a single
ply solution in each separated ply in a composite plate according to its shape. The
disadvantage of using mollifier function is the sacrifice of satisfaction on field conditions
(CE) and (EE) but the satisfaction of (IDC) may improve the accuracy of whole
approximation. Satisfaction with complete conditions of various schemes studied here and
suggested scheme is summarized in TABLE 6-1.
129
lo
4 1- .1
equivalent modulus \
/
3 ~ _.
direct single ply solution
2 exact ‘
1 1
0 l l 1 J 1 1 l
o 1 2 3 4 5 TI
5 1 1 1 r 1 o
lo
4 e equivalent modulus
3 .—
\ direct single ply solution
2 r exact .
1 1— —1
0 l J_ L l l l 1 l 1
5 1'I
0 1 2 3 4
Figure 6.6 Comparison of the flexure (top) and barreling (bottom) onset prediction
curves for symmetric tri-layer with (B,or)=(0.5,0.5) as generated by the exact
scheme, the equivalent modulus scheme, and the variational scheme with direct
single ply solutions. The upper bound property of variational scheme is evident.
130
Table 6-1 Summary of complete conditions and satisfaction for various schemes
CEl EEI FSCI ITC I IDCI
Exact solution yes yes yes yes yes
Equivalent modulus scheme no2 yes yes yes yes
Variational scheme yes yes yes yes no3
(combined single ply solution)
Variational scheme no2 yes yes yes yes
(direct single ply solution)
Variational scheme - suggested no5 no5 possible4 possible4 yes
(mollified single ply solution)
Condition (6.1)..
IBI'ITIS.
2. satisfied only in a volume averaged sense, not pointwise.
1. CE: Constitutive Equation (2.13); EE: Equilibrium Equation (2.6); FSC: Free Surface
Condition (2.9); ITC: Interface Traction Condition (6.1),; IDC: Interface Displacement
3. however expect the result to be close since the mode shapes should be well approxi-
mated by the single layer theory.
4. these interface conditions can be ensured for mollifier functions of sufficiently many
5. greater sacrifice of CE and EE will yield more improvement in FSC and ITC.
6.5 Summary
Three approximate schemes to determine the critical buckling load for three-ply
neo-Hookean plate were investigated. Each scheme satisfies with parts of the perfect
conditions and gives reliable values of critical load. The schemes developed here can
estimate the critical buckling load of general multi-ply plate in a simpler manner.
13]
CHAPTER 7
CONCLUSIONS AND RECOMMENDATIONS
7.1 Conclusions of the Thesis
The elastic stability analysis for the post-buckled and homogeneous deformations
of single-ply neo-Hookean plate and the approximate schemes for buckling load of multi-
ply neo-Hookean plate have been investigated. The elastic stability near buckling onset
gives the prediction for physically existed deformation and can be evaluated by comparing
the energy of all possible deformations. The buckling behavior are understood by the
bifurcation theory in mathematical terminology.
Under these basis, two dimensional nonlinear boundary value problem for single
ply neo-Hookean rectangular plate was generated in the context of finite elasticity. The
perturbation expansion method was then applied to analyze the nonlinear problem into the
set of linear equations by the order of a. Each set of linear equations gives rise to a
nonhomogeneous boundary value problem (except for the first order) and the solutions are
related to the solutions and parameters of previous order. The first order equation is
homogeneous and gives the thrust on buckling onset which is the critical buckling load.
Also the load parameter that is barometer for thrust, are determined by Fredholm
Alternative Theorem as a solvability conditions. The process for obtaining the solutions
and load parameter for each order was developed in a systematic way and the more
accurate results can be obtained from investigation of higher order equations. The
deformations and their behavior at buckling initiation was at first obtained from the first
order problem and shows the same results as other works.
132
For the stability evaluation of post-buckled deformation, the solutions and load
parameters of second and third orders were obtained to analyze the energy difference
which has the fourth order as the first appearing term (see Equation (4.159)). Since
formulations is quite complex, a numerical analysis was applied at final evaluation of
stability parameter (Rs in (5.14)). The numerical analysis shows that for flexural
deformation, the buckled deformation has less energy than the homogeneous (unbuckled)
deformation when 11<11c=0.6443 and 1.305<11<1.6283 and the opposite is true when
11c<11<1.305 and 11>1.6283. For barreling deformation, the buckled deformation has
always less energy than the homogeneous deformation (see Section 5.3).
The complexity of formulation and non-explicity of load parameter lead to an
asymptotic analysis for post-buckled deformation. The analysis on extreme shape gives
the same critical load as Euler buckling. Also for extreme case of flexural deformation, the
buckled deformation has less energy than the homogeneous defamation when 11<1.l95
and the opposite is true when 11>1.195 (see Equation (5.55) and Figure 5.6). This
asymptotic analysis is valid when 11 is extremely small. Hence the analytic results of
energy comparison for the flexural deformation are true for limiting case of small 11.
In this study, the perturbation expansion approach was used for evaluating the
stability of homogeneous deformation. As a comparative work, Sawyers and Rivlin
(1974) applied the variational methods to determine the stability of homogeneous
deformation near the critical state. Their results are that the homogeneous deformation of
flexural deformation is stable (has less energy) when 11<0.32 and unstable elsewhere and
the homogeneous deformation of barreling deformation is always stable. Hence the results
of their works and this study have the opposite pattern. Also the transitional points 11
133
obtained by Sawyers and Rivlin are different from those found here. The perturbation
method used in this study is more direct and reasonable compared to the variational
methods but the procedure is equally complex. One possible source of the difference in
results stems from the different predictions that may be expected in general, as discussed
in Appendix A in the context of some simpler problems. A second possible source of
difference may be due to the difference in norrnalizations as explained in (4.59).
