Você está na página 1de 9

Iznttim'ertn~t %'Ill./u~cs \ 4,1 ? n "-.L~ i. pp S?

I qT~L H~9~
, l~9X t..Ise~ler Science I.td
,\11 rL~hls rescr,,ed Printed in (;real Bril~lm
PII: SO14!-0296(97)00067-9 HI.II-.H296/gX $1U(X) - OAX)
EI.S[iV I[-R

Non-linear buckling and


postbuckling of elastic arches
Yong-Lin Pi and N. S. Trahair
The Centre Ira" Advam'ed Structural I-n.~,,ineerin~. The University ~71Svdm, v. NSW 2(X)6.
Australia
( Received ,hmuarv 199 7; revised version a¢'cepted Mar~h 1997)

This paper investigates the non-linear in-plane buckling and post-


buckling behaviour of elastic arches using a curved finite element
model for the non-linear analysis of elastic arches, which includes
the effects of prebuckling deformations and second-order terms in
the deformed curvatures and bending strains. It is found that the
effects of prebuckling deformations on the buckling of shallow
arches are significant, that the existence of a linear bifurcation
buckling load is not a sufficient condition for linear bifurcation
buckling to occur, and that the non-linear buckling loads of shallow
arches may be much lower than their linear buckling loads. The
effects of the higher-order terms in the deformed curvature on the
buckling and postbuckling behaviour of arches are also investi-
gated. It is found that these effects are important, particularly for
deep arches. © 1998 Elsevier Science Ltd.

Keywords: arch, buckling, elastic, nonlinear, postbuckling

I. Introduction Winkler-Bach theory and used a Runge-Kutta scheme to


analyze the large deflection behaviour of deep slender
Classical buckling theory assumes that the effects of pre- arches. Elias and Chen ~' proposed a non-linear shallow
buckling deformations on the displacement and geometrical curved element based on the exact deformed shape rather
stiffnesses can be ignored, and leads to linearized buckling than on polynomial shape functions. Wen and Suhendro H
solutions for elastic arches. This linearized buckling has proposed non-linear curved elements fi)r the moderate dis-
been investigated extensively ~ ". The linearized in-plane placement analysis of arches, and improved the accuracy
buckling solutions provide good predictions of the in-plane by averaging the non-linear parts of the axial strain. Kang
buckling for deep arches because the prebuckling defor- and Yoo ~ applied their non-linear finite element model to
mations are relatively small compared with their rises. the in-plane buckling analysis of elastic arches. Their
However, the prebuckling deformations of shallow arches results do not appear to include the effects of the prebuck-
are not small compared with their rises and may affect the ling deformations.
in-plane buckling behaviour, so that their effects should be DaDeppo and Schmidt" ~: considered higher-order terms
taken into account, in addition to this, linearized buckling in the deformed curvatures caused by bending actions in
analysis does not allow the investigation of the postbuck- their analytical model for elastic arches subjected to con-
ling behaviour at finite displacements. To include the centrated loads, and found that these higher-order terms are
effects of prebuckling deformations and perform postbuck- important for deep arches. However, the higher-order cur-
ling investigation, non-linear buckling and postbuckling vature terms does not appear to have been included in finite
analyses have to be used. element models for arches. In all the above finite element
Early research on the non-linear analysis of elastic arches m~xlels.O ~5, the non-linear strains consist of non-linear
using analytical methods has been summarized in the Guide membrane and linear bending strains. The higher-order
to Stability Design 7.x and the Handbook of Structural Stab- bending strain components caused by the higher-order cur-
ility". Recent research on the non-linear analysis of elastic vature terms have not been considered. Arches are struc-
arches using finite element methods has concentrated on tures which combine the bending member function of trans-
the membrane locking problem. Noor and Peters"' used mitting transverse forces with the compression member
selective reduced integration in their mixed model. Stolar- function of transmitting axial forces. In addition to the
ski and Belytschko ~ used reduced integration to solve the higher-order curvature components due to bending, the
membrane locking problem caused by using low-order axial deformations of the arches may affect the deformed
membrane displacement fields. Calhoun and DaDeppo ~2 curvatures, and may yield additional higher-order curvature
developed a curved non-linear arch element based on the terms. Pi and Trahair ~xconsidered these higher-order curva-

571
572 Non-linear buckling and postbuckling o f arches: Yong-Lin Pi and N. S. Trahair

