Vous êtes sur la page 1sur 19

Composite Structures 67 (2005) 37–55

www.elsevier.com/locate/compstruct

Buckling analysis of tri-axial woven fabric composite structures.


Part I: Non-linear finite element formulation
Duosheng Xu, Rajamohan Ganesan, Suong V. Hoa *

Concordia Center for Composites, Department of Mechanical and Industrial Engineering,


Concordia University, Montreal, Quebec, Canada H3G 1M8
Available online 7 February 2004

Abstract
This paper deals with the stability analysis of curved woven tri-axial composite tow structures. A non-linear finite element
analysis is presented. Based on the continuum mechanics principles, the finite element updated Langrangian incremental formu-
lation for non-linear analysis is derived. Example problems are solved and comparison of the present results with the results
available in existing reference has been made. The effectiveness of the formulation and the validity of the corresponding computer
code are demonstrated. The buckling analysis of tri-axial composite straight beam and curved beam that is a part of woven fabric
tri-axial composite tow structures subjected to different boundary conditions is performed. In order to further confirm the accuracy
of the numerical solutions, approximate analytical solutions corresponding to the different structures have been derived. The
numerical results obtained are in very good agreement with the analytical solutions for the straight beam, curved beam and curved
beam structures. The effect of the resin on the curved woven tri-axial composite tow structures is also investigated.
 2004 Elsevier Ltd. All rights reserved.

Keywords: Non-linear analysis; Buckling; Tri-axial composites

1. Introduction For design purposes, it is of interest to understand the


mechanical behavior of the tri-axial composite structure.
Due to their attractive properties such as light weight, Works on the characterization of the mechanical prop-
high strength, high stiffness, low density, etc., composite erties of the materials such as the determination of
materials have been increasingly used in many engi- elastic modulus, deflection due to bending and thermal
neering fields such as aerospace, automobile, sports expansion coefficient have been carried out by col-
equipment, and marine structures. One of the interesting leagues of the authors of the present paper at the same
applications of composites due to their light weight institution [1–5]. Due to the complex nature of the
properties is the satellite dish made of tri-axial weaves. structure, it is of great interest to understand its stability
In this application, a single layer of tri-axial fabric is when subjected to compressive loads. This is the objec-
used to make up the thickness of the dish of the satellite. tive of the present study, namely the buckling behavior
The tri-axial composite material is made by weaving of the tri-axial composite structure subjected to in-plane
three tows along three directions making an angle of 60 compressive loading.
or 120 with each other. A portion of this tri-axial The tri-axial structure consists of three yarns inter-
structure is shown in Fig. 1. The cross-section of a tow is lacing one over the other along the three directions. In
about 0.2 mm · 0.84 mm. It can also be seen that holes order to study this structure subjected to in-plane
occupy more than 50% of the surface area of the loading condition, it is appropriate to examine the
structure. These holes reduce the mass of the material behavior of a representative unit cell of the structure.
and at the same time allow the escape of impacting air, The understanding of the behavior of this unit cell may
which in turn reduces the load to be supported. then be extended to understand the behavior of the
larger structure. A representative unit of the tri-axial
structure can be the one shown in Fig. 2. It consists of
*
Corresponding author. three tows interlaced over each other along the corre-
E-mail address: mieng@vax2.concordia.ca (S.V. Hoa). sponding directions. The curvature of the tow is
0263-8223/$ - see front matter  2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruct.2004.01.004
38 D. Xu et al. / Composite Structures 67 (2005) 37–55

x3
x2
P P
o
x1

Fig. 3. Simply supported curved tri-axial beam.

Fig. 1. Tri-axial woven fabric composite structure. x B3


B
x3
P′
xC 3 x B1 x2 60° A PA
PB A x1
Resin
C B
x3 P2 B
x2 P1′ PB
60° A PA L x B1
A
60 ° x1
P3′ Fig. 4. Tri-axial tow structure with two intersected curved tows.

resin
x B3 and this connection is positioned at mid-lengths of the
C beams for the sake of symmetry. The initial configura-
B
PC tions of the beams follow half of a sine curve. This sit-
xC1 uation is not representative of the unit cell in Fig. 4, but
L it serves to give some insight into the interaction be-
tween two beams. The lateral load P0 is still used to
Fig. 2. Woven tri-axial tow structure with three intersected curved initiate the instability. The values of the axial loads for
tows. the two beams will be kept to be the same, that are PA
and PB .
The analysis of the tow structure consisting of three
maintained. The resin layer at the crossover of the tows curved beams as shown in Fig. 2 is conducted next. The
can be represented by bars. The ends of the tows are values of the axial loads, PA , PB and PC are kept to be
represented by pins at one end of the tow and by pin- equal and they are increased by the same increment each
on-rollers at the other end of the tow. The tows are time. The values of the three lateral loads P01 , P02 and P03
constrained at the ends such that the reaction forces at are also kept to be equal throughout.
the constraints are perpendicular to the orientation of The results of the more complicated investigations
the tows at these ends. Loads are applied along the will be presented in a series of articles. This paper pre-
directions of the center lines of the tows. sents the analytical formulation for the single curved
In order to arrive at the analysis of this structure, it is beam element and also the numerical analysis for a
necessary to analyze first simpler structures to build up single curved beam, and two intersected and three
the experience and the knowledge base. Also it would be intersected beam structures.
easier to check the results at different steps. The basic
unit in the unit cell is a curved beam with resin con-
nection with other beams at locations situated at 1/4 and
3/4 of its length. To model this beam, the analysis of a 2. Stability analysis of a single curved composite beam
curved beam as shown in Fig. 3 will be first conducted.
Lateral load P0 is applied to simulate the connection From the dimensions of a tow in a unit cell (0.2
with other beams, and also to provide the initial mm · 0.84 mm cross-section with length of about 9.16
imperfections for the instability analysis. The critical mm), the effective length of the basic unit cell of the
values of the axial load P will be determined. structure is much larger than its cross-sectional dimen-
Subsequently the analysis of two curved beams con- sions. Since the ratio of the dimension of the cross-sec-
nected with each other by a bar that represents the tion of a tow to its length is small, we can consider the
adhesive layer is conducted. This structure is shown in individual tow to behave like a beam when performing
Fig. 4. There is only one connection between two beams buckling analysis. These tows will be modeled as curved
D. Xu et al. / Composite Structures 67 (2005) 37–55 39

beams with six degrees-of-freedom. The displacements x2 t +∆ t


and rotations of the structure could be very large and P( x1 , t +∆ t x 2 , t +∆t x3 )
the kinematic relations between displacements and
t
rotations of the structure and loads applied will no P( x1 ,t x2 ,t x3 )
Configuration at time t+∆t
longer be linear. A non-linear analysis will be used.
Non-linear analyses can be classified into three types: P(
0
x1 , 0 x 2 , 0 x3 )
materially-non-linear analysis, geometrically-non-linear Configuration at time t
analysis and combination of both. As a start, we will
only perform the geometrically-non-linear buckling
analysis. Configuration at time 0
For the curved beam model, the cross-section shape x1
t+ t t +∆ t t +∆ t
of the tow is assumed to be rectangular. This is slightly t
xi = xi + u i ,
0 t
xi = xi +
0
ui , ui = ui t
u i , i = 1,2,3
different from the actual cross-section of the tow which x3
is elliptic in reality. However, the real cross-section is Fig. 5. Motion of body in Cartesian co-ordinate system.
fairly flat and for buckling analysis, the approximation
of rectangular cross-section may be acceptable. This
simplifies the analysis greatly. The torsional rigidity of strains. The aim is to evaluate the equilibrium positions
the tow is not taken into consideration. This is not taken of the body at the discrete time points 0; Dt; 2 Dt; 3 Dt; . . .,
into account to imply negligible effects of stress com- where Dt is an increment in time. To develop the solu-
ponents corresponding to deformation of the cross-sec- tion strategy, assume that the solutions measured in the
tion in its own plane. In modeling, we also assume that co-ordinate system corresponding to all time steps from
the plane sections originally normal to the centerline time 0 to time t, inclusive, have been obtained. Then the
axis remain plane and undistorted under deformation solution process for the next required equilibrium posi-
but not necessarily normal to this axis. This kinematic tion corresponding to time t þ Dt is typical and is ap-
assumption does not allow for warping effects in torsion. plied repetitively until the complete solution path has
Buckling analysis of curved beam structures has been been solved for. Hence, in the analysis we follow all the
conducted by many researchers during the past decades. particles of the body in their motion, from original to
Many methods proposed have shown their effectiveness the final configuration of the body.
to perform the analysis. For example, Bathe and Bo- Consider the equilibrium of the body in Fig. 5 again
lourchi [6] presented the curved beam element model for at time t þ Dt. The principle of virtual displacements
large displacement analysis. Hu et al. [7] used constant requires that
and linear strain curved beam elements to perform the Z
tþDt tþDt tþDt
buckling analysis of arched beams. Most of them were t Sij d t eij d V ¼ d tþDt W ð1Þ
tþDt V
based on incremental continuum Lagrangian formula-
tion. Bathe and Bolourchi [6] has shown that the up- where t tþDt Sij are Cartesian components of second
dated Lagrangian formulation and total Lagrangian Piola–Kirchoff stress tensor and t tþDt eij are Cartesian
formulation are consistent in formulating the large components of the Green–Lagrange strain tensor cor-
rotation non-linear three-dimensional beam elements, responding to the deformatiom from the configuration
but the updated Lagrangian-based element is compu- at time t to the configuration at time t þ Dt and referring
tationally more effective. We will use this method in this to the configuration at time t. tþDt W is the total external
paper to carry out the buckling analysis of tri-axial virtual work. The left-hand side of Eq. (1) includes the
woven fabric composite structures. The theoretical contributions from both resin and tows of the struc-
development that follows the formulation in Bathe [8] is tures. Finite element models for both of them will be
presented in the following. presented in the following two sections, respectively. Eq.
(1) cannot be solved directly since the configuration at
time t þ Dt is unknown.
3. Theoretical development for a general curved beam for
buckling analysis 3.2. Formulation for tow elements

