Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
and
JOHN ARGYRIS
University of Stuttgart,
Germany
Preface ix
Acknowledgments xi
4 Composites 37
4.1 Fundamental concepts 37
4.2 Basic material unit . . 39
4.3 Laminates....... 41
4.3.1 Special laminates . 42
4.4 Micromechanics and macromechanics . 44
vi CONTENTS
Appendices 321
Bibliography 327
Index 335
Preface
. f).T
11m
A8-+0
- - T n,
f).S - lim
A8-+0 u
~S = 0, (1.1)
where Tn is a vector, called the traction or the stress vector across the
surface at P. In general Tn will not be in the direction n, but will have
components in each of the three coordinate directions. We write
The quantity (J'xx, and the analogous (J'yy, (J'zz are called normal (or direct)
stresses; (J'xy, (J'xz etc. are called shear stresses.
It is found that the forces acting on a volume ~ V of a continuum may
be divided into surface forces caused by the tractions described above, and
body forces. The most common example of the latter is provided by gravity,
which exerts a downward force (p 9 ~ V) on an element of volume ~ V and
mass density p. We shall assume that if the body forces acting on a small
volume element surrounding a point P are equivalent to a force ~R at P
and a moment ~M, then
~R F
11m . ~M
AV = ,
~ V --+0 L.l. ~V
hm
--+0
AV
L.l.
= O. (1.4)
The vector F is called the body force per unit volume. We shall assume
that F is a continuous function of the coordinates (x, y, z), and write
Consider a small rectangular box with sides 2a, 2b, 2c parallel to the
axes and centered at P. The moment in the z-direction, of the forces acting
on the element, is provided by the moment ~Mz of the body forces and
the moments of those surface stresses shown in Fig. 1.1. Thus equilibrium
of the element demands that
~Mz
~V - (J'yx + (J'xy = 0, (1.7)
The symmetry of the other shear stresses may be proved in a similar way,
giving
1.1 The analysis of stress 3
2a
<1yz ___ _
2b
~~~~~~====;1~::::::::::~<1~~/~/~/5/~----~~ x
__-1______<1YZ /
"y/
2c
/
/
y z
z
nM
oJ [>x
nonnal
y y
~ ________-r__________ ~t>X
Figure 1.1: Surface and volume forces on rectangular and tetrahedron ele-
ments
4 Some results from continuum mechanics
The surface stresses may, when hand !:l.S are small, be approximated by
their values at P. Divide equation (1.10) by !:l.S and consider the limit as
!:l.V -t 0
3
Ti = LaijVj. (1.14)
j=1
1.1 The analysis of stress 5
Ti = O'ijVj. (1.15)
This equation allows us to show that the O'ij are the components of a tensor
of order 2. If we change from one orthonormal triad of axes iI, i 2, i 3, to
another, i~, i~, i~, where, in double suffix notation,
'ti
./ = lij'tj,
' (1.16)
r r r a2Ui
is O'ijVj dS + iv Fi dV = iv P at2 dV. (1.19)
where VI, V2, V3 are the components of the outward drawn normal on the
surface.
Now use Gauss' divergence theorem. In its usual, vectorial form it states
that
1S
Av
] ] dS -1
-
V
A] ,. dV
], (1.21)
6 Some results from continuum mechanics
In the same manner, a load applied in the y direction causes strain yy and
a load applied in the Z direction causes strain zz. We have
OU x OUy oU z
xx -- ---,
ox yy -
-~,
oy zz = oz (1.26)
1.2 The analysis of strain 7
I _~ru, I
L.
I [ I I
I ,hou ,om, I
du
1= x co 1
dy
I_ <=1
I"'
dx
Now suppose that the left vertical side of the rectangle displaces in
the x direction by dux and the bottom horizontal side moves in the y
direction by du y as shown in Fig. 1.2. As a result, the left vertical side
rotates counterclockwise by an angle (h while the vertical displacement du y
induces a clockwise rotation (h. Due to these rotations the original1r/2
angle becomes 7jJ, and the change of angle is
8 Some results from continuum mechanics
7r
'Y = "2 - 7/J = 01 + O2 (1.27)
OU
oy ~ O2 ,
tan02 = OU
tan 01 = -OX
y '" 01 x
- , (1.28)
1 (ou x ouy)
Exy =2 oy + ox '
1 (ou x ou z ) (1.29)
Exz =2 oz + ox '
1 (Ou y ou z )
Eyz =2 oz + oy .
The two shear strains Exz , Eyz are obtained considering angular displace-
ments in the xz and yz planes, respectively. We may now show that the
nine quantities
Ell = Exx , E12 = Exy , E13 = Exz , ... ,E33 = Ezz , (1.30)
We note that the strain tensor, like the stress tensor, is symmetrical, i.e.
Note also that we may write equations (1.27), (1.30) in the succinct form
1
Eij = 2(Ui,j +Uj,i)' (1.33)
1.3 Strain energy 9
so that
The dynamic equilibrium equation (1.23) allows us to write the first integral
as
and we notice that this is the rate of change of the kinetic energy
1 { .2
T = "2 }VPUi dV, (1.39)
dE
dt
- -dT -
dt
-1 V
-Ui J- dV
a-ZJ, (1.40)
. 1. -+ 1 . {1- (u-
. - +u--
.)}
ZJ Z,J- -- -a--u-
a--u- 2 ZJ Z,J 2 JZ J,Z- -- a--
-a--u- ZJ 2 Z,J J,Z
(1.41)
= aijEij,
where we have used the symmetry of the stresses (aij = aji) and the relation
(1.33). Thus
dE
dt
-dT
- -
dt
-1 V
a-ZJ-iJ- dV (1.42)
Up to this point we have used the stress tensor with components aij and
strain tensor strain tensor with components Eij; both of these are symmet-
ric, and they have 6 independent components. For practical computations
it is convenient to string these components into vectors:
Note that the components of e corresponding to the shear strains are dou-
bled, thus E4, E5, E6 are the engineering shear strains, denoted by ')'12, ')'13,
')'23 respectively
8u x 8u y
4 = 2E12 = -
8y
+ -8x = ')'12,
8u x 8u z
E5 = 2E13 = -
8z
+ -8x = ')'13, (1.44)
_ _ 8u y 8u z _
E6 - 2E23 - - +- - ')'23
8z 8y
This renaming has an important advantage: it allows the sum aijij in the
integrand in (1.42) to be written as an inner product of 0' and E
1.3 Strain energy 11
t
q e = 0'11:1.+ . .
0'21:2 + ... + 0'61:6
= 0'11 El1 + 0'22 E22 + 0'33 E33 + 20'12 E12 + 20'13E13 + 20'23E23
= O'ijl:ij (1.45)
There is one drawback to the use of q and e. These quantities are not
physical vectors, i.e. they do not change as a vector should (i.e. according
to O'~ = lijO'j) when the axes are changed.
With the new notation, we may write equation (1.42) as
of
O'i = Ol:i' (1.47)
E=T+U, (1.49)
where
U= iFdV. (1.50)
F is called the strain energy density, and U the strain energy of the vol-
ume. In infinitesimal elasticity theory F(l:l' 1:2, ... ,1:6) is a homogeneous
quadratic form in 1:1, 1:2, ... ,1:6 so that Euler's theorem on homogeneous
functions gives
12 Some results from continuum mechanics
8F
2F = -Ei = (7iEi, (1.51 )
8Ei
thus
1 f 1 f t
(1.52)
U = 2" 1v (7i Ei dV = 2" 1v u edV.
Reversing the steps from (1.52) to (1.34) we may deduce that, in the
static case,
1ft If If
U = 2" 1v u edV = 2" 1v FiUi dv + 2" ls TiUi dS. (1.53)
We prefer to use a different notation for the body forces and surface trac-
tions:
An analysis exactly similar to that given in section 1.3 shows that we can
write
6W = Iv u t 6g dV. (1.57)
6U=6W
This equation is the principle of virtual work. In words, this states that
Definition 1 The virtual work done by the body forces and surface trac-
tions on the virtual displacements, is equal to the work done by the stresses
on the virtual strains.
lit
F = -C""E"E"
2 tJ t J = -g
2 Cg ' (1.59)
6W = 6U. (1.60)
For
14 Some results from continuum mechanics
oU = o(~ Iv utgdV)
= ~ Iv utogdV + ~ Iv outgdV,
where, to find OU, we neglect the second order terms. But equations (1.51)
and (1.59) give
8F
(Ti = 8i = Cijj, (1.62)
so that
u = Cg. (1.63)
Thus
so that
OU = Iv UtOg, (1.65)
oW = -oUe . (1.66)
where the suffix i denotes the internal strain energy. This states that the
total potential, of external stresses, body forces and surface tractions, is
stationary, i.e.
i5(Ui + Ue ) = O. (1.68)
The quantity
U = Ui + Ue , (1.69)
For the stable structures considered in this book the potential energy
in this stationary configuration is always a minimum. For further details
on this and other energy theorems, the reader is referred to the books of
Przemieniecki [2], Tauchert [3] and Gould [4].
Chapter 2
World War II, with the advent of jet propulsion, there suddenly arose the
necessity of developing high subsonic speed fighters and fighter bombers.
At the same time there appeared the first electromechanical calculating
devices; the advent of such machines in the United States, Great Britain
and Germany was a revelation to the engineers who were active in aircraft
engineering and were faced with the necessity of analyzing and designing
swept-back wings for the first modern high-speed fighters. In those days,
practically all aircraft calculations were performed by some kind of force
method in which forces or stresses appeared as the major unknowns. When
the swept-back planes were to be designed it became apparent that in
a structure which had practically no two parts at right angles, the force
method was most inconvenient and unpractical.
In those days, John Argyris had heard of the Meteor fighter designed
by Gloucester Aircraft as the answer to the German ME-262. The ME-262
had good aerodynamic performance, but suffered many failures due to the
poor structural design of its swept back wings; this was a problem which
the manufacturers could not overcome at that time. In those far-away days
of 1944 John Argyris was working wit~ HL Cox at the National Physical
Laboratory at Teddington, London. HL Cox was then the prime figure in
aircraft technology in Great Britain. In spite of the strict security precau-
tions, they spoke about the Meteor wings and the difficulty of analyzing
them. So, during the course of three brainstorming days and nights, Ar-
gyris realized that the force method was not suitable for this problem due
to the great difficulty in developing the self-equilibtating systems. Conse-
quently, he toyed with the idea of the displacement method and it suddenly
occurred to him that the triangle was incredibly well suited to these odd
swept-back wing structures -by using a number of triangles it was possi-
ble to discretize such high-speed subsonic structures! As the next step, an
elementary matrix code was designed. The first structure studied using
this new method was a simple wing model described by a set of 64 linear
equations. It was analyzed on the electromechanical calculator at the Na-
tional Physical Laboratory similar to a computing device being developed
at Harvard University in the United States. Needless to say, no software
was available to solve for the unknowns, so that simple, elementary solu-
tion techniques were conceived. In 2-3 days of frenzied work, displacements,
strains and stresses for a loading case were obtained. HL Cox, who was
aware of the on-going work, advised a direct comparison with experimental
tests for a simple wing model. All were surprised when it was discovered
that the maximum stresses did not deviate more than 8 to 9% from the
2.2 The finite element method 19
Courant
Courant [5] developed the idea of the minimization of a functional using
linear approximation over subregions. He specified the values at discrete
points similar to the node points of a finite element mesh. In his paper
he shows a mesh subdivision up to 9 approximate points to solve the St
Venant's torsion of a square hollow box. He considered the potential energy
U of the system and used the condition of stationary potential energy which
he expressed as
c5{V + U} = O. {2.1}
8{V + U} = O. {2.2}
8ai
Courant also applied the Rayleigh-Ritz method to create a functional for
cp with two unknowns.
Courant did not fully clarify his piecewise linear approximation to the
cp surface nor did he give any mathematical details of this approximation.
He clearly indicated, however, a procedure which could be used for the
minimization of the total potential energy for the torsion problem.
elements. They developed the concepts of flexibility and stiffness and pro-
vided equations which have become standard in structural mechanics. They
include
in which [f], [k] are constitutive relations and [b]' [a] are stress-force and
strain-displacement relationships, respectively. They applied their theory
to a rectangular panel for the case of nodal displacements which vary lin-
early along the edges of the element and calculated its stiffness matrix; the
first element in plane stress using interpolation functions in terms of nodal
displacements was developed! The panel had an 8 x 8 stiffness matrix which
was expressed as a sum of shear and direct strain matrices:
Thus Argyris and Kelsey developed the rectangular panel stiffness matrix
in plane stress using element interpolation functions in terms of nodal dis-
placements. They also showed that, in the context of aircraft skin models,
the triangular panel behavior may well be approximated by using energy
minimization procedures which were shown later to provide a firm basis for
finite element formulation.
will be used as the basic building block for calculating stiffness matrices for
plates of arbitrary shape. It then proceeds to the study of the triangular
element in plane stress. It starts by assuming constant strains which are
integrated to yield the displacements u, v. Then, it expresses the relation
between stresses O"xx, O"yy and O"xy and the nodal displacements in the form
The nodal forces were obtained from the three stresses via
Clough
The name finite element method is attributed to Clough [13], [11], who
was a coauthor of the original paper by Turner [12]. In [11], [13], he out-
lined how he first invented the name finite element method because he
wished to show the distinction between the continuum analysis and the
matrix method of structural analysis. In [20] he outlined the research pro-
gram undertaken at Boeing Company in 1952-1953 for the calculation of
the flexibility coefficients for low aspect ratio wing structures for dynamic
analysis. He extended Turner's work from 1957 onwards; continued conver-
gence studies on stress components; and popularized the ideas of the finite
element method.
this paper they set up the approximation to the functional in terms of the
nodal values of the triangular domain, into which the region is subdivided.
The function minimization techniques, originally discussed by Courant,
were clarified by Zienkiewicz in 1965; this opened the way to the analysis
of field problems by the FEM. A first book on the FEM was published in
1957 [19] with new editions appearing in recent years [22].
Zienkiewicz expressed his personal views on the origins, milestones and
directions of the finite element method in [23]. He stated that the finite
element method was made possible only by the advent of the electronic
digital computer, and discussed the "variational" approaches via extremum
principles.
Thus if the strain field in an elastic continuum is defined by a suitable
operator S acting on the displacements u as
e=Su, (2.9)
with the corresponding stresses given as
(T = De, (2.10)
where D is a matrix of elastic constants, then the finite element solution
sought could be obtained by the minimization of the potential energy
uh = N-u, (2.12)
Ku=!, (2.13)
2.4 The basic ideas of FEM 25
where
K=LKe, Ke = r (SNfD(SN)dn,
Jn e
(2.14)
and
The domains ne and r e correspond to elements into which the whole con-
tinuum problem is divided. Equations (2.14), (2.15) are used to generate
element stiffness coefficients and forces providing that the approximation
shape functions of Eq. (2.12) are defined on a local basis. These equa-
tions summarize the basic philosophy of classical finite element methods.
Zienkiewicz points out that such a derivation of the finite element procedure
is a particular case of the approach introduced much earlier by Rayleigh [24]
and Ritz [25]. The main difference lies in the use of the local shape func-
tions N which yield a banded structure of the assembled stiffness matrix
K and preserve the local assembly structure of the matrix equations.
Chapter 3
wu2 n
:~];,;.
WI
PI] _
U = [ w! wv2 Wu
[
P2
-
W p.
(3xn)(nxl)
(3.3)
wI
w ww2
r = A p =?p=A-1r=ar. {3.4}
(nxl) (nxn)(nxl)
Therefore
u w p = ware, {3.5}
(3xl) (3xn)(nxl) (3xn)(nxn)(nxl)
C=wa. {3.6}
The elemental natural modes p will be always split in two parts; the nat-
ural rigid body modes Po and the natural straining modes p N. The latter
describe pure straining deformation.
p t 8p = P't8pl = p'ta8p
-+ P at pI = a~Po + a~PN (3.9)
PN = aNP, Po = Ab P ,
PN A~P, P=a~PN' pI = Atp. (3.11)
From equation (3.11) we see that the connection between natural and carte-
sian displacements as well as natural and global load vectors is provided by
matrices a, A. Usually these matrices are functions only of the geometry of
the element and can be easily constructed. Ignoring the presence of initial
strains, we write in the natural coordinate system
g = aNPN, (3.12)
where g is the elastic strain vector, PN is the vector of natural modes, and
aN is a strain operator matrix. Inserting (3.12) in the first integrand of
the expression of virtual work (1.58) we obtain
The right hand side includes surface, volume and concentrated loads refer-
ring to the natural coordinate system. Substituting equation (3.12) in the
middle and right integrands above
30 Natural modes for finite elements
Po=o, (3.15)
Equation (3.15) states that the rigid body forces acting on an element must
equal zero.
