Professional Documents
Culture Documents
(Asce) 0733-9399 (1989) 115 5 (1094)
(Asce) 0733-9399 (1989) 115 5 (1094)
(Asce) 0733-9399 (1989) 115 5 (1094)
ABSTRACT: The thin-walled curved beam equations are formulated using the prin-
Downloaded from ascelibrary.org by UNB - Universidade de Brasilia on 03/09/23. Copyright ASCE. For personal use only; all rights reserved.
ciple of virtual work. The Euler Lagrange equilibrium equations and the associated
static and kinematic boundary conditions thus obtained are compared with Vlasov,
Chai Hang Yoo, and Papengelis and Trahair. In and out of plane buckling of
curved beam problems are solved and compared to published results.
INTRODUCTION
ASSUMPTIONS
1. The length of the member is much larger than the depth and the lateral
dimensions are larger than the thickness:
t d
-=£0.1; -<0.1
d I
2. The displacements are large and the strains are small.
3. The projection of the cross section on a plane normal to the longitudinal
axis does not distort during deformation.
4. The shear strains due to shear stress in equilibrium with the changes of
normal stress is small and can be neglected (Euler Bernoulli's hypothesis).
5. The shear strains in the planes normal to the middle surface of the thin
wall are small and can be neglected.
Consider a circular curved member with the center of curvature o' as shown
in Fig. 1. The well-known three-dimensional small strain and large displace-
ment relations in cylindrical coordinate system may be written in terms of
cartesian system X(Z), Y(-r), Z(0) by taking the origin at arbitrary point O
and selecting D as sectorial centroid (see Fig. 2) as
2-,
du du
6.v =
(la)
dx dx
_ J AV L—
FIG. 2. Definition
dv 1 du\2 (dvx ^
(lb)
dy 2l\dy dy/ J
1 3vv v 1 dv 'du
«e _ h W (Ic)
z
r ae + —; ae vae,
r 2r 2 1 dv dv W dv 1 du du
= 7x9 = — + - — + + (Id)
dx r1 30
3v r dx 59 r dx r dx 30
1 3v w dw 1 dv dv w dv 1 3K du
lyz = 7^6 = - — - +— + (le)
r ae r 3y r dy 30 r 3;y r ay 30
a« av du du dv dv
— +— + _l_ (1/)
dy dx dx dy dx dy
where u, v, and w denote the displacements in X, Y, and Z directions, re-
spectively. It is seen from Eq. 1 that the nonlinear terms associated with
dw/dx, dw/dy, and dw/dz are neglected since w is much smaller compared
to u and v. As per the assumptions of open section, shear strain in the plane
normal to the middle surface of the wall is small and can be neglected. yes
1096
w0 (2c)
+ { vk H J sin $
"o
\p = $ (3)
i?
Substituting Eq. 2 into Eq. 1 and $ for sin $ and 1 - $ 2 / 2 for cos $ and
(R — y) for r we get the nonzero strain components at the point on the cross
section as axial ez and shear ysB given by Padhmanabhan (1984)
'vo + W(A
1 i? «o 2 Vo
w 0 - - + + — + —l- I - y Vo + u'i<t>
(i - y) I 2R R
R
Vo + w 0
R / w0\ <E> v0$
2R 2R 2/f3 2i?J
vK + - 1$
*'2
"o
- co $ " - — + (x2 + y2) + $V)
/?
xy\v'Q + — W
y(x2 + y 2 ) $ ' 2 yX2®2 xyv 0 $ y2«o<I)'
_) 1 .— 1_ (4a)
2R 2R6 R R" R
and
1097
0>z
J*n <*zy &zz
p
**> ;_ I ,„ „< (f^o)' (M*v°y (M*w°y ^ w ^ ' v^y
-Mx hi -(PVQ) : 1 • ;
2
R [ R R R R R
-,«,*,--<^;+^+5+^+«}=„ m
—P H M 1- —T- H r Vo H H
R [R R2 R2 \ Rj R
WJ& Txy<P' rsv4>]
1098
Mv MA> W,„
- WL — T' + ' + MX - Myv'i
R
*u+5
1
.TV '
Wj vg + —°
R
(.MJb'y - wjvr . T^
2 3
R R R
Eqs. 8a-d represent precisely the same requirement for total equilibrium.