3.1. Principle of virtual displacements 3.2.1. Description of element geometry


Consider a general three-dimensional curved beam of
The basis of the displacement-based finite element rectangular cross-section as shown in Fig. 6. Co-ordi-
solution is the principle of virtual displacements. Con- nates x1 ; x2 ; x3 form the global co-ordinate system,
sider the motion of a general body in a fixed Cartesian g1 ; g2 ; g3 represent the local co-ordinate system attached
co-ordinate system as shown in Fig. 5, and assume that to the beam (so-called body attached co-ordinate sys-
the body can experience large displacements and large tem), and r; s; q is the natural co-ordinate system. The
40 D. Xu et al. / Composite Structures 67 (2005) 37–55

b2 point P ðr; s; qÞ within the element for a m-noded element


at time t are interpolated in terms of the nodal transla-
tions t uki , i ¼ 1; 2; 3 along the global co-ordinate axes
a2 and the rotations t hki , i ¼ 1; 2; 3 about global co-ordinate
2
r ,η1 axes using
4 Xm
qX m
t
ui ¼ t x i  0 x i ¼ hk t uki þ ak hk ðt Vqik  0 Vqik Þ
s,ηη2 2
b1 k¼1 k¼1
q,η 3
3 sX m
Node þ bk hk ðt Vsik  0 Vsik Þ ð3Þ
a1 1 2 k¼1
θ3 t
Vs1
ui ¼ tþDt ui  t ui
x3 ,u 3 X m
qX m
sX m
t
Vq1 ¼ hk uki þ ak hk Vqik þ bk hk Vsik ð4Þ
k¼1
2 k¼1
2 k¼1

x 2 ,u 2 θ2 where
x1 ,u1 Vqik ¼ tþDt Vqik  t Vqik ð5Þ
θ1
Vsik ¼ tþDt
Vsik  t
Vsik ð6Þ
Fig. 6. Three-dimensional curved beam element. are the increments of unit directional vectors at node k
and can be obtained by using the following second-order
section dimensions can be specified by ak and bk and a approximations
set of vectors t Vkq , t Vks and t Vkr at a node k at time t
1
(where the left superscript t could also be 0 or t þ Dt, Vkq ¼ hk  t Vkq þ hk  ðhk  t Vkq Þ ð7Þ
referring to time 0 or time t þ Dt, respectively. The same 2
notation will be employed in the following sections). The 1
Vks ¼ hk  t Vks þ hk  ðhk  t Vks Þ ð8Þ
directions of t Vkq ; t Vks can be conveniently selected to be 2
the q; s directions at time t. In this case the Cartesian co- By substituting Eqs. (5)–(8) into Eq. (4), incremental
ordinates (x1 ; x2 ; x3 ) of a point P ðr; s; qÞ within the ele- displacements can be expressed as

ui ¼ tþDt ui  t ui
X m
qX m
sX m
¼ hk uki þ ak hk ðhkiþ1 t Vqðiþ2Þ
k
 hkiþ2 t Vqðiþ1Þ
k
Þþ bk hk ðhkiþ1 t Vsðiþ2Þ
k
 hkiþ2 t Vsðiþ1Þ
k
Þ
k¼1
2 k¼1
2 k¼1

qX m
 ak hk ðððhkiþ1 Þ2 þ ðhkiþ2 Þ2 Þ t Vqik  hki hkiþ1 t Vqðiþ1Þ
k
 hki hkiþ2 t Vqðiþ2Þ
k
Þ
4 k¼1

sX m
 bk hk ðððhkiþ1 Þ2 þ ðhkiþ2 Þ2 Þ t Vsik  hki hkiþ1 t Vsðiþ1Þ
k
 hki hkiþ2 t Vsðiþ2Þ
k
Þ
4 k¼1

¼ uLi þ uNi ð9Þ

ment for a m-noded element at time t can be written as where i ¼ 1; 2; 3, and i þ 1 takes the value of 1 when i ¼
X 3 and i  1 takes the value of 3 when i ¼ 1; uLi , uNi are
m
qX m
sX m
t
xi ¼ hk t xki þ ak hk t Vqik þ bk hk t Vsik linear terms and non-linear terms specified by single
k¼1
2 k¼1 2 k¼1 underlining and double underlining in Eq. (9), respec-
i ¼ 1; 2; 3 ð2Þ tively.

3.2.3. Strain–displacement relationship


3.2.2. Displacement approximation of element
Green–Lagrangian strain tensor in the configuration
The displacement field follows from the assumptions
at time t referred to the initial configuration is defined as
that the cross-section normals remain straight during
deformation. In the iso-parametric element solution, the 1
t
0e ¼ ð0t XT 0t X  IÞ ð10Þ
displacements approximation and their increment at a 2
D. Xu et al. / Composite Structures 67 (2005) 37–55 41

where deformation gradient at time t is 3.3. Formulation for resin element


2t t t
3
0 x1;1 0 x1;2 0 x1;3 3.3.1. Basic assumptions
t
0 X ¼
4 t t
0 x2;1 0 x2;2
t
0 x2;3
5 ð11Þ
t t t
The resin is placed between the two interlaced tows
0 x3;1 0 x3;2 0 x3;3 and is used to bond the tows together. The bonded tows
in which comma subscript denotes the differentiation constitute an integral body. The thickness of the resin is
with respect to 0 xi . Considering the strain increments about one-third of the thickness of the tow and has the
tþDt
eij , the following relations hold value of 0.067 mm. The horizontal cross-section of the
t
resin is a parallelogram as shown in Fig. 7. Each side is
tþDt along one of the interlaced tows, but is intersected with
t eij ¼ t eij ð12Þ
another tow at an angle of 60. Its dimensions of the
t eij ¼ t eij þ t gij ð13Þ both sides are 0.87 · 0.87 mm. For the strains of the
resin, only out-of-plane normal strain and in-plane shear
in which the linear parts of strain increments are strain are assumed to be non-zero. Details are as fol-
1 lows.
t eij ¼ ðt ui;j þ t uj;i Þ ð14Þ First, the strains of resin, egR2 gR2 and egR3 gR3 , in local
2
resin co-ordinate system shown in Fig. 10 can be ne-
the non-linear parts of strain increments are glected. Actually, no strains of the resin at the interfaces
between resin and tow 1 along y1 direction and between
1
t gij ¼ t uk;i t uk;j ð15Þ resin and tow 2 along y2 direction, corresponding to ey1 y1
2 and ey2 y2 in local beam co-ordinate systems O1 x1 y1 z1 and
where the comma subscript denotes the differentiation O2 x2 y2 z2 in Fig. 7, will be allowed since the strains of the
with respect to t xi . interlaced tows in these two directions are neglected due
to their higher order nature as stated in the previous
section. The top tow tends to stretch the resin if it de-
3.2.4. Stress–strain relationship
forms downward; while the bottom tow tries to com-
Stresses and strains employed in updated Lagrangian
press the resin, and vice versa. Because the thickness of
formulation are second Piola–Kirchoff stresses, Cauchy
the resin is very small compared with the horizontal
stresses and Green–Lagrangian strains.
dimensions of the resin, deformation due to compres-
The relation between the second Piola–Kirchoff stress
sion of bottom part of the resin may cancel a part of the
tensor in the configuration at time t and measured in the
deformation due to tension of the top part of the resin
configuration at time 0 and Green–Lagrangian strain
shown in Fig. 8. Therefore, the strains of the resin in
tensor at time t referred to the initial configuration can
be expressed as
t
0 Sij ¼ t0 Cijrs t0 ers ð16Þ
Tow 2
where 0 t Cijrs is the constitutive tensor.
The relation between the second Piola–Kirchoff Resin 60 ˚
y1
stresses and Cauchy stresses is as follows:
t o1 Tow 1
t q
smn ¼ 0 t0 xm;i t0 Sij t0 xn;j ð17Þ z1
x1
q
where t q and 0 q are mass densities of the element in the x2
configuration at time t and time 0, respectively. The
relation between them is as follows: y2
o2
t
z2
q 1
¼ ð18Þ
0q detð0t XÞ Fig. 7. Tows and resin.