tures and developed a curved tinite element model for the S developedlengthof an arch
three-dimensional non-linear analysis of elastic arches. lSl matrix of cross-section coordinates of a point /'
In this paper, a curved finite element model is developed t~ ,1~ thicknesses of the flange and web of an I-section
for the in-plane non-linear buckling and postbuckling I',ll' displacements of centroid in the o y , o s directions
analysis of elastic arches. The effects of the higher-order I'c,tt' < central displacements of centroid in the ov.o~
terms in the deformed curvatures and bending strains and directions
of the prebuckling deformations are included in the model. | 'q. li'q displacements of load point m the or-, o~-direc-
The inclusion of the higher-order curvature terms allows tions
the use of bending strains which are consistent with the V volume of lllelnber
membrane strains, and the same low-order polynomials can p, : I" + wlR
be used for both the axial and the radial displacements. As 17" : w' - v l R
a result, membrane locking problems are avoided. Neither %. coordinate in the or-direction
reduced-selective integration nor higher-order interpolation Yq v-coordinate of point on or-axis at which load q,
polynomials are needed. The model is used to investigate acts
the effects of prebuckling deformations and of higher-order first variation
curvature and bending strain terms on the in-plane non- second variation
linear buckling and postbuckling behaviour of elastic arch- I,~1 strain tensor
es. normal strain in the or-direction
~_~., E.~ shear strains
E.. longitudinal nomml strain
Notation
#(11 absolute vah,e of initial curvature of the a r c h cen-
A cross-section area troidal axis (= I I R )
[AI matrix for relationship between load point and K curvature after deformation
centroid displacements (1". longitudinal normal stress
B width of the flange of an I-section {,p} displacernents of centroid
matrix for strain-displacement relationships o arch subtended angle
b width of rectangular section 0m rotation of a point at which moment acts
D overall height of an l-section
IDI tangent modulus constitutive matrix
2. Finite element model
E Young's modulus of elasticity
f rise of an arch
2. ]. Rofotion matrix and curvature
h distance between the centroids of flanges of an I-
section, or height of rectangular section The Euler-Bernoulli hypothesis that plane sections that are
I/I unit matrix normal to ccntroidal axis remain plane and nornlal to the
major axis second moment o f area deformed centroidal axis after defomlation is used ill this
i,.,i.. basis vectors of point P paper. Three sets of axes are used to describe the geometry
basis vectors of displaced point Pt of the arch as shown in I"igure I. The first set is the lixed
J,d:
displacement stiffness matrix axes O Y Z with the basis vectors (P~,P,,). T h e second set is
Ikl
tangent stiffness matrix a body attached axis system ors which is used to describe
[/,-ll
[kI<; geometric stiffness matrix the geometry of the undeformed arch. The origin o of the
IkI,~ load stiffness matrix axis system ov.s is at the centroidal axis of the arch, the o.v-
L span of an arch axis coincides with the centroidal axis and the or-axis is
M,m concentrated and distributed moments towards the centre of the arch. The unit vector p, is along
N = qR the or-axis and the unit vector p. is tangential to the o.~-axis.
Nl-r ,Nii.~ linear buckling loads
[NI displacement shape function matrix
kZ
() centroid of cross-section
P an arbitrary point in cross-section
PI position of point P after deformation
PY,PI basis vectors of tixed space axis system O Y Z
P,P.. basis vectors of undeformed axis system ors
Q,Q,.,Q, concentrated loads
q,,q.- body attached moving basis vectors
q , q , ,q , distributed loads
R arch radius (= I/Ko) 0 ~ SPz ' _ ~ =Z
R position vector of point PI
{k} stress resultants P~
[RI translbrmation matrix
It,, position vector of point P
{r} nodal displacements
r position vector of deformed centroid oj
ro pos!tion vector of centroid o iy
r~ = \..I, IA
S coordinate around the os direction
SI coordinate around the OlSr direction Figure I Position vectors
Non-linear buckling and postbuckling of arches: Yong-Lin Pi and N. S. Trahair 573

After deformation, the origin o displaces radially v and q:'q:=q,'q,= 1 + ~'-" (9)
tangentially w to o~. and the oy- and os-axes move to oty,
and o,s,, respectively. The third set of axes o , y : , is In spite of this, the basis vectors are still assumed to be
attached to the arch and moves with the arch during the preserved as unit vectors in the conventional analysis, and
deformation. The body attached moving orthogonal unit the curvature after deformation becomes
vectors q, and q: are used to describe the motion of the
axes o~y,s,. Before deformation, they coincide with the unit K = d' + K,) (10)
vectors p, and p... During deformation, the vector q,. follows
the direction of the o~y,-axis and the vector q. follows the which does not include the second- and higher-order terms.
tangential direction of the axis o,s,. The rotations from the This expression has been widely used t 2 ' ' ' : a .
vectors p,.p. to the vectors q,.q. can be described using the
orthogonal rotation matrix R] as 2.2. Strains
The strain tensor Iel at a displaced point P, can be
{P,,P.}' = IR I {q,,q}' (I) expressed in terms of the basis vectors j,,j_ after defor-
mation and i,.,i.,before deformation as:'-
where

IR,,
I/¢1= R,
R,:]
R (2)
lel =
l' ']
~,, E
='9
'J¸ "1 I'" "])
-'LjJ, j:j.
-
ii, ii-
(II
The elements of the matrix [R] can be obtained from where
Figure I as
i. dR,. dRl=_ ~ I dR dR
I + ,i"' ~' = th -p,.i th + ~K,,,p.j : d~ : q , . a n d j d~ • II + , - ~ l l ~e:~q

R,,=R = l+e a n d R , "= - R . , =" I+e {3)


12
where /~' = v' + WE(,, ,7" = w' - VKo, ( 1 + e) =
It follows that
'J(l + ~i"')2+ ~,_,. and K,)= I/R is the absolute value of the
initial curvature of centroidal axis of the arch of radius R.
I R] is a skew-symmetric matrix which satisfies the orthog- ~,, = E,: = E,., = 0 13
onal conditions for a rotation matrix that [ R][R]r = [ I] and
det[RI = I'". where I l l is the unit matrix. After rotation. and
the basis vectors are preserved as unit vectors, so that
l C,, + I ~i.,z
~:_. = ~,i"' + ' 9 '9
q "q:=q,'q, = I, q,'q:=q:'q,.=0 (4)
_ [ g'( 1 + ~-" I - f%i-'"
The curvature I< of the centroidal axis after deformation can 14)
v Ll(I + ~" )-~ + e'~l ~
be obtained from >
+ K,,I( I + ~i"')2 + 9': I~ - K,,I
(1 + e)K = ( R I , . - R,:Ko)R.., + (R[. + R,,K )R.. (5) J

where v"e' = v" + w'K., W' = w" - V'K., and the small terms
Substituting equation (3) into equation (5) leads to
containing y~ are omitted. The strains e., include the
second-order membrane strains and the second- and
F'( 1 + g") - ~'vV' K,)
higher-order bending strains.
K = I(I + ¢"); + 9'2] 3;2 + [( l + v[") 2 + v'21 "~ (6)
The variation BE.. of the strain can be written as

which includes the second- and higher-order terms, and can ~e_. {S}lel{&o}
= = {S}IBIINI{Sr} (15J
be reduced to the accurate expression for the curvature of
a deformed straight beam (K~)= 0) 2~ where {S}={1.y}, the matrix [BI is as given in Pi and
~tt Trahair 2-', and