They are approximate to the extent imposed by the assumptions of thin-
walled beam theory; however, they are consistent with classical beam equa-
1099
tions. In classical beam theory it is assumed that the stress resultants con-
tained within the braces of the Eq. 8 may be determined by statics in the
initial undeformed configuration of the member. The remaining primary stress
resultants are expressed in terms of the displacements of the reference axes
by assuming a linear relationship between strain and the displacement of the
reference axes. Here ez, yS2 are given by
1 «0
w0 y[ v° + j XUg co d>" - .. (10a) •
i-y
"o
•y„ = 2T| $' - (106)
buy
CU}>
-(<*>" - ul/R))
and
(13)
R R
n,=
{ xydA
V =
(jyydA
1 wxdA
i -y
R R
\ xx22dA
dA I u2dA
JA JA
Iy = /' = —
1-y
R
nl
T
dA 2+ (14)
}A(R-y + nl) {R - yf (R - y)J
If the values of the stress resultants obtained in Eq. 12 are substituted into
Eq. 8 one gets the Euler Lagrange equations in terms of the displacements
referred to any arbitrary point O and references axes OX, OY passing through
O. When the beam becomes inelastic the centroid and shear center lose their
meaning and hence these equations are more general. Upon specializing O
to be the centroid and shear center (for doubly symmetric section) and ori-
enting X, Y along the principal axes the equilibrium equations become
«o
GK'T\ $ '
Ell R
Ei;[ u'i + ^
R
R(W
K K
+ \ -{Pu'ay + {MM - -^- + -^- = Qx (15a)
- (MyQ)" -
R R R
1101
/?,
Wo
+ +
^^ 7T,+ rt R2 R2 R2
(15c)
£/;( «g + -
«0 .
£/:!<*>"--I + + Mxu'i - Myv'i
R
Wo
v0 +
M^ _K / ; _v t ^ /w _ (Mp,<D'y 7V»
+
* T ^ + ZJ - ^ * ' > ' " (r„)-
T?
SOLUTIONS
(16a)
R R2 R3
or
v'i + wo'
EL
R Pw0
-P\Vo + ^)+EIx[vi + ^) + = 0, (16ft)
Papangelis and Trahair (1987) have assumed for inextensional buckling w'
— v/R = 0 and simplified the equations to calculate the buckling load. Since
P - -qR = EA(w' - v/R) (see equation 28 of Papangelis and Trahair 1987)
1102
w' can not be assumed to be equal to v/R. Neglecting the effects of axial
displacements w0 on in-plane buckling, one gets
This result agrees with Vlasov (1961), Timoshenko (1961), and Chai Hang
Yoo (1984), but disagrees with Papangelis and Trahair (1987).
But for linear formulation, equation for inplane buckling may be written
as
M'
-P' + — = 0 (21)
R
where
1103
*K + J
Vo
Mx = -EIA v0' + (23b)
R2
Hence
M'x
-P' + — = 0 (24a)
R
Vo
w0 (24b)
R
WIT II \\ (I /I'lr"
nV \ {2n3TT3\/EIx
-P 3 + IR2
I R^n-n + l\EIx) I3 R*
nit\ EL.
+ = 0 (25a)
I W
2 I2
R2 + nWR'
qR = -P„ = EIX (25b)
1 - +
n2jr2R2 nVR'
For the circular ring of thin-walled open cross section used by Chai Hang
Yoo (1982) Pcr = 806.5 kips, which is 8% less than that obtained by Vlasov
or Eq. 20.