We are now in the position to define the elemental natural elastic stiff-
ness kN as
so that,
Po=o,
kNPN = PN (3.19)
Figures (3.3-3.3) show the straining modes for the selected finite elements.
We stress that once the natural stiffness matrices are computed, simple
congruent matrix multiplications transform them to the local and global
cartesian coordinates.
3.3 Natural modes for selected finite elements 33
3D truss element
("2' v 2 'W 2)
2
2
PN extension
z.w
x, v
plane stress triangle
I ( "I' vI)
x, "
2/
\3
(")' v)
2/ /
) ' ,~ ==--- )3
3~'Y1II
1/2'Yry
Figure 3.1: Strain modes for truss and plane stress triangular finite elements
34 Natural modes for finite elements
1I2p
I NlI
~
~
EI
1/2pNl
I extension
r~
1I2p N2
a symmetrical bending in x-z
[ Mode 3
II2P~
B_-----;Jt..__ y
8
,../ ---j ,/ ' ,
1/2PN4 symmetrical bending in x-y
1I2PN4 --- ,/' (........ --i>x
<
[ ModeS
112PN5
.",/ ,-- _--_
\
1 ;'
,,/
~Y
112PN5 8 antisymmetrical bending in x-y
--'---------
< ,"
-----
\
'-
fI"~ ~I2P
i 8
1/2PN6 Mode 6
~
torsion about x
l/
1I2Y,a
~ --<t,:
)3'Ytjl ...{')-- { {
i ;
::::;;::000
1I2Yta
1/2YI"(
1I2"';a
Figure 3.3: Strain modes for a plate and shell triangular finite element
36 Natural modes for finite elements
Z,W
x, v
2
(u 2,v2 ,w2)
.x.u
1
(ul'vl' WI)
2 2 2
2
2 2
Composites
Reinforcing fibers
In fiber reinforced composites, the fibers provide virtually all the strength
and stiffness. The most widely used fibers include the following:
Carbon: Carbon fibers are the most popular and widely used rein-
forcing fibers in high-performance composite structures. They are
commonly made from rayon and polyacrylonitrile precursors (PAN).
4.2 Basic material unit 39
Boron: Boron fibers were first developed in the early 1960s and were
the first reinforcements for advanced composites. They are of high
modulus and high strength, but expensive due to the high raw mate-
rial costs. The United States Air Force F-15 and Navy F-145 fighter
planes both use boron fiber reinforced composites as structural com-
ponents.
Matrix materials
Common matrix materials are as follows:
Epoxy: Epoxy thermoset resins are the most commonly used matrices
in advanced composite materials. Although limited to a maximum
service temperature of 120C, they provide an almost unbeatable
combination of characteristics such as handling, processing flexibility,
mechanical properties and cost.
........
3,z
local coordinate
/2~x +9,))
x
fiber coordinate - e
Turn to Fig. 4.1. We shall always choose the coordinate system 123
so that the 1 axis lies along the fiber direction and axes 2 and 3 are di-
rected perpendicular to it. This coordinate system will be called the ma-
terial coordinate system and will differ for each layer, depending upon the
orientation of the fibers. When multiple plies are present -as for a compos-
ite laminate- a second coordinate system is chosen with respect to which
all material transformation and homogenization procedures (through the
thickness integrations etc.) are performed. This coordinate system will be
labeled xyz, and called the local coordinate system (when a global coordi-
nate system is also used the local coordinate system will be x'y' z'). For a
beam element the local coordinate system is placed at its center. Then, as
shown in Fig. 4.1, the material axis 1 forms, for each ply k, an angle Ok
with the local axis x, the latter being common to all layers. This angle is
4.3 Laminates 41
k~~~~7 9=-45
,/\\\\\7
9=45
.x
y~ 7 9=0
////7 9=90
4.3 Laminates
A structural laminate is created by stacking together many plies of
fiber reinforced layers. The different layers of lamina are permanently
bonded under heat and pressure using a hot press or autoclave. There are
many types of composite laminates; carbon matrix-carbon fiber compos-
ites (CCC), metal-matrix composites (MMC), ceramic-matrix composites
(CMC), titanium-matrix composites (TMC), etc. Some of these laminates
combine the advantages of both metallic and non-metallic technology. For
example, silicon carbide, boron carbide, alumina and graphite fibers can be
combined with a metal matrix made of aluminum or magnesium to form a
metal matrix composite stronger, stiffer and lighter than its constituents.
The stiffness of the laminate is obtained from the properties of the con-
stituent layers; the laminate can resist loads in several directions. Using
various theoretical methods such as the classical lamination theory (CLT)
42 Composites
[27], we may derive the stiffness coefficients of the laminate from the stiff-
nesses of the individual plies. Usually the stress strain relations for a thin
lamina of an orthotropic material are written down, and subsequent trans-
formation and integration procedures yield the stiffness entries of the whole
laminate. Figure 4.2 displays an unbonded (-45/45/0/90) laminate made
from four layers with different fiber orientations.
The intrinsic mechanical behavior of composite laminates offers tremen-
dous possibilities. Figure 4.3 shows two composite laminates, a (0/90/90/0),
or simply (0/90)5, and a (90/0/90/0), or (90/0)s. Both consist of the
same geometry, four layers of the same thickness, and are subjected to the
same axial force N and end moments M. The figure shows that when
the same axial load is applied, both stretch the same amount. However,
when the same moment is applied, the laminates behave differently, with
the (0/90/90/0) exhibiting a stiffer response than the (90/0/0/90) plate.
This illustrative and interesting example shows the great advantages of
structures made from composite materials; by varying the fiber orientation
we can alter and optimize the mechanical response of the structure un-
der certain loadings. For example, the three laminates (+45/ - 45/0/90)5'
(+45/0/ - 45/90)5, and (+45/90/ - 45/0)5, have the same number of lay-
ers and fiber orientations but different lamination schemes; the different
laminations will lead to different behaviors under various loads.
a (0/90/90/0) laminate
M
N~
..... \
-- ... .
.. . .
.. . . ....
....N
N
....N
N
.. ..
M
M( (0/90/90/0)
)
i-"'i
M
)
differences in bending deformations
.--.
Figure 4.3: Axial and bending deformations of two cross-ply laminates
Angle-ply: If every layer has either angle () or angle -(), the lami-
nate is said to be an angle-ply laminate. An example of an angle-ply
44 Composites
These questions show that there are two basic problems: determine the
lamina properties by the procedures of micromechanics; use these proper-
ties in macromechanics for a global analysis of the continuum. In the first
(micro) problem many difficulties arise due to uncertainties which may ne-
cessitate the adoption of stochastical or statistical methods. The objective
of most approaches is to determine the moduli of a composite material from
the moduli of the constituent materials, or determine the strengths of the
4.4 Micromechanics and macro mechanics 45
unit cell ~
matrill
~ interface
fiber
1
l--I~=--)
I~
~ supersonic tmnspon
~ ~~'
slruClure l
Figure 4.4: From micro to macro-structures
5.1 Introduction
Beams are autonomous or secondary load bearing members of many struc-
tures. They are used extensively in the formation of linkages, shafts and
frames and as reinforcements on plate and shell panels. They are also used
in robotics and high speed machinery. Laminated composite beams are in-
tegral parts of lightweight structures that require high strength over weight
ratios. In general, laminated beams have been studied to a lesser degree
than laminated plates and shells. This may not come as a surprise consid-
ering the fact that some aspects of the theory of beams are more intricate
than aspects of plate and shell theory. For example, the geometrical stiff-
ness of a shell element, required for buckling and large deflection analyses,
is sufficiently developed using membrane stresses. However, for beam ele-
ments other forces or moments may contribute to the geometrical stiffness,
albeit not significantly. In the latter case, the geometrical stiffness is more
difficult to derive than for plate and shell elements.
The analysis of beams is usually based on four approaches; classical
theory of elasticity [30]; theory of mechanics of materials [31], [32]; and
variational methods and strain energy statements [29]. For practical engi-
neering computations of three dimensional beam structures, these models
are usually followed by associated finite element implementations; these
lead to various beam elements with or without transverse shear deforma-
tion. Most of the literature is devoted to the development of isotropic beam
elements. Laminated beams have received less attention. Various straight
and curved laminated beam finite elements were developed by Venkatesh
and Rao [33], Yuan and Miller [34], Oral [35], Chan and Yang [36].
Most isotropic and composite beam finite elements are based on clas-
sical finite element methods in conjunction with classical, first or higher
order lamination theories. These models often lead to parasitic phenomena
such as shear locking, rigid body failure, zero-energy modes, etc. Most
require some sort of stabilization. These phenomena also occur with high
anisotropy and the low ratios of axial to shear moduli encountered in com-
posite materials. When higher order approximations are adopted some of
these problems may vanish, but then the delicate balance between economy
and accuracy is often sacrificed.
In developing a beam finite element we follow a different approach than
classical methods by adopting invariant deformation measures, the so-called
natural modes, to describe the elemental kinematics. The natural modes
include both rigid body and straining components [37]. We assume that
the 12 cartesian freedoms produce both rigid body motion and pure defor-
mation. Only the modes contributing to the pure deformation are used in
the elemental stiffness matrix. Thus, the order of the problem is automati-
cally reduced. Some similar underlying concepts applied to the analysis of
composite structures can be found in [16], [38]. For the composite beam
finite element in space we will select 6 straining modes (equal to the total
cartesian degrees of freedom minus the rigid body freedoms), and using
concise energy and work statements will develop a 6x6 sparse natural stiff-
ness matrix. Subsequent congruent transformations will refer this matrix
to the local and global cartesian coordinates. This approach leads to great
savings in computing time since most CPU time is spent on algebraic op-
erations such as vector and matrix multiplications [39]. The approach has
also significant potential for vectorization/parallelization.
5.2 Kinematics
Figure 5.1 displays a cross section of a beam element under the action
of edge moments. Our basic assumption is that the vertical displacement
w varies only with x. Then, symmetrical considerations lead to v = 0.
Using these assumptions, we express the displacement field within the beam
element as
u = u{x, z),
B
v == 0, By = 0, (5.1)
w = w{x).
5.2 Kinematics 49
For a beam element one may also adopt the usual assumption a yy = azz =
o.
We begin by considering only bending in which the normals remain
straight and normal after deformation, and derive expressions for the axial
displacement u and the axial strain 'Yxx due to an applied moment M. Turn
to Fig. 5.1 and observe that at a distance z from the middle surface the
beam is displaced horizontally by the amount
u= d = -ztan(}. (5.2)
!:1w
tan{l = . x + !:1uo (5.3)
Dividing the top and bottom of the right hand side by !:1x we find
8w
tan {I = Ox,
1+ x
10 (5.4)
or
8w
tan {I = Ox 8w (5.5)
1 + xo ::: 8x
8w
{I 1, then {I = 8x. (5.6)
0 P Uo
~ x,U
' w(x)
M
( .....w = W(X)
................. 11-- M
y, v
z, w ~ /1 --....::.:::- r l i - .
)
d = z tane
original length
z=o
o ( P Q x, U
.ix
wi Iw+.iw
U o P'
----- 7/ ~
U o +.iuo
final length
z, w
or
5.3 The beam element 51
xO -
_ ~uo '"
~x -
_!2 (~W)2
~x
= O((p),
I
(5.10)
Uo 10 -
1
1___
2
J(~W)2
_
~x
dx.
o
We deduce from these equations that the strain xO of the middle plane is
of order (j2 and therefore quite small in the context of linear theory. In
classical theory we assume that Uo ~ 0 (~x* = ~x). Now the expression
for U becomes
dw dw
u( x, z) = Uo - z tan 0 ~ Uo - z dx ~ - z dx . (5.11)
ou(i, z) d2w(x)
'Yxx = ox = -z dx 2 (5.12)
P ~ p ~ Po ,PN PN = P- Po {5.15}
(12x1) (12x1) (6x1) (6x1)
'V
y. "
e x. u
z'.3
z', w
2
Z'k+1
"I
u narural coordinale
~ _ _ _ _,----.
~ _
~."..,- ~ eL
~~ " .t'
To the rigid body modes correspond rigid body forces and moments com-
ponents of the vector
The three translational modes are deduced from a perusal of Fig. 5.3 as
follows:
54 Composite beam element
~, EI
, ,/ '
El
z
2
I
X"
2E1
2
~ 2
Figure 5.3: Natural rigid body modes for the beam element
5.4 Natural rigid body modes 55
Ul = POI,
U2 = POl,
1
VI = P02 - "2P06,
1 (5.19)
V2 = P02 + "2P06,
1
WI = P03 + "2P05,
1
W2 = P03 - "2P05,
1
POI = "2(Ul + U2)
1
P02 = "2(Vl + V2)
1
P03 = "2(Wl + W2) (5.20)
1
P05 = y(Wl - W2)
1
P06 = y(V2 - vd
The fourth rigid body mode is a rotation with respect to the x axis (a
twisting movement). It is expressed as
1
P04 = "2(91 + 92 ) (5.21)
From these relations we construct the matrix ao which extracts the natural
rigid body modes from the local cartesian degrees of freedom
56 Composite beam element
1 1
2 2
1 1
2 2
1 1
2 2
1 1
(6xl2)
iio = '". 1
2
-7
1
2 1 (5.22)
1
7 1
-7 7
or using compact matrix notation
Po = iio p, (5.23)
(6xl) (6xl2)(l2xl)
where p is the vector of the local cartesian degrees of freedom. Thus, when
the local nodal degrees of freedom are available (they can be easily deduced
from the global displacements and rotations) we can explicitly define the
rigid body modes. The element rigid body modes can also be extracted
from the global nodal degrees of freedom. If
p = T04 p, (5.24)
(l2xl) (l2xl2)(l2xl)
where p, T04 represent the vector of global nodal displacements and matrix
of direction cosines, respectively, equation (5.23) becomes
It is thus possible to extract the natural rigid body modes from both the
local and global cartesian degrees of freedom.
The natural straining modes and forces are related via the natural stiffness
matrix,
I PN = kNPN f (5.28)
(= 2x I I
-1 ~ (~1, --2-
< x <-.
-2
(5.29)
1'
Also
o{) o{) o( 0 0 2
- = --::::} - = - - . (5.30)
ox o( ox ox o( 1
The natural stiffness of the beam element will be based only on defor-
mations and not on rigid body motions. The stiffness kN corresponding to
these deformations is denoted the natural stiffness matrix.
1
u{() = "2 (PNI. (5.31)
58 Composite beam element
112pNl r::::7i
~
B
I I 112pNl
I I extension
I
a
Mode2
1!2p N2
symmetrical bending in x-z
i
I Mode3
II2P~
112P~4::\ b ~~
~
antisyrnmetriCalbe~dinginx-z
\:Y
112PN2 _ transversesheanng
+ ~
'l'---~
Il2pJ.n.
1!2PN4
////)
I
_----
Mode4
_----.tly
<",/
I
,/' --....
C",
1I2PN4
a symmetrical bending in x-y
B
1I2PN5
-- )
..,.......