As an extended work from the stability of homogeneous deformation, the
analytical way of the determination of critical load for composite plate was also studied
(see Section 6.2). The critical load which can be found from the first order equation, plays
a major role in the design of load-bearing structures. However for multiple stack
composite plates, the determination of critical load requires the solution of a nonlinear
eigenvalue problem for a 4Nx4N matrix. As stack grows, the equation will be so complex
that we need simpler albeit approximate methods. In a beginning stage, three-ply
symmetric plate was analyzed for the critical load in three different schemes. Compared
with exact critical load in previous study by Qiu, et a1. (1993), the results are quite close
(see Figure 6.5 and Figure 6.6). These schemes can be expanded to more general ply
plates.
7.2 Recommendations for the Future Work
Direct energy comparison for the stability of homogeneous deformation was
performed and the results showed the differences compared with those of previous works.
Hence as a verification of results, other limiting cases of infinite 11 for flexure, near zero 11
for barreling and infinite 11 for barreling should be investigated.
134
Though the computations on the values of higher order will give more accurate
behaviors of post-buckling for hyperelastic plate, this will also be confined in local
behavior because of the limitation of perturbation method. However the local stability
criterion such as the stability near critical buckling load will be ensured.
135
APPENDICES
136
APPENDIX A
STABILITY EVALUATION BY PERTURBATION EXPANSION METHODS
A.1. Introduction
The analysis for the stability of buckled deformations in the vicinity of buckling
initiation for a neo-Hookean plate requires complex mathematical calculations, so that the
clarity of evaluations for each step is confused. In this Appendix, more simplified
examples than considered problem in the context will be investigated in order to obtain the
thorough acknowledgments for the concept of stability.
The well-known elastica problem will be considered at first as the simplest model.
A modified elastica problem will be considered next to investigate the relation between
the post-buckling behavior which is dependent upon a second order term in load
parameter and the stability. Finally a higher order problem, which in certain ways
resembles the neo-Hookean plate problem, will be examined. For the methods on stability
evaluation near the buckling initiation, the perturbation analysis (PA) which has been used
in the context and the analysis method adopted by Sawyers and Rivlin (SR) in (1982), will
be used for the comparison. In particular, it is shown that these methods can give different
stability predictions.
These examples will consider an energy equation E(th) where 71 and u are load
parameter and buckled deformation in the buckling problem or eigenvalue and solution in
the mathematical bifurcation problem, respectively. Then the first variation of the energy
equation gives a governing equation in a domain 9 and boundary conditions on a
boundary F,
137
111).. = 0 in :2, 6(1):. = 0 on r, (A.l)
where F and G are differential operators. The statement in (A1) constitutes a nonlinear
boundary value problem.
With respect to (Al), it is assumed that there is an obvious trivial solution um, for
all values of 7.. Thus u = u,,,vOt) which is the family of trivial solutions. We now seek
additional solutions (competitors) that bifurcate from this trivial solution. These additional
solutions would also depend on 3., say u = u,,/(1.), so that a continuous parametric
dependence on 7t also defines a family or branch of those additional solutions. Unlike the
trivial solutions, the family u,,/(71.) may exist for only a restricted range of 7.. Now the
solution family u,,-10.) is said to bifurcate from the trivial family u,,-,0.) at the value 3.0 if
quO) = ”611(10)-
The stability evaluation is well explainable under the concept of energy
minimization. If, at a given load parameter A, the energy of the one equilibrium solution is
less than that of another competing equilibrium solution, then the original solution is
energetically preferable to that of the competitor (it is more stable). The energy difference
between the trivial solution and the buckled solution at certain load level 1..
415(1) = £14.51). 111-150....0). 1) (42)
will be considered in the following analysis. According to the energy rrrinirnization
scheme, if ADO, then the state corresponding to the trivial solution is stable.
138
A.2. The Perturbation Expansion Method
One of the well-established approaches to solve the nonlinear boundary value
problem is by using the perturbation method. This approach utilizes an expanded solution
which is perturbed from the trivial solution u0=u,,,-, with a small parameter s such as
u = u0+sul +82u,+83u3+ (A.3)
The load parameter A is also expanded accordingly,
7» = lo+sll +szl,+s3k3+ (A.4)
Here 3.0 is the critical load for bifurcation initiation so that the bifurcated solution u =ubI-f
in (A.3) is branched from the bifurcation initiation u, = u,,-,0.) at k=1, Substituting (A.3)
and (A.4) into (Al) and (A2), and collecting along the same order of a, give the set of
linearized boundary value problem for each order of a. By solving each set of equations,
we can construct a complete expansion on u and 1.. Budiansky (1974) also used the similar
procedures to this study for post-buckling analysis.
For the purpose of stability evaluation, the energy difference AE compares the
energy between the bifurcated and trivial path at a load level it as shown in (A2). The
general solution (A.3) in this approach is an expansion in the vicinity of buckling initiation
so that the results will be limited to the local analysis. Again by using the expansions (A.3)
and (A.4), the energy difference (A.2) becomes along the order of e as
AE = 22E, + 23E, + 24E, + (A.5)
Here E1 is vanished automatically if we substitute the equilibrium solution. For i-th set of
the order 8, the solution u,- can be solved by the conventional way used in the analysis of
linear differential equation and the eigenvalue X,,, can be obtained through the Fredholm
139
Alternate Theorem (FAT). The solutions and eigenvalues obtained are substituted into the
energy equation then EH, can be determined. Since a is small, the first nonzero term on
energy equation (A.5) becomes the leading term on energy difference. If AE becomes
positive then E(ube) > E(uI,,-,,) so that the trivial solution is energetically preferable at the
same load level 71.. For the opposite case, the bifurcated solution is preferable. The
schematic diagram of the procedure is shown in Figure A.1. During the process, it is
sometimes hard to find all the solutions uI-. Instead of direct application, we introduce a
certain orthogonal condition to make some terms in energy equation vanish. This
condition can be obtained by using the integration by parts to the linearized equation.
Detailed calculation will be explained later for a specific examples.