The incremental stress decomposition is


tþDt
t Sij ¼ t sij þ t Sij ð19Þ Tow 1
Tension
where t sij is Cartesian components of the Cauchy stress Resin
Compression
tensor and t Sij is Cartesian components of Second Piola– Tow 2
Kirchoff stress increment tensor referred to the config-
uration at time t. Fig. 8. Deformation of cross-section of tows and resin.
42 D. Xu et al. / Composite Structures 67 (2005) 37–55

horizontal plane, corresponding to egR2 gR2 and egR3 gR3 in Consider a general three-dimensional two-node bar of
local resin co-ordinate system shown in Fig. 10, should rectangular cross-section as shown in Fig. 10. Co-ordi-
be of small order of magnitude and will be neglected in nates gR1 ; gR2 ; gR3 represent the mapped local co-ordi-
the present case. nate system attached to the bar. The cross-sectional
Second, the shear strains, egR1 gR2 and egR1 gR3 , are neg- dimensions can be specified by aRk and bRk , which are the
ligible. This is because the shear deformations of the dimensions of the parallelogram, and a set of vectors
t k
tows in their local co-ordinate planes O1 y1 z1 and O2 y2 z2 VRq , t VkRs and t VkRr at a node k at time t. The Cartesian
corresponding to tow 1 and tow 2, respectively, are ne- co-ordinates (x1 ; x2 ; x3 ) of a point P ðr; s; qÞ within the
glected. These planes approximately are corresponding element for a m-noded element at time t can be written as
to the two local vertical co-ordinate planes, Og gR1 gR2 Xm
qX m
sX m
t
and Og gR1 gR3 , of the resin. Therefore, the shear strains, xi ¼ hk t xki þ aRk hk t VRqi
k
þ bRk hk t VRsi
k
i ¼ 1;2;3
2 k¼1 2 k¼1
egR1 gR2 and egR1 gR3 , may be of small order of magnitude k¼1

and can be neglected. ð21Þ


Third, uniform shear strain, egR2 gR3 , (or no warping where the nodal co-ordinates t k
are also the co-ordi-
xi
due to torsion) will be assumed due to small thickness of nates of the nodes corresponding to the related tows.
the resin. The equation used to express the relation be- This is because the above mapping did not change the
tween the shear stress and shear strain is SgR2 gR3 ¼ scaling in xR1 direction and the co-ordinates of each
GR egR2 gR3 . center of the cross-section.
Finally, the linear stress–strain relation of the resin, In the iso-parametric element solution, the displace-
SgR1 gR1 ¼ ER egR1 gR1 , in local co-ordinate system of the ments approximation and their increment at a point
resin is assumed. The displacements of the two inter- P ðr; s; qÞ within the element for a m-noded element at
secting points between the center lines of the two time t are interpolated in terms of the nodal translations
interlaced tows and the resin may be different. It will t k
ui , i ¼ 1; 2; 3 along the global co-ordinate axes and the
cause deformation, either stretching or compressing the rotations t hki , i ¼ 1; 2; 3 about global co-ordinate axes
resin along its thickness. Due to small thickness and using
large cross-sectional dimensions of the resin, corre-
X m
qX m
spondingly, the linear stress–strain relation of the resin t
ui ¼ t x i  0 x i ¼ hk t uki þ aRk hk ðt VRmi
k
 0 VRmi
k
Þ
will be employed in the present analysis. k¼1
2 k¼1
sX m
3.3.2. Element geometry and displacement description þ bRk hk ðt VRsi
k
 0 VRsi
k
Þ ð22Þ
2 k¼1
In order to employ the geometric interpolation
functions used in the case of a rectangular cross-sec-
tional beam, the following mapping, which maps the
parallelogram on the plane of xR2 xR3 in the local co- r ,η R1
ordinate system xR1 xR2 xR3 into a rectangular on the
plane of gR2 gR3 in co-ordinate system gR1 gR2 gR3 (body 2
attached, so-called mapped local co-ordinate system) as oη
shown in Fig. 9 (xR1 and gR1 are not shown in Fig. 9), is
used. After the strains are found in global co-ordinate θ3 q, η R 3
system, they will be transformed into the mapped local x3 , u3 Node 1
s,η R 2 t 1 VRq
t 1

system and local system consecutively at Gauss inte- bR1


a R1 VRs
gration points.
8 9 2 38 9 o
< gR1 = 1 0 0 < xR1 = x2 , u 2 θ
2
g ¼ 4 0 1  p1ffiffi3 5 xR2 or gR ¼ TM xR x1 , u1
: R2 ; : ; θ1
gR3 0 0 1 xR3
ð20Þ Fig. 10. Two-node resin element.

ηR 3
xR 3 x R 3 = bR / 2 (– aR / 2, bR / 2) (aR / 2, bR / 2)
x R 3 = 3 ( x R 2 + a R / 2) x R 3 = 3 ( x R 2 + a R / 2)

oR o
600 xR 2 ηR2
x R3 = – bR / 2 (– aR / 2,– bR / 2) (aR / 2,– bR / 2)

Fig. 9. Mapping from oR xR1 xR2 xR3 to og gR1 gR2 gR3 .


D. Xu et al. / Composite Structures 67 (2005) 37–55 43

ui ¼ tþDt ui  t ui Since only the strain along the thickness in xR1 direction
and shear strain in xR2 xR3 plane are considered, the
X m
qX m
sX m
strain vector in local co-ordinate system can be written
¼ hk uki þ k
aRk hk VRqi þ k
bRk hk VRsi ð23Þ
k¼1
2 k¼1
2 k¼1
as
 
where ^ t
eR11

e
t R ¼ ð33Þ
k 2 t eR23
VRqi ¼ tþDt VRqi
k
 t VRqi
k
ð24Þ
k where the over-bar of the quantities corresponds to their
VRsi ¼ tþDt VRsi
k
 t VRsi
k
ð25Þ
local values.
are the increments of unit directional vectors at node k Transformation matrix of strains from the global co-
and can be obtained by using the following first-order ordinate system into the mapped local co-ordinate sys-
approximations tem is given by

2 3
l211 l212 l213 l11 l12 l12 l13 l11 l13
6 l2 l222 l223 l21 l22 l22 l23 l21 l23 7
6 21 7
6 l2 l232 l233 l31 l32 l32 l33 l31 l33 7
Te ¼ 6 31
6 2l11 l12
7 ð34Þ
6 2l12 l22 2l13 l23 l12 l21 þ l11 l22 l13 l22 þ l12 l23 l13 l21 þ l11 l23 7
7
4 2l21 l31 2l22 l32 2l23 l33 l22 l31 þ l21 l32 l23 l32 þ l22 l33 l23 l31 þ l21 l33 5
2l11 l31 2l12 l32 2l13 l33 l12 l31 þ l11 l32 l13 l32 þ l12 l33 l13 l31 þ l11 l33

where lij are the direction cosines of the global co-


VkRq ¼ hk  t VkRq ð26Þ ordinate axes in mapped local co-ordinate system at
Gauss integration points.
VkRs ¼ hk  t VkRs ð27Þ
Transformation matrix of strains from the mapped
By substituting Eqs. (24)–(27) into Eq. (23), incremental local co-ordinate system into the local co-ordinate sys-
displacements can be expressed as tem is given by
 