-----\
1 /
,"
,,"
~y
112PN5
'> ....
a antisymmetrical bending in x-y
---
< I _;a
\ ...... _--
a
1I2PN6 Mode 6
fI "JI ~I2PN6
~ t",," ~
torsion about x
[M~
B
FN FN
.q-- --i> extension
I M~21
B
t~3
symmeUicaI bending in x-z
(~ ;>qt XW." MA
a
M~_=) _____----___('",~
r~~~]
symmeUical bending in x-y
[~-;;;;;J
r: 2
MA
<
........ j . . _----\ MA~
\. . _-----'>_ . . '
I
2
8 antisymmeUicaI bending in x-y
r~6J
8
~\ l
torsion about x
au au a( _ PNI = a f PNl, 1
I~x = ax = a( ax - l a -
- -l
(5.32)
l l
w(x) = ax 2 + bx + c, x: -2' --+ 2" (5.33)
l l
w(-2') =0, w(2') =0,
(5.34)
'( l 1 , l) 1
w -2') = 2'PN2, w (2' = -2' PN2,
we have
1 2 l
w( x) = - 2l x PN2 + 8 PN2 (5.35)
Using
l
x = 2''' (5.36)
w(()
l
= 8(1- ( 2) PN2 (5.37)
5.5 Natural straining modes 61
b + s
PN4 = PN4 PN4' (5.38)
1 1
w(-2)=0, w(2) = 0,
(5.40)
'( l) 1 '( 1) 1 b
= PN3,
w -2 = 2 PN3 , w 2 = 2 PN3 , PN3
Superposing the elastic displacement curves due to the second and third
modes we obtain
1 2 1 3
w(() = g{1- ( )PN2 + g (( - () PN3 (5.42)
We shall use equation (5.12) to estimate the strain arising from the sym-
metrical and antisymmetrical bending modes. We have
62 Composite beam element
{Pw z z
'xx = -Z--2
ax = -1 PN2 - 3- (PN3
1
(5.43)
Q= 2MA (5.44)
1 .
We assign two antisymmetrical moments of equal magnitude MA on the
two beam edges as shown in Fig. 5.6. From elementary beam theory we
may express the angle on the left edge due to the action of the moment MA
on the same edge as
III MAL
__ ,
Ub = 3EI (5.45)
MAL
O~ = - 6EI' (5.46)
MAL
Ob = ot + O~ = 6EI' (5.47)
For transverse shearing (significant for moderately thick and thick cross
sections), the angle of straining is
(1 Q 2MA
Os =, = G = (GA) = l(GA)' (5.48)
5.5 Natural straining modes 63
":>+
Mtt~ El .MA)
I ~Q
MA( k
9
======-=-
2
~ )A M
~ ~
~1>
t 9s
~
ilikkbo= "'1~Q*
Figure 5.6: Bending and shearing as induced by the antisymmetrical mode
The total angle due to antisymmetrical mode components P'N3 and PN3 is
MAl 2MA)
PN3 = P'N3 + PN3 = 2{fh + Os) = 2 ( 6EI + l{GA) .
(5.49)
3(E!)
M - PN3 = 1 + \2(EI)
A- 2(I 2)
6(EI) + I(GA)
I
I (GA)
PN3, (5.50)
or
(5.51)
where>. is the beam shearing coefficient. Equation (5.51) shows the series
connection between the antisymmetrical bending and the transverse shear-
ing and naturally leads to the circumvention of the transverse shear locking
phenomenon frequently encountered in classical finite element methods. For
a thin beam
64 Composite beam element
1 1 3(ElN)
UN3 = 2MA PN3 = 2PN3 ", , ,\ PN3, (5.53)
where
(ED)
(ElN) = (El) - (EC)z5, Zo = (EC) (5.54)
k 3,3 _ 3(ElN)
(5.55)
N - 1(1 + >.)
1 2
v(() = 8(1- ( )PN4. (5.56)
Z,W
x,u
........ .........
..- .... ..- ........
..-"-
..- .... ..-
..-"-
y. v
..-"-
..-
..-"-
..-"-
.""...""..".,
is not accounted for. Note that in the absence oftransverse shearing forces,
the elemental equilibrium may not be completely satisfied since the anti-
symmetrical moment M1 is not equilibrated. This moment, however, for
a laminated beam is small and its contribution to the overall deformation
is negligible. We superpose the displacement curves stemming from the
fourth and fifth modes to obtain
l
v{() = 8(1- ( 2 )PN4 + 8l (( 3 - () PN5 (5.57)
The fourth and fifth natural modes give rise to an axial strain which is
derived by using once again (5.12). We have
a2v y y
'Yxx = -y ax 2 = yPN4 - 3 y (PN5 (5.58)
half to the right. Its linear variation with respect to the non-dimensional
coordinate ( is
1
O(() = 2( PN6 (5.59)
The displacement field due to this twisting angle is (see Fig. 5.7)
1
u((, y, z) = O,x w(y, z) = 1 w(y, z) PN6,
1 1
v((, z) = -0 z = -2( z PN6, w((, y) = Oy = 2( Y PN6 (5.61)
In order to derive the strain field arising from this mode we need to express
the field as a function of the coordinate x. This is accomplished by using
(5.29) and leads to
1
u(x, y, z) = 1 w(y, z) PN6,
x X
(5.62)
v(x, z) = -y Z PN6, w(x, y) = yYPN6.
1 (QUi
lij =2 QXj
QUj )
+ QXi '
i,j = X,y,z. (5.63)
We need to insert a v'2 in front of all shear strains and stresses and derive
a consistent expression fro the strain energy via
1 1 1 (QU' QU')
U = 2(J"ij/ij = 2[v'2(J"ij][v'2, ij] = 2(J"ij QX; + QX~ .
(5.64)
5.6 Natural stiffness matrix 67
au av aw
'xx = ax = 0, ,yy = ay = 0, ,zz = az = 0,
1
v'22,xy = v'2 (au 1(1
av) = v'2 l aw
ay + ax ay - lz) PN6,
In
y2,xz
1(auaz + aw)
= v'2 1(1l awaz + ly)
ax = v'2 PN6,
(5.65)
In
y2,yz 1 (av
= v'2 az + aw) 1 (x
ay = v'2 x) PN6 = 0.
-y + Y
Grouping the non-zero strains due to the torsional natural mode we find
'xx = 0, y
(aw(y,
In
ay z) - z )PN6,
2,xy
1 1
= v'2l
(5.66)
In 1 1 (aw(y,z) )
y 2,xz = v'2 l az + y PN6
We put this strain field in the elegant matrix form
[Ji;x
v'2
,xz
y] =!
l
11.
v2 ay z) - z )
(aw(y, PN6, (5.67)
~ (aw~~, z) + y)
or using compact matrix notation
T
IT = a PN6 (5.68)
(3xl) (3xl)(lxl)
h
were 'Yxx,
E Sb 'Yxx,
'Yxx, T are the straIns
Ab 'Yxx, . d ue t 0 t h e axIa
. IstraInIng,
. symmet-
rical bending, antisymmetrical bending and torsional modes, respectively.
We point out that the in-plane shear strain 'Yxy arises solely from the tor-
sional mode.
Let us collect the expressions for the axial strain 'Yxx. We have
E
'Yxx = 'Yxx + 'Yxx
Sb
+ 'Yxx
Ab
+ 'Yxx
T
or in matrix form
I = aN PN (5.72)
(3xl) (3x6)(6xl)
El 1/12E2
0"11
[ 0"22
1 _1_
2E2
[ 1/12E2 E2 . E ] ['Y11 1
'Y22,
V20"12 k 1- 1/12 El . 2G12 (1 - 1/f2 E~} k V2'Y12 k
(5.74)
[V20"13]
V20"23 k =
[2G13
.
.]
2G23 k
[V2'Y13]
V2'Y23 k
For complete isotropy, 1/12 = 1/21 = 1/, El = E2 = E, G 12 = G, and these
equations reduce to
'Y11
[ 'Y22
1,
V2'Y12
(5.75)
where
70 Composite beam element
A= [
C~
So2
S2
o
..,fisoco
-..,fisoco
C2
o
1
-..,fisoco ..,fisoco (C~ - S~) k
(5.77)
The angle 0 represents the angle formed between the fiber axis 1 and and
local axis x' (see Fig. 5.2). Since we account only for strains "{xx, "{xy and
"{xz, the transformed constitutive relation takes the form
[
1
(7xx [Exx Exy
..,fi(7xy = Exy 2G xy . . 1[..,fi"{Xy
"{xx 1 =} U = K, .
..,fi(7xz . 2G xz ..,fi"{xz (5.78)
u= 21 ju t ,dV. (5.79)
v
Using equations (5.71) and (5.78), namely
PNt [j OtN
u = -21 (lx6) t K OtNdV ] PN (5.81)
(6x3)(3x3)(3x6) (6xl)
v
''------....,-----
natural stiffness matrix
5.6 Natural stiffness matrix 71
kN = (5.82)
(6x6)
~JI
v
symm. 9y2( 2Exx -3y(Exy (\fI _ ) I dV.
\12 ,Y Z
G Xy (\fI,y_Z)2
+G xz (\fI,z+y) 2
This full natural stiffness matrix represents the most general stiffness for-
mulation for an anisotropic beam element.
Simplifications
Some of the entries in this full stiffness matrix represent very small to
negligible contributions and therefore introduce unnecessary complications
in practical engineering computations. In particular, the coupling term Exy
is very small (identically zero for isotropic materials). Therefore, we can
omit some of these terms without a significant loss of accuracy. In addition,
we shall assume that some of the natural straining modes do not interact.
This is illustrated via
Exy ~ 0,
PNI !><l PN3, PNl!><l PN5, PN2!><l PN3,
(5.83)
PN2 !><l PN5, PN3 !><l PN4, PN3!><l PN5,
P N 4 !><l P N5, !><l: no interaction.
1
kN = - (5.84)
(6x6) Z2
-Exx zExx yExx
z 2 E xx zyExx
9Z 2( 2Exx
y 2Exx
l b ~
k~l = Z; / Exx{z)dV = Z; / dx / dy / Exx{z)dz
v 0 0 _l!
2
b N (5.85)
= Y2)Zk - zk+dE;x
k=l
11 b(h h ) EA
kN =y "2-(-"2) E=-Z-' (5.86)
Similarly,
5.6 Natural stiffness matrix 73
1 b ~
k~2 = l; / zExx{z)dV = l; / dx / dy / zExx{z)dz
v 0 0 _l!
2
N (5.87)
1 2 '"
= ~! L..,.{Zk2 - 2
zk+1)E k
xx
k=1
1/ z Exx{z)dV = l/l dx
kN22 = l2 2 l2 /b dy /~2z Exx{z)dz
v 0 0 _l!
2
N (5.89)
_bl", 3 3 k
- [3 L..,.{Zk - zk+1)Exx
k=1
k N2,2 -_ Ebh 3 EI
-
12Z -
-
-z-,
yy (5.90)
where Iyy is the bending moment of inertia with respect to the y axis. With
the evaluation of the integrals k~l, k~2, we are in the position to define the
quantities (EC), (ED) and (EI) that enter (5.54). They are
74 Composite beam element
N
(Ee) = L(Zk - Zk+1)E;x,
k=l
(ED) = 2"1"
N
2 2) k (5.91)
~(Zk - zk+l Exx,
k=l
N
(EI) = 31"
~(Zk3 - 3) k
Zk+l Exx
k=l
The evaluation process continues with term (2,4) which is
kN24 1
= [2 I yzExx(z)dV = 0 (5.92)
v
kN44 = 1
[2
I y2 Exx(z)dV = [21 Il Ibdx Y2 dy I~ Exx(z)dz
v 0 0 _l:!
2
1 b3 N
(5.93)
= 12 T L(Zk - zk+1)E;x
k=l
k~5 = ~ I
v
9y2(2 Exx(z)dV
36
=F II x dx Ib y dy I~ Exx(z)dz
2 2
(5.95)
o 0 _l!
2
1 b3 N
= 4T ~)Zk - zk+dE!x
k=l
k~6 = l; I [z
v
2 GXY + y 2 Gxz ] dV. (5.98)
N
6,6a _ 1 b ""' 3 3 k
kN - 31 ~(Zk - Zk+l)G xy ,
k=l
N
(5.99)
b3
6,6b _ 1 ""'
- 12 T ~(Zk - Zk+l)G xz
k
kN
k=l
76 Composite beam element
J=kbh 3 , (5.101)
where
1
O'xx = Exx'Yxx = E xxy 'I!PN6 => O'xx,x = 0, (5.104)
f) 2 'I!
O'xy = Gxy'Yxy = G XYy1 (f)'I!
!l - Z
)
PN6 => O'xy,y =
1
yGXY {j2PN6,
uy Y (5.105)
2
O'xz = Gxz'Yxz = G xz -1Z (f)'I! ) 1 f) 'I!
! l + Y PN6 => O'xz,z = -Z Gxzf.l2"PN6.
uZ uZ (5.106)
1
( yG 82lJ! 1 82lJ!)
XY 8y2 + yGxz 8z2 PN6 = 0, (5.107)
The warping function must satisfy (5.108). In the absence of warping, the
natural stiffness matrix becomes
Exx zExx
z2E xx
9Z 2( 2E xx
kN =~J y 2Exx dV.
(6x6) I v
symm. 9y2(2 Exx (5.109)
z2Gxy+y2Gxz
This form of the stiffness matrix does not include the effect of transverse
shear deformation. In order to account for the latter, we replace term (3,3)
with the previously derived expression (5.55) and write
Exx zExx
z 2 E xx
kN~~ Ii dV,
(6x6) I V y2Exx
(5.110)
symm. 9y2( 2Exx
z2G xy +y2G xz
where
1 z
G xz dz
k=l
N
= 2)Zk - Zk+1)G~z' (5.111)
EA
-[-
Elyy
[
[2
kN = Elzz
(6x6) [ (5.112)
3Elzz
symm. [
GJ
-[-
G = Be/G, (5.113)
where Be/ denotes the shear correction factor. To the shear moduli corre-
spond transverse shear energies, viz.
Us = Be/Us. (5.114)
From elementary beam theory it is known that the transverse shear stress
f7 q at position z is given by [40]
f7 q =Q
Rg(z), (5.115)
where
!
h/2
and
z
g(z) =- / Exx(z - zo)2dz, shear stress shape function.
-h/2 (5.117)
h h
2 2
1/ (72 1 Q2 / g2 (z)
Us = "2 G(:) dz = "2 R2 G(z) dz. (5.118)
h h
-2 -2
/2 iG(z)idz.
h
_ 1
Us = "2 (5.119)
h
-2
Using
-
, =
{7q
G - hG'
-!L (5.120)
h
_ 1 Q2 2
Us = "2 h2G2 / G(z)dz. (5.121)
h
2
By setting
h
2
hG = / G(z)dz, (5.122)
h
-2
_ 1 Q2
US = ---"
2 l!
(5.123)
!
2
G(z)dz
_l!
2
- 2
S _ Us _ R
(5.124)
! !
e! - U - h/2 h/2
g2(z)
G(z)dz G(z) dz
-h/2 -h/2
In order to proceed with the elaborate computation of the shear correction
factor Se!, the z coordinate of the neutral plane must be defined. This is
given by
!
h/2
zE!x dz
-h/2 (ED)
ZOa = h/2 - (EC)'
(5.125)
! -h/2
E!x dz
h/2
=
[- !
Se! h/2 ]2
(5.126)
E!x(z - zOa)2dz
h/2
Ga
!
-h/2
h/2
Gka
dz
5.7 Local and global stiffness matrices 81
where
! G~dz = L G~zhk'
h/2 N
Ga = (5.127)
-h/2 k=1
G xz = ScI G xz (5.128)
PN = aN P (5.129)
(6xl) (6xI2)(12xl)
The connection matrix aN contains only zeros, unit numbers and geomet-
rical parameters. The contribution of the vertical nodal displacements to
the antisymmetrical mode is illustrated in Fig. 5.8. We now express the
natural modes as functions of the local cartesian coordinates, viz.
PNI = U2 - UI
PN2 = <P2 - <PI
PN3 = -(<PI + <P2) + 2(WI :- W2)
(5.130)
PN4 = tPl - tP2
PN5 = (tPl + tP2) + 2(VI :- V2)
PN6 = (h - (h
82 Composite beam element
z
3
w I
o~I ~~
~_
~
10
1= "I
112
-1
-1 1
2 2
-1 -7 -1
aN 7
-1
(6x12) (5.131)
2 2
7 -7 1
-1
[(:~)l = (5.132)
(6x12)
1 1
2 2
1 1
2 2
1 1
2 2
1 1
2 2
1 1
1 -1
1 1
-1 1
-1
-1 1
2 2 "":1
1 -1 -1
1 -1
2 2
1 1 -1 1
-1 1
[
1 [aD 1 [ aD 1
(6x1)
Po
(6x12)
~; = ~~ (~~) = ~; (1~~i2)(lfx1)'
(6x12)
(5.133)
(6x1) (6x12) (6x12)
Conversely,
P
(1"'1)
= T- 1
(12:2)
AD
[(12 X 6)
I A
(1':;;)
1PO
1 ;~ .
