I Paper 24
I Paper 24
www.elsevier.com/locate/engstruct
Received 16 July 2001; received in revised form 9 January 2002; accepted 18 January 2002
Abstract
Lateral buckling of a laminated composite beam with I-section is studied. A general analytical model applicable to the lateral
buckling of an I-section composite beam subjected to various types of loadings is developed. This model is based on the classical
lamination theory, and accounts for the material coupling for arbitrary laminate stacking sequence configuration and various bound-
ary conditions. The effects of the location of applied loading on the buckling capacity are also included in the analysis. A displace-
ment-based one-dimensional finite element model is developed to predict critical loads and corresponding buckling modes for a
thin-walled composite beam with arbitrary boundary conditions. Numerical results are obtained for thin-walled composites under
central point load, uniformly-distributed load, and pure bending with angle-ply laminates. The effects of fiber orientation, location
of applied load, and types of loads on the critical buckling loads are parametrically studied. 2002 Elsevier Science Ltd. All
rights reserved.
Keywords: thin-walled structures; laminated composites; finite element method; lateral bucking
0141-0296/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 1 - 0 2 9 6 ( 0 2 ) 0 0 0 1 6 - 0
956 J. Lee et al. / Engineering Structures 24 (2002) 955–964
lytical solution for predicting the lateral bulking capacity along the contour line of the cross section. The (n, s, z)
of composite channel-section beams. More recently, Lee and (x, y, z) coordinate systems are related through an
[16] gave a closed-form expression for the location of angle of orientation ⍜ as defined in Fig. 1. The third
center of gravity and shear center as a function of lami- coordinate set is the contour coordinate s along the pro-
nation stacking sequence as well as sectional properties. file of the section with its origin at any point O on the
In the present study, a general analytical model appli- profile section. Point P is called the pole axis, through
cable to the lateral buckling of a I-section composite which the axis parallel to the z axis is called the pole
beam subjected to various types of loadings is axis.
developed. This model is based on the classical lami- To derive the analytical model for a thin-walled com-
nation theory [17], and accounts for the material coup- posite beam; the following assumptions are made:
ling for arbitrary laminate stacking sequence configur-
ation, i.e. unsymmetric as well as symmetric, and various 1. The contour of the thin wall does not deform in its
boundary conditions. The effects of the location of own plane.
applied loading on the buckling capacity are also 2. The shear strain of the middle surface is zero in
included in the analysis. A displacement-based one- each clement.
dimensional finite element, model is developed to predict 3. The Kirchhoff–Love assumption in classical plate
critical loads and corresponding buckling modes for a theory remains valid for laminated composite thin-
thin-walled composite beam with arbitrary boundary walled beams.
conditions. Governing buckling equations are derived 4. The time-dependent behavior is neglected.
from the principle of the stationary value of total poten-
tial energy. Numerical results are obtained for thin- According to assumption 1, the midsurface displacement
walled composites under central point load, uniformly- components ū, v̄ at a point A in the contour coordinate
distributed load, and pure bending with angle-ply lami- system can be expressed in terms of a displacements U,
nates. The effects of fiber orientation, location of applied V of the pole P in the x, y directions, respectively, and
load, and types of loads on the critical buckling loads the rotation angle ⌽ about the pole axis,
are parametrically studied. u(s,z) ⫽ U(z)sin⌰(s)⫺V(z)cos⌰(s)⫺⌽(z)q(s) (1a)