ui ¼ tþDt ui  t ui 1 0 0 0 0 0
TL ¼ 0  p2ffiffi 0 0 1 0 ð35Þ
X m
qX m
3
¼ hk uki þ aRk hk ðhkiþ1 t VRqðiþ2Þ
k
 hkiþ2 t VRqðiþ1Þ
k
Þ
k¼1
2 k¼1 The incremental Piola–Kirchoff stress–strain relation
sX m in local co-ordinate system at time t is as follows:
þ bRk hk ðhkiþ1 t VRsðiþ2Þ
k
 hkiþ2 t VRsðiþ1Þ
k
Þ ð28Þ     
2 k¼1 t SR11 ER 0 t eR11
¼ ð36Þ
t SR23 0 GR 2 t eR23
where i ¼ 1; 2; 3 and i þ 1 takes the value of 1 when i ¼ 3
and i  1 takes the value of 3 when i ¼ 1. where the ER and GR are Young’s modulus and shear
modulus of elasticity of the resin.
3.3.3. Strain–displacement and stress–strain relationships
The incremental decomposition of Piola–Kirchoff 3.4. Equilibrium equation
stress of the resin is
tþDt Substituting the stress–strain relation of both tow and
t SRij ¼ t sRij þ t SRij ð29Þ
resin into Eq. (1), and considering that all variables in
where t sRij is Cauchy stress of the resin at time t. Eq. (1) are referred to the configuration at time t, i.e. the
Considering that only the linear strain terms are in- updated configuration of the body, one obtains the
cluded, the Green–Lagrangian strain of the resin can be equilibrium equation of body in global co-ordinate
expressed as system at time t þ Dt as follows
tþDt
Z Z Z
t eRij ¼ t eRij ¼ t eRij ð30Þ t t t t
S d e
t ij t ij d V T þ s ij dg ij d V T þ t SRij d t eRij d VR
tV tV tV
Stress–strain relation in incremental form is given by T
Z T
Z R

tþDt t t t
t SRij ¼ t CRijrs t eRrs ð31Þ ¼d W  sij d t eij d VT  sRij d t eRij d t VR
tV tV
T R
The incremental strain–displacement relation in the
ð37Þ
tensor form in global co-ordinate system is given by
where the subscript T of V denotes the volume of the
1 tow and R of V the volume of the resin. Substituting the
t eij ¼ ðt ui;j þ t uj;i Þ ð32Þ
2 stress–strain and strain–displacement relations and the
44 D. Xu et al. / Composite Structures 67 (2005) 37–55

displacement interpolation into Eq. (37), one can obtain where ds0 and ds are the undeformed and deformed
the formulation of iso-parametric finite elements. This lengths of elements on the reference line.
procedure has been presented in Bathe [8]. In order to Since
confirm the numerical results of Eq. (37), approximate 
ðdsÞ2 ¼ ðdx1 þ du1 Þ2 þ ðdu3 Þ2
analytical solutions will be given first. ð40Þ
ðds0 Þ2 ¼ ðdx1 Þ2 þ ðdu30 Þ2
and are approximated to the second order by using
Taylor series expansion, one obtains
4. Approximate solutions for curved beam structures 
8  2  2 
>
> du1 1 du1 1 du3
< ds ¼ dx1 1 þ dx1 þ 2 dx1 þ 2 dx1
4.1. Simply-supported full sinusoidal beam   2  ð41Þ
>
> 1 1 1 du30
: ds0 ¼ dx1 1  2 dx1
Suppose that the curved beam takes the form of a full
sinusoidal beam that is simply-supported at both ends
Substituting Eq. (41) into Eq. (39) and considering that
under loads shown in Fig. 11. 1 2 3 2
ðdu Þ ðdu Þ , the reference plane extensional strain, e0 ,
Assume that the plane sections remain plane after dx1 dx1

bending; the effect of transverse shear is negligible; the can be further written in the following form:
loads and the bending moments act in a plane passing  2  2
du1 1 du3 1 du30
through a principal axis of inertia of the cross-section; e0 ¼ þ 
dx1 2 dx1 2 dx1
the initial rise of the beam is not large as compared to  2  2 !
the cross-sectional dimensions; deflections are small as 1 du30 du1 1 du3
 þ ð42Þ
compared to the cross-sectional dimensions; the mate- 2 dx1 dx1 2 dx1
rial points on the undeformed midline (mid-plane) are
characterized by u30 ðx1 Þ shown in Fig. 11. Let u1 ðx1 Þ and The general expression for the change in the curvature
u3 ðx1 Þ denote the location of material points on the de- of a curved beam, j, is given by
0 1
formed midline. On the basis of these assumptions, the 2 2
d u3 d u30
strain at any material point is given by B C
B dx21 dx21 C
j ¼ B   3    3 C ð43Þ
e ¼ e0 þ x03 j ð38Þ @  2 2  2 2 A
1 þ du
dx1
3
1 þ dudx1
30

where e0 and j denote the reference plane (mid-plane)


extensional strain and change in curvature, respectively. For small initial curvature, the expression for the change
Let D0 and D denote the undeformed and deformed in curvature, j, after Taylor series expansion of Eq. (43)
positions of a material point of the reference line. The and accurate to the second order of the slope of the
co-ordinates of D0 and D are ðx1 ; u30 Þ and ðx1 þ u1 ; u3 Þ, deflection, can be written as
respectively. For small strains, the reference plane  2 
d u3 d2 u30
extensional strain, e0 , is given by j¼ 
dx21 dx21
 2  2 !
ds  ds0 3 d2 u3 du3 d2 u30 du30
e0 ¼ ð39Þ þ  ð44Þ
ds0 2 dx21 dx1 dx21 dx1

x3
q

x1
D Qi
Deformed configuration
u3 u3-u30
x3′
u30 D0 A Mj ML
Mo u1

P x1 x2′
( x1 , u30)
A
x1i A-A Section
Undeformed configuration
x1 j
L

Fig. 11. Sinusoidal beam.


D. Xu et al. / Composite Structures 67 (2005) 37–55 45

and    2
d2 u3 d2 u30 d2 u3 du3
 2   3 
d2 u3 d2 u30 d2 u3 d2 u30 dx21 dx21 dx21 dx1
j2 ¼   3 
dx21 dx21 dx21 dx21 !!1
   2 !  2  2 !
d2 u3 du3
2
d2 u30 du30 d2 u30 du30 A 1þ1 du3
  ð45Þ  dx1
dx21 dx1 dx21 dx1 dx21 dx1 2 dx1

Strain energy of the beam is given by Z L X


Z Z  qðu3  u30 Þ þ Qi dðx1  x1i Þðu3  u30 Þ
1 1
U¼ re dV ¼ Ee2 dV ð46Þ 0 i
2 V 2 V
X  !
where the stress–strain relation, r ¼ Ee, is employed. du3 du30
þ Mj gðx1  x1j Þ  dx1
Assume that the u1 , u3 are functions of x1 , then j
dx1 dx1
Z
"L 
x03 dA ¼ 0 ð47Þ du3 du30 ""
A þM   Pu1 jL0 ð52Þ
dx1 dx1 "0
Substitution of Eqs. (38) and (47) into Eq. (46) leads to
Z

1 S0 This is the general form of the total potential of a


U¼ ðEAe20 þ EIj2 Þ ds
ð48Þ
2 0 beam. In the present case, if only the compressive axial
load is considered, the total potential takes the following
where I is the moment of inertia of the cross-section, S0

form
is the length of the reference line of the beam and the
finite element length ds
of the beam can be approxi- 0
Z L  2  2
mated as being accurate to the second order 1 @ EA du1 þ 1 du3 1 du30
UT ¼ 
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  2 ! 2 0 dx1 2 dx1 2 dx1
2 2 1 du3
ds ¼ ðdx1 Þ þ ðdu3 Þ dx1 1 þ ð49Þ
2 dx1  2  2 !!2
1 du30 du1 1 du3
 þ
Substituting from Eq. (49), Eq. (48) can be written as 2 dx1 dx1 2 dx1
Z  2 !
1 L 1 du 3  2  
U¼ ðEAe20 þ EIj2 Þ 1 þ dx1 ð50Þ d2 u3 d2 u30
d2 u3 d2 u30
2 0 2 dx1 þ EI 3  
dx21 dx21
dx21 dx21
Work done by external forces is given by 1
 2  2 !!
Z L X d2 u3 du3 d2 u30 du30
  A
W ¼ qðu3  u30 Þ þ Qi dðx1  x1i Þðu3  u30 Þ dx21 dx1 dx21 dx1
0 i