Governing Equation ‘__. Energy Expression
Boundary Condition 5
u = 110+ 8111+ azu,+ e3u3 +...
‘ 2 3 .
l=lo+skl+8 k,+8 1.3+...
BVP on First Order (a) —> Mu. —> E,(u,,>.,)
BVP on Second Order (82) ‘V 7m u, —> E3(u,,[u,], 7(0, 7+1)
/
BVP on Third Order (83) ‘> 12, [113] —> E4(u1,u2.[u3l,
A'09 kl, A'2)
Figure A.1 The procedure for a perturbation method. Here the variables with a bracket
are not necessary if we introduce a certain orthogonal conditions.
140
A.3. The Analysis Scheme by Sawyers and Rivlin
The approach to evaluate the stability of fundamental state which is the buckling
initiation for the buckling problem, makes reference to the work of Koiter (1981). The
energy of the admissible displacement u from the fundamental state characterize the
stability of the fundamental state as the second variation to the energy is positive definite.
Later Sawyers and Rivlin (1982) applied this approach to evaluate the stability of buckling
initiation for the neo-Hookean rectangular plate.
If the second variation P,(u) for the potential energy difference P(u) which is
eventually the same as AE in (A2) under the dead loading condition, is deterrrrined as
positive definite then the fundamental state is regarded as stable. The neutral equilibrium
solution u which becomes buckling deformations, can be obtained by a zero minimum of
the second variation P,(u). Here P,(u) represents a linear version in the whole energy
analysis so that P,(u) vanishes with the neutral equilibrium solution. Hence for the further
condition on stability, the bifurcated path u is decomposed into a linear version of the
neutral equilibrium it and an additional term v. Substituting this new solution into the
energy equation P(u) leads to a new energy equation P(v). Again by solving the
equilibrium solution for v and substituting the solution into P(v). one can evaluate whether
P(v) as well as P(u) is positive definite. This approach is also based on energy
minimization scheme but the objects for competition is difi‘erent from the perturbation
expansion scheme.
A.4. Example 1: Elastica Problem
An elastica problem in which a long slender beam is compressed axially, has been
a model for the buckling analysis (Thomson and Hunt, 1969). The vertical deformation or
141
the buckled shape u is described mathematically as the bifurcation from the unbuckled
deformation at a critical load parameter 110. The potential energy for a slender beam with a
The primes denotes the differentiation with respect to x. The governing equation and
boundary condition of (A6) through the first variation lead to
u" + ksinu = 0, u'(O) = u'(1) = 0, (A7)
where l. is an eigenvalue which represents the load parameter applied to the beam axially
and the boundary conditions are characterized as a natural condition. Clearly one solution
to the problem (A.7), valid for all possible A, is u(7t) = 0. This is therefore the trivial
solution to this problem, um, =0.
The linearized version for the nonlinear boundary value problem (A7) is stated as
u"+?tu = 0, u'(0) = u'(l) = 0. (A8)
For a variable v, the inner product (u" + ku, v) = 0 gives the adjoint problem to the
linearized problem (A.8) through the integration by parts. The adjoint problem with
respect to v has the same differential operator as that in (A8) so that the linear differential
operator in (A8) is self-adjoint. The bracket used in inner product is defined as
14g) = III/mane. (A9)
The solution to the linearized problem (A.8), u,,-,I becomes
u,,-,I = Acosmrtx, A = (M102, m = 1,2,3, (A.10)
142
where A is an arbitrary constant. The linear solution reveals that there is an infinite
sequence of bifurcation load (one for each m).
A.4.l Perturbation Analysis (PA)
We now seek additional solutions u= u,,,f, bifurcated from uo = u,,II, =0 in (A.3)
such as
u = 8u1+82u,+s3u3+... (A.11)
Here the perturbation parameter a is defined as e= with an orthogonality condition
(u,, uI) = 0, i1: 1. (A.12)
The bracket denotes the inner product defined in (A9). Substituting the expanded solution
(All) and load parameter (A.4) into the boundary value problem (A.7) leads to the set of
governing equations and boundary conditions which form the separate boundary value
problems that originate from the various order of 8. Each set of governing equation
consists of a linear differential operator with i-th order deformation u, in the left hand side.
The right hand side of i—th equation involves all of the previous solutions including u,,, and
X,,]. The boundary conditions of each equation becomes
uI'(O) = uI'(1) = 0, i = 1,2,... (A.13)
For the coefficient of s, the first order equation becomes
u," + 1.0a, = O. (A.14)
The solutions of (AM) with the normalized coefficient are
u1 = J2cosm11x, 1.0 = (mn)2, m = 1,2,3, (A.15)
143
The first order equation (A. 14) and the solutions (A. 15) are in fact the same as those found
in linearized problem (A8) and (A.10) except for the subscript. Hence the first order
equation represents the linearized problem. There is an infinite sequence of bifurcation
loads according to the eigenvalue parameters 2.0 in (A.15) but, from now on, we will
consider the lowest value, that is the case of m=1. The second order equation is shown as
u,” + kou, = 4.114,. (A.16)
Since the differential operator in the left hand side of (A.16) is same as that in (A.14), the
solutions u, will exist only if the right hand side of (A.16) is orthogonal to the solutions of
(A.14), namely (—}.Iu1, 111) = 0 according to the Fredholm Alternate Theorem (FAT).