K=(I +v'-') ~2 (7) {,~} = {v,v',v",w,w',w"}' = INI{,} (16)

if the effect of axial extension w' is ignored. is introduced, in which IN] is the shape function matrix
Conventionally. the elements of the matrix [RI are whose elemcnts are functions of s. and {r} are the nodal
given b.'v,~ displacements

R,., =R... = 1 and R , : = - R . , . = 9 ' (8) (17)

In this case. [R] is still a skew-symmetric matrix, but does Because the strains include the higher-order deformed
not satisfy the orthogonal conditions for a rotation matrix curvature terms caused by the effects of axial delor-
that [R][R]r= [/] and det[R] = 1. After rotation, the basis mations, the order of the bending strains is consistent with
vectors are not preserved as unit vectors because that of the membrane strains, so that the same low-order
574 Non-linear buckling and postbuckling of arches: Yong-Lin Pi and N. S. Trahair

cubic polynomials can be used as the shape functions fl)r and the load stiffness matrix [kl,.~ is given by
both the v and w displacements. As a result, the membrane
locking problem is avoided. Neither selective reduction l
integrations nor higher-order interpolation polynomials
are needed.
Using the conventional rotation matrix and ignoring the
[kh, : -
f II
INI'IMI,,INIds- ~ INI'IMh.,INI
1.2
126)
small term ~"2/2 lead to the conventional strain e
The elements of the matrices [M],,. IM],, and [M),., are
given in Pi and "l'rahair 22. The tangent modulus matrix l 1)]
'~- = if" + 21 ~,2 _ w-e, (18) is given by'

which includes the second-order membrane strains, but D I = f , {.s}' t:{S}da ~27)
does not include the second- and higher-order bending
strains. Another drawback of using the conventional
rotation matrix is that it leads to non-zero shear strains ~,. An incremental-iterativc procedure is used Io solve
and E.,. which contradicts the Euler-Bernoulli hypothesis. equation (21 ) as

2.3. Loading and di.wlacenlents at a load point [k, I,{..X,-}~= {Ap~.}, + {&,,}l ' ~ 2;"; )
General distributed loads per unit length {q} and concen-
trated loads {Q} can be written as where the subscript i denotes the load step. the superscript
j denotes the iterative cycle. {Ap,.}, = the incremental forces
of the current load step i. and {Ap,~-I = the unbalanced
{q} = {q,, q,. m}' and {Q} = {Q,. Q,. M}' (19)
forces in the iteration (j - I ) of the current load step i. The
unbalanced forces can be calculated from
where q, and Q,. are the distributed and concentrated forces
parallel to the oy-axis, q, and Q, are the distributed and
concentrated forces tangential to the os-axis, and m and M
are the distributed and concentrated amounts.
The variation {au}q of the displacements at the load 1291
points can be written as
where the stress resultants {R} are given by
{Su}q = {By,,. 8w,,. 80,,,}'= [A 1{,8,¢} = IAIINI{&}
( 2o )
{R} = f.x IS']~r:dA and ~v = E E 130 )
where the matrix IA] is given in Pi and Trahair':.
The details of solving the incremental-itcrati~e
2.4. Non-linear equilibrium equation ( 2 8 ) a r k given in Pi and Trahair:'.
The non-linear incremental equilibrium equations can be
obtained by applying the principle of the virtual work to
3. Verification
two successive equilibrium states as 22
Comparisons of experimental, analytical and tinite element
Ikl.,{kr} = {&~} (211 results are used to verify the present model. In the follow-
ing numerical investigations, cubic polynomials are used to
where the increment of the equivalent element external interpolate both the v and w displacements. The lirsl
forces {Ap} is given by example is a shallow arch tested by Gjelsvik and Bodner 24.
which was also referred to in the Handbook of Structural
fl. Stability". The arch was made from aluminium alloy with
{AI,} = INI'[AI], {kq}ds + ~ [NI'IAI~,{AQ}. a Young's modulus E = 6.895 x I(P MPa ( 10; psil. The
o 1.2 cross-section was rectangular with the width B = 25.4 nun
(22) ( 1 in) and the depth h-- 4.7625 mm (3/16 in). The span of
the arches was L = 863.6 mm (34 in). Both ends were tixed
and the tangent stiffness matrix [k],. is given by and the arch was subjected to a central concentrated load.
The variations of the central deflection with the load for
[kit = [kl + [ k l , + [klo (23) the arch with the parameter 2J/h -- 11.62 (where the rise of
the arch./'= 27.66 m m ( 1.089 in)) by the present model are
where the displacement stiffness matrix [k] is given by compared with the experimental curve of Gjelsvik and
Bodner 24 in kTgure2. Eight elements were used. Also
L shown in Figure 2. are the results of Belytschko and
Ik] =
f o
[Nf"IBI"IDI[BI[N]ds (24) Glaum >, Bathe et al. 2~', and Mallett and Berke :~.
Belytschko and Glaum 2~ used 16 curved elements and
Bathe et al. '-~ used 12 two-dimensional isoparametric
the geometrical stiffness matrix [k](; is given by
elements with 8 nodes per elements. The arch undergoes
snap-through buckling. The buckling load and the load-
L