X  !  2 !
du3 du30 1 du3
þ Mj gðx1  x1j Þ  dx1  1þ dx1  Pu1 jL0 ð53Þ
j
dx1 dx1 2 dx1
 "L
du3 du30 ""
M  þ Pu1 jL0 ð51Þ Boundary conditions for a simply-supported beam are
dx1 dx1 "0
where dðx1  x1i Þ is the Dirac d-function and gðx1  x1j Þ u3 ð0Þ ¼ u3 ðLÞ ¼ 0; u003 ð0Þ ¼ u003 ðLÞ ¼ 0;
is the Doublet function. u1 ð0Þ ¼ 0 ðroller supportÞ ð54Þ
The total potential of the beam is given by
UT ¼ U  W Assume that the initial configuration of the beam is
Substitution of Eqs. (42), (45) and (51) yields  
2p
0 u30 ¼ Z sin x1 0 6 x1 6 L ð55Þ
Z  2  2 L
1 L@ du1 1 du3 1 du30
UT ¼ EA þ 
2 0 dx1 2 dx1 2 dx1 where the Z is the initial rise parameter.
 2  2 !!2 The deflection of the beam may be represented by an
1 du30 du1 1 du3 infinite sine series, each term of which satisfies the
 þ
2 dx1 dx1 2 dx1 boundary conditions (54),
 2 2  np 
d u3 d2 u30 X
1
þ EI  u3 ¼ u30 þ an sin x1 0 6 x1 6 L ð56Þ
dx21 dx21 L
n¼1
46 D. Xu et al. / Composite Structures 67 (2005) 37–55

where an , n ¼ 1; 2; . . . are the undetermined coefficients 1  p 2 X 1


1  p 4 X1
of the deflection of the beam. EA b an n2 þ EAL a21 an n2
2 L 8 L
The axial displacement of the material points may be n¼2
!
n¼2

represented by, which satisfies the boundary conditions X1


1  p 4 X
1
5  p 6
3 4
(54), þ an n þ EIL an n4  EIL
n¼2
2 L n¼2 8 L
  p  !
X
1 X
1 X
1
u1 ¼ b 1  cos x1 0 6 x1 6 L ð57Þ  a21 2 2
an n ð1 þ n Þ þ a3n n6 þ 4Z 2
an n 4
2L
n¼2 n¼2 n¼2

where b is the undetermined coefficient of axial dis- ¼ 0 m ¼ 2; 3; 4; . . . ð61Þ


placement of the beam.
Substituting Eqs. (55) through (57) into the exp-  2 !
EA 61 2 Z
ression for the total potential, Eq. (53), and perform- Pþ b 1 p
L 63 L
ing the corresponding integration and neglecting !
the higher order terms of undetermined coefficients 1  p 2 2 X 1

yields þ EA a1 þ a2n n2 ¼0 ð62Þ


4 L n¼2
 2 !
EA 2 61 2 Z There are two possible cases that result from Eqs. (60)–
UT ¼ Pb þ b 1 p
2L 63 L (62).
! Case I: a1 6¼ 0 and an ¼ 0 for n ¼ 2; 3; 4; . . ..
1  p 2 2
X1
2 2 Case II: a1 6¼ 0, am 6¼ 0 and an ¼ 0 for n ¼ 2; 3; 4; . . .
þ EA b a1 þ an n
4 L n¼2
except n ¼ m.
! Case I: Substituting an ¼ 0 for n ¼ 2; 3; 4; . . . into
1  p 4 X 1 X1
þ EAL a41 þ 2a21 a2n n2 þ a4n n4 Eqs. (60) and (62) and rearranging them, one obtains
32 L n¼2 n¼2
!    2 !
1  p 4 X1 P 5  p 2 2 2 61 2 Z
þ EIL 2
a1 þ 2 4
an n ¼ 1 ða1 þ 4Z Þ 1 p
4 L p2 EI=L2 4 L 63 L
n¼2
 
61 EA p 2 2 2
5  p 6 X1  Z a1 ð63Þ
 EIL a41 þ 2a21 a2n n2 ð1 þ n2 Þ 252 EI L
32 L n¼2
!! The load reaches maximum when a1 approaches zero.
X
1 X1
þ a4n n6 þ 8Z 2 a21 þ a2n n4 ð58Þ The buckling load for this case is
n¼2 n¼2  2  4
Pcr1 376 2 Z 305 4 Z
We are interested in finding buckling load at which ¼1 p þ p ð64Þ
p2 EI=L2 63 L 63 L
instability is possible. We find it by first writing the
equilibrium equations and then studying the character Case II: Eliminating b and am from Eqs. (60) through
of these static equilibrium equations. To find the static (62) and rearranging them, one obtains
equilibrium positions, we use the principle of the sta-
tionary value of the total potential, or  2 !
P 61 2 Z 5
¼ 1 p m4  p2
( oUT p EI=L2
2 63 L 4
¼ 0; n ¼ 1; 2; 3; . . .
oan
ð59Þ  a 2  2 !!
oUT
¼0 2 4 1 Z
ob  ð1 þ m þ m Þ þ4 m4
L L
This leads to  a 2
61 EAZ 2 2 1
!  p ð1 þ m2 Þ
252 EI L
1  p 2 1  p 4 X1
EA ba1 þ EAL a31 þ a1 a2n n2  2 !
2 L 8 L n¼2 Z 2 61 EAZ 2
þ4 ðm  1Þ  ð1  m2 Þ
 p 4  p 6 L 315 EI
1 5
þ EIL a1  EIL
2 L 8 L ð65Þ
!
X
1
 a31 þ a1 a2n n2 ð1 þ n2 Þ þ 4a1 Z 2 ¼0 ð60Þ The maximum load corresponds to a1 approaching zero.
n¼2 This yields
D. Xu et al. / Composite Structures 67 (2005) 37–55 47

 2 !  2 ! The total potential of the resin is given by, assum-


P Z 61 Z
¼ m4 1  5p2 1  p2 ing that the resin acts like a stretching spring and its
p2 EI=L2 L 63 L
shear effect is neglected due to its small order of mag-
 2 !
61 EAZ 2 2 1 2 Z
nitude,
þ ðm  1Þ p ð66Þ
63 EI 5 L     2
1 aR bR LA LB
UR ¼ E R DuA3  DuB3 ð72Þ
The lowest value of P with respect to values of integer m 2 tR 2 2
corresponds to m ¼ 2 and where
 2 !  2 ! ( $ % $ % $ %
Pcrm 2 Z 61 2 Z
¼ 16 1  5p 1 p DuA3 L2A ¼ uA3 L2A  uA30 L2A
p2 EI=L2 L 63 L $ % $ % $ % ð73Þ
  ! DuB3 L2A ¼ uB3 L2B  uB30 L2B
2
61 EAZ 2 1 Z
þ  p2 ð67Þ are the displacements of the beams AA and BB at their
21 EI 5 L
midpoints.
Comparing the Case I with Case II, the critical load W is the work done by the external forces, which is
should be the smaller one of them, which is given by Eq. given in the present case by
(64).
From Eq. (64), two observations can be seen: first, the W ¼ PA uA1 ðLA Þ  PB uB1 ðLB Þ ð74Þ
change of buckling load for a curved beam with respect
to the Euler buckling load for straight beam is propor- The formulae for strain and change of curvature of the
tional to the square of the initial rise of the beam. Sec- reference line are given by Eqs. (42) and (44). Substi-
ond, when the initial rise, Z, of the beam is very small, tuting Eqs. (42), (44), (70) through (74) into Eq. (69),
i.e., Z 1, one obtains from Eq. (64) by neglecting the one can obtain the formula for the total potential in the
last two terms that the buckling load is the same as the same form as Eq. (53).
Euler’s buckling load for a straight beam. This means Boundary conditions:
that the buckling load for a curved beam is the same as Beam AA : DuA3 ð0Þ ¼ DuA3 ðLA Þ ¼ 0;
the one for a straight beam. Actually, a beam with small
initial rise corresponds to a shallow curved beam. In this Du00A3 ð0Þ ¼ Du00A3 ðLA Þ ¼ 0; uA1 ð0Þ ¼ 0
case, initial deflection may be considered as a kind of ð75Þ
small imperfection of the beam. As we have known that
Beam BB : DuB3 ð0Þ ¼ DuB3 ðLB Þ ¼ 0;
the small imperfection due to small deflection does not
change Euler buckling load. Du00B3 ð0Þ ¼ Du00B3 ðLB Þ ¼ 0; uB1 ð0Þ ¼ 0
Substituting Z ¼ 0:133 mm and L ¼ 4:58 mm into Eq.
ð76Þ
(64) in the present case, one obtains
Pcr Initial shape:
¼ 0:9507 ð68Þ  
p2 EI=L2 p
Beam AA : uA30 ¼ ZA sin xA1 0 6 xA1 6 LA
It is seen that there is 4.93% decrease for the buckling LA
load with respect to the Euler buckling load. ð77Þ
 
p
4.2. Two intersected half sinusoidal beam structure Beam AA : uB30 ¼ ZB sin xB1 0 6 xB1 6 LB
LB
ð78Þ
As shown in Fig. 4, the two intersected beam structure
is constituted by two half-sinusoidal beams and resin, Deflections:
which bonds the beams together at their midpoints.
Beam AA : uA3
The total potential is given by  
X
1
nA p
UT ¼ UAT þ UBT þ UR ð69Þ ¼ uA30 þ aAn sin xA1 0 6 xA1 6 LA
n¼1
LA
where the total potentials of beams AA and BB are gi-
ven, respectively, by ð79Þ
Z LA Beam BB : uB3
1
UAT ¼ ðEA AA eA 02 þ EA IA j2A Þ dx1 þ PA u1 jL0 A ð70Þ X
1  
2 0 nB p
Z ¼ uB30 þ aBn sin xB1 0 6 xB1 6 LB
1 LB
n¼1
LB
UBT ¼ ðEB AB e2B0 þ EB IB j2B Þ dxB1 þ PB uB1 jL0 B ð71Þ
2 0 ð80Þ
48 D. Xu et al. / Composite Structures 67 (2005) 37–55