u(s,z) ⫽ U(z)cos⌰(s) ⫹ V(z)sin⌰(s)⫺⌽(z)r(s) (1b)
2. Kinematics These equations apply to the whole contour. The out-of-
plane shell displacement can now be found from the
The theoretical developments presented in this paper assumption 2 (also known as Vlasov assumption) as
require three sets of coordinate systems which are mutu-
ally interrelated. The first coordinate system is the w(s,z) ⫽ W(z)⫺U’(z)x(s)⫺V’(z)y(s)⫺⌽’(z)w(s) (2)
orthogonal Cartesian coordinate system (x, y, z), for where differentiation with respect to the axial coordinate
which the x and y axes lie in the plane of the cross sec- z is denoted by primes (‘); W represents the average axial
tion and the z axis parallel to the longitudinal axis of the displacement of the beam in the z direction; x and y are
beam. The second coordinate system is the local plate the coordinates of the contour in the (x, y, z) coordinate
coordinate (n, s, z) as shown in Fig. 1: wherein the n system; and w is the so-called sectorial coordinate or
axis is normal to the middle surface of a plate element, warping function given by
the s axis is tangent to the middle surface and is directed
冕
w(s)⫺ r(s)ds (3)
es ⫽ es ⫹ nks
ez ⫽ ez ⫹ nkZ
(5a)
(5b)
冕
dU ⫽ {sz[d苸zo ⫹ (x ⫹ nsin⌰)dy ⫹ (y
v
∂u 2
∂u
2
∂u
2
⫹ Mtdsz其dz
ks ⫽ ⫺ 2 ,kz ⫽ ⫺ 2 ,ksz ⫽ ⫺2 (6b)
∂s ∂z ∂s∂z Nz, Mx, My and Mw are axial force, bending moments in
All the other strains are identically zero. In eq. (6), ēs the x and y directions, and warping moment (bimoment)
and ¯ s are assumed to be zero, and ēz, ¯ z, and ¯ sz are with respect to the centroid, respectively, defined by
midsurface axial strain and biaxial curvatures of the integrating over the cross-sectional area A as:
冕
shell, respectively. The above shell strains can be con-
verted to beam strain components by substituting eqs. Nz ⫽ szdsdn (13a)
(1) and(2) into eq. (6)
A
ez ⫽ eoz ⫹ xKy ⫹ yKx ⫹ wKw
冕
(7a)
Kz ⫽ Kysin⌰⫺Kxcos⌰⫺Kwq (7b) My ⫽ sz(x ⫹ nsin⌰)dsdn (13b)
Ksz ⫽ Ksz (7c) A
x ⫽ ⫺V⬙
冕
(8b)
y ⫽ ⫺U⬙ (8c) Mt ⫽ szsndsdn (13e)
w ⫽ ⫺⌽⬙ (8d) A
sz ⫽ 2⌽’ (8e) In eq. (10), V is the potential of transverse load acting
on the cross section at a point a distance ā above the
The resulting strains can be obtained from eqs. (5) and shear center; and the variation of the potential of trans-
(7) as verse loads at shear center becomes [4]
ez ⫽ eoz ⫹ (x ⫹ nsin⌰)y ⫹ (y⫺ncos⌰)x ⫹ (w
冕
(9a) l
The total potential energy of the system can be stated, d⌸ ⫽ d(u ⫹ v) ⫽ 0 (15)
in its buckled shape, as Substituting eqs. (12) and (14) into eq. (15), the follow-
⌸⫽U⫹V (10) ing weak statement is obtained:
冕
l
where U is the strain energy
冕
0 ⫽ {NzdW’⫺MydU⬙⫺MxdV⬙⫺Mwd⌽⬙ (16)
1
U⫽ (s 苸 ⫹ szsgzs)dv, (11)
2 v z z 0
⫹ 2Mtd⌽’⫺Mb(⌽dU⬙ ⫹ U⬙d⌽)⫺ap⌽d⌽}dz
The variation of the strain energy is circulated by substi-
tuting eq. (7) into eq. (11) In eq. (16), Mb and p are the buckling moment and trans-
958 J. Lee et al. / Engineering Structures 24 (2002) 955–964
verse load, and can be written for various types of load- b3(3) 3
ing as E33 ⫽ [A11aya(2)⫺2B11aya ⫹ D11a]ba ⫹ A (21j)
12 11
Mb ⫽ lf(z) (17a) b3(3) 3
E34 ⫽ ⫺ B (21k)
p ⫽ lg(z) (17b) 12 11
where λ is a buckling parameter and f (z) and g(z) are E35 ⫽ Ba16yaba⫺Da16ba (21l)
polynomial functions which depend on the loading pat-
ba(3)
tern. E44 ⫽ [A11aya(2)⫺2B11aya ⫹ D11a] (21m)
12
b3(3) 3
⫹ D
4. Constitutive equations 12 11
冕
force Nz can now be expressed with respect to the gen-
eralized strains by combining eqs. (13a), (18) and (9) (Aij,Bij,Dij) ⫽ Qij(1,n,n2)dn (22)
冕
Nz ⫽ {Q11[ezo ⫹ (x ⫹ nsin⌰)y ⫹ (y
A
(19) In eq. (21), the superscript in the parenthesis () denotes
the power, and repeated index denotes summation. Index
k varies from 1 to 3 whereas a varies from 1 to 2
⫺ncos⌰)x ⫹ (w⫺nq)w] ⫹ Q16nsz其 implying 4 top and bottom flanges (1,2) and web (3) as
Similarly, the other stress resultants (My, Mx, Mw, Mt) shown in Fig. 2. bk denotes width of flanges and web.
can also be written in terms of the generalized strains.