The existence condition and orthogonality (A.12) give
rI = 0. (A17)
Introducing (A. 17) into (A. 16), the solutions with the orthogonality (A. 12) become
u, = 0. (A18)
In a similar way, the third order BVP is expressed as
u3"+}\.0u3 = éloU?—xlu2‘}\.2ul. (A.19)
Applying the FAT again to the right hand side of (A.19) and the first order solutions in
(A.14) gives
l.2 = -)l‘0' (A.20)
With the eigenvalues (A.17) and (A20) into (A.19), the solution for the third order
becomes
144
u, = ~36‘71cos3m11x. ~ (A.21)
For the fourth order problem, the similar approaches give the boundary value problem as
u,," + 2.0114 = 52.011111, — k,(u3 — £11?) — 74,21, — 13m, (A22)
and the results are
A, = O and u,, = 0. (A23)
The higher order solutions can be obtained in a similar way. Then the buckled solution and
load parameter up to the fourth order are summarized as follows.
u = u,,If = (J2cos1tx)e — (gai- cos3rtx) e3 + 0(85), (A.24)
)4 = 2.0 + £71,082 + 0(34), 10 = 112. (A25)
The equation (A.25) represents the post-buckled solution path in A-e curve.
To evaluate the stability of the equilibrium solution near the buckling initiation,
the energy minimization between the bifurcated and trivial solutions at the load level it, is
used. The energy difference (A.2) becomes
II 1
AB = (II-2.152 + A(cosu—1)]dx. (A.26)
If AE >0, then the trivial solution is stable. Substitution the bifurcated solution 14be in
(A.24) and eigenvalues (A.25) into (A.26) give the even-ordered series of energy
difference
145
AE = 82E2 + 84E4 + (A.27)
The odd-order terms vanish automatically. The second order term, E2 in (A27) is
_ 1 J1 :2 2
With the solutions u] in (A.15), E2 vanishes. The fourth order term, E4 is
E =f[u'u'-kuu—lkuf+iku4]dx (A29)
4 0 1 3 0 l 3 2 2 24 O l ' '
Substituting all the solutions and eigenvalues in (A.24) and (A25) into (A.29) gives
In fact, it is not necessary to know u3 explicitly to evaluate (A.29). To see this, multiplying
the first order problem (A.14) by uk, k=1,2,3,... and integrating over the domain 0O in (A.21) so that the load is increased when the bifurcation grows. Hence,
considering the local behavior near the first bifurcation, m=1, there exists only one
solution before bifurcation initiation, which is trivial and stable. After bifurcation
initiation, there exist three local solutions of which one solution is trivial and unstable, and
two other solutions follow the bifurcated path and are stable.
A.4.2 The Sawyers and Rivlin (SR) Scheme
Now consider the elastica problem with the approach used in the paper by Sawyers
and Rivlin (1982). Starting from the potential energy equation for the admissible
displacement u defined in (A.6). The energy difference which is in (A26) is now rewritten
as
P(u) = E(u,l)—E(u,,,v,k) = £{éu'2+k(cosu— 1)}dx, (A.34)
where u = um], the bifurcated branch of solutions that we seek to construct. With the power
series expansion of cos u, equation (A.34) becomes
P(u) = i1 lu’2+k(— lu2+-l—u“- ) dx (A 35)
o 2 2 24 ' '
The second variational term in the expansion (A35) is
P2 = %£(u'2—lu2)dx. (A36)
147
The necessary condition for stability of trivial solution is non-negative P2 2 0. To find the
stationary P2, set 8P2=0
8P2 = £(u’5u’—Au5u)dx = O. (A.37)
After integration by parts, equation (A.3 7) yields
u" + Au = 0, u'(0) = u’( l) = 0. (A.38)
The solutions of (A38) are
u = Acosmnx, X = (mn)2, m = 1,2,3, , (A.39)
which retrieves the bifurcation initiation values A0 and u, previously given in (A.15) and
also the linearized solution u,,-n in (A.10). Again we will stick to the lowest bifurcation
value m=1. Here u is only a linearized version of the solution branch u,,,f near the point of
bifurcation initiation. By substituting (A.39) into (A.36), the result is
P2 = O. (A.40)
Hence the state for which P2 has a stationery value, is regarded as a state of neutral
equilibrium.
Now we decompose u = u,,,f into the linearized solution (A39) and the remainder
term v as
u = a£4+v it = cosnx, (A.41)
where it = A cos nx comes from the linearized u in (A39) and a is the coefficient given
by a = (fl, u)/ (fl, 1?). The solution components 2": and v have the orthogonality
condition
148
(a, v) = 0. (A.42)
Substituting the bifurcation solution u,,,f as given in (A.41) into the energy difference
(A.35) gives
P(u) = £{éazfiz' - 1545+ a(a'v' — 7m) + 5.0/2 — W) + ilzuaa + v)4 + }dx.
(A.43)
It is to be noted that, unlike the PA method, the SR scheme always use 7L = 1.0 associated
with bifurcation initiation. Now consider the multiplication v to the equation (A3 8) with
changed variable 2? and integration over the domain O0, then from (A.61) and (A.64), A2O. Since E4 is the dominant term in
energy difference, the trivial solution has smaller energy than the bifurcated solution, that
is, energetically stable. If a<0, there exist three solutions before bifurcation initiation and
A2>O and E40, P(u) becomes negative so that the trivial solution is unstable. This result is
opposite to the PA method.
A.6. Example 3: A Higher Order Problem Represented Neo-Hookean Buckling
The buckling and post-buckling problem for a neo-Hookean plate considered in
the context by using the perturbation analysis, have a slightly different form from the
previous two examples. Namely, the neo-Hookean plate involves u2 at 0 (sec (4.128)) and
A1=0 (p,=0, (4.138))so that the expansion for the deformation u and eigenvalue A become
u = 8u1+82uZ+83U3+ ...,
A.69
A = A0+32A2+84A4+.... ( )
In order to understand this type of expansion, it is useful to examine a simpler model than
the neo-Hookean plate buckling problem. Such a simple model may be provided by
considering the following energy expression
2 1 '2 1 2 1 ' 2 '2
E(u, A) Mia 7M +5u(u +u )]dx, (A70)
for a function u obeying u(0) = u( 1) = O. The governing Euler equation for the energy
equation (A.70) are given by the first variation as
u"+Au+u'u" = 0. (A71)
154
The trivial solution for the problem (A.71) is obviously u,,,v=0. The linearized problem to
(A71) is stated as
u"+Au = O, u(O) = u(1) = 0. (A.72)
The boundary condition in (A.72) is different from previous two examples. However this
does not affect the self-adjointness for the linear operator. The linear solution is obtained
by solving (A.72) directly
u = Asimrx, A = n2. (A.73)
Here we considered the lowest value of bifurcation (m=1).