1104
where
M P^r2
(27a)
EL P^r'
Ky. K7 (21b)
R R
Ml2 Pf
K-n — + P$,r2 (27c)
GKT +
P* = {21 d)
nWEL
Py = (lie)
I2
1105
I2
1 1
IT2/?2 1V 2 /? 2 R
+ EIyP^r2 — - 1 = 0 (28)
ir2R2
Note that when the subtended angle 0 = 180; / = TTR the coefficients of
each term in the characteristic equation vanishes which proves that the semi-
circular beam will rotate about the diameter forming the ends. Hence the
values obtained by Vlasov and Thimoshenko in table 2(a) can not be viewed
correct. For other angles of 0
AT + bM + c = 0 • (29)
where
(Ely + Ptf2)
b = (30a)
R
I
c - EIyP«r2 y- - - (30&)
which is same as the obtained by Vlasov (1961) whereas the values obtained
by Papangelis and Trahair (1987) are close to Vlasov but not the same as
Vlasov.
For a quarter circular beam / = TT/?/2
(EIy + P^r2) (Ely + P^r2)2 3£W 2 l l / 2
M, 2 + 2 (32)
2R AR I
The positive moment and negative moment are shown in Fig. 7(a) and 1(b),
respectively. The beam considered by Chai Hang Yoo (1984) was analyzed
for out-of-plane buckling. The critical moments obtained by various re-
searchers are given in Table 2. Rajasekaran and Ramm (1984, 1985) ideal-
ized the curve beam into nine 9-node degenerated plate/shell elements for
the top and bottom flanges and twelve 16-node degenerated plate elements
for the web portion and solved using NISA 80 (1980). Note that the critical
moments (+/—) gives out of plane buckling mode whereas (—/+) gives the
local flange buckling mode which cannot be considered in the classical anal-
ysis. The results from NISA for the ( + / - ) agree with Vlasov with an error
of 5.64%, whereas the values obtained in this analysis are in close agreement
with Vlasov as shown in Table 2. From the finite element analysis Raja-
sekaran and Ramm (1984) have already proved that Vlasov's formulation
1106
for the member buckling of curved beams subjected to end moments is cor-
rect whereas the formulation of Chai Hang Yoo (1984) is in error. This is
confirmed by Papengelis and Trahair (1987).
1107
^ u K12
= {0} • (33)
K2\ K22
Downloaded from ascelibrary.org by UNB - Universidade de Brasilia on 03/09/23. Copyright ASCE. For personal use only; all rights reserved.
in which
/V2
Ku = P + P, + (34a)
P<s,rl
^12 — (34b)
^ n and ^T12 are the same as those obtained in Eq. 50 of Papengelis and
Trahair (1987) but differ in the K22 term of Eq. 34c. The characteristic equa-
tion is written as
P2 + bP + c 0 • (35)
where
I + r'
b = Px I + - TT4R2r: + P^ (36a)
~pT
c = PvP,i,[ I - (36fc)
When R -> 00 the critical load for the straight beam will be
* rr •* V) * cr -* <I . (37)
1108
substituted the curvature terms of the thin walled curved beam in the thin-
walled straight beam equilibrium equations whereas Chai Hang Yoo substi-
tuted the curvature terms of the thin-walled curved beam in the potential
expression of the thin-walled straight beam to get the equations for the curved
beam. Even Papengelis and Trahair's (1987) derivation are also not based
on first principles since nonlinear terms are obtained by considering the de-
formed shape to be in equilibrium. There is always a possibility of omission
of certain terms in such approach as indicated in this paper. In this paper it
has been demonstrated that the expression for equilibrium equations arise in
a completely natural way from the nonlinear terms of the strain-displacement
equations for virtual strains. This approach to the derivation and the asso-
ciated physical interpretation is precisely associated with the "geometric
stiffness" (or stability) matrices which arise in finite element continuum for-
mulations. Inclusion of these nonlinear terms eliminates the necessity of going
through rather complex and seemingly adhoc visualizations required to ac-
count for changes in geometry, when attempting to formulate beam equilib-
rium equations from free body diagrams.