(6x1)
[ (5.134)
(6x1)
Obviously,
[at 1 = A. (5.135)
[ Ao : AN ] (5.136)
(12x6) (12x6)
r
1
1 -2
I
1 -2
I
1 -2
1
-2
1 1
1 -2 ~2
1 1
1 2 2
1
1 2
1
1 2
I
1 -2
1
1 2
1 1
2 -2
1 1
1 -2 2
Now the expression for the strain energy takes the form
u= ~/ ut,dV
v
= -I PN t [/t
aN K, aN dV ] PN
2 (lx6) (6x3) (3x3)(3x6) (6xl)
v
....
natural stiffness matrix
(5.137)
= -1 pt [ii}y[ / a}y K, aN dV]aN
] P
2 (lx12) (12x6) (6x3)(3x3)(3x6) (6x12) (12xl)
v
....
natural stiffness matrix
~------~v~--------~
local cartesian stiffness matrix
The local elemental vector p is related to the global elemental vector P via
p = T04P, (5.138)
5.7 Local and global stiffness matrices 85
=
[TO. To
T04
To
: ] = rTo To To Toj,
(12x12) : (5.139)
To
with
In equation (5.140), Cx'x, denotes the cosine of the angle formed between
the local cartesian axis x' and global cartesian axis x. The same convention
is used for all other entries (see Appendix A). Using
u=-1 PtIt [-
T04 aN t
2 (lx12 (12x12) (6x12) v
[I ON t K, ON dV ]
(6x3)(3x3) (3x6)
- ] T04
aN 1 p.
(6x12) (12x12) 12xl)(5.142)
v ~
The matrix equilibrium equations in the natural, local and global coordi-
nate systems are
Next, we consider transformation rules for the stress resultants, i.e. the
generalized natural and cartesian forces and moments in the various coor-
dinate systems. Using the equivalence of work
- -
P- t_P = ptNPN = ptNaNP, (5.144)
we deduce
- -t - t -
P = aN PN, PN = AN P ,
(12xl) (12x6)(6xl) (6xl) (6x12)(12xl)
PN
(6xl)
= A~ T06 P,
(6x12)(12x12)(12xl)
P
(12xl)
= Tal ak PN.
(12x12)(12x6)(6xl)
(5.145)
u w P (5.147)
(6xl) (6x12)(12xl)
Ps = W PSI, (5.148)
(12xl) (6x12)(12xl)
where PSI includes the nodal values of the distributed load. Substituting
equations (5.147), (5.148) in equation (5.146) we obtain
Wp =-II
2
S
PsudS
t =- 1[ tit
PSI
2 (lx12)
S
W W
(12x6)(6x12)
dS T04 ] P
(12x12) (12xl)
1 t (5.149)
= 2PsP ,
from which we deduce the global vector of distributed loads
Ps =
(12xl)
Tb4
(12x12)
[I t w w dS] PSI
(12x6)(6x12) (12xl)
(5.150)
S
Note that in order for (5.150) to be valid, the distributed load vector PSI
must always be given with respect to the local elemental cartesian coordi-
nate system x'y'Z'. We embark now on the computation of the term wtw.
For this purpose we use the non-dimensional coordinate 'f/
x
'f/: 0 -+ 1, 'f/ = l' dx = l d'f/. (5.151)
W
(6x12) (5.153)
1-"7 "7
1-"7 "7
1-"7 "7
1-"7 "7
1-"7 "7
1-"7 "7
Then, a straightforward line integration yields
n
(12x12)
=j wtwdx
x
1 1
'3 6
1 1
'3 6
1 1
'3 6
1 1
'3 6
1 1
'3 6
= 11
1
'3
1
11 (5.154)
6 '3
1 1
6 '3
1 1
6 '3
1 1
6 '3
1 1
6 '3
1 1
6 '3
The complete global load vector becomes
This relation provides us with the global nodal loading vector when local
distributed loads are prescribed. For three commonly used loading cases,
the equivalent nodal loads in the local coordinate are
triangular load: load at left node kql; load at right node ~ql
trapezoidal load: load at left node {~ql + kq2)l; load at right node
(kql + ~q2)l
These are line loadings. If the loads are also applied across the width of
the beam b, the above values must be multiplied by b.
1 1
To = "2{Tt + n), Tl = -,;,{Tt - n), (5.157)
in which Tl, Tl, Tl, T~, represent the nodal temperatures at the top and
bottom surfaces at nodes 1,2, respectively. As for the elastic stiffness, the
thermal load vector will be computed first in the natural coordinate and
then transformed to the local and global coordinates before assembled.
The natural thermal load vector J N is written as the sum of two vectors,
namely
90 Composite beam element
JN = J NO + J Nl , (5.159)
(12xl) (12xl) (12xl)
where J NO, J Nl are the thermal loads due to uniform and linear through
the thickness temperature distributions, respectively. To this end we define
the thermoelastic coefficients for every layer in both material and local
coordinates via
at
12
[ O:'tll
.
.] ,
O:'t22
at = axx
[ a yy
1 , (5.160)
V2a xy
where, O:'tu, at22 are the coefficients of thermal expansion along and perpen-
dicular to the fiber direction, respectively. The thermoelastic coefficients in
the local coordinate system are obtained from the thermoelastic coefficients
of the material coordinate system using
The total natural strain is the sum of the elastic and thermal strains, re-
spectively, namely
I ak~aNdV 1
= pk [
= Ue + ut ,
V
natural ...'
, elastic stiffness
PN - [
------
l11t~aNdV 1PN
v
natural thermal vector
(5.164)
Jk = / l1t~aNdV. (5.165)
v
Using the basic definition of thermal strain
Jk = Jko +Jkl
(lx6)
I
= a~~aNTo(z,()dV +
V
I za~~aNTl(z,()dV
V
(5.167)
11
JtJ11 = 1 T(z,()ExxcxxxdV. (5.168)
v
92 Composite beam element
N
J NO
l,l _
- 4b (T2t +Tb
2
+
TIl) "'(
t +Tb L...J Zk - Zk+l
)Ek k
xxaxx
k=l
N
b (5.169)
J Nl
l,1
= 4h
(T2 fTl2
t -.Lb
Tl
+
fTll) " ' ( 2 2 )Ek k
t -.Lb L...J Zk - Zk+l xxaxx
k=1
The addition of these two entries yields the first component of the natural
thermal vector. For an isotropic beam of area A and Young's modulus E
under equal temperatures on the top and bottom surfaces, respectively, the
addition of the two integrals yields
J NO
1,1 + J Nl
1 ,1
-- (AE) aT, (5.170)
which is a well known expression of basic beam theory. The second entry
in the initial natural thermal load vector produced by the second straining
natural mode is of the form
N
J NO
2,1 _
- 8"b (T2t fTl2
+.Lb + Tl fTll) " ' ( 2 2 )Ek k
t +.Lb L...J Zk - Zk+l xxaxx,
k=l
N
J Nl b (T2t 5.172)
2,1
= 6h fTl2
-.Lb + Tl fTll) " ' ( 3 3 )Ek k
t -.Lb L...J Zk - Zk+l xxaxx,
k=l
5.9 Initial load due to temperature 93
31 =
J'; -1 1/ 3( zT(z,()ExxaxxdV. (5.173)
v
After some algebraic manipulations this term gives
N
J NO
3,1 _
- -8"b( - T2t '1"12
-.Lb + Tlt '1"11) " ' ( 2
+.Lb L...J zk
2) k k
- zk+l Exxa xx ,
k=1
N
J 3,1 = _~ (_7'.2 + T,2 + Tl _ T,1) ' " (z3 _ z3 )Ek ak (~.174)
Nl 6h t b t b L...J k k+ 1 xx xx
k=1
For isotropic materials under the same applied temperature, the sum of
these two entries is zero. Entries (4,1) and (5,1) represent the effect of the
symmetrical and antisymmetrical natural straining in the x - y plane and
are
41
J'; =11/ yT(z,()ExxaxxdV=O, (5.175)
v
51
J'; = -1 1/ 3(yT(z,()Exxa xx dV = O. (5.176)
v
We are only left with entry (6,1) -the direct result of the action of the
torsional natural straining mode. The corresponding integral is
1/
J,;61 = 1 [
T(z,() 2G xy (B'I!(Y,z)
By - z ) a xy ] dV.
(5.177)
v
94 Composite beam element
IN
61 = -Y 1! 2zT(z, ()GxylxydV. (5.178)
v
Using piecewise summation we derive the expressions
N
JNO
6,1
= -4b (fTl2
.Lt
fTl2 fTl1
+.Lb +.Lt +.Lb
fTl1) ~(2
~ Zk -
2 )Gk
Zk+1
k
xylxy
k=1
N
J N1
6,1 _ _ ~(fTl2_fTl2 fTl1_fTl1)~(3_ 3 )Gk k (~.179)
- 3h.Lt .Lb +.Lt .Lb ~ Zk Zk+1 xylxy,
k=1
For isotropic materials under the same applied temperature the addition of
these two terms gives zero.
Note that the natural thermal load vector has been completely defined.
Equilibrium in the natural coordinate yields the equation
The natural initial load due to temperature is transformed to the local and
global coordinate systems via
- ]t t-
J = [aN T04 JN = T04 J
(12x1) (12x12)(12x12) (12x1) (12x12)(12x1)
-- t
aN J N
(5.181)
(12x12)(12x1)
I KEr~P+J I'
(5.182)
where
The solution of (5.182) yields the global nodal displacements and rotations.
5.10 Postprocessing
5.10.1 Computation of forces and moments
To gain a better insight into the behavior of the beam element, we look
at the forces, moments, stresses and strain energies. In our method they
are all obtained in a postprocessing mode following the computation of the
global displacements.
We use equation (5.141), namely
to extract the natural modes PN, from the global cartesian vector P the
natural modes PN. This is performed for every element. Then we multiply
the vector of natural modes by the natural stiffness to obtain the generalized
natural forces and moments
PN =kNPN (5.185)
We use the following sign convention for natural forces and moments:
PLANE x-z
PLANE x-y
1
Axial straining energy in x - x : -FNPNl,
2
1 1
Symmetrical bending energy in x - z : -MSPNl,
2
1 2
Symmetrical bending energy in x - y : -MSPN4,
2
1 1
Antisymmetrical natural energy in x - z : -MA PN3, (5.186)
2
1 3
Antisymmetrical natural energy in x - y : -MA PN5,
2
1
Torsional natural energy in y - z : -MTPN6.
2
k
u xx = Ekxx'Yxx
k (5.187)
where
k PNI Zk Zk
'Yxx = -l- + TPN2 - 3 T (PN3. (5.188)
k
U yy = V12 Ekxx'Yxx'
k (5.189)
The in-plane shear stress u xy arises solely from the torsional straining mode
PN6 Therefore
Z
'Yxy = U,y +v'x = -yPN6' (5.190)
98 Composite beam element
and
k
u xy = Gkxy 'Yxy
k (5.191)
In our theory the transverse shear stresses are constant through the beam
thickness. For an accurate computation of the through the thickness trans-
verse shear stresses u xz we shall proceed as follows: we first state equilib-
rium in the x direction via
auxx auxz _ 0
ax + az - , (5.192)
or
auxz auxx
az - ax ' u xz -
__ ! auaxz
xx dz + Cl. (5.193)
h
-2"
!
h
2"
zExx(z)
auxx aMz aM ~ (z - zo),
ax ax ax
!
I ~
(5.194)
Exx(z - zo)2dz
h
-2"
!
h
2"
zExx(z)
! ax
z h
!
= -
(5.195)
-~ Exx(z - zo)2dz
h
-2"
5.11 Geometrical stiffness 99
aM 2MA
q=-=--, (5.196)
ax l
On account of (5.196), equation (5.195) yields
!
h
2"
zExx{z)
! 2~A
z
h
CT xz =_ h -2" {z - zo)dz + CI
!
2" (5.197)
_l!:.
2
Exx{z - zo)2dz
h
-2"
Since the transverse shear stresses vanish at the top and bottom beam
surfaces, the constant CI is zero.
-
P = -t
aN
P N. (5.198)
(12xl) (12x6)(6xl)
dP = da~P N + a~ dP N. (5.199)
The term da~ represents the effect of change of geometry when the natural
forces are not varied and this is the main contributor to the geometrical
stiffness, while a~ dPN stems from a variation of PN, the latter being
usually an initial end load. In our theory we assume that only the rigid body
rotations contribute appreciably to the geometrical stiffness. Therefore the
term da~ in equation (5.199) will be derived based only on this assumption.
We consider the vector of rigid body rotations
These rotational variations will induce geometrical forces. First they give
rise to the incremental rigid body moments
Now a small change of the cartesian nodal loads produces rigid body mo-
ments
- t - - t -t
dPOR = AORdP = AORdaNPN, (5.203)
102 Composite beam element
- t -t
AORaN = 0, (5.204)
we have
- t -t - t -t
dAoRaN + AORdaN = o. (5.205)
dPOR = -dAoRaNPN.
- t -t
(5.206)
The incremental rigid body rotational moments give rise to the geometrical
cartesian forces
- G = aOR
dP -t dP OR -t dA- t -t P
= -aOR ORaN N, (5.207)
(12x3) (3xl)
whereby
-t -t -t
- t OAOR OAOR OAOR
dAoR = -!:}-dP04 + -!:}-dP05 + -!:}-dP06' (5.208)
UP04 UP05 UP06
Note that in the three dimensional case we consider the rotations POR to be
applicable only over a small range, so that they can be added vectorially.
Substituting equation (5.208) in equation (5.207) we get
-
oAt
OR -t P ] - d-
dP G = - aOR OP04 aN N a04 P
-t [
-t [
oAtOR -t P ] - d-
- aOR OP05 aN N a05 P (5.209)
-t [
oAtOR -t P ] - d-
- aOR OP06 aN N a06 P
= [a&RkGRaOR] dp = kGdp
5.11 Geometrical stiffness 103
in which
-t
kCR4 = OAOR]
[- - -t
aN PN,
OP04 (12x6)(6xl)
(3x12)
-t
kCR5
OAOR]
= [- - -t
aN PN, (5.211)
OP05 (12x6)(6xl)
(3x12)
-t
kCR6 = OAOR]
[- - -t
aN PN
OP06 (12x6)(6xl)
(3x12)
R =
c xx ,
[ cyx'
Cxy'
Cyy'
c xz']
cyz' =
[1 -P06
P06
1
-P05]
P04 .
Czx' czy' Czz' P05 - P04 1 (5.212)
Let us derive one of these matrix entries. We choose
OX (7r ) .
Cxz = OZ' = cos 2" - P05 = - smp05 = -P05, (5.213)
for P05 small. All other entries are defined in a similar way. We state now
the incremental rigid body rotation matrix
. dP06
-dP05 ]
dR = [ -dp06 dP04 , (5.214)
dP05 -dP04
104 Composite beam element
:
P06
(5.215)
. 1.
with
dE4= [
dR
dR :1= rdR dR dR dR j.
(5.217)
dR
- - t - - - t -
da = adR4, dA=dR4 A, dAoR = dR4 AOR
(5.218)
Differentiating the last expression in (5.218) with respect to the rigid body
modes P04, Pos, P06, we obtain
- t t-
oA OR = A OR R 4,
OP04
- t t-
oA OR = AORRs , (5.219)
opos
- t t-
oA OR = AOR~'
OP06
where
R4=r R 4 R4 ~ R4j,
Rs = r Rs Rs Rs Rsj, (5.220)
R6=r~ ~ ~ ~j.