[klc; =
f [NI'[MI,,[N]ds (25) deflection curve predicted by the present model are close
to the experimental results. The other finite element results
Non-linear buckling and postbuckling of arches: Yong-Li.n Pi and N. S. Trahair 575
40~
/ : I deformed curvature of equation (6) without considering the
i
o
15. r . effect of the axial extension ~" on the curvature changes.
Langhaar et al. 2~ used a centre line inextensional theory to
"*" w- -d • • obtain analytically a non-linear buckling load of
25~ ;f \ ,' ' i Q = 18.77 N (4.22 Ib), which is even higher. The value of
5.
C'
,17 -... /o / Q = 19.48 N (4.37 lb) of Wen and Lange 2'~ (eight elements)
20~/ \ ~ /." q
was reported as accounting for the effect of prebuckling
/. o

15. i deformations, but appears to be too high.


The third example is the symmetric snap-through behav-
~ Expcrlrm:ntal results of Gjelsvlk arid Bodner :~ iour of a 60 ° (f/L ~- O. 134) fixed circular arch subjected to
" Pre~ent mo~el
5
I/ • Altematwe mo~l
Bel tschko and Glaum :~
:. a central concentrated load as shown in Figure 4. The
' B~etal ~ I pound = 4 448 N J results for the load-deflection relationships are compared
1 -'--- Malletl and Berke": I llx:h = 254 mrn
0t with those of Calhoun and DaDeppo 12 and Wen and Suhen-
0 02 04 06 08 1 12 14 16 Lg
Central deflection v~ (ir,~hes)
dro '4. The present results, using two and four elements,
agree very well with those of Calhoun and DaDeppo 12 with
Figure 2 C o m p a r i s o n w i t h e x p e r i m e n t a l results
16 elements, and of Wen and Suhendro ~" with 8 elements.
It can also be seen that the present m~xlel can predict the
predict higher buckling loads and stiffer load-deflection postbuckling behaviour, while the results of both Calhoun
curves. and DaDeppo I-~ and Wen and Suhendro I~ did not pass the
The second example is the antisymmetric bifurcation snap-through buckling point. The snap-through buckling
buckling of a semi-circular pin-ended arch (0 = 180 °, f / L = load obtained analytically by DaDeppo and Schmidt 17 is
0.5) subjected to a central concentrated load. The proper- identical to the results using the Handbook of Structural
ties of the arch are I , = 2 . 5 8 5 m m 4 ( 6 . 2 1 x 10 ~in4), Stability ~and the Guide to Stability i)esign ~, and is slightly
A = 0 . 5 m m ~ ( 0 . 7 7 5 x 10-~in-~), R = 2 5 4 m m ( 1 0 i n ) and higher than the present result due to the use of an inexten-
E=7.171×I(PMPa ( l . 0 4 x l 0 7 p s i ) . To induce bifur- sional assumption. The results of DaDeppo and Schmidt ~7
cation buckling, a small initial horizontal load (0.001 N ) i s and Calhoun and DaDeppo I-~ were based on the accurate
applied at the crown. deformed curvature of equation (6) without considering the
The results for the load-deflection relationships obtained effect of the axial extension ~" on the curvature changes.
by the present model are shown in Figure 3. Six elements The comparisons with the experimental, analytical and
were used. The bifurcation buckling loads were determined finite element results demonstrate that the present model
both from the load-deflection curve and at the sign change produces more accurate results and needs fewer elements
of the determinant (71" the tangent stiffness matrix. The for convergence than other finite element models.
bifurcation buckling load Q and the corresponding central
vertical deflection v, of the present model are Q = 15.05 N
4. Non-linear buckling and postbuckling of steel
(3.38 lb) and ~; =42.67 mm (1.68 in).
Also shown in Figure 3 are the values of the bifurcation
arches
buckling load obtained by other methods. The values Uniformly distributed radial loads acting on a circular arch
obtained by Elias and Chen I-~ using 18 elements are cause uniform compression without bending when the pre-
Q = 15.43 N (3.47 lb) and v, =45.47 mm (1.79 in), which buckling deformations can be ignored, and the arch may
are a little higher than those of the present model. The buckle antisymmetrically under uniform compression.
element used by Elias and Chen ~ was based on the exact When the prebuckling deflections can be ignored, the buck-
deformed shape of a locally shallow curved element. The ling loads obtained using a linearized theory agree well
analytical values of DaDeppo and Schmidt I" and the results with the results of non-linear analysis. However, when they
using the Handbook of Structural Stability '~ and the Guide cannot be ignored, as for example for a shallow arch where
to Stability Design 7 are Q = 16.81 N (3.78 lb) and the prebuckling deformations reduce the stiffness of the
v, = 49.53 mm ( 1.95 in), which are even higher. The result arch, then the buckling loads predicted by a linearized
of DaDeppo and SchmidP" was based on the accurate analysis may be substantially different from those obtained

20f. . . . . .
.Q
4 5 :~ . 7 . R=SO~mc~O~nl
i AiI~'- '\ ,~ ,,~ = Z5 8064 ~tO ~(40m:l
18t . . . . . .
16 ~O ~i- f* -"~,.-. -. ~o~. ,..ss.97..m ~. m,.',
.0- . . . .