In summary the formulation technique presented in this paper leads di-
rectly to the open thin-walled curved beam equations which are independent
of constitutive relation and for which the nature of approximation is clearly
evident. Based on the equations derived in this paper a new curved beam
finite element could be developed. Since there are always imperfections present,
a realistic analysis would be to carry out the nonlinear load-deflection re-
sponse.
ACKNOWLEDGMENTS
The writers thank the Management of P.S.G. Institutions and Dr. A. Shan-
mugasundaram, principal, P.S.G. College of Technology for giving facilities
to the writers to carry out the research work. This research work is a part
of the continued research work in the field of thin walled curved beams of
open cross section at P.S.G. College of Technology funded by the Aero-
nautical Research and Development Board.
APPENDIX I. REFERENCES
Chai Hang Yoo. (1982). "Flexural torsional stability of curved beams." J. Engrg.
Mech. Div., ASCE, 108(6), 1351-1369.
Hafner, L., et al. (1980). "Nonlinear structural analysis—NISA 80, program doc-
umentation." Institut fur Baustatik, Univ. of Stuttgart, W. Germany.
Murray, D. W., and Rajasekaran, S. (1975). "Technique for formulating beam equa-
tions." J. Engrg. Mech. Div., ASCE, 101(5), 561-573.
Papangelis, J. P., and Trahair, N. S. (1987). "Flexural torsional buckling of arches."
J. Struct. Engrg., ASCE, 113(4), 889-906.
Padmanabhan, S. (1984). "Analysis of thin-walled curved beams of open cross sec-
tion," thesis submitted to the Bharathiar University in partial fulfillment of the
requirements for the degree of Master of Engineering.
1109
148.
Rajasekaran, S., and Ramm, E. (1985). "Local and member stability of thin-walled
straight and curved beams of open section." Proc. 2d Int. Conf. Computer Aided
Analysis and Design in Civil Engineering, Roorkee, India.
Timoshenko, S. P. (1961). Theory of elastic stability. McGraw-Hill, New York,
N.Y.
Usami, T., and Koh, S. Y. (1980). "Large displacement theory of thin-walled curved
members and its application to lateral torsional buckling analysis of circular arches."
J. Solids and Struct., 16, 71-95.
Vlasov, V. Z. (1961). Thin-walled elastic beams. 2d ed., Israel Program for Sci-
entific Translation, Jerusalem.
A = cross-sectional area;
A' = modified cross-sectional area;
A'Y = modified first moment of area about x axis;
A'X = modified first moment of area about y axis;
d = depth of the cross section;
E = modulus of elasticity;
G = rigidity modulus;
Ii = modified sectorial moment of inertia;
II, I'y = modified moments of inertia about x and y axes, respec-
tively;
KT = St. Venant torsional constant of the section;
/ = arc length of the arch;
Mx, My = bending moments about x and y axes, respectively;
M* = stress resultant defined as in Eq. 9;
Mx, M'y = modified bending moments about x and y axes;
n = number of buckled half sine waves;
n = normal coordinate at any point on the contour;
O = arbitrary reference point;
P = axial force;
P„ = critical load;
P4, = torsional buckling load;
Py = Euler bucklin load;
qx = uniformly distributed load in x direction;
qy = uniformly distributed load in y direction;
q2 = uniformly distributed load in z direction;
q = uniformly distributed radical load;
R = initial radius of arch center line;
r = radial coordinate;
S*x S«,y = sectorial moments about x and y axes;
s — tangential coordinate along the contour;
t = thickness of the thin walled section;
Tx, Ty = stress resultants defined in Eq. 12;
1110
e = axial strain;
0 = angle subtended by arch;
Qx 0 V 0 2 = rotations of the section about x, y , z axes;
X. = critical load factor;
cr = normal stress;
T = shear stress;
<£> = twist of the cross section; and
ca = normalized warping coordinate.
1111