A.6.] PA Scheme
The bifurcated solution u=u,,,-f is an expansion from the trivial solution u,,,v=0.
Substituting the expansions in (A.69) into the problem (A.71), the resulting boundary
value problems for each order become
0(8): u,"+A0ul = 0, (A74)
C(82): u2"+A0u2 = —A.u, -u,'u,", (A.75)
0(83): u3"+Aou3 = -A1u§-A2u, —u,'u2"—u2'u1", (A.76)
and so on. The boundary conditions are
u,.(0) = u,(1) = 0 i= 1,2,... (A.77)
Since the trivial solution for this problem is u,,-50, the energy difference between the
bifurcated and trivial solutions is the same as (A.70). By using the expansions (A.69), this
energy difference AE can be expanded accordingly.
155
AE = E282 + E383 + E484 + 0(85), (A.78)
where the components are
1
E2 = vi:§(u,'2—A0u12)dx, (A.79)
_ f I I 1 1 I l '
E3 — O[u, u, —A0ulu2+§ —x,u,2+§ul .4430.l )3 ]dx, (A.80)
E4 = £%{2(u1'u3' — A0u1u3) + (142')2 - 7‘0“? ‘ 211“]“2 — A2u%}dx
1 1 (A.81)
+ 2 Ji){§(2“1u1'u2 + “12“2') + “1 'uz'}dx,
and so on. Now we consider the first order equation. By solving directly the solution
becomes for the lowest value of bifurcation (m=1) as
u, = Jisinnx, A0 = 1:2. (A.82)
In a similar way in (A.32), consider the following integration
kw," + A0u1)ukdx = O = 1, 2, 3, (A.83)
Integration by part once and applying the boundary condition in (A.77) lead to the
equation (A.32). This equation can apply to E2 in (A.79) and the first two terms in (A.80)
and (A.81) so that it is not necessary to solve 143 in (A.81).
The solution in (A.82) is substituted into the second order solution and the results
become
x
u2 = —Tosin21tx, A, = o. (A.84)
156
Then applying the inner product with the right hand side of (A.76) and ul and the FAT
gives the next order eigenvalue as
1
With the previous solutions (A.82), (AM) and (A.85), the third order solution for the
equation (A.76) becomes
A
u3 = isin3nx. (A.86)
4f2
However the third order solution (A.86) is not necessary to evaluate E4 since there are no
u3 terms in (A.81) after considering the equation (A.83)]. Substituting all the solutions and
eigenvalues obtained previously into the energy equation, then E2 in (A.79) and E3 in
(A.80) vanish and E4 in (A.81) gives
E4 = —A3, (A.87)
so that the energy difference becomes
AE = 115138” 0(85). (A88)
In this case, A2 is always negative and then AE is always positive so that the trivial solution
near the bifurcation initiation is always stable.
1. This parallels the results in Section 4.7.2 where the u”) terms are eliminated in
the energy expression (4.164) by using the boundary conditions (3.9)2 for k=1.
157
A.6.2 SR Scheme
The energy equation for the neo-Hookean plate is expressed in (A.70). The second
variation in energy P2 to (A.70) is given by
P = f (l(u')2 -1Au2)dx (A 89)
2 o 2 2 ' '
After first variation to P2, the linearized governing equation becomes
u" + Au = O, u(O) = u(1) = 0. (A.90)
The solution to the boundary value problem (A90) is
u = Asinnx A = n2. (A.91)
Again we stick to the first mode m=1. Substituting the solution (A.91) into (A.89) gives
P, = 0. (A92)
Then the decomposition of u=u,,,~finto the linearized solution u=u,,,, (A91) and remainder
term as
u = ail + v, (A.93)
where a is a small coefficient and the following orthogonality holds
(it, v) = 0. (A.94)
After applying the new solution (A.93) to the energy equation (A70) and set v = 022-4,
the energy equation after ignoring the order higher than 4 becomes
P(u)=a4£[{%fi'(&§+£¢"+§) +504 +(“'))+2((u )2 Au )H . (A95)
158
The first variation for the equation (A.95) gives the governing equation about {I as
17" + A2} = —z‘4'£4" 27(0) = 5(1) = O. (A.96)
The solution for (A.96) with orthogonality condition (A.94) becomes
- A2 .
u = Knsrn(2nx). (A.97)
Substitution the solution (A.97) and linearized solution (A91) to the energy equation
(A.95) gives
9.2.44 4
P = - 48 a . (A98)
The equation (A.98) reveals that P is negative so that the trivial solution is regarded as
unstable.
A.7. Discussions
In this Appendix, we examined the stability in the vicinity of buckling initiation
for more simplified problems than the buckling problem of neo-Hookean plate by using
the perturbation method and the method used in the paper by Sawyers and Rivlin (1982).
The whole procedure and their corresponding equations are summarized in Table 1. and 2.
The perturbation approach compares the energy between the bifurcation and the
trivial solutions on the same load level of A and on the first mode (m=1). The sign of an
energy difference determines the lower energy level of two competing solutions which
shows more energetically preferable solution. Also the results tell the relation between the
post-buckling behavior (A2) and the energy difference (E4). From the results on A2 and E4,
(A20) and (A30) for the elastica problem, (A61) and (A.64) for the modified elastica
159
problem and (A85) and (A.87) for the simplified neo-Hookean buckling problem, it can
be shown that
E, = "1,. (A99)
Therefore the signs of A2 and E4 are different so that, if A2 is negative then the trivial
solution is energetically preferable. Otherwise the bifurcated solution is preferable.