Consequently, the constitutive equations for a thin-
walled laminated composite are obtained as
{NzMyMxMwMt其 ⫽ [E11E12E13E14E15E22E23E24E25 (20)
E33E34E35E44E45sym.E55]{e 其
z
o
y x w sz
5. Governing equations ity, (EIx)com and (EIY)com represent flexural rigidities with
respect to x and y axis, (EIw)com and (GJ)com represent
The lateral buckling equations of the present study can warping and torsional rigidities of the thin-walled com-
be derived by integrating the derivatives of the varied posite, respectively, written as
quantities by parts and collecting the coefficients of dU,
(EA)com ⫽ E11 (27a)
dV, dW and d⌽:
(EIy)com ⫽ E22 (27b)
N⬘z ⫽ 0 (23a)
(EIx)com ⫽ E33 (27c)
M⬙y ⫹ (Mb⌽)⬙ ⫽ 0 (23b)
(EIw)com ⫽ E44 (27d)
M⬙x ⫽ 0 (23c)
(GJ)com ⫽ 4E55 (27e)
For the above equations, it is well known that the criti-
Mw ⫹ 2M’l ⫹ MbU⬙ ⫹ ap⌽ ⫽ 0 (23d) cal buckling moment for pure bending is given by the
The natural boundary conditions are of the form: closed form solution for simply-supported boundary
conditions[1]:
冋 册
dW:Nz (24a)
冪
p2(EIy)com p2(EIw)com
dU:M1y ⫹ (Mb⌽)’ (24b) Mcr ⫽ ⫹ (GJ)com (28a)
2 2
dU’:My ⫹ Mb⌽ (24c)
’
dV:M x (24d)
6. Finite element model
dV’:Mx (24e)
The present theory for thin-walled composite beams
d⌽:M1w ⫹ 2Mt (24f) described in the previous section was implemented via
d⌽’:Mw (24g) a displacement based finite element method. The gen-
eralized displacements are expressed over each element
By substituting eqs. (20) and (8) into eq. (23), the as a linear combination of the one-dimensional Lagrange
explicit form of the governing equations yield: interpolation function ⌿j and Hermite-cubic interp-
E11W⬙⫺E12U⫺E13V ⫹ 2E15⌽⬙ ⫽ 0 (25a) olation function yj associated with node j and the
nodal values;
E12W⫺E22U ⫺E24⌽ ⫹ 2E25⌽ ⫹ (Mb⌽)⬙
iv iv
(25b)
冘
n
⫽0 W⫽ wj⌿j (29a)
E13W⫺E23U ⫺E33V ⫺E34⌽ ⫹ 2E35⌽ ⫽ 0
iv iv iv
(25c) j⫽1
冘
n
⫺E24Uiv⫺E34Viv⫺E44⌽iv ⫹ 2E15W⬙⫺2E25U (25d) U⫽ ujyj (29b)
⫺2E35V ⫹ 4E55⌽⬙ ⫹ MbU⬙ ⫹ ap⌽ ⫽ 0 j⫽1
冘
n
The above equations are the most general form for lateral V⫽ vjyj (29c)
buckling of a thin-walled laminated composite with an j⫽1
I-section, and the dependent variables, U, V, W and ⌽
冘
n
are fully-coupled.
⌽⫽ fjyj (29d)
If the stacking sequence of the web is symmetric: E12
j⫽1
= E34 = 0, and the thin-walled composite is symmetric
with respect to z axis, E13 = E15 = E24 = 0. Further, if Substituting these expressions into the weak statement
both the web and flange are balanced laminates (±q pairs in eq. (16), the finite element model of a. typical clement
of layers), Aki6 ⫹ Dki6 ⫽ 0, and thus, E25 = E35 = 0. can be expressed as the standard eigenvalue problem:
Finally, eq. (23) can be simplified to the uncoupled dif- ([K]⫺l[G]){⌬} ⫽ {0} (30)
ferential equations using eq. (17) as:
where [K] is the element stiffness matrix
(EA)comW⬙ ⫽ 0 (26a)
冤 冥
⫺(EIy)comUiv ⫹ l(f⌽)⬙ ⫽ 0 (26b) K11 K12K13K14
⫺(EIx)comViv ⫽ 0 (26c) K22K23K24
[K] ⫽ (31)
K33K34
⫺(EIw)com⌽ ⫹ (GJ)com ⫹ l(fU⬙ag⌽) ⫽ 0
iv
(26d)
sym. K44
From the above equations, (EA)com represents axial rigid-
960 J. Lee et al. / Engineering Structures 24 (2002) 955–964
and [G] is the element geometric stiffness matrix All others are zero.