5.11 Geometrical stiffness 105
kGR
(3x3)
= {5.221}
- [A~R ~ a}y PN
(3"'12) (12'" 12) (12"'12) (12'" 1)
A~R Rs a}y PN
(3"'12) (12"'12) (12:1: 12) (12"'1)
A~R ~ a}y PN]
(3:1:12)(12:1:12)(12:1: 12) (12:1:1)
When all natural forces are included in {5.221} the rotational geomet-
rical stiffness kOR is unsymmetric and the resulting cartesian geometrical
stiffness k also displays asymmetry. The asymmetry can be explained by
the fact that higher order rotation terms are involved which are not commu-
tative. Using more mathematical phraseology, we may say that successive
finite rotations referred to fixed axes introduce path dependent work and
as such they are nonconservative. A first approximation would be to sym-
metrize the geometrical stiffness via
- s 1 - -t
kG = 2[kG + kG] {5.222}
When only the axial natural force FN enters the rotational geometrical
stiffness the latter is always symmetric. It is given by
kGR =
(3x3)
- [A~R R4 a}y FN
(3:1:12) (12:1:12) (12:1:1) (l:t 1)
A~R Rs a}y FN A~R ~ a}y FN] {5.223}
(3:1:12) (12:1: 12) (12:tl) (1", 1) (3"'12) (12'" 12) (12'" 1) (l:t 1)
t -
kG = T04kG T o4. (5.225)
If, however, the rotation is not sufficiently small, nonlinear terms influence
(5.226). We approximate the sin by
3
. PN2 PN2 PN2
s l l l T = T - 48' (5.227)
so that
3 I 2 I 2
PN2 PN2) I PN2 PN2
r ( T - 48 = 2" ~ rpN2 - = rpN2 24 = 24'
(5.228)
This equation expresses the change of the beam length f:11 produced by
the natural mode PN2. Now the axial natural force FN operating on this
displacement produces work
5.11 Geometrical stiffness 107
12
W=FN PN2 {5.230}
24 .
1 FNl 2 FNl 2
W NG = 24 PN2 + 24 PN4' {5.231}
2 FNl 2 FNl 2
W NG = 40 PN3 + 40 PN5' {5.232}
Note that the energy W~G is usually very small and can be neglected. From
{5.231-5.232} we deduce the natural geometrical stiffness via
or explicitly
1
12
1
kNG = FNl 20
1
(12x12) 12 {5.234}
1
20
1
= FNlr 12 20 12 20
~ ~ ~ .J
108 Composite beam element
Alternatively, one may add the natural geometrical stiffness to the simpli-
fied geometrical stiffness and then proceed to global transformation proce-
dures.
P=a~PN' (5.237)
through the small angles PORt:.. We require P Gt:. in terms of PORt:.. The
nodal force increments are
. POM -P05t:.]
Rt:. = [ -POM P04t:. , (5.239)
P05t:. - P04t:.
5.12 Partly simplified geometrical stiffness 109
and
R~
R~
R~
R~e= (5.240)
R~
R~
R~
Noting that
as
P3 -P2
-P3 PI
P2 -H
Pij -P5
-Pij P4
PG~= I
P5 -P4
Pg R
- a I aORP~
(5.243)
-Pg P7
Pa -P7
Pl2 -Pn
-P12 PlO
Pn -PlO
This equation may be written for each natural mode separately by replacing
P r by PNpaNpr in the above matrix. Thus
110 Composite beam element
aNp3 -aNp2
-aNp3 aNpl
aNp2 -aNpl
aNp6 -aNp5
-aNp6 aNp4
-1
1
1 1
I -I
1 1
I -I
kGl = PNl
1 1 (5.246)
-I I
1 1
-I I
1-1
= PN2 (5.247)
A
kG2 aOR
-1
1 1
-I 1
1 1
-2" -2"
kG2 = PN21
(5.248)
1 1
:I
1 -I
1 1
2" 2"
1 1
1 -I
1 1
-2" -2"
I
kG4 = PN4 .
(5.249)
1 1
:I
-I 1
1 1
2" 2"
-1
1
kG6 = PN6 aOR (5.250)
1
-1
Carrying out the multiplication with aOR
1
Y -y1
1
-y1
kG6 = PN6 Y
(5.251)
-y1 1
1 Y 1
-y Y
we observe that the geometrical stiffnesses due to PN2, PN4, PN6 are
unsymmetric. We examine now whether symmetrization is feasible. By
adding the terms
Yl PN2
= +-1-(()2 - )
()l ,
PN2
Y 2 = --1-(()2 - ()l ) ,
(5.252)
PN2 )
Mxl = +-2-('l/J2 - 'l/Jl ,
PN2 )
Mx2 = +-2-('l/J2 - 'l/Jl ,
114 Composite beam element
matrix kC2 becomes symmetrical. The same holds for PN 4. These terms
are proportional to the deformations (fh - (h) and {'l/J2 - 'l/Jd which are
small compared to deformations {V2 - vd, (lh + ( 2) which are proportional
to rigid body movements according to equation (5.20). Likewise, we may
symmetrize kC6 by adding the terms
PN6 )
YI = +-l-{<PI - <P2 ,
PN6
Y2 = --l-{<PI - <P2 ) ,
(5.253)
}JN6 )
ZI = +-l-{'l/JI - 'l/J2 ,
}JN6 )
Z2 = --{'l/JI-'l/J2 .
1
These terms are proportional to the deformations (<PI - <P2) and ('l/JI - 'l/J2)
and are small compared to rigid body deformations -(V2 - VI), (WI - W2).
Therefore, we may use the symmetrical parts of kC2' kC4 and kC6.
The contributions of the antisymmetrical moments PN3, PN4 to the
partly simplified geometrical stiffness are negligible and they will be ig-
nored.
-Ite
p
1 . d A
~
elements dof Kirchhoff shear deformation
1 6 0.0120 0.0120104
2 12 0.0120 0.0120104
~::J
shear correction factor 0.8333
E= 90.109
[ J V =0.3
A m I = 0.30
~ r>J b =h =0.01
Wo
E = 90.109
v= 0.3
I = 0.30
b = h = 0.01
Wo= ION/m
-6
Kirchhoff 2.9883 x 10-5 2.5268 x 10
4 elements
shear def. -5 -6
23 dof 2.9979 x 10 2.5499 x 10
-6
analytic 3.1000 x 10-5 2.5312 x 10
A
P
1
~
1 E= 90.109
v =0.3
I = 0.30
b = h =0.01
a = 1/3
I<l I>j.:::J I>j.:::J [>j
-3
Kirchhoff 1.1111 x 10-3 1.0000 x 10
3 elements
-3
analytic UIII x 10-3 1.0000 x 10
Figure 5.10: Deflection and slope comparison for two isotropic beams
text of Meek [42]. It examines a plane frame under the action of a horizontal
force as shown in Fig. 5.13. For all structural members, values for the
axial forces, moments and shear forces were given in [42]. They are shown
together with the present results in Fig. 5.14. The convention assigned
5.13 Computational experiments 117
E= 90.109
1
w w
C-]
v =0.3
A BCD IE / = 0.20
im ,Id, b = h =0.01
w= 10N/m
a =/12
-5 -6
Kirchhoff 9.0122 x 10 8.3946 x 10
48
shear def. -5 -6
9.0123 x 10 8.4124 x 10
-5 -6
Kirchhoff 8.9331 x 10 8.3553 x 10
84
shear def. -5 -6
8.9335 x 10 8.3730 x 10
-5 -6
analytic 8.8889 x 10 8.3333 x 10
B D
moment diagram
M _ 2
analytic - IV II 12 = - 0.005
Figure 5.11: Deflection, slope and moment comparison with the analytic
solution for an isotropic beam
w= 30kN/m
eI 'I F
h -O.IOm
w=30N/m
0.036 m
a =112
H
DI,,
141<N 141<N
0.4 m
dof deflection at C
strain energies in the individual finite elements. They are all expressed on a
percentage basis. From these bar diagrams one can immediately see how the
various components of the strain energy are allocated during deformation.
Figure 5.16 illustrates a four storey frame which was studied also in [42]
where reference values for the moments were reported (see Fig. 5.16). The
values for the axial, bending and shearing forces as obtained by our beam
element are shown in Fig. 5.17. The moment values agree very well with
the solution of Meek [42]. Figures 5.18, 5.19 display the deformation of
the isotropic frame under the action of the horizontal force and a vertical
5.13 Computational experiments 119
10
21
1 12 elements 1 10
66 degrees of freedom
~ ~
tr
20
4, f4
21 .60 ' 2.14
ft4~~r4
10.91 10.91
t1
0.23
4.93{ f ' 4
0.23
t7
10.45
4.93
10.45
4
I 2.14
fit
21.13
21._~ 21.13
3.51
3.511~
d.
15.84 t
2.14
IS.
211.17
ll4
,----=--:...J -==------------.
Figure 5.13: Isotropic frame under horizontal force; elemental forces and
moments
5.13 Computational experiments 121
I ---.
- 4.93
-
+Qt 5M
----t:>+N
0
+ 2.14 - 2.14
M Q
~t !t-N
,... M
L.........c=
- 21.4
o
reference sol. - 28.50 reference sol. 28.50
-2.14
+5.07
(!] +4.93
Figure 5.14: Reference force and moment diagrams for the isotropic frame
122 Composite beam element
100
energy (%)
80
\0
21 60
10
40
20
k. t>l
o
Elem I Elem2 Elem3 Elem4
energy (%)
100
80
;mlisymmelrical mode
20
o VI ~i
==
~ elem.4 elem.9
80
~ elem.3 elem. \0
60
~ elem.2 elem. II 40
~ ~
-
elem. 12 20
element I
o VL..d ~I
10 kips
32 elements
132 in.
168 dof
1,0 kips ......
132 iii.
10 kips
132 in.
1,0 kips
E= 3,0 x I,oin.- kip
132 in.
1= lOOOin4
f<l .
,olD.
r>j
Ifoundation1,0 insettlement
V W=
reference solution for memeber moments (Meek)
513.6
1
1940.9 1940.9
. ..
- 5.00
3.40 - 3 40
o
+ 10.00 - 5.00 -J( 00
+ 20.13 - 5.00
J
- 2( 13
l
+ 31.00
- 5.00
......
- 31 00
513.7 5137
J
o
1945 1941.5
- 3.42
+ 5.00 + 5.00
~
+ 15.00 - 10.05 + 15.00
+2O.Q3 + 20.03
- 10.93
Figure 5.17: Present force and moment diagrams for the 4 storey isotropic
frame
5.13 Computational experiments 125
...,.--___l
==-=t
I
--'
/
/
1/----~11
I I~ I
Figure 5.18: Displacement of the 4 storey isotropic frame under the action
of a horizontal force
126 Composite beam element
----
--
---
a O. llb
EI - 3.107 psi
2 - 3. 106 psi
G12=G n =Gn = 106 p'i
v I2 =vJ3=vn =0.3
45
_ 45 y
:r _ 45
45
-7 -7
u 0.5113 x 10-7 0.51 \3 x 10 0.5113 x 10 0.5113 x 10-7
-5
v 0.51 \3 x 10-5 0.5113 x 10-5 0.51 \3 x 10 0.51 \3 x 10-5
Figure 5.20: Comparison of the tip displacements for a (45/ -45) s composite
beam with analytical and numerical solutions
restrain the bottom boundary while applying compressive loads at the top.
An eigenvalue analysis is conducted using the simplified geometrical stiff-
ness. The satellite structure bifurcates at a critical load of Per = 19079.6
Pa. The critical buckling mode is shown in Fig. 5.24. This example clearly
illustrates the applicability of our composite beam finite element to the
study of three dimensional beam structures.
128 Composite beam element
l!h l!h
05 0.5
0.25
1
CJ
xx
7
Z4~1 0.25
_ 45
/
~~I / -4~1
0
-0.25
7
45
-0.51/1 1 -0.5
-3 -2 -1 0 2 3 4 -1 -0.5 0 0.5 (pSI.) 1
(psi)
l!h
r-------~:;:>'..::.i 0.5
;/' 10.25
I { I0
""" 1-0.25
1- =::"'" -0.5
-10 -8 -6 -4 -2 0
~ J1~ p
i<' ~ ~
hI~
I" b .. I
- h=O.1
Calcote
Figure 5.22: Critical buckling loads for isotropic and (45/ - 45/0/90)8'
(30/ - 30h composite beam columns
130 Composite beam element
I:)
,.0
-s
.....
"'""
~
@
o
00
Q)
Q)
Q)
~
+
+
Figure 5.24: First buckling mode of the composite satellite at a critical top
compressive load of Per = 19079.6 Pa; the bottom boundary is clamped
132 Composite beam element
5.14 Problems
1. What is a natural rigid body and a natural straining mode ? How do
they relate to the various coordinate systems defined in this chapter?
2. Under which displacement assumptions can rigid body motion occur
in a beam element? Provide an example.
3. Write the expression for the elastic curve due to symmetrical and
antisymmetrical bending modes as a function of the coordinate x,
where x varies from 0 to l. Write also this expression when x varies
from -l/2 to 1/2. Using these coordinates, what would the equation
for the elastic curve be due to the fourth and fifth natural straining
modes?
4. Using elementary beam theory, prove expressions (5.45) and (5.46).
5. Explicitly evaluate the integrals k~l_k~5.
6. A beam element has length 10 cm. Under the action of external
loadings it displaces by Ul = 2, U2 = 3, VI = 3, V2 = 2, WI = 0.2,
W2 = -0.05, (h = 1.5, (h = -2, </>1 = 0.2, </>2 = -0.05, and "pI = 0.5,
"p2 = -0.001 in the local coordinate x'y'z'.
12. Express the equation of the elastic curve as a function of the beam
coordinate x when the latter varies between 0 and t. What is the
strain expression 'Yxx for this case?
13. What is the physical meaning of the strain operator matrix? Define
this matrix for a rod element under the action of the natural mode
PNl
14. Write the strain energy expressions using compact matrix expressions
for the beam element in the natural, local and global cartesian coor-
dinates. Show the dimensions of the matrices used.
16. In which areas do you expect the linear through the thickness tem-
perature variation of (4.171) to hold more accurately?
18. What is the physical meaning of the geometrical forces? Why are
rigid body rotations important in elastic buckling?
19. Follow the methodology of section 5.10.3 and derive an expression for
the through-the-thickness stress (1zz.
21. Starting from the matrix equation A = a-I, provide cross relations
for the submatrices ao, aN, Ao, AN, ao, aN, Ao, AN' Prove that
AORa N = o.
- t -t
22. Show in a small example that successive finite rotations produce path
dependent work.
23. Provide examples in which you expect the partly simplified geomet-
rical stiffness to be more accurate than the simplified geometrical
stiffness.
24. Suppose that a two node beam finite element is under a sinusoidal
load of half period. Provide the kinematically equivalent nodal loads.
Chapter 6
6.1 Introduction
Shells are used in many modern structures because of their efficiency and
economy, their ability to retain their form, and because of other features
stemming from their reaction to certain loads. Their shape allows certain
membrane stress systems to develop parallel to their tangential plane and
become prime carriers of the deformation. Indeed the analysis of many
thin shells is solely based on the membrane theory of shells which neglects
their bending rigidity. On the other hand, bending becomes important in
the presence of rapidly changing loads (e.g. concentrated loads, line loads
etc.), near edge constraints, near discontinuities in the shell geometry and
in nonlinear deformations. The development of a general membrane and
bending theory as well as related numerical implementations are subjects
of intense research efforts which aim at providing deeper understanding of
the mechanics of load carrying. The literature on this subject has been
growing at a rapid pace over the last decades.
When the shell thickness is comparable to its radii of curvature, trans-
verse shear deformation cannot be ignored (especially in multilayer com-
posite shells) and must be treated in the context of a rather general and
rigorous theory. Ultimately, finite element models must be developed for
the analysis of shell structures, whatever their complexity. Furthermore,
the introduction of composite materials and their ever expanding use in
high-performance lightweight structures has called for a unification of con-
cepts involved in the structural analysis of isotropic and anisotropic shells
[55], [56]. These elements were applied to numerous linear and nonlinear
computations of isotropic and composite structures and have shown superb
behavior with respect to both accuracy and economy.