Axial displacement: Pcr


¼ 1:0105; 1:0019 and 1:0106
p2 EI=L2
Beam AA : uA1
  
p It is seen that Eq. (85) is a very good approximation.
¼ bA 1  cos xA1 0 6 xA1 6 LA From Eq. (83) or (85) one observes that the buckling
2LA
load of two intersected half-sinusoidal beam structure
ð81Þ
with flat cross-section not only depends on the factor of
initial rise parameter, (Z=L2 ) but also depends on the
Beam BB : uB1
   ratio of the tension stiffness (EA) to the bending stiffness
p (EI) of the beam.
¼ bB 1  cos xB1 0 6 xB1 6 LB
2LB
ð82Þ 4.3. Three intersected full sinusoidal beam structure

Substituting all the assumed initial shapes, deflections As shown in Fig. 2, the total potential of unit cell is
and axial displacements into total potential expression, given by
using the same material properties and geometries for
UT ¼ UA þ UB þ UC þ UR  W ð86Þ
the two beams but different initial rise parameters, i.e.
ZA and ZB , setting ZB ¼ ZA ¼ Z and performing the where UA ; UB ; UC are strain energies of beams AA, BB
same procedure of operation as in Section 4.1, one and CC, respectively. UA is given by
obtains the buckling loads as follows: Z
1 LA
UA ¼ ðEA AA e2A0 þ EA IA j2A Þ dx1 ð87Þ
Pcr 2 0
$ p %2
EI L UB and UC can be obtained simply by changing the
2 subscript A in Eq. (87) into B and C.
 2  2  2
13 2 Z p2 Z 4 169 EA UR is the strain energy of the resin, which is given by
¼1 p þ Z4
60 L 8 L 1800 EI     2
1 aR bR L L
 2 !2 U R ¼ ER DuA3  DuB3
13 2 Z 2 tR 4 4
þ 11 1  p     2
60 L 1 aR bR 3L 3L
3 þ ER DuC3  DuA3
 2 ! ," 2 tR 4 4
143 EA 2 13 2 Z 5 13 EA 2     2
þ Z 1 p Z 1 aR bR 3L L
60 EI 60 L 60 EI þ ER DuB3  DuC3 ð88Þ
 2 !# 2 tR 4 4
13 2 Z
þ5 1 p ð83Þ W is the work done by the external forces, which is given
60 L in the present case by
For accuracy to the second order of the initial rise, W ¼ PA uA1 ðLA Þ  PB uB1 ðLB Þ  PC uC1 ðLC Þ ð89Þ
buckling load in Eq. (83) can be approximated by
 2 Boundary conditions:
Pcr 7 2 Z
$ %2 1 þ p ð84Þ Beam AA : DuA3 ð0Þ ¼ DuA3 ðLA Þ ¼ 0;
EI p 120 L
L
Du00A3 ð0Þ ¼ Du00A3 ðLA Þ ¼ 0; uA1 ð0Þ ¼ 0
If the beam is very flat and the initial rise has the ð90Þ
same order as the height of the cross-section of the
beam, the factor, EAZ 2 =EI, will not be a small quantity. Beam BB : DuB3 ð0Þ ¼ DuB3 ðLB Þ ¼ 0;
In this case the buckling load can be approximated by Du00B3 ð0Þ ¼ Du00B3 ðLB Þ ¼ 0; uB1 ð0Þ ¼ 0
 2 ( "  2 ð91Þ
Pcr pZ 13 1 169 EAZ 2
$ %2 ¼ 1 þ  þ Beam CC : DuC3 ð0Þ ¼ DuC3 ðLC Þ ¼ 0;
EI pL 2L 15 2 1800 EI
#, ) Du00C3 ð0Þ ¼ Du00C3 ðLC Þ ¼ 0; uC1 ð0Þ ¼ 0
143 EAZ 2 13 EAZ 2
þ 11 þ þ5 ð85Þ ð92Þ
60 EI 60 EI
Initial shape:
Substituting Z ¼ 0:133 mm and L ¼ 2:29 mm into Eqs.  
2p
(83)–(85) in the present case, one obtains the unitless Beam AA : uA30 ¼ ZA sin xA1 0 6 xA1 6 LA
LA
buckling loads in these three cases, respectively, as fol-
lows ð93Þ
D. Xu et al. / Composite Structures 67 (2005) 37–55 49

 
2p 5. Buckling analysis of tri-axial woven fabric composite
Beam BB : uB30 ¼ ZB sin xB1 0 6 xB1 6 LB structures
LB
ð94Þ
Using the above formulation, the buckling behavior
  of a few configurations of the curved composite beam
2p
Beam CC : uC30 ¼ ZC sin xC1 0 6 xC1 6 LC will be presented in the following. The material prop-
LC erties of the beam and resin used in the following cases
ð95Þ are given in Table 1:
Deflections: The geometric parameters of the tow are as
X
1   follows:
nA p
Beam AA: uA3 ¼ uA30 þ aAn sin xA1 06xA1 6LA Height of cross-section of tow, a ¼ 0:2 mm
n¼1
LA
Width of cross-section of tow, b ¼ 0:84 mm
ð96Þ The geometric parameters of the resin are
  Width · Length · Thickness ¼ 0.87 · 0.87 · 0.067
X
1
nB p
Beam BB: uB3 ¼ uB30 þ aBn sin xB1 06 xB1 6 LB mm
n¼1
LB Transverse loads used in the following case,
ð97Þ P 0 ¼ 0:005 N.

X
1  
nC p 5.1. Buckling loads of single isotropic arch beam with
Beam CC: uC3 ¼ uC30 þ aCn sin xC1 06xC1 6LC
LC clamped ends
n¼1

ð98Þ
The single isotropic circular arch beam is shown in
Axial displacement: Fig. 12. It is clamped at both ends with a single load at
   the apex. The material of the arch is assumed to be
p isotropic and linearly elastic. The arch is idealized using
Beam AA: uA1 ¼ bA 1  cos xA1 0 6 xA1 6 LA
2LA eight equal curved beam elements. The maximum
ð99Þ deflection w is measured from the apex of the configu-
   ration before loading.
p
Beam BB: uB1 ¼ bB 1  cos xB1 06 xB1 6 LB Geometry and material properties
2LB
of the beam are as follows:
ð100Þ
The radius of the arch, R ¼ 3381:1
   mm (133.114 in.)
p
Beam CC: uC1 ¼ bC 1  cos xC1 0 6 xC1 6 LC Height of the cross-section, a ¼ 4:8
2LC
mm (3/16 in.)
ð101Þ Width of the cross-section,
Substituting all the assumed initial shapes, deflections b ¼ 25:4 mm (1.0 in.)
and axial displacements into total potential expression, Length of the arch, L ¼ 863:6 mm
using the same material properties and geometries for (34.0 in.)
the three beams but different initial parameters, i.e. ZA , Young’s modulus of elasticity,
ZB and ZC , setting ZC ¼ ZB ¼ ZA ¼ Z and performing E ¼ 68:95 GPa (107 lb/in2 .)
the same procedure of operation as in Section 1, one Poison’s ratio, m ¼ 0:2.
obtains the buckling load
Comparison of the results with the one given in Ref.
 2  4
Pcr 376 2 Z 305 4 Z [7] is shown in Fig. 13. It can be seen that very good
¼1 p þ p ð102Þ agreement is obtained, proving the accuracy of the for-
p2 EI=L2 63 L 63 L
mulation and also of the computer program developed
Substituting Z ¼ 0:133 mm and L ¼ 4:58 mm into Eq. for the analysis.
(102) in the present case, one obtains
Pcr 5.2. Simply-supported straight beam made of tri-axial
¼ 0:9507 composite
p2 EI=L2
It is seen that nearly 5% of decrease for the critical load Simply-supported straight composite beam with a
with respect to the Euler critical load for single straight pair of static compressive loads at both ends and a small
beam case is predicted. lateral downward static load at the center is shown in
50 D. Xu et al. / Composite Structures 67 (2005) 37–55

Table 1
Material properties of tri-axial composite tow and resina
Material EL (GPa) ET (GPa) GLT (GPa) GTT (GPa) mLT
Composite tow 500.0 40.0 24.0 14.3 0.26
Resin 3.5 3.5 1.3 1.3 0.35
a
Subscript L denotes longitudinal; subscript T denotes transverse.