In eq. (30), {⌬} is the eigenvector of nodal displace-
冤 冥
G11 G12G13G14 ments corresponding to an eigenvalue
G22G23G24 {⌬其 ⫽ {w u v f其T (35)
[G] ⫽ (32)
G33G34
sym. G44 7. Numerical results and discussion
The explicit form of [K] and [G] are given by For verification purpose, an isotropic simply-sup-
ported I-section beam with l = 8 m is considered (Fig.
冕
l
3). The material properties are assumed to be:
ij ⫽
K11 E11⌿’i ⌿’j dz (33a)
E ⫽ 1GPa,n ⫽ 0.34 (36)
0
The results by the present approach are compared to
冕
l
those of closed-form solutions in Table 1 for various
ij ⫽ ⫺ E12⌿i yj dz
K12 ’ ’’
(33b) types of loading conditions.
0
The functions f(z) and g(z) in eq. (26) are given as
follows for various types of loading:
冕
l
K ⫽ ⫺ E13⌿’iy’’j dz
13
ij (33c) 앫 pure bending
0 f(z) ⫽ 1; g(z) ⫽ 0 (37)
冕
l
앫 uniformly distributed load
ij ⫽
K14 2E15⌿’iyjdz (33d) 1 l2
f(z) ⫽ ( ⫺z2); g(z) ⫽ 1 (38)
0 24
앫 central point load
冕
l
1 l
f(z) ⫽ ( ⫺z); g(z) ⫽ 0 (39)
K ⫽ E22y⬙jy⬙jdz
22
ij (33e) 22
0
冕
l
ij ⫽ ⫺ (E24y⬙iy⬙j⫺2E25y⬙iy⬘j)dz
K24 (33f)
0
冕
l
ij ⫽
K33 E33y⬙iy⬙jdz (33g)
0
冕
l
ij ⫽
K34 (E34y⬙iy⬙j⫺2E35y⬙iy⬘j)dz (33h)
0
冕
l
K ⫽ (E44y⬙iy⬙j ⫹ 4E55y⬙iy⬙j)dz
44
ij (33i)
0
冕
l
G ⫽ fy⬙iyjdz
24
ij (34a)
0
冕
l
ij ⫽
G44 agyiyjdz (34b)
Fig. 3. Geometry of a thin-walled composite beam [10].
0
J. Lee et al. / Engineering Structures 24 (2002) 955–964 961
Table 1
Lateral buckling moments and loads of a simply-supported I-section beam for various loading cases
Load acts at Pure bending Uniformly distributed load Central point load
Table 2
Comparison of Lateral buckling moments and loads of a simply-supported I-section beam
E1/G12 Pure bending (kNm) Uniformly distributed load(kN) Central point load( kN / m)
Fig. 7. Lateral buckling load with respect to the fiber angle of a sim-
ply-supported I-section composite beam under uniformly distributed
load.
8. Concluding remarks
Acknowledgements
References
[10] Lin ZM, Polyzois D, Shah A. Stability of thin-walled pultruded [14] Lee J, Kim S. Flexural-torsional buckling of thin-walled I-section
structural members by the finite element method. Thin-walled composites. Computers and Structures 2001;79(10):987–95.
Structures 1996;24:1–18. [15] Kabir MZ, Sherbourne AN. Optimal fibre orientation in lateral
[11] Davalos JF, Qiao P. Analytical and experimental study of lateral stability of laminated channel section beams. Composites Part B
and distorsional buckling of FRP wide-flange beams. Journal of 1998;29B:81–7.
Composites for Construction, ASCE 1997;November:150–9. [16] Lee J. Center of gravity and shear center of thin-walled open-
[12] Brooks RJ, Turvey GJ. Lateral buckling of pultruded GRP I-sec- section composite beams. Composite Structures
tion cantilever. Composite Structures 1995;32:203–15. 2001;52(2):255–60.
[13] Turvey GJ. Effects of load position on the lateral buckling [17] Jones RM. Mechanics of composite materials. New York: Hemi-
response of pultruded GRP cantilevers-comparisons between sphere Publishing Corp, 1975.
theory and experiment. Composite Structures 1996;35:33–47.