Nevertheless these composite shell elements did not bring perfection to
the theory of natural modes. The method of physical lumping was still
in use and explicit constraints were needed: when angles greater than 90
were involved the elements yielded negative stiffnesses. The authors then
decided to pursue further research to circumvent this and other constraints.
This was followed with diligence and persistence and led to a new theory
with no constraints. With the evolution of new concepts a unification of
the natural mode method was accomplished.
Definition 4 The x'y' z' coordinate system is referred to as the local ele-
mental cartesian coordinate system and is placed at the barycenter of the
triangle.
The z' coordinates of the plies are shown in Fig. 6.2. The cartesian strain
tensor is defined as
138 Composite plate and shell element
composite laminate
r
I ~I .. X'
~ a. x'
L
y'
ry
P~-- -x' r x'
.x
x'
a
~ p
'~l
global coord. local coord.
x'
l:' x
~ x'
k=1
lk 1
k=2
recursion formula zk+1 2
IW2
ZI =+h12 k=1 t h
zk =zk_l- hk k=2 ...... .N+ 1
N
k= N+l
_l(i i) .. _ (6.2)
"tiJ' - -2 u.'3 +u.,
"
~,J - X,Y,Z,
With these strain and stress definitions, a consistent expression for the
strain energy is (assuming O'zz = 0)
140 Composite plate and shell element
1
E: U =-,tu
2
1
=2" (O"XXIXX + O"YYIYY + v'2O"xyv'21XY + v'2O"xzv'21xz + v'2O"yzv'21YZ)
1 1
=2" (O"XXIXX + O"YYIYY + O"Xyv'2v'22" (u,y + v,x)
1 1
+ O"xzv'2v'22" (u,z + w,x) + O"yzv'2v'22" (v,z + w,y))
(6.4)
1 e
=2"(O"XXIXX + O"YYIYY + O"xy Ixy
~
engineering shear strain
e
+ O"xz I~Z +O"yz Iyz ).
~ ~
engineering shear strain engineering shear strain
The three direct total strains will be measured parallel to the edges of
the triangular element. We measure these strains if we place a triangular
strain rossete on an aircraft panel (see Fig. 6.3) and measure three strains
, namely Ita, It(3 and It'Y along the rossete edges. These direct total strains
implicitly contain all the plane cartesian strains.
Remark 1 The total natural strains are equivalent to the three in-plane
cartesian strains in the sense that they contain the same information.
'ta]
It = [It(3 , (6.5)
It'Y
IXX ]
I' = [ 'Vy , bzz = 0). (6.6)
V2,xY
6.2 Natural kinematics of the shell element TRIC 141
o
strain rosette
"fry
"fla
++
++
The concept of total strain allows us to use the triangle (alternatively the
tetrahedron in three dimensions) as the fundamental block on which equi-
librium can be prescribed (see Fig. 6.4).
Turn to Fig. 6.4. Shown is a square element S in a state of plain strain
i.e. TZZ = TXZ = TYZ = O. During the course of an elastic deformation,
the element strain state can be fully described by the three cartesian strain
measures, namely TXX, TYY and TXY If we ignore any rigid body motion
occurring during deformation, the square element deforms into a parallelo-
gram with sides of length ~s(l + TXX) and ~s(l + TYY), respectively. Our
objective is to derive an expression for the strain Tt(a) along an arbitrary
direction forming an angle a with axis x. For this purpose we consider the
right triangle t::.ABC and the oblique triangle t::.A' B'G' into which the first
is deformed. Applying the law of cosines to t::.A' B'G'
Y&[J
oL
As
6= 90-2yxy
Yxy =112 (6 1+6 2 )
x
B B'
oLA x
lu C A' Ax (1+ Yxx)
90 + 2 Yxy
l)
E
~
~s2{1 + ')'t{a))2 = ~s2 cos2 a{l + ')'xx)2 + ~s2 sin2 a{l + ')'yy)2
- 2~s2 sina cos a{l + ')'xx)(l + ')'yy)cos{ i+2')'xY~6.11)
[ [
It = Bt,' ~ rt(31 C~""
rta
C2
a",'
s2
a y'
S~y, V2s (3"" c(3y'
1
V2 s a '" ,Ca Y ,] ['"II/xx
ryy .
rt"( c"("" s"(y' V2s"(",,c'Yy' V2r xy (6.19)
The angles ax" a y', etc. are shown in Fig.6.1. As long as one of the
triangular angles is not zero, equation (6.19) can be inverted to read
It = B t r , => I '=
B - t It, (6.20)
B- t = [Bttl. (6.21)
6.2 Natural kinematics of the shell element TRIC 145
Remark 2 Three total natural strains suffice for the definition of the in-
plane cartesian strain field.
ca
Uc = [uu c{3
] (6.22)
Uc-y
Uxx ]
u' = [ U yy . (6.23)
../2uxy
The square root is introduced here in order to define the shear stress using
tensorial notation [51]. Using this notation, we can write the strain energy
expression for plane stress as
p
x
p
0.
I
0.
n'y
r=rc~+ 'q + rca r="p + '/"(+ ',a
r=,x+ ry
,x
0. p p
0.
2
Ym =l!Jalla
l!Jp= l!J, =0
Po.
CJ,u
Ik~I-L..J....-.......- I
'!-It.....J'-.1-L...L...J.wl aen
CJcb
Po.
Po.
P,
Po.
Pp
We express the equivalence of the strain energy density in the natural and
local cartesian coordinates, viz.
where e is an energy operator symbol. Note that there are three pairs of
stresses and strains to operate with, these being, U Clt , Utlc' U I , where
I c are the component strains which we did not apply here. As a result, the
expression for the strain energy may be expressed in the alternative forms
1 tIt 1 t
e: 21tUc = 2/cUt = 21 u. (6.26)
U e = B -1 U , {:::::::} U' =
B U c. (6.27)
To this end we regroup the derived expressions for strains and stresses as
follows:
/~
'Yqa b /. -' '
\ G
tr/, ,tt,
qa ..........
Qa
'.2
q~
Qr
Itt[;~=fJ
a,
USOl.]
[Us /3 COl. X'
[ c/3x'
SOI.X']
s/3x'
[UXZ] {:} Us
u yz k
= Tsu~. (6.30)
U s'"( k Cux' Sux' k
The two previous equations which are derived from pure geometrical argu-
ments can be grouped together as follows:
[~~l ~ [Bt B- 1
Ts
J[~l (6.32)
In the natural coordinate, the constitutive relation for each layer (k) reads
in which Xij are the transverse shear stiffness coefficients. Using compact
notation we write
We observe from equations (6.33-6.34) that the axial stresses remain un-
coupled from the transverse shearing strains.
With respect to the material coordinate system, the constitutive rela-
tions read
E1 E2
1- V12V21 V 12 1 _ V12 V 21
[
0"11
0"22
1 - E2 E2
[
1
1'I'll
22 ,
V20"12 k
V12 1- V12 V 21 1- V12V21 V21'12 k (6.35)
2G12.J k
[V20"13]
V20"23 k =
[2G 13
. 2G 23
.] [V21'13]
k V21'23 k
(6.36)
V12 V21
E1 = E2
E2 E1
(6.38)
V12
(1 - V12v21)
= V21 (1 - V12V21
)'
El V12E2
1 V12 E 2 E2
~12 = 2 E2
1- v 12 El (6.39)
2 E2)
2(i12(1 - V 12 El
1
~12 = - - vE
[E vE
E
1- v 2 2G(1:- V2J
E E
v-- (6.40)
1- v2 I- v2
E E
v--
1- v 2 I- v2
2(i
c: 21'12t 0"12 = 1 It ,
-"V 1
2 ,t
t
(6.41)
------- -----
2 0"
I O"c
~
material coord. local coord. natural coord.
The following relation holds between strains and stresses, in the material
and local coordinate systems [51]
6.3 Constitutive relation 153
['12]
IS
[A. As.] ["]
IS
(6.42)
[0"12]
O"s
= [~12
.
.] ['IS12] = [~12A.
GS
.] [,:]
GsAs IS
.
(6.43)
[u'v' ] - [-sinO
cos 0 0] [u]v
sin
cosO -
[C(J S(J]
-S(J C(J
[u]v ' (6.44)
Similarly,
I
which implies
-ax -ay
ax' ax '
[ ax ay = [~!(J ~:]. (6.47)
ay' ay'
Using the chain rule of differentiation we have
154 Composite plate and shell element
a]
ax' = [aax ax'
ax + oya ax'
oy]
[
-
a -
a -ax+ a-oy -
. oy' ax oy' oy oy'
:][:]
ax
[ (6.48)
= ::
oy' oy' oy
= [co so]
-so Co
[:x].
a
oy
Now the strain expressions in the local coordinates read
au' = -(cou
'YX'X' = -
a + sov)
ax' ax'
= (co! +so~)(cou+sov)
20U 20V (OV au (6.49)
= co- oy + coso -ax +-
ax +so- oy
2"xy
(6.50)
2"xy
We can replace indices xyz with 123 in equations (6.49-6.51) and obtain the
local cartesian strains from the material strains 'Yn, 'Y22 and 'Y12. Matrices
A, As are easily deduced as
c2
o s~ V2S0CO]
A= [
So2
c~ -V2s oco
-V2soco V2soco (c~ - s~)
(6.52)
Co so]
[-so
As = Co
Using A, As, we find the expression for the strain energy in th~ material
coordinates 123
156 Composite plate and shell element
"
c- :
1 t
2'120'12 2' It[At K12,
{6.42},{6.35} 1
=
A]'
1 't
= -,
2
[K'lT'
(6.53)
= ![Bt,,]tKct[Bt,']
2
= ~,/t[BKctBtlT"
Thus, for each layer (k)
Kk = [AtKI2A]k = [BKctBt]k
(6.54)
K~t = [B-l[AtKI2A]B-tL
1 t 1 't ,
& : 2'sO's = 2'sO's
1 't [ t ] ,
= 2's AsGsAs,s (6.55)
1 't [ '] ,
= 2's Gs 'S,
where
G's = [2Gxx
2G xy
2Gxy]
2G yy . (6.57)
The next step is to compute the natural transverse shear stiffness. Ob-
serving (6.29) and aiming at solving for the transverse shear stresses as
functions of the natural shearing strains we realize that we have at our
6.3 Constitutive relation 157
= cJ.lXI'Yxz + cJ.lyl'Yyz,
'YJ.I
(6.58)
'Yv = CVX''Yxz + cvy''Yyz, jL, v = Ct', {3, 'Y.
Solving the above system of equations for 'Yxz, 'Yyz we find
, _ ['Yxz] -_ ~ [ l"(yo:
IS- -lo:Y"(] ['Y"(] = W "(0:1"(0:'
'Yyz 20 -l"(xo: lo:x"( 'Yo:
We use the principle of virtual work in order to generate the natural trans-
verse shear stiffnesses. We have
8Us = ! u 'st , dV =
8,s ! 't,
IS , dV.
G s 8,s (6.62)
v v
On account of equation equation (6.61) we may write equation (6.62) as
158 Composite plate and shell element
Proceeding in a similar manner for the other two shear strains we obtain
the following three expressions for the transverse shear stiffnesses:
Xll X12]
Xl = W~.BGsWQ.B = [ b ~2'
Xl Xl
Xll X12]
X2 = W~'YGs W.B'Y = [ I2
X2
~2'
X2
(6.64)
X3 = W~QGsW'YQ = [ Xll
~2
X12]
~2
X3 X3
Both axial and transverse shear stiffnesses K,ct, Xs have now been defined
in the natural coordinate system.
= n!
h/2
Pa = 1
2CJca h ah la k
CJcadz,
-h/2
1.
P {3 = 2CJc{3h{3h = l{3n! h/2
CJck{3dz, (6.66)
-h/2
P -y = 21 CJe-yh-yh
k n!
= l-y
h/2
k
CJc-ydz,
-h/2
1
MSa = 2mSa h a h = la n! h/2
k
zCJcadz,
-h/2
MS{3 1
= 2ms{3h{3h = l{3 n! h/2
k
zCJc{3dz, (6.67)
-h/2
Ms-y 1
= 2ms-yh-yh = l-y n! h/2
k
zCJe-ydz.
-h/2
where mSa, mS{3 and ms-y, are the moments per unit length. The nodal
force and moment resultants are shown in Fig. 6.8.
po. W ,p.
MsfI.
( 2
)Sa
M
cb
MS')'~
Py
N"
MS'( Q.,
M"~ ND
... ( :::4'1
pi
M
< IN"
i4N
"( M:
"
a: M"
\'..i,/
:I
N"
p.4 M((lt?>2?!))
.10. MStl M
. p
MStl A"
Pa:::: P a
P,8::::P,8
P')'::::P')'
MSa:::: MSa
Ms,8:::: Ms,8
Ms')' ::::Ms')'
(6.68)
Qa- 2MAa =0
la
Q,8 - 2MA,8 = 0
l,8
Q,),_ 2MA')' =0
l')'
Ma +M,8 +M')' =0
PN ={PNa PN,8 PN,), MSa MAa Ms,8 MA,8 Ms')' MA,), Ma M,8 M')'}.
(12xl) , " (6 6~)
generalized natural forces .
PN = [PN']
PN,8
PN,),
= r. l
P,8l,8.]
P')'l')'
Fe = [ N
xx
Nyy ] ,
../2Nxy
(6.70)
Ms =
[MS,]
Ms,8 Me= [ M .. ] ,
Myy (6.71)
Ms')' ../2Mxy
QN ~ [~~] Q e-
_ [Qxz]
Qyz . (6.72)
6.5 Natural modes and stiffness 163
ZI', WI
.'v'
~
<2', .W2'
'V'
'V '2
loW
x', u'
0'
'V
y. v
x, u 0'3
o
Figure 6.10: Multilayer triangular shell element with local cartesian free-
doms; local and global coordinates
The relation between the natural and cartesian stress resultants reads
, PO]
(6xl)
(l~:l) = [ PN
(6.75)
.
(12xl)
Po, PN represent the rigid body and straining modes, respectively, with
corresponding entries
PN = h~ 'Y$ 'Y~ 'l/Jso: 'l/JAo: 'l/Js/3 'l/JA/3 'l/Js'Y 'l/JA'Y 'l/Jo: 'I/J/3 'I/J'Y}. (6.76)
(12xl)
l::
a1= 112 ('!'Sa + '!'Aa)
symmetric mode
aI s =1I2'!'sa
1I2Msa
antisymmetric mode
aI A =1I2'!'Aa
1I2MAa
PN = o,NP (6.89)
p = T 06 P, (6.90)
Wi = '2
1 (,2
-t + Yi )'l/JSi 1(
+ '2 Yi -
3,2 + 2,3)
z:- II 'l/JAi,
, ~ , y ,
_ .8Wi
Ui = Uio -z 8Yi (6.93)
8Ui 8uo 82 w i
'Yti = 8Yi = 8Yi - z 8Yl
o 'l/JSi
='Yi +zT+ z
(3Z;- q
6Yi) .
'l/JA~
(6.94)
The terms coupling the total axial strain with the antisymmetrical bending
mode are very small and will be neglected. With this assumption, equation
(6.94) yields
168 Composite plate and shell element
'1"
z'
z
x'
z' z'
@
~
[9 z'
o
Wo
Wo
x'
~ ~/;
)}1/fJ ....{'J' 0')
I ==- I
1I21,u
11211'(
12V~
,,
IV;yl
,,
112V;u
Ita = I~ + z 7/JSa
la
It(J = 1$ + z 7/JS(J {6.95}
l(J
It-y = I~ + z 7/JS-y
l-y
We shall need relations connecting the total strains ft' the natural dis-
placement vector PN, the local elemental cartesian vector p, and the global
cartesian vector p. They are
(6.91) _ _ (6.91) _
ft = ONPN = ONaNP = oNaNT o6 P {6.96}
z
1 10
= [. z
J
ON 1 1{3
(3xI2) . 1 z {6.97}
r:;
8U = ! 0'~8f
v
tdV. {6.98}
fJU = ! t
u cfJft dV =
(6.34)! t
ft""ctfJftdV
v v
V
natural stiffness matrix
from which we deduce an expression for the natural stiffness matrix con-
taining contributions from the axial and symmetrical bending modes
kN([~' 1/JS) = !