j . / /
12 " R II I
¢, ,°
..
3
-- Pmt~*l~mod~l (6 elemen;~ 15 32 e l e r n ~ ~ ""
- AIt~matave model (10elemcms)
o Elias and C ~ m ' 1( 18 elea~.ms)
• Guide to Stability De~l~,~ I F~L • Candc tO Stabihty DCSt~.7
Handbook of Su~cmnd-Sc~ility of Japan? Ir Hma,x~okof Sm~ctuzS~tabilJ~of Japan.
~
' and D a l ~ l ~ and S d ~ U d ~ ' * 0 5~ l n d AtudyUcM b~dd~ll Iold of D I d ~ Schnu~ '~
.... l..,mlluu~ et ~J.~
. . . . . - - Wen ~¢:1 ~ ~ (8 elemems) [ ~ B~mB ~ o ~ n l ~ ( i 6 elerc~nts )
01 , , , , ,
,o io ~o ,~ ~ ;o 70 80 0 20O 40O 600 8(30 1(300

Ccnlral vcnica] deflection vc (ram) Central vet~cal d~flccuon v< ( ~ )

Figure 3 C o m p a r i s o n w i t h analytical and f i n i t e e l e m e n t results Figure 4 C o m p a r i s o n w i t h analytical and finite e l e m e n t results


for a 180 ° p i n - e n d e d arch for a 60 ° fixed arch
576 Non-linear buckling and postbuckling of arches: Yong-Lin Pi and N. S. Trahair
I2! 12[~-- ........

._._,_ , _, .__--- .... ...........

I t - o . . . . . . . . . • a.,-. • - - - ='-"" - . . . . . . . . . . . . . . . . .
t .....

,,~-:-- / _--:-:-~::..:,.; ..........


08
"a

06 ...... ! 06
\\:.//:

0.4 L
.... _L
O4

- - 0= 180 °
- - 0= 180" O BifttrcaUonbucldmg ....... 8 = 150"
02 ..... O= 150" + Maximum load 02 ..... 0 = 120"
--- 0=90" .............. 0=90 ¢
.......... 0=50" o B ffurcauon h u c k h n g
+ Maximum load
0
o~: 06: oi~ o~ 0"2 o~'~ o.13 e3: o.,. o,,~ o: 005 Ol { I~ 02 025 -0~ ()'~5 o)i o;'5 ;7.
Dlmensmnless cenlral vemcal dcflecuon vdf Dimensionless ccntzal v e r n c a J dcllecuon v i i

Figure 5 Deeper arches of constant slenderness Figure 7 Deeper arches of constant radius

from a nonlinear analysis. The differences are investigated [ - - 0 = 4o ° o~,T=c=;o, b,~u,.~


" 0 = 35 ~ &Snap-through buckling
for two groups A and B of pin-ended arches with different 09I.-
:
0=33 ~
8 - 22"
subtended angles 0 subjected to uniformly distributed radial 08- q

loads. The cross-section used for the arches is a 10UB29 0,! /N


(BHP3')). Group A arches have a constant slenderness ratio
of 0.5S/r, = 50 and Group B arches have a constant radius oor / ..,,\
o51- i ' .., x, _ L
o f R = I0 m. A small initial horizontal load (about 0.003 N~,
where N~, = rreEl,/(0.SS): is the antisymmetric buckling
load of a corresponding column) is applied at crown of
each arch to induce antisymmetric bifurcation buckling.
The typical variations of the dimensionless central verti-
cal deflection r , / f with the dimensionless axial load N/N),
are shown in Figure 5 for the constant slenderness Group A 0 05 I I} 2 25
arches with subtended angles 180, 150, 90 and 50 ° , and in Dimensionless cenu~J vcrucad d~l]ectmn vdf
Figure 6 for arches with subtended angles 40, 30, 20, 15
and 8 °, where N = qR is the nominal axial load. The typical Figure8 Shallower arches of constant radius
variations of the dimensionless central vertical deflections
r , / f with the dimensionless axial load N/NE~ are shown in bifurcation buckling, but the postbuckling h)ads decrease.
Figure 7 for constant radius Group B arches with subtended and the buckling loads are the maximum loads. The con-
angles 180, 150, 120 and 90 °, and in Figure 8 for arches stant slenderness Group A arches with subtended angles 90.
with subtended angles 40, 35. 33 and 22 °. 50, 30 and 20 ° and the constant radius Grot, p B arches with
Five types of behaviour can be seen in Figures 5-8 and subtended angles 90, 40 ° are of this type.
are summarized in Figure 9. For the first type, the arch For the third type, the arch buckles by symmetric snap-
undergoes antisymmetric bifurcation buckling. After buck- through under constant load. Under deflection corer.l, these
ling, the load carrying capacity of the arch increases arches bifurcate unsymmetrically on the descending branch
slightly and the maximum load is slightly higher than the of the load-deflection curve. The Group A arch with it sub-
buckling load. The deep constant slenderness Group A tended angle of 20 ° and the Group B arch with subtended
arches with subtended angles 180 and 150 ° and the deep angle of 35 ° are of this type. For the fourth type, the arch
constant radius Group B arches with subtended angles 180, buckles by symmetric snap-through without bifurcation.
150 and 120 ° are of this type. The Group A archwith a subtended angle of 15 ° and the
For the second type, the arch undergoes antisymmetric Group B arch with it subtended angle of 3Y' are of this
type. For the last type, there is no buckling. "i'hc Group A
12 r arch with a subtended angle of 8" and the Group B arch
- -- 0 = 40 ° o Bffxu~auon b u d d i n g
I . . . . 0 = 30° ASn~-~rough buckJm~ ! with a subtended angle of 22 ° are of this type.
_ . 0=20 °
I . . . . . . . . . O= 15" -' It can also be observed that the dimensionless prebt,ck-
I __0=8o
ling deflections v,,'f increase as the subtended angle ~t
decreases and become substantial for shallow arches. For
:..,.".... 07 :/, example, the prebuckling deflection at buckling is about
0 . 2 I f for constant slenderness Group A arch with a sub-
tended angle of 30 ° . In this case, the profile of the arch is
no longer circular so that the uniformly distributed radial
loads produce bending as well as axial compression. This
reduces the stiffness of the arch which buckles at a load
which is much lower than the linear buckling load.
The numerical investigation of the buckling behaviour of
0 05 I 15 2 25
arches with different subtended angles # shows thal bifur-
DLn~nslonless c c n t n d venJc~1deflection vdf
cation buckling governs (Type 1 or 2) when ~ >-- 2 1 for
Figure6 Shallower arches of constant slenderness the constant slenderness Group A arches and ~ :- 36 for
Non-linear buckling and postbuckling of arches: Yong-Lin Pi and N. S. Trahair 577