In the SR method, the eigenvalue is not expanded so that the energy difference
used in this method just compares the energy on the bifurcation initiation. This will give a
confusion that on the eigenvalue (load parameter) at the bifurcation initiation. It also can
generate opposite conclusions as to the stability of the bifurcated branch of solutions.
160
Table A.1 Summary of the procedures and their corresponding equations for the
perturbation expansion.
Modified Simplified Neo-Hookean
Step Elastica Elastica neo-Hookean plate (context)
1. Nonlinear (A6), (A26) (A55),(A.62) (A.7l) (2.28), (2.30)
BVP, Energy
equation
2. Applying the (A. 1 1),(A.4) (A. 1 1),(A.4) (A. l 1),(A.4) (3.1)
expansion
3-1. lst order (A.14) (A57) (A.74) (3.8), (3.9),
BVP (3.10)
4-1. A0 (A.15) (A.61) (A.82) (4.62)
5-1. u, (A.15) (A.60) (A.82) (4.68)-(4.70)
6-1. E2 (A.28), E2=0 (A.64), E2=0 (A.88), E2=O (4.155),E2=O
3-2. 2nd order (A.16) (A58) (A.75) (3.8), (3.9),
BVP (3.11)
4-2. A, (A.17), AI=O (A.61), A1=0 (A.84), AI=O (4.138), p,=0
5-2. uz (A.18), u2=0 (A.60), u2=0 (A.84) (4.126)
6-2. E3 E3=O E3=O (A.88),E3=0 (4.158),E3=O
3-3. 3rd order (A.19) (A59) (A.76) (3.8),(3.9),
BVP (3.12)
4-3. A2 (A20), (A61), (A.85),
A2=1/4 A0 A2=-3/2 0er A2=-1/3 A02
5-3. u3 (A.21) (A.60) (A.86) N/A
6-3. E4 (A.30), (A.64), (A.87),
E4=-1/16 A0 < E4=3/8 0er E4=l/12 A02> O
O
161
Table A2 Summary of the procedures and their corresponding equations for the
Sawyers-Rivlin method.
Modified Simplified Neo-Hookean
Step Elastica Elastica neo-Hookean plate (1984)
1. Energy (A.35) (A65) (A70) (2.12)
equation, P(u)
2. Second vari- (A.36) (A.36) (A.89) (3.1)
ation, P2
3. Solving for (A.37) - (A.39) (A.37) - (A.90) - (A.91) (3.8)
neutral CQUlilb- (A.39), (A.66) (314) (3.16)
rium ’
4. Define new (A.4l) (A.41) (A.93) (4.1)
solution
5. Energy with (A.48) (A.67) (A.95) (4.7), (4.10)
new solution
6. Solving for (A.49) - (A52) (A.49) - (A52) (A.97) (4.13)
new solution (5.5)
7. Evaluate (A53), (A.68), (A.98), (6.1)
energy “,9 P=1/64 AA > o P=-3/32 (1AA =-1/48 2.2.44 Fig.3, Fig.4
new solution
(4th order)
162
APPENDIX B
COEFFICIENTS FOR STABILITY PARAMETER
The coefficient functions Y[i,j] used in equations (5.11) and (5.12) for stability
parameter R3 are defined as follows:
1 _ _ _ _
Y“, 1] = mlymao)+N1)’1,2(7~o)-N2.Y1,30~0)+N3y1,4(}~o)-VN4)’1,5(7¥0)]
Ylll, 2] = 3—2ngb’1,1(‘xo)—N|yl,2(—xo)-Nzy1,3(‘xo)—VN4)’1,4(-}t-o) + N3J’I, 5910)]
”193] = figiyraao)+N1Y1,7()~0)—VN4}’1,8(7~0)], i
Y[l, 4] = -§;T§D’1,6(—7\o)—N1y1,7(-7~0) +N3Y1,8(-7~0)],
O
1’11. 51 = gi—at—zwo — A3)(1 + 317,)- 1-)-’3)’1,9(7~o)‘r mum—m].
v _ ._ _
Y[I, 6] = —m[3210(1- 1110+ 16M“ — Ad)(1+ 2523-193‘ N3y1,1o(}~o) + VN4y1,1o(-7¥o)la
(1+Ao) — — -
”1,71 = "Wb’mlafi-ZWNWI,12(10)‘24N1+N2V1,13(7~0)}],
l—A _ _ _
Y[l, 8] = “%§18L)U1,11(4~o)—Zi—N3J’1,12(-}~o)+ 24N1+N2y1,13(—}~o)}],
V(1+ Ao)
Y[1,9] = ———41—8—{3VM—st(l 40)},
Y[l 10] = M9137 +Ns(1+A )}
, 41-8 3 2 0 9
111,11] = 0,
N132
1,12 = — l-A2 .
Yl ] 2x0( 0)
163
1'12. 01 = —1—2§73(1—7~8)2{—N3(1—ko)+vF/4(1 + m}
+(1+A%)
W{16(1+s4x3)+52(3x3+ 58A3 + 3)},
1'12, 11 = —%}:gly2,1(7~o) + N1y2,2(?~o) + 1852,30.) + 7852,4001].
112. 21 = —1—6V-,7,-1yz,.(—xo)-N1y2,z(—xo) + My. 3(—x0)—vK/4y2, 46-14)].
1'12. 31 = 5362,41.» N1y2,.(xo)— vN4y2,7(xo)1.
Y12. 41 = fi—gm, .(-xo)—N1y2,.(—xo) + NW2, 76710)].
1'12, 51 = gg—igtso — W3 + 6418 — 313) — 6131mm.) - VN4y2,s(-}~o)}],
Y[2, 6] = .3:—:B[s(1 - x3)(3 + 32x3 — 313 + 32:21.3) + 6{]V3y2,8(A0) — mum—1.0)} ],
Y[2, 7] = 64—17L3-[y2’9(AO)+ 1927\7.(1— A8) + Nzyz,10(A0)+ VIV4y2, ”(A0)],
1’12, 81 = gawk—ka—wzml — x8) + N212, Mao-My. ”Hon.