V
Ot.hr""ctOt.NdV (6.101)
We now transform the natural stiffness matrix first to the local coordinate
and then to the global coordinate systems. Substituting equation (6.91)
in equation (6.100) we generate the following sequence of congruent oper-
ations: :
/jU = f
P'" [T~ [a'" [ a'" .."aNdV JaN 1T06] PN (6.102)
v
natural coord. (12x12)
v
local coord. (18x18)
, '
global coord. (18x18)
kN =0 (6.103)
(12x12)
-kQQ kQP kQ'Y l:kQQ r kQf3 ~kQ'Y
f3 1"1
kN =0 (6.104)
(12x12)
- kQQ
kQf3 kQ'Y
kf3f3 kf3'Y
k'Y'Y
%2 %2 k %2
tr01 kQQ IQlf3 Qf3 IQI'YkQ'Y
kQQ k Qf3 kQ'Y
A A A
%2 %2
tr kf3f3 1f3 1'Y kf3'Y
f3
k f3f3 k f3 'Y
A A
%2
symm. trk'Y'Y
"I
k'Y'Y
A
k';Q k,;f3 k';'Y
k~f3 k~'Y
k'p
For isotropic materials only 18 entries need to be computed since the coef-
ficients corresponding to the azimuth rotations are arbitrarily assigned.
We now define the entries referring to the axial and symmetrical bend-
ing terms. Piecewise summation is used to account for the fact that the
6.7 Axial and symmetrical bending stiffness terms 173
J
h
2" N
kaa = fi.aadz = L fi.~a(Zk - Zk+1),
h k=l
-2"
J
h
2" N
ka(3 = fi.a(3dz = L fi.~(3(Zk - Zk+l),
h k=l
-2"
h
J
2" N
ka"{ = fi.a"{dz =L fi.~"{(Zk - Zk+1),
h k=l
-2"
(6.105)
J
h
2" N
J
h
2" N
k(3"{ = fi.(3"{dz = L fi.~"{(Zk - Zk+1),
h k=l
-2"
h
J
2" N
k"{"{ = fi."{"{dz =L fi.~"{(Zk - Zk+1),
h k=l
-2"
J
2" N
zkaa = Zfi.aadZ = "2 L....t fi.aa(Zk2 -
1", k 2)
Zk+l ,
h k=l
-2"
J
h
2" N
zka(3 = Zfi.a(3dz = ~ L fi.~(3(Zf - Zf+1)' (6.106)
h k=l
-2"
h
J
2" N
1", k
zka"{ = Zfi.a"{dz = "2 L....t fi.a,,{(Zk2 - 2)
Zk+1 ,
h k=l
-2"
174 Composite plate and shell element
J
"2 N
1~
zkf3f3 = ZK.f3f3dz = 2 L- K.f3f3(Zk
k 2
-
2
zk+l)'
h k=l
-"2
h
J
"2 N
1~
zkf3'Y = ZK.f3'Y dz = 2 L- K.f3'Y(Zk
k 2
-
2)
Zk+1 , (6.107)
h k=l
-"2
h
J
"2 N
1~
zk'Y'Y = ZK.'Y'Y dz = 2 L- K.'Y'Y(Zk
k 2
-
2)
Zk+1 ,
h k=l
-"2
J
h
"2 N
z 2kaa = z 2K.aadz = ~ ~ K.~a(z~ - Z~+1)'
h k=l
-"2
J
h
"2 N
z 2kaf3 = Z2K.af3 dz = ~ ~ K.~f3(z~ - z~+d, (6.108)
h k=l
-"2
h
J
"2 N
z2 ka'Y = z2 K.a'Ydz = ~ ~ K.~'Y (z~ - z~+1)'
h k=l
-"2
J
h
"2 N
z 2kf3f3 = z 2K.f3f3dz = ~ ~ K.~f3(z~ - z~+1)'
h k=l
-"2
J
h
"2 N
J
"2 N
z2k'Y'Y = Z2K.'Y'Y dz = ~ ~ K.~'Y(z~ - z~+1).
h k=l
-"2
6.8 Antisymmetrical bending and shearing stiffness terms 175
Now all terms due to axial and symmetrical bending modes have been
defined.
MA = KA1/JA' (6.111)
In matrix form
Kll
A
K12
A
K13
A
[MAO]
MA/3
K12
A
K22
A
K23
A
[~AO]
'l/JA/3 . (6.112)
MA'Y 'l/JA'Y
K13 K23 K33
A A A
MA = K~1/J~, (6.113)
MA = K~1/J~. (6.114)
~/.b
'f'A b ]-1 MA,
= [KA
(6.115)
1/J~ = [K~rlMA.
Substituting equation (6.115) in equation (6.110) we get
1
WAa = "2la({3(,Y(({3 - (-y)'l/JAa
1
WA{3 = "2l{3(-y(a((-y - (a)'l/JA{3 (6.118)
1
WA-y = "2l-y(a({3((a - ({3)'l/JA-y
In the process of deriving the elastic stiffness matrix, the following deriva-
tive matrix is needed:
8WA
8t;
= I8WAP
8(a
8WAP
8(p
8WAP
8(-y
8e
8."
= I8(p
8"'a
8(p
8",p
8(p
8",-y = [~l 1 -1]
-1
1
.
. (6.121)
18wAa 18wAa
uaa=-zla 8"'0 u a(3 = -z 1(3 8"'(3 ua'Y = -z ~
1'Y 8WAa
8.,,'Y l6.125)
1 8WA(3 1 8WA(3 ~ 8WA(3 l6.126}
u(3a = -z 10 8"'0 u(3(3 = -z 1(3 8"'(3 u(3'Y = - Z 1'Y 8.,,'Y
1 8WA'Y 1 8WA'Y ~ 8WA'Y .{6.127}
u'Ya = -z 10 8"'0 u'Y(3=-zl(3 8"'(3 un = -z 1'Y 8.,,'Y
One more differentiation yields the expressions for the total strains, namely
b _ 1 8u aa 1 8u(3a 1 8u'Ya
Ita - -zf2a;;- - zf2a;;- - zf2a;;-
a '/0 a '/0 a '/0
b 1 8u a(3 1 8u(3(3 1 8u'Y(3
It(3 = -z 1~ 8"'(3 - z 1~ 8"'(3 - z 1~ 8"'(3
{6,128}
b 1 8u a'Y 1 8u(3'Y 1 8u n
It'Y = -z [2 a;;-
'Y '/'Y
- z 12'Y a;;-
'/'Y
- z 12'Y a;;-
'/'Y
We define now the above terms, Using the chain rule of differentiation we
write
u aa = - z1 -8w-Aa 1 [1 1 2
- = -z-l -2la({3 + "2la('Y -
2 1 r r ] .J,b
2 a'>{3'>'Y 'f'Aa'
la 8"'a a
u a{3 = - z1- 8w-Aa- = -z- [1
1 --la({32 + la({3('Y ] .J,b
'f'Aa'
l{3 8"'{3 l{3 2
(6.132)
1 8w Aa 1 [1 2
ua'Y = -z--8-- = -zZ -"2la ('Y + la({3('Y 'f'Aa'
] .J,b
l'Y "''Y 'Y
u{3a = - z1-8w-A{3 1 [1
- = -z- -"2l{3(a2 + 1{3'>'Y,>a
r r ] .J,b
'f'A{3'
la 8"'a la
u{3{3 = - z1 -8w-A{3
- = -z-1 [1-l{3(a2 1 2 - 2l{3,>a,>'Y
+ "2l{3('Y r r ] .J,b
'f'A{3'
l{3 8"'a l{3 2
(6.133)
u{3'Y = - z1- 8w-A{3
- = -z-1 [1
-"2l{3('Y2 + 1{3'>'Y,>a
r r ] .J,b
'f'A{3'
l'Y 8",'Y l'Y
U'Ya = - z1-8WA'Y
- - = -z- 1 [1
--l'Y({32 + l'Y'>'Y,>a
r r ] .J,b
'f'A'Y'
la 8"'a la 2
u'Y{3 = - z1-8w -A'Y
- = -z-1 [--l'Y({3
1 2 + l'Y,>a,>{3
r r ] .J,b
'f'A'Y'
l{3 8"'{3 l{3 2
(6.134)
U'Y'Y = 1 8w A 'Y
_Z _ _ _ = -z-1 [1-l'Y({32 + "2l'Y(a
1 2 - 21'Y'>{3,>a
r r ] .J,b
'f'A'Y.
l'Y 8",'Y l'Y 2
180 Composite plate and shell element
=- 1 [
z 1~ b + 31'Y(({J - (o.)"p A'Y
bAo. - ('Y1{J"p A{J
('Y10."p b ].
6.8 Antisymmetrical bending and shearing stiffness terms 181
The matrix connecting the total strains and the antisymmetrical bending
modes is
Itb = CtNb'YA,
"I.b
~ ~1= fl-
l ..
1 1 I-I 1-IJ (6.139)
1= [ : a fJ 7'
and,
8U = ! 0'~8,~dV ! b~tKct8b~]dV
v
=
v
[.p~]t [{ f
h
=
__ n
1[.p~],
[z'IAhr'""I-' Ahl] dndz (6.141)
2
v
natural antisymmetric bending stiffness
from which we deduce the stiffness terms due to the antisymmetrical bend-
ing mode, viz.
182 Composite plate and shell element
h
2"
K~ = / / [z2IAhl-2Kctr2Ahl] dndz
n _l!
2
All terms included in K~ are integrated explicitly using the symbolic sys-
tem MACSYMA [57] with the aid of the integration formula
1/ 2!p!q!r!
0, (~($(~dn= (2+p+q+r)!" (6.143)
n
h h h
2 2" 2 2" 2 2"
K A11 = OlQ
6[4
/ z 2 1'i.'Y'Y dz - 6[2[2
OlQ / z 2 1'i.{3'Y dz + OlQ
6[4
/ z 2 1'i.{3{3dz
'Y h {3'Y h {3 h
-2" -2" -2"
h h h
0, /2" 0, /2" 30, /2" (6.~44)
+~2 ~~~+~2 ~~~+~ ~~~
'Yh {3h h
-2" -2" -2"
6.8 Antisymmetrical bending and shearing stiffness terms 183
h h h
2 2 2 2
22
~=~ 0,[{3 / 2 0 / z~~+~
z~~+m2 2 30 / z~~
2
h h h
2 2 2 2
33 30 / Z~~+~2
~=+~2 2 0 / z~~+~
2 O,[-y / 2
z~~
-Yh {3h {3h
-2 -2 -2
h h h
o /2 0,[2 /2 0,[2 /2 (6.146)
+ 2[2 z2/'i,a-ydz - 6l 2 ;2 z2/'i,a{3dz + 6l J Z2/'i,aa dz
a h a{3 h a h
-2 -2 -2
!
h h
2 2
O,[a l {3!
K A12 =- ~
2
z /'i,-y-ydz -
O,[a
6l{3[2
2
z /'i,{3-ydz
-y h -y h
-2 -2
h h
2 2 (6.147)
0,[{3 /
- 6l l2
2
Z /'i,a-ydz -
50 /
6fT 2
z /'i,a{3dz
a-y a{3
h h
-2 -2
184 Composite plate and shell element
J J
h h
2 2
13
K A = - 6l 2 l
nIa 2
z K,/1,,"(dz -
nIcJy
"6l4 2
z K,/1/1dz
/1""( h /1 h
-2 -2
J J
h h
2 2 (6.148)
50 2 nI""( 2
- 6la l""( Z K,a""(dz - 6lal~ Z K,a/1dz
h h
-2 -2
J J
h h
2 2
K A23 =- 501
-6l z 2 K,a""(dz - nI/1
[2"l z 2 K,a""(dz
/1""( 6 a , "(
h h
-2 -2
J J
h h
2 2 (6.149)
- 6lnI""(
2l
2 nI/1 l""(
z K,a/1dz - """"6z4 2
z K,aa dZ
a/1 a
h h
-2 -2
h
2" N
/ Z
2
/'i,OI.OI.dz = 'l~k{3
3 .l....J /'i,0I.0I. Zk 3)
- Zk+l ,
h k=l
-2"
h
2" N
/ Z
2
/'i,{3{3dz = 'l~k{3
3 .l....J /'i,{3{3 Zk 3)
- Zk+1 ,
h k=l
-2"
h
2" N
/ Z
2
/'i,n dz = 'l~k{3
3 .l....J /'i,'Y'Y Zk 3)
- Zk+1 ,
h k=l
-2"
h
(6.150)
2" N
/ Z
2
/'i,OI.{3dz = 'l~k{3
3 .l....J /'i,0I.{3 Zk - 3)
Zk+l ,
h k=l
-2"
h
2" N
/ Z
2
/'i,OI.-ydz = 'l~k{3
3 .l....J /'i,0I.-y Zk 3)
- Zk+1 '
h k=l
-2"
h
2" N
/ Z
2
/'i,{3-ydz = 'l~k{3
3 .l....J /'i,{3-y Zk 3)
- Zk+1 ,
h k=l
-2"
Kll K12]
Kl = i[W~{3G'sW OI.{3]dV = [K[2 K?2 ,
K3
Kll
= i[W~OI.G'sW'YOI.ldV = [K12
K12]
K~2 .
186 Composite plate and shell element
1
Kll1
+ K22
3
K121 K12
3 ]
Ks-
- - [' K12
1 KF +Kr2 K12
3 K12
3
K122 KJ1 ~ Ki 2
Kll K12
s s K;3] (6.152)
= K12
[
s
K22
s
K23
s .
K13
s
K23
2
K33
s
-21
h
(~.153)
Gxydz(x,y, + Xpyp) 1
-2
-21
h
(~.154)
Gxydz(xaYa + X,y,) 1
-2
6.8 Antisymmetrical bending and shearing stiffness terms 187
-2{ (~.155)
h
G""dz(xm +X.Y.) 1
-2
-1~nlaZ.B [ J J
~ ~
K;2 = (YaYp) Gxxdz + (XaXp) Gyydz
h h
-2" -2
h
6.156}
- (x.YP + xpV.) { Gxvdz 1
-2
-1~nlal" [
~ ~
K;3 = (YaY,,) / Gxxdz + (XaX,,) / Gyydz
h h
-2 -2
h I'
1(6.157)
- (x.Y, + x,V.) { Gx.dz 1
-2"
188 Composite plate and shell element
= -1~nl{3l'Y [ ! !
~ ~
K;3 (Y{3Y'Y) Gxxdz + (x{3x'Y) Gyydz
h h
-"2 -"2
h
(6.158)
- (XPY7 + X7YP) { G",dz 1
-"2
where
!
h
"2 N
Gxxdz = L G~x(Zk - Zk+1), etc. (6.159)
h k=l
-"2
For a right triangle with la = l{3 = 1 and isotropic material of shear modulus
G, thickness h, and area A, the shear stiffness K s simplifies to
K _
s -
Gh
12A
[3 . 3
v'v'22] . (6.160)
V2V24
The expression for the transverse shear strain energy is
1 t
Us = 2'sKs's (6.161)
1 s
Is = 2'I/JA- (6.162)
which provides us with a new expression for the transverse shearing matrix,
namely
.ii"n [(12
= [K12 [(13]
- 1
Ks = -Ks
4
[(22 .ii"23 . (6.164)
[(13 [(23 K33
R2
Us = !
h/2 2 )
(6.165)
!
Sci = U h/2 9 (z dz
G(z)dz G(z)
-h/2 -h/2
In order to proceed with the computation of the three shear correction fac-
tors, the coordinates of the neutral planes along the three natural directions
o!, /3, "I must be defined. They are
h/2
[! K~a(Z _ zoa)2dZ] 2
Aa = ___-_h..:.../_2- ; : J . ; - - - - - - -
! K~a(Z
h/2
(6.167)
h/2 [- - Zoa)2dZf
Ga !