P
Node 7 10 13 16
4 19
1 4 5 22
3 6 24
2 7
1 8
Element w
R

Fig. 12. Arch beam with clamped ends.

Fig. 15. Load versus deflection curve for simply-supported straight


tow.

The load versus maximum deflection curve of the


beam in non-linear analysis is shown in Fig. 15. In this
figure and in similar figures that follow, the word pi
denotes p.
Solving the eigenvalue problem, one obtains that the
non-linear finite element solution for the unitless buck-
ling load in this case is Pcr =ðp2 EI=L2 Þ ¼ 0:9648. This
result is very close to the result obtained using well
Fig. 13. Load versus maximum deflection curve for the arch.
known Euler’s beam-column formula, that is,
Pcr =ðp2 EI=L2 Þ ¼ 1. There is about 3.5% difference that is
predicted. The reason for this is that the shear effect is
included in the present case while the shear effect is not
Node
1 P′ 10 13 considered in Euler case. The shear effect will ‘‘soften’’
4 7 P
the beam and decrease the buckling load. Therefore, a
1 2 3 4 very good consistency is obtained.
Element w
L
5.3. Straight cantilever beam made of tri-axial composite
Fig. 14. Simply-supported straight tri-axial composite beam.
Straight cantilever composite beam with a single
static compressive load at the free end and a small lat-
Fig. 14. The beam is composed of a single composite eral downward static load at the same end is shown in
tow which is assumed to be orthotropic and linearly Fig. 16. It was analyzed using four elements. The length
elastic. The same assumption holds in the following of the beam is 4.58 mm. The maximum deflection w is
sections. The length of the beam is 4.58 mm. It is ana- measured from the free end corresponding to the un-
lyzed using four elements. The maximum deflection w loaded configuration. The load versus maximum
was measured from the center corresponding to the deflection curve of the beam in non-linear analysis is
unloaded configuration. shown in Fig. 17.
D. Xu et al. / Composite Structures 67 (2005) 37–55 51

Node curved beam is on ox1 x3 plane and the equation of its


1 4 7 10 13 P ′ center line can be expressed as:
1 P 
2 3 4 x3 ¼ Z sinð2px=LÞ
Element ð103Þ
L w x2 ¼ 0
where the maximum value of the co-ordinate in z
Fig. 16. Straight cantilever composite beam made of tri-ax. direction is Z ¼ 0:133 mm. This value is obtained from
observation of the microphotograph of a section of the
tri-axial tow as shown in Fig. 18.
Load versus maximum deflection curve of simply-
supported single curved composite beam is shown in
Fig. 19.
Solving the eigenvalue problem, one obtains that the
non-linear finite element solution for the unitless buck-
ling load of a simply-supported individual curved tow is
Pcr =ðp2 EI=L2 Þ ¼ 0:9185. It is 4.8% less than the finite
element result in straight beam case and 3.4% less than
analytical result. It is also worthy to note that the dif-
ference (4.8%) of the non-linear finite element solution
for buckling loads between straight beam and curved
beam almost has the same amount of difference (4.9%)
from analytical solution for the two beams. It shows
that the change of the buckling load due to curving the

Fig. 17. Load versus deflection curve for clamped straight tow.

Solving the eigenvalue problem, one obtains that the


non-linear finite element solution for the unitless buck-
ling load in this case is Pcr =ðp2 EI=L2 Þ ¼ 0:2429. This
result is very close to the result obtained using Euler’s
beam-column theory given by Pcr =ðp2 EI=L2 Þ ¼ 0:25.
There is about 2.84% difference between them. The
reason for this is also the shear effect. The difference has
the same order as the simply-supported straight beam. It
can be seen that excellent agreement is also obtained. Fig. 18. Configuration of the fiber tow and resin.
It is worthy to mention that the ratio of the critical
loads corresponding to the simply-supported beam and
cantilever beam obtained using Euler’s theory is 4. In
the present non-linear finite element analysis the ratio is
3.972, which is very close to 4. The agreement is there-
fore confirmed. This further proves the accuracy of the
formulation and the computer program developed for
the analysis.

5.4. Simply-supported curved beam made of tri-axial


composite

Simply-supported curved tri-axial composite beam


with a pair of static compressive loads at both the ends
and a small lateral downward static load at the middle
of the beam is shown in Fig. 3. It was analyzed using
four elements. The span of the beam is 4.58 mm. The
maximum deflection w is measured from its center line Fig. 19. Load versus deflection curve for simply-supported curved
corresponding to the unloaded configuration. The tow.
52 D. Xu et al. / Composite Structures 67 (2005) 37–55

Fig. 20. Deformed and undeformed shapes of simply-supported Fig. 22. Deformed and undeformed shapes of cantilever curved beam.
curved composite tow.

beam in finite element solution obeys the same law given larger deformation rate of the curved beam compared
by Eq. (64) if the buckling load is measured based on the with the cantilever straight beam case in Section 5.3 at a
straight beam buckling load. Next, from Figs. 15 and 19 certain amount of axial compressive load. We did not
we can see that the slope of the load versus maximum draw the curve together with the curves of other cases
deflection curve of a curved beam before buckling is less due to this reason. This further confirms the observa-
than the slope of the load versus maximum deflection tions made in Section 5.4.
curve of a straight beam before buckling. It means that Solving the eigenvalue problem, one obtains that the
curving the beam will increase the deforming rate of the non-linear finite element solution for the unitless buck-
beam before buckling and as a result, as we have ex- ling load of a cantilever single curved tow is
pected, the deformation of a curved beam will be larger Pcr =ðp2 EI=L2 Þ ¼ 0:2201. The ratio of the buckling load
than the deformation of a straight beam at the same for simply-supported curved beam to the one in the
amount of axial compressive load. Therefore, good present case is 4.17. It is very close to 4. It shows that the
agreement is thus confirmed. Euler buckling load relation between a simply-sup-
Deformed and undeformed shapes of simply-sup- ported beam and a cantilever beam also holds for the
ported single curved composite tow are shown in Fig. 20 present case. Deformed and undeformed shapes of
when load is equal to 120 N. cantilever single curved composite tow are shown in Fig.
22 when load is equal to 27 N. It may be noted here that
5.5. Cantilever curved tri-axial composite beam for the curved beam the original configuration is ori-
ented at an angle with respect to x1 -axis at both the fixed
Using the same beam configuration, the same mate- and free ends.
rial properties and the same finite element mesh, using
the clamped boundary condition at the left end of the
5.6. Simply-supported two intersected curved tri-axial
beam and free end condition at the other end and per-
composite tow structure
forming the non-linear analysis, one obtains the load
versus maximum deflection curve as shown in Fig. 21.
Simply supported curved tri-axial composite tow
We see from Fig. 21 a much larger deflection and much
structure with two pairs of static compressive loads
along the connection lines of corresponding supported
ends and a small lateral downward static load at the
apex is shown in Fig. 4. The structure is made of two
intersected curved composite tows at angle of 60, which
are bonded together by resin at the midpoints of each
tow. It is analyzed using four elements. The maximum
deflection w is measured from its central line corre-
sponding to its unloaded configuration. The dotted lines
in the figure constitute the x1 x2 plane of co-ordinate
system. The length of the beam is 2.29 mm. Geometries
of the structure are as follows:
The beam AA is in x1 x3 plane and the equation of its
central line is given by
 $ %
x3 ¼ Z sin pL x1
ð104Þ
x2 ¼ 0

Fig. 21. Load versus maximum deflection curve for cantilever curved The beam BB is in xB1 xB3 plane, where xB3 is parallel
tow. to x3 -axis and xB1 is in ox1 x2 plane and is intersected with
D. Xu et al. / Composite Structures 67 (2005) 37–55 53

ox1 -axis at angle of 60, as shown in Fig. 4. The equation shear effect into consideration while the analytical
of central line of beam BB is as follows: approximate solution does not.
p  In order to know the effect of the resin on the buck-
xB3 ¼ Z sin xB1 ð105Þ ling behavior of simply-supported two oppositely curved
L
tow structure, 10 times of variation, both increasing and
Load versus maximum deflection curve is plotted in Fig. decreasing, of Young’s modulus of resin is made.
23. The unitless buckling load obtained from the non- Buckling loads at different Young’s modulus values are
linear finite element solution in this case is Pcr =ðp2 EI= plotted in Fig. 24. From the figure one observes that
L2 Þ ¼ 0:9862. It is 6.86% larger than the single curved buckling load increases as the Young’s modulus be-
beam case and 2.22% larger than the single straight comes smaller and this is because smaller Young’s
beam case. It shows that the resin used for bonding two modulus leads to smaller stiffness of the resin as a bar,
oppositely curved tows may have the effect to resist correspondingly, the resin is deformed more easily. This
further deformation of the structure. The result obtained leads to further increase in curvature of each beam when
from the approximation solution given in Section 4.2 is loads are smaller. Note that each beam is curved along
1.01. It is 1% larger than the Euler buckling load in different directions. In order to make the two beams to
straight beam case. This further confirms the effect of buckle together, great effort needs to be exerted in order
the resin on resisting the deformation of the structure. to reverse the curvature of one beam. Therefore, larger
Comparing the two results, the non-linear finite element load P is needed to buckle the structure if the two beams
solution is seen to be 2.36% less than the approximate are to buckle on the same side, either up or down. For
solution. It is a very good agreement. The reason for the the case of stiff resin, the two tows are held together
difference between the two solutions is the shear effect of more rigidly. As such they tend to deform together at
the beams. Non-linear finite element solution takes the the initial low loads. That makes buckling easier. This

Fig. 23. Load versus maximum deflection curve of simply-supported Fig. 24. Buckling load for two-tow structure versus Young’s modulus
two intersected tow structure. of the resin.