1-- ---- Tylx 1


. . . . . . I 1
Type 1 ( 1 ) when fir, > 4.6, the arches are governed by bifurcation
-I
buckling (Types 1 and 2),
J (2) when 4.1 < f i r , < 4.6, the arches are governed by
J symmetric snap-through buckling with bifurcation
point on the descending branch of the curve (Type 3):
i
(3) when 1 . 7 < f / r , < 4 . 1 , the arches are governed by
0 0!4 04
vJf wolf
symmetric snap-through buckling (Type 4):
1 Type 2 , 1 Type 2
(4) when .f/r, < 1.7. the arches do not buckle (type 5).

The buckling loads of the two groups of arches are com-


pared in Figure I0 with the linear buckling loads obtained
by other researchers. For a circular arch subjected to uni-
formly distributed loads around the arch, Simitses ~ found
0 2 O! 0.2 that the elastic buckling load depends on the load con-
ve./f wc/f ditions assumed, if the loads remain normal to the deflected
l ...... ~ 3 . --- L~ T y p e 3 reference axis and the effects of the tangential displace-
J
t
nnents on in-plane buckling are ignored, then the linear
buckling load is given by N~, = El( rF/(0.SS)-' - I IR ~-)'.e,'*.''
If the loads are assumed to remain parallel to their original

v,:Jf wJf
t
0.02
directions and the centre line of an arch is inextensional
(w'-r/R=0). then the linear buckling load is given by
NL, = rreEll(0.5S) >. This buckling load is the same its the
antisymmetric linear buckling load of a column with the
TvP ¢ 4 Tgpe 4
same length and boundary conditions. The linear buckling
/ J 'l loads N~:~ are lower than N~.:. It can be seen from Figtov I0
thal the non-linear buckling loads are lower than the linear

vdf
.
3
(

wc/f
i
002
buckling loads N~.~ and that when the subtended angle O is
less than 60 °, the non-linear buckling loads decrease dra-
nnatically due to the increase in the effect of the prebuckling
deformations, and become substantially lower than the lin-
1.... Type 5 . _ 1 __ Type 5 ear buckling loads NH and NN~ at sunall subtended angles.
At very low values of 0, there is no buckling. This indicates
] that the effects of prebuckling deformations on the in-plane
buckling of arches are significant, that existence of a linear
bifurcation buckling load is not a sufficient condition for
linear bifurcation buckling to take place, and that the ram-
1
0 ~ 00 0.02. linear buckling loads of shallow arches are much lower than
vcJf wJ r
the linear buckling loads. At high values of O, the ram-
linear bifurcation buckling loads lie between the linearized
o Btfurcataon buckling & Snap-through buckling + Maximum load
predictions N~:~ and N~:,.
Figure9 Types of nonlinear behaviour of pin-ended arches

5. Effect of high-order curvature terms


To investigate the effects of the higher-order curvature and
the constant radius Group B arches. Symmetric snap- bending strain teruns on the buckling and postbuckling
through buckling is followed by bifurcation on the hehaviour of arches, an alternative model based on the con-
descending branch of the load-deflection curve (Type 3) ventional strain of equati(m ( l 8) is also used for compari-
when 1~.)~<7 0.<-21 ¢ for the Group A arches and
34 ° < 0 ~:- 3~¢' for the Group B arches. The arches are gov- 12 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
erned b~, symmetric snap-through buckling (Type 4) when
g' <:- 0 "-, 19" for the Group A arches and 22 ° <- 0 < 34 ¢
I. . . . . . . . . . .
for the Group B arches. There is no buckling ( T y p e 5 ) a

when 0 -: 8 ° for the Group A arches, and (9 <_ 22 ° for the g


<18" +
Group B arches. The angles distinguishing the different 2
q

cases for the Group A arches are different from those for Z

the Group B arches, so the subtended angle # cannot be _s

used as a single criterion to distinguish the different unodes. f


L
o 4, o
Further investigation shows that the ratio.l/~:, of the rise to
the radius of gyration may be used as a single criterion.
0 2-
For the (.}roup A arches, the limits of 21, 19 and 8 ° corre- Linear buckhng load ~ '
• Linear buckhng load n : i 4
spond to 17r, = 4.57.4.14 and 1.74 which are almost equal o
*
N o n l i n e a r b ~ c k l i n g Imu:l for
Nonh~ar buckling load
Group A archc',
for Group B alches
OL-- . . . . . . . . . . . . . . . . . .
to the ~altles fir, = 4.53. 4.0b and 1.70 for the Group B 20 40 60 80 ILI4} 120 140 ]64) Ig()
arches of 36. 34 and 22 ~'. Thus, the following criteria are S u b l e n d e d angle 8 {degree~

suggested for distinguishing the different types of behav- Figure 70 Buckling of pin-ended arches subjected to uniformly
iour: distributed radial loads
578 Non-linear buckling and postbuckling of arches: Yong-Lin Pi and N. S. Trahair
12[
son. Cubic polynomials are also used to interpolate both , - - P r e ~ n t m o ~ l for O = 5 0 " o BiiemcatJon b u c k l i n g i
I - - - Altcrrlauvemodtlfoi'0=50 ~ + M~xlrnumload
the v and w displacements for the alternative model. • ----Pr¢~nt~lforO= 120"