1’12. 91 = —-4—‘,:—8 Uz, no.) + vN4y2,13(7»o) + My), I.01..)1.
1’12, 101 = 7&8 Dz, 12(-7~o)-N3y2, ”(280) + N216, ”(4491.
Y12. 111 = 71-0.
112,12] = —‘2—41.0(1 —21.g).
164
Here the 27 notations yum) used in coefficient functions Y[i, j] are defined as;
yi,i(€) = -S§(1-€2){4+52(1-2§+5€2)},
yi,2(C) = 85C(1+§)(10+§+C2),
yi,3(C) = 24S(I+€)(1+3§),
yi,4(€) = 8{SZC(1+5€+2§2+4€3)-3(3+€)},
yi,s(C) = -8(1+C){(3-C)+52€2(7-5€)},
yi,o(C) = 52(1-€)(1+C)3~
yi,7(€) = -8(1+ €)(1-€)(6+C),
yi,s(€) = 85(1+§)(1+24),
yi,9(€) = -(1-€)(3 +23C-3CZ+€3),
yi,io(€) = 5(1-€)(5-31C-13€2-€3),
yi,u(€) = -52C(1-C)(1+€)2,
yi,12(C) = -S(3 + 134-762-519,
yi,13(§) = 852C(1+€+C2),
y2,i(€) = S(1+C){(7+5€+17CZ+3€3)+SZCZ(15-19C+41C2-5C3)}.
y2,2(€) = 128€(1-€)(1+§)(2+3€-€2).
y2,3(€) = -125(1-€)(1 + C)(1+ 3C),
y2,4(€) = -6(1-§){2(3+€)-52€(1 +3€)(1+C2)}.
y2,5(C) = 2S2§(1-§)(1-3C-5€2+ 3C3),
y2.o(€) = -6(1 + 92(2 + 3C - C2),
yz,7(§) = -3S(1 +€)(1-3€)(1+€2),
y2,s(§) = (1-€)(1+6€-16€2-6€3-§4),
165
y2,9(C) = s2(3 +34C-19Q2—4Q3+173Q4—62§5+_3§6),
y2,io(€) = -9652€(1-C2)(1+€2).
y2,n(§) = 125(1-C2)2(1+€).
y2,iz(€) = 28C(1-€)(1+2§-C2),
y2,13(C) = -3(1 +€)(3 -§),
y2,14(C) = 35(1 +C)(1-C)(1-3§).
166
BIBLIOGRAPHY
167
BIBLIOGRAPHY
Antman, Stuart S. (1995). Nonlinear Problems of Elasticity, Springer-Verlag, New York.
Bajenitchev, A. (1996). A numerical procedure for a non-linear elastic problem for
incompressible material based on a perturbation method, Comput. Methods Appl.
Engrg., 131, 31-39.
Ball, J. M. (1982). Discontinuous equilibrium solutions and cavitation in nonlinear
elasticity, Phil. Trans. R. Soc. Lond. A, 306, 557-611.
Ball, J. M. and Schaeffer, D. G. (1983). Bifurcation and stability of homogeneous
equilibrium configurations of an elastic body under dead-load tractions, Math. Proc.
Camb. Phil. Soc., 94, 315-339.
Bauer, L and Reiss, E. L. (1965). Nonlinear buckling of rectangular plates, J. Soc. Indust.
Appl. Math. (SIAM, 13, 603-626.
Beatty, M. F. (1965). Some static and dynamic implications of the general theory of
elastic stability. Arch. Rational Mech. Anal, 19, 167-188.
Beatty, M. F. (1971). Estimation of ultimate safe loads in elastic stability theory, J. Elast,
1, 95-120.
Beatty, M. F. (1987). Topics in finite elasticity: Hyperelasticity of rubber, elastomers, and
biological tissues - with examples, Appl. Mech. Rev, 40, 1699-1734.
Beatty, M. F. and Dadras, P. (1976). Some experiments on the elastic stability of some
highly elastic bodies. Int. J. Engng. Sci, 14, 233-238.
Beatty, M.F. and Pan, F. (1998). Stability of an internally constrained, hyperelastic slab,
Int. J. Non-Linear Mechanics, 33, 867-906.
Biot, M. A. (1963). Exact theory of buckling of a thick slab, Appl. Sci. Res, 12, 183-198.
Budiansky, B. (1974). Theory of buckling and post-buckling behavior of elastic
structures, Advances in Applied Mechanics, Vol. 14, Academic Press, New York, 1-65.
Burgess, I. W. and Levinson, M. (1972). The instability of slightly compressible
rectangular rubberlike solids under biaxial loadings, Int. J. Solids Structures, 8, 133-
148.
Dym, Clive L. (1974). Stability Theory and Its Applications to Structual Mechanics,
Noordhoff.
168
Elgindi, M. B. M., Yen, D. H. Y. and Wang, C. Y. (1992). Deformation of a thin-walled
cylindrical tube submerged in a liquid, J. Fluids Struct., 6, 353-370.
Erbay, H. A. (1997). On the asymptotic membrane theory of thin hyperelastic plates, Int.
J. Engng Sci, 35, 151-170.
Haughton, D. M. (1992). Boundary layer solutions for incompressible elastic cylinders,
Int. J. Engng Sci, 30, 1027-1039.
He, L. W. and Cheng, C. J. (1992). Nonlinear buckled states of rectangular sandwich
plates and their stabilities, Int. J. Non-Linear Mechanics, 27, 423-435.
Hill, R. (1957). On uniqueness and stability in the theory of finite elastic strain, J. Mech.
Phys. Solids, 5, 229-241.
Holden, J. T. (1964). Estimation of critical loads in elastic stability theory, Arch. Rat.
Mech. Anal, 17, 171-183.