PB PB PB
Deformed shapes
Deformed shapes Deformed shapes
P P P
PA PA PA

Undeformed configurations
Undeformed configurations Undeformed configurations

1.Deformed configuration 2.Deformed configuration 3. Buckled configuration


when axial loads are small when axial loads are large

Fig. 25. Deformation patterns of two-tow structure.


54 D. Xu et al. / Composite Structures 67 (2005) 37–55

procedure can be described by Fig. 25. The structure in


Fig. 25 is buckled downward due to a downward small
transverse load P0 .

5.7. Simply-supported three intersected curved tri-axial


composite structure

Simply supported curved tri-axial composite tow


structure with three pairs of static compressive loads at
the six supported ends and three small lateral downward
static loads at the three apexes is shown in Fig. 2. The
structure is made of three composite tows woven at
angles of 0, 60 and )60 with each other and bonded
together by resin. It is modeled using 12 elements with
four elements for each tow. The maximum deflection w
is measured from its central line corresponding to its
Fig. 27. Buckling load for three-tow structure versus Young’s modu-
unloaded configuration. The dotted lines in the figure
lus of the resin.
constitute the x1 x2 plane of co-ordinate system. Geom-
etries of the structure are as follows:  
The beam AA is in x1 x3 plane and the equation of its 2p
xB3 ¼ Z sin xB1 ð109Þ
central line is given by L
 
2p The load and maximum deflection curve of the structure
x3 ¼ Z sin x1 ð106Þ is shown in Fig. 26. The buckling load from non-linear
L
finite element solution in this case is Pcr =ðp2 EI=L2 Þ ¼
x2 ¼ 0 ð107Þ
0:9564. The result obtained from approximate solution
The beam CC is in xC1 xC3 plane, where xC3 is parallel to in Section 4.3 is 0.9507. Comparing the two solutions,
x3 -axis and xC1 is in ox1 x2 plane and is intersected with the difference of the buckling load for a unit cell between
ox1 -axis at angle of )60. The equation of central line of them is 0.6%. It is an excellent agreement. The effect of
beam CC is given by the resin on three intersected tow structure is shown in
  Fig. 27.
2p
xC3 ¼ Z sin xC1 ð108Þ
L
The beam BB is in xB1 xB3 plane, where xB3 is parallel to 6. Conclusion
x3 -axis and xB1 is in ox1 x2 plane and is intersected with
ox1 -axis at angle of 60. The equation of central line of A non-linear formulation for the buckling analysis of
beam BB is given by tri-axial composite curved beam structures has been
developed. Corresponding approximate analytical solu-
tions to the structures are also presented. The accuracy
of the numerical solutions has been confirmed by the
approximate analytical solutions for corresponding
structures. Therefore, the non-linear formulation can be
used for analyzing the buckling behavior of the com-
plicated woven fabric tri-axial composite tow structures,
which will be presented in a later paper.

7. List of symbols

Notation
The following convention for tensor and vector sub-
scripts and superscripts is employed:
A left superscript denotes the time of the configura-
tion in which the quantity occurs.
A left subscript denotes the time of the configuration,
Fig. 26. Load versus maximum deflection curve of simply-supported in which the co-ordinate is measured with respect to
three intersected tow structure. which is differentiated, if the quantity considered is a
D. Xu et al. / Composite Structures 67 (2005) 37–55 55

t k tþDt k
derivative; otherwise the left subscript denotes the time xi ; xi Cartesian co-ordinate of nodal point k in con-
of the configuration in which the quantity is measured. figuration at time t and t þ Dt
0 t
Right lower case subscripts denote the components of t xi;j ; 0 xi;j derivative of co-ordinate in configuration at
a tensor or vector. Differentiation is denoted by a sub- time 0 and t with respect to co-ordinates 0 xj and
t
script following a comma, with the subscript indicating xj
t
the co-ordinate with respect to which is differentiated. 0 X deformation gradient of the configuration at
Right upper case subscript R corresponds to the re- time t referred to configuration at time 0
sin. d denoting variation
t
0 e ij component of Green–Lagrange strain tensor in
ak ; bk height and width of cross-sectional dimensions configuration at time t, referred to configura-
of the beam at nodal point k tion at time 0
t tþDt
0 C ijrs component of constitutive tensor at time t re- t eij component of Green–Lagrange strain tensor in
ferred to configuration at time 0 and t configuration at time t þ Dt, referred to config-
t Cijrs component of tangent constitutive tensor at uration at time t
time t referred to configuration at time t t eij component of strain increment Green–Lag-
t eij component of linear part of strain increment of range tensor referred to configuration at time t
configuration at time t t gij component of non-linear part of strain incre-
EL ; ET longitudinal and transverse Young’s moduli ment of configuration at time t.
0
GLT ; GTT longitudinal–transverse and transverse–trans- q; t q specific mass of body in configuration at time 0
verse shear moduli and t
t
hi finite element interpolation function associated sij component of Cauchy stress tensor in configu-
with nodal point k ration at time t matrices
P; P0 axial load and transverse load t
s Cauchy stress matrix and vector of in configu-
r; s; q natural element co-ordinates ration at time t
t
0 Sij component of second Piola–Kirchoff stress
tensor in configuration at time t referred to
configuration at time 0
tþDt References
t Sij component of second Piola–Kirchoff stress
tensor in configuration at time t þ Dt referred to
configuration at time t [1] Zhao Q, Hoa SV. Tri-axial woven fabric (TWF) composite with
open holes (Part I): Finite element models for composites.
t Sij component of second Piola–Kirchoff stress
J Compos Mater 2003;37(9):763–91.
increment at time t [2] Zhao Q, Hoa SV, Ouellette P. Tri-axial woven fabric (TWF)
t; t þ Dt time composite with open holes (Part II): Verification of the finite
t
ui ; tþDt ui component of displacement vector from initial element model. J Compos Mater 2003;37(10):849–75.
position at time 0 to configuration at time t and [3] Zhao Q, Hoa SV. Thermal deformation behavior of triaxial woven
fabric (TWF) composites with open holes. J Compos Mater
t þ Dt
2003;37(18):1629–49.
ui increment in displacement component, ui ¼ [4] Zhao Q, Hoa SV, Moudrik R. Finite element modeling of a
tDt
ui  t ui membrane sector of an art em reflector. J Compos Mater, accepted
t k
ui displacement component of nodal point k in for publication.
configuration at time t [5] Zhao Q, Hoa SV, Ouellette P. Progressive failure of triaxial woven
fabric (TWF) composites with open holes. J Compos Struct, in
uki increment in t uki
press.
t i;j tþDt i;j derivative of displacement increment with
u ; u [6] Bathe KJ, Bolourchi S. Large displacement analysis of three-
respect to co-ordinate t xj and tþDt xj dimensional beam structure. Int J Numer Meth Eng 1979;14:961–
t
V ; tþDt V volume of body in configuration at time t and 86.
t þ Dt [7] Hu N, Hu B, Yan B, Funkunga H, Sekine H. Two kinds of
tþDt C 0 -type elements for buckling analysis of thin-walled curved
W external virtual work expression corresponding
beams. Comput Meth Appl Mech Eng 1999;171(1–2):87–
to configuration at time t þ Dt 108.
t tþDt
x; x Cartesian co-ordinate in configuration at time t [8] Bathe KJ. Finite element procedures. Englewood Cliffs, NJ:
and t þ Dt Prentice Hall; 1996.

Vous aimerez peut-être aussi