Fixed and pin-ended arches are used to investigate the


effects of high-order curvature and bending strain terms on
the non-linear buckling and postbuckling behaviours of 0., .... IT ~.~ -~'-<.L " "'

arches. The arches are subjected to uniformly distributed i i i= / :I


radial loads. The cross-section used for these arches is the
10UB29 ~° (the flange width B = 0.146 m, the overall depth
of section D = 0.2559 m, the flange thickness t~ = 0.0109 m,
and the web thickness t,, =0.0064 m). The developed ' y /"
length and the slenderness ratio of the arches are S = 10.8 m 02[ ." ;-
and 0.5S/r, = 50.
The first group of arches consists ot" a fixed shallow arch 00 L : ..... 0_0 5 Z _ _0 I 0 15
'" 02 0 25 . . . . . . . ~3
( 0 = 5 0 ° , f / L ~ 0.11085) and a fixed deep arch ( 0 = 120°, D t m e n s i o n l e s s central honzontal d i s p l a c e m e n t w d [
/'/L ~ 0.2887). The snap-through buckling of these arches
is investigated. The variations of the dimensionless central Figure 12 Bifurcation buckling of pin-ended arches
vertical deflections v,/f with the dimensionless axial loads
N/N~ are shown in Figure 11, where N = qR is the nominal
axial load and NE = 7r2EI/(0.35S): is the buckling load of
the corresponding column with the same length and bound- 6. Conclusions
ary conditions. For both arches, the alternative model pre-
dicts higher snap-through buckling loads and postbuckling A curved linite clement model for the non-linear analysis
behaviour than the present model. of elastic arches has been developed based on an accurate
The second group of arches consists of a pin-ended shal- rotation matrix which satisfies the conditions [RIIRI* =
low arch ( 0 = 5 0 °, f/L--~ 0.11085) and a pin-ended deep IR]rlR]=[II and d e t [ R l : l . The model includc~, the
arch (0 = 120° (ilL ~ 0.2887). The bifurcation buckling of effects of high-order terms in the deformed curvature, so
the arches are investigated. To induce bifurcation buckling, that the deformations of arches are predicted more acct*-
a small initial horizontal load (about 0.003Nt. where ratcly. The higher-order terms also allows the order ,)t" the
N~:= rr2EI,/(0.5S) 2 is the antisymmetric buckling load of bending strains to be consistent with that of membrane
strains, so that the same low-order polynomials can bc used
the corresponding column with the same length and bound-
to interpolate both the radial and axial displacements,. As
ary conditions) is applied at the crown. The variations of
a result, membrane locking problems are avoided. Neither
the dimensionless central horizontal displacements w,/J
reduced-selective integrations nor higher-order inter-
with the dimensionless axial loads N/Nt: are shown in
Figure 12. For both arches, the alternative model predicts polation polynomials are needed. Comparisons with the
higher bifurcation buckling loads and postbuckling behav- experimental, analytical and finite element results have
demonstrated that the present model is very efficient and
tour than the present model.
The buckling and postbuckling behaviour of the effective in terms of accuracy and the number of elements
used for convergence.
examples in the early verification section are also investi-
gated using the alternative model. As shown in Figures 2-- The non-linear buckling and postbuckling bchaviours of
elastic pin-ended arches subjected to uniformly distributed
4, the alternative model predicts higher buckling loads and
radial loads have also been investigated. The effects of pro-
postbuckling behaviour than the present model. These
buckling deformations on the buckling loads are significant
examples show that the higher-order curvature and bending
strain terms are important for the buckling and postbuckling for shallow arches. The non-linear buckling loads of shal-
of arches, particularly for deep arches, and so they should low arches are much lower than the linear bifurcation buck-
be considered in a finite element model when the accurate ling loads due to the effect of the prebuckling deformations.
The existence of a linear bifurcation buckling load is not
predictions of the elastic buckling load and postbuckling
a sufficient condition for bifurcation buckling to take place.
behaviour of arches are required.
The non-linear buckling mode of the arches studied i.,, by
bifurcation only when the ratioJ7r, of the rise to the radius
14 r ." • .
of gyration is greater than a certain value. Other arches arc
i
governed by symmetric snap-through buckling, or may not
lzL ' q buckle. The non-linear buckling loads of deep arches are
higher than the linear buckling loads NE~,. but lower than
,[, ,, ] the linear buckling loads N~,_.
The effects of high-order terms in the deformed ct, rva-
ture on the non-linear buckling and postbuckling behaviour
of arches have also been investigated. It is found that the

°°/
e5 Ot "'""
J effects are important for the buckling and postbuckling
analysis, particularly for deep arches.