Horgan, C. O., Lang, K. W. and Nemat-Nasser, S. (1978). Harmonic waves in layered
composites: New bounds on eigenfrequencies, J. Appl. Mech., 45, 829-833.
Horgan. C. O. and Pence, T. J. (1989). Void nucleation in tensile dead-loading of a
composite incompressible nonlinearly elastic sphere. J. Elast, 21, 61-82.
Horgan. C. O. and Polignone, D. A. (1995). Cavitation in nonlinearly elastic solids: A
review, Appl. Mech. Rev., 48, 471-485.
James, R. D. and Spector, S. J. (1991). The formation of filamentary voids in solids. J
Mech. Phys. Solids, 39, 783-813.
Keller, Joseph B. and Antman, S. Eds. (1967). Bifurcation Theory and Nonlinear
Eigenvalue Problems, Courant Institute of Mathematical Sciences, New York.
Kirchgassner, K. (1975). Bifurcations in nonlinear hydrodynamic stability, SIAM Rev.,
17, 652-683.
Koiter, W. T. (1981 ). Elastic stability, buckling, and post-buckling behavior, Proceedings
of the I UT AM Symposium on Finite Elasticity, Carlson DE. and Shield R.T. Eds.,
Martinus Nijhoff Publishers, 13-24.
Lang, K. W. and Nemat-Nasser, S. (1977). Vibration and buckling of composite beams. J.
Struct. Mech., 5(4), 395-419.
Lazopoulos, K. A. (1996). Second order displacement fields of a thick elastic plate under
thrust - the incompressible case, Int. J. Solids. Structures, 33, 433-449.
169
Levinson, M. (1968). Stability of a compressed neo-Hookean rectangular parallelepiped,
.1. Mech. Phys. Solids, 16, 403-415.
MacSithigh, G. P. and Chen, Y. C. (1992). Bifurcation and stability of an incompressible
elastic body under homogeneous dead loads with symmetry Part 1: general isotropic
materials and Part II: Mooney-Rivlin material, Quart. J. Mech. app]. Math, 45, 277-
313.
Matkowsky, B. J. and Putnick, L. J. (1974). Multiple buckled states of rectangular plates,
Int. J. Non-Linear Mechanics, 9, 89-103.
Meijers, P. (1987). Stability of a rubber half-space, J. Elast., 17, 143-158.
Nemat-Nasser, S. and Minagawa, S. (1975). Harmonic waves in layered composites:
comparison among several schemes, J. Appl. Mech., 699-704.
Nowinski, J. L. (1969). On the elastic stability of thick columns, Acta Mech., 7, 279-286.
Ogden, R. W. (1992). Nonlinear elasticity: Incremental equations and bifurcation
phenomena, Nonlinear Equations in the Applied Sciences, 437-468.
Ogden, R. W. (1995). Nonlinear effects associated with waves in pre-stressed elastic
solids near bifurcation points, Nonlinear Waves in Solids, ASME Pub., 109-113.
Parker, R. G. and Mote, C. D. Jr. (1996). Exact boundary condition perturbation solutions
in eigenvalue problems, J. Appl. Mech., 63, 128-135.
Pearson, C. E. (1955). General theory of elastic stability, Quart. Appl. Math, 14, 133-
144.
Pence, T. J. and Song, J. (1991). Buckling instabilities in a thick elastic three-ply
composite plate under thrust, Int. J. Solids Structures, 14, 1809-1828.
Polignone, D. A. and Horgan, C. O. (1993). Effects of material anisotropy and
inhomogeneity on cavitation for composite incompressible anisotropic nonlinearly
elastic spheres, Int. J. Solids Structures, 30, 3381-3461.
Qiu, Y., Kim, S. and Pence, T. J. (1994). Plane strain buckling and wrinkling of neo-
Hookean laminates, Int. J. Solids Structures, 31, 1149-1178.
Reddy, J. N. (1986). Applied Functional Analysis and Variational Methods in
Engineering, McGraw-Hill, New York.
Rivlin, R. S. (1974). Stability of pure homogeneous deformations of an elastic cube under
dead loading, Q. Appl. Math, 32, 265-271.
170
Sagan, H. (1961). Boundary and Eigenvalue Problems in Mathematical Physics, Dover,
New York.
Sawyers, K. N. (1976). Stability of an elastic cube under dead loading: two equal forces,
Int. J. Non-Linear Mechanics, 11, 11-23.
Sawyers, K. N. and Rivlin, R. S. (1974). Bifurcation conditions for a thick elastic plate
under thrust, Int. J. Solids Structures, 10, 483-501.
Sawyers, K. N. and Rivlin, R. S. (1982). Stability of a thick elastic plate under thrust, Int.
J. Solids Structures, 12, 101-124.
Sawyers, K. N. (1996). On the thickness limitation for Euler buckling, Nonlinear eflects
in fluids and solids, Carroll, M.M and Hayes, M., Eds, Plenum Press, 297-320.
Song, J. and Pence, T. J. (1992). On the design of three-ply nonlinearly elastic composite
plates with optimal resistance to buckling, Struct. Optimization, 5, 45-54.
Tagg, Randall (1994). The Coutte-Taylor problem, Nonlinear Science Today, 4, 1-25.
Thompson, J. M. T. and Hunt, G. W. (1969). Comparative perturbation studies of the
elastica, Int. J. Mech. Sci, 11, 999-1014.
Timoshenko, S. P. and Gere, J. M. (1963). Theory of Elastic Stability, McGraw-Hill.
Weinberger, H. F. (1965). A first course in partial diflerential equations, with complex
variables and transfer methods, Dover Publications.
Wu, C. H. and Widera, O. E. (1969). Stability of a thick rubber solid subject to pressure
loads, Int. J. Solids Structures, 5, 1107-1117.
171
UV.
MICHIGAN STATE NI
lllllllllllllllllllllllllWIN
31293020
RRRIES
llllllll
1'7
LIB
7410