{
Altl~mltls~m o d e l foc O = 120" |

02 0.4 06 05 1.2 14
Acknowledgements
Oa/lledlr~onJe.~scsmlnJ vcrtlclfl dasplacerl~ntsvdf
This work has been supported by a research grant provided
Figure 11 Snap-through buckling of fixed arches by the Australian Research Council.
Non-linear buckling and postbuckling of arches: Yong-Lin Pi and N. S. Trahair 579

arches under a concentrated load'..I. ,4ppl. Mech.. ASME 1969,


References
36 (6). 325- 327
I Vlaso',. V Z. lhm-walh,d ela~tk heams. 2nd edn. Israel Program for 17 DaDeppo. D. A. and Schmidt, R. "Large dellections and stability of
Scicntilic lrans[ation, Jerusalem. Israel, 1961 hingeless circular arches under interacting loads', .I. AppL Mech..
2 Timoshenko, S. P. and Gere. J. M. Theory of elastic stabilitr. 2nd ASME 1974, 41 (4), 989-994
edn. McGraw-Hill, New York. 1961 18 Pi. Y.-L. and Trahair, N. S. 'Three-dimensional nonlinear analysis of
3 Simitscs, (;. J. Atz introthwtion to the elastic stahilit.v of structure.~. elastic arches', Engng Strm't. 1996. 18(I). 49-63
Prentice-Hall. Englewo~M Cliffs. New Jersey, [JSA, 1976 19 G~db~xty. A. M. Cartesian tensors with application to mechatm',~.
5 Yang. Y.-B. and Kuo. l,.-J. 'El'leer of curvature on stability of curved fluid mechanics and elasticity. Ellis tlorwo~t, ('hichester. England.
beams'. J Struct. k-,'tgng. AS('E 1987, 113 (6), 821-841 1982
5 t"apangeli,,, J. P. and Trahair. N. S. 'l:lexural-torsional buckling of 21) l,ove, A. E, H. A treatise on the matlzemati~al tlu'orv ~!l elaxticilv.
arches', J. Strm't. Engng. A.~'('E 1987. 113(4). 889-906 4th edn, l)over Publications. Ncv. York. 1944
6 Rajasekaran. S. and Padmanabhan. S. 'Pkluations of curved beams', 21 Thompson, J. M. T. and tlunt, G. W. ~ twnera/ theot~v ~!/ ehtsti~
J. Engn,g Mech. AS('E 1989. 115 (5), 1094-11 I I stability. Wiley. New York, 1973
7 Guide t, stability de.sign criteria ]br metal structures, 3rd cdn. cd. 22 Pi. Y.-I,. and Trahair. N. S. 'Effects of curvature on thc mmlincar
B. G. Johnston, Wile},. New York, 1976 behaviour of elastic arches', Rex. Rcp. no. R720, School Civ. and
8 (;uide to stability de.~ign <'riteria fiJr metal stru<'ture~. 4th edn. ed. Min. Engrg. Univ. Sydne}. NSW, Australia. 1996
"I'. V. Galamb~>s. Wiley. New York, 1988 23 Pi. Y.-I.. and Trahair. N. S. "Nonlinear inelastic analysis of steel
9 tt,ndhoo~ <~fstru<'tural stability, ed. Column Research Committee of beam-columns'. J. Strm't. Engng. AS('E 1994. 120 ( 7 I, 2041 2(~1
Japan. Corona Publishing Company, Tokyo, Japan, 1971 24 Gjelsvik. A. and B~xtncr, S. R. "The energ,, critcrion and snap buck-
I(I N~w,r, A. K. and Peters, J. M. "Mixed unodel and reduced/selective ling of arches'..L Engng Mech. I)iv.. ASCE 1962, 88(51, 87.. 134
integration displacement models for nonlinear analysis of curved 25 Belytschko, T. and Glaum, I,. W. 'Application of higher order co-
beams'. Int. J. Numer. Method Engng 1981. 17 (4). 615-631 rotation stretch theories to nonlinear linite element analysis', ('otnlL
II Stolarski. H. and Bclytschko. T. "Membrane locking and reduced Struct. 1979. 1 0 ( I - 2 ) . 175-182
integration for curved elements'. J. Appl. Mech. "l)'ans. ASMI- 1982. 26 Bathe, K., Rarnm, I-. and Wilson, E. "l'mitc element formulations for
4 9 ( I ) . 172 -176 large deflection dynamic analysis', Int. J. Numer. Method.~ Engng
12 ('alhoun. P. R. and l)aDeppo. D. A. 'Nonlinear tinite element analysis 1975, 9(2), 353.-386
of clamped arches'. J. Strmt. Engng. ASCE 1983. 109 (3). 599-612 27 Mallett. R. and Berke, I,. "Automated method for the large deflection
13 Elias. Z. M. and Chen. K.-L. 'Nonlinear shallow curved-beam linite and instability analysis of three-dimensional truss and frame assem-
element'. J. l:'ngng Mech.. AS('E 1988, 114 (6), 1076-1087 blies'. AFFDI, TR-66-102 (AD-810499). 1966
14 Wen. R. K. and Suhendro, Bambang "Nonlinear curved-beam clement 28 l,anghaar. H. 1,.. Boresi. A. P. and Car'.er, I). R. "Energy theory of
for arch structures', J. Strm't. Engng. ASCE 1991, 117(111. 3496-. buckling of circular elastic rings and arches'. Pro~. Second U.S. Nat.
3515 ('ong. of Appl. Mech.. ASME 1954, pp437-443
15 Kang. Y. J. and Ytx~. C. H. 'Thin-walled curved beams. II: analytical 29 Wen, R. K. and Lange. J. 'Curved beam element for arch buckling
stflutions for buckling of arches'. J. Engng Mech., ASCE 1994, analysis'. J. Struct. I)iv.. AS('E 19111. 107 (S'I'I I ), 2053--2(h59
120¢ I0). 2102 -2125 30 BHP rolled sections and plates. The Broken Hill Proprietary Co. lad.
16 l)aDcpl~. I). A. and Schmidt. R. 'Sidcsway buckling of deep circular Mclbournc, Australia. 1972

Você também pode gostar