-h/2
-h/2
Gk
a
dz
where
and
! G~dz = L G~hk.
h/2 N
Ga = (6.169)
-h/2 k=l
The shear corrections factors for the other two natural directions are ob-
tained by replacing a with {3, 'Y. The new corrected transverse shear stiff-
ness is
6.10 Simulative azimuth stiffnesses 191
kz 1 / z 2Yi.OlOldz,
= n max [[2 1/2
l2 1 / z 2Yi."dz ] .
z Yi.{3{3, dz, l2
01 {3 , (6.172)
Note that in most numerical tests performed this somewhat arbitrary stiff-
ness posed no problem; in most cases fast convergence was obtained. When
much higher modeling accuracy is required, the useful concept of "vertex
rotation" (or drilling freedoms) can be adopted following Allman [58], [59],
[60].
localsiff. k
(18x18)
global stiff. K E
(18x18)
'Yto: = Xo:
p(u3 - U2)
Yo:
+ p(V3 - V2)
0: 0:
Xf3 Yf3
'Ytf3 = r(UI - U3) + r(VI - V3) (6.175)
f3 f3
X-y Yf3
'Yt-y = p(U2 - ud + r(V2 - vd
-y -y
Xo: Yo:
"pso: = -I (fh - ( 3) + p(<P2 - <P3)
0: 0:
Xf3 Yf3
"pSf3 = -1 (03 - ( 1) + r(<P3 - <PI) (6.176)
f3 {3
X-y (
"ps-y = -1 01 - O2) + r(<Pl
Yf3
- <P2)
-y -y
6.11 Local and global cartesian stiffnesses 193
-11 -12
aN aN aN
aN =
[
(6x6) (6x6)
_
13
(6x6)
-23
1 .
(6.179)
(12x18) a21 a22
N N aN
(6x6) (6x6) (6x6)
~
l{3
?1{3
_'!IT
-11 _
aN -
(6x6)
I_~
P
'1
rr'1
(6.180)
_'MP.. ~
l{3 1{3
194 Composite plate and shell element
-f-f
I", -~
I",
~
1"'( t
1"'(
-12
aN = lh..
I",
-~
I",
.1 , (6.181)
(6x6)
2 lh.. -~
I", I", I",
f-f
I", ~
I",
-~
1{3
-l-
1{3
-13 _ _lh..
aN - ~ . I , (6.182)
I", I",
(6x6)
2 lh.. -~
-I", I", I",
1I.i! -~
1{3 1{3
2 1I.i! -~
-lfj 1{3 1{3
1b:. -~
1"'( 1"'(
2 1b:. -~
-21 r:; 1"'( 1"'(
I~
aN (6.183)
(6x6) lh..
4!l 4!l 1
~ lh..
4!l 4!l
~ lh..
4!l 4!l
6.11 Local and global cartesian stifInesses 195
(6.184)
(6.185)
(6.186)
t -
KE = T06KTo6. (6.187)
with
196 Composite plate and shell element
In equation (6.189), Cx'x etc., denote the cosine ofthe angle formed between
the local cartesian axis x' and global cartesian axis x. The same convention
is adopted for all other entries.
oU=oV, (6.190)
oV = I
S
pto'U dV + F~ori + M~o8i' (6.191)
in which P are the applied distributed surface loads and F, M are applied
concentrated forces and moments, respectively. Then, at a point within the
triangle the displacement 'U is interpolated from the nodal values via
(6x18)
w = [(016 (/316 (,y I 6]' (6.194)
6.12 Kinematically equivalent nodal loads 197
p = W PSI. (6.196)
(6xl) (6x18)(18xl)
Substituting equations (6.193) and (6.196) in equation (6.191) we obtain
pI
(lx18)
= ptSI IwtwdS + F~z + M~.z (6.198)
S
The integration is carried out explicitly to yield
pI = P~I 12
(lx18)
n [216 16
16 216 16 +F~+M~
16 16 216
h] (6.199)
R = Tb6 Pl . (6.200)
Following assembly of both elastic stiffnesses and applied loads, the matrix
global equilibrium is
Kr=R. (6.201)
The solution of this system of linear equations provides us with the global
nodal unknowns.
198 Composite plate and shell element
The parameters To, TI in (6.202) are related to the temperatures at the top
and bottom surfaces of a laminate T t , n, respectively, via
1 1
To = 2(Tt + n), TI = X(Tt - n), (6.203)
1 3 1 3
Tt=a LTt , n=a LT:. (6.204)
n=I n=I
The thermal load vector will be computed first in the natural coordinate
and then transformed to the local and global coordinate systems. The
natural thermal load vector J N is written as the sum of the two vectors
JN = J NO + J NI, (6.206)
(12xI)
where J NO, J NI, are the thermal loads due to uniform and linear through
the thickness temperature distributions, respectively. To this end we define
for every layer the thermoelastic coefficients in the three coordinate sys-
tems, namely the material, the local cartesian and the natural coordinate
as
6.13 Initial load due to temperature 199
Ctt
123 _
-
[atll
.
.] ,
at22 Ct~ = [
aXX
a yy
V2a xy
]
, Ctt = l
ata]
at{3 .
at'Y (6.207)
In (6.207), atll' at22' are the coefficients of thermal expansion along and
perpendicular to the fiber direction, respectively. These coefficients are
transformed (for every layer k) to the local coordinates using matrix As
{see (6.52)), viz.
The entries of Ctt are obtained from Ct~ using matrix B defined in (6.19) as
follows:
[ca""
l ata]
at{3
at'Y
C{3""
C'Y""
Say'
S(3y'
S'Yy'
V2 S a",'Ca y ,] [ a xx ]
V2S{3""C{3y'
V2s'Y",' C-yy'
a yy .
V2a xy
(6.209)
Ctt = B t Ct t, (6.21O)
The total natural strain is the sum of the elastic and thermal strains
Substituting equations (6.211) and (6.212) into the principle of virtual work
we obtain
200 Composite plate and shell element
8U = ! (T~81'tdV = ! b~ - '1~]
V V
K.ct 81't dV
J}y = ! '1~K.ctoNdV.
v
(6.214)
J tN -Jt Jt
NO + Nl
! [ToO~K.ct]ONdV + ! [ZTIO~K.ct]ONdV
-
=
(6.216)
v v
We now define all integrals included in the natural thermal load vector. We
begin with
[(~)] = n !
h
2"
[Too}YK.ctot]dz
h
-2"
/'i,o{3 (6.217)
N [/'i,OO
= no}y LTo{Zk - Zk+1) /'i,o{3 /'i,{3{3 /'i,07l
/'i,{37 [atol
at{3 ,
k=l /'i,07 /'i,{37 /'i,77 k at7 k
6.13 Initial load due to temperature 201
[~]
(~:) = n ~ [zToal.""at]dz
1 N [Kaa
= na}.r"2 LTo(z~ - z~+l)l-l "'af3
"'af3
"'f3f3
K...,]
"'f3'Y
[aatf3tat218)
.
k=l "'a'Y "'f3'Y "''Y'Y k at'Y k
Vector J NO is
JNO [ J1] .
(3xl)
= Jg (6.219)
(12xl) (::)
Similarly,
[J:::
11
(3xl) ]
=n 2"
, t
[zT1aN""at]dz
N
= na}.r"21 LTo(z~ r~
- z~+l) "'af3
"'af3 K...,]
"'f3'Y
[aatf3ta ] (6.220)
"'f3f3
k=l "'a'Y "'f3'Y "''Y'Y k a t'Y k
[03 ] ~
r
(~) = n [z'Tlal.",,,a,Jdz
1 N
= na}.r3 LTl(ZZ -
["aa
ZZ+l)Z-l "'af3
"'af3
"'f3f3
7
"'f3'Y ]
"a [aatf3
ta
,
221
)
, k=l "'a'Y "'f3'Y "''Y'Y k at'Y k
and
Jl]
(3xl)
JNl
(12xl)
= [(::)
Ji . (6.222)
202 Composite plate and shell element
Now the natural thermal load vector is completely defined from its vec-
tor components, and we may write down the equilibrium equation in the
natural coordinates via
kNPN=PN+JN. (6.223)
J
(18xl)
= [aN T06
(12x18)(18x18)
t JN
(12xl)
= akr JN
(18x12)(12xl) (6.224)
KEr=R+J (6.225)
! !
h h
"2 "2
h h
2 "2
PN = [J ~ctdzl'Y~+ [J z~ctdz]l-l1/Js
h h
-"2 -"2
h h
2 2
MN = [J z~ctdz]"Y~ + [J z2~ctdz]I-11/Js (6.228)
h h
-2 -2
If the transverse shear forces per unit length are required, the last of equa-
tions (6.228) becomes
2MA
qN=n- (6.229)
The stress resultants are transformed to the local cartesian coordinate x'y' z'
using
Remark 5 The local cartesian forces and moments can be easily referred
to the three triangular nodes. This requires only the separate projection
of the the various natural nodal axial forces as well as symmetrical and
antisymmetrical nodal moments to the three local coordinate systems at the
vertices, and superposition of all components in the local directions.
204 Composite plate and shell element
k k k
U e = Ketlt (6.231)
Uk=
, B U ek , (6.232)
{)Uea + {)Usa _ 0
(6.233)
{)X a {)z - ,
--J
z
{)U sa {)U ea
--=---,
{)z {)x a Usa - {)u ea
{)X a dz + Cl (6.234)
h
-"2
where ZOa is the Z coordinate of the neutral axis (see also (6.166)) and
kNB is the bending stiffness in this direction. A combination of these two
relations yields
6.14 Computation of stresses and stress resultants 205
a so. =- Jz aMA
axo. %
kNB{Z) (z - zoa,}dz + CI.
-% J h
k%B{z - zOo)2dz
(6.236)
-"2
We have already provided the definition for the shear force per unit length,
namely
aMA 2MAo.
qo = axo = -0-' (6.237)
I
Z
kNB{Z)
a so =- qo % (z-zOo.)dZ+CI
-% Ih
k%B(Z - zOo)2dz
{6.238}
-"2
Since the transverse shear stiffnesses vanish at the top and bottom sur-
faces, constant CI also vanishes. Now appropriate stiffnesses kNB must be
assigned. For this purpose we use the natural stiffness K.ct and invoke the
physical lumping method [16], [51]. We assign the following values to the
k~B
206 Composite plate and shell element
z
oazz
--=---,
oz
oa so.
oxo. azz -
--j oaso. dz
oxo.
+ CIZ + C2,
h (6.241)
-2"
=
jz jZ oxo.
oqo.
~
kNB{Z) {z _ zOo.)dz + CIZ + C2.
(6.242)
j
azz
_l!
2
_l!2 NB {Z - ZOo. )2dz
ko.
h
-2"
6.14 Computation of stresses and stress resultants 207
8qo.
- ----,
p- 8xo. (6.243)
JJ
z z
kNB{Z)
!
a zz =- P ~2 ) dz + Cl Z + C
( -zZOo.
h
-2" -2"
h
k):,B(Z _ z(}a)'dz ,. (6.244)
-~
2
h h
at Z = --2 -+ a zz = 0, at Z = + -2 -+ a zz = -po
(6.245)
JJ
z z
a zz = - P kNB ()
Z (z - ZOo. )dZ - !!'z-'!!.
h 2
J
h
h h 2" 2 (~.246)
-2" -2"
kNB(z - ZOo.) dz
-~
a- zz -_1(0.
3" a zz + a f3zz + a 'Y)
zz . (6.247)
208 Composite plate and shell element
Definition 10 Matrix 0.0 relates the natural rigid body modes Po to unit
cartesian nodal displacements and rotations.
UI + U2 + U3 = 3po, VI + V2 + V3 = 3qo, WI + W2 + W3 = 3r o
(6.249)
1
POI = Po = "3(Ul + U2 + U3)
1
P02 = qo = "3 (VI + V2 + V3) (6.250)
1
P03 = TO = "3 (WI + W2 + W3)
1
P06 = - 40 (XaUl + YaVl + Xf3 U 2 + Yf3 V 2 + X-y U 3 + Y-y V 3)
6.252)
Turn now to Fig. 6.14 which shows a rotation (h. The idea here is to
project this rotation on the x, Y axes. The angle (h shown in Fig. 6.14 is
o WI wlla
(6.253)
1 = ha = 20 .
It projects on the local cartesian coordinates via
Xf3 X-y
02x = 20 W2, O3x -- -W3,
20 (6.255)
Yf3 Y-y
O2y -- -W2,
20 03y = 20 W3 (6.256)
210 Composite plate and shell element
*-
X
/ a: ' IC>
2 p(J4
2
(xt Y2)
3
(x)' Y3 ) ~3
Xo xf3 X"(
P04 = ()lx + ()2x + ()3x = 20 WI + 20 W2 + 20 W3
Yo Yf3 Y"(
P05 = () ly + ()2y + ()3y = 20 WI + 20 W2 + 20 W3 (6.257)
ao
[-11
(3x6)
a-12
o
(3x6) -13]
ao
(3x6) ,
i:io (6.258)
(6xI8) -21 -22 -23
ao ao ao
(3x6) (3x6) (3x6)
with
-11
ao
(3x6)
-12
ao
(3x6)
-13
ao
(3x6) [~ 3"
1
1
3" J (6.259)
6.15 The simplified geometrical stiffness 211
~
211
-21
ao = 1k
211 . I , (6.260)
(3x6)
-~ _1k
411 411
~
211
-22
ao = ~
211 . I , (6.261)
(3x6)
-~
411
-~
411
!1..
211
-23
ao = h
211 (6.262)
(3x6)
_!1.. _h
411 411
_ ao ]
(6x18)
(18~18) = [ aN .
(6.263)
(12x18)
in which Ao is defined as
-t
Ao =
[-1
Ao -2 A-3]o ,
Ao (6.265)
(6xI8) (6x6) (6x6) (6x6)
and
212 Composite plate and shell element
- 1
AO = Yl 1
., , {6.266}
(6x6)
-Xl 1
-Yl Xl 1
., ,
1
AO
- 2
(6x6)
= I 1
Y2 1
{6.267}
-X2 1
-Y2 X2 1
1
- 1
AO = Y3 1
{6.268}
(6x6)
-X3 1
-Y3 X3 1
Incidentally,
o
W.Z Wp=W -XPOS+YP04
P~
01
I
Up=U
o
'YP06 +ZPOS
WI
I
------...J
o
U I
/
/
/
/
/
/
o
/ 0
/ v
/
/
/
X. U
In concert with the principles of our theory, rigid body motion should not
change the shape of an element and therefore we can use the linear matrix
aN to estimate the natural deformations which arise from a small addi-
tional nodal displacement 8p. A coordinate transformation T{po) yields
the position of the element after rigid body motion. Initially for small
displacements (see Fig. 6.15)
[P~6
-P06
+ P05]
-~04
[X]
;
=UO+TR'X,
[:] = [::] -P05 (6.271)
P04
where
214 Composite plate and shell element
pos]
[P~6
-P06
TR = -~04 , (6.272)
-Pos P04
and
TCJt = [: : ~1]'
. 1 .
TCJf = [.. .. 1]. , TC: = [. -1 .]
-1 . .
1
.
.
.
.
.
.
(6.273)
-pos]
[-~06
P06
RR = P04 , (6.274)
Pos -P04
aN = aN T 06 (6.276)
We effect a rigid body rotation increment POR' The global forces rotating
with the element, originally P, become P + P A and a rigid body rota-
tion initially A~R becomes A~R + A~RA' Since the resultants of all forces
produced by rigid body motion must vanish, we write for the rigid body
moments
-t -t-t
POR = AOR P = [AOR + AORA][P + P A] = O.
(3xl) (3x18)(18xl)
(6.279)
-t -t -t -t
AORP 06.+ AORAP = 0 =} AORP A = -AORAP = PORA'
(6.280)
Hence
- t - t -t
PORA = -AORAP = -AORA aN PN, (6.281)
(3x18) (3x18) (18x12)(12xl)
which is a measure of the effect of the natural forces due to the change of
geometry. Equations (6.281), (6.277) can be combined to yield
- t -t
PORA = -AORAaNPN
(3xl)
-- - [A- OR
t
6 aN p
R(4)-t 1
N P04/). - [A- OR
t
6 aN p
R(5)-t 1
N P05/). - [A- OR
t
6 aN p
R(6)-t 1
N P06/).
-- kGRPORA' (6.282)
(3x3) (3xl)
whereupon
R6 - fR Oi
(i) - R R Oi
R R Oi
R R Oi
R R Oi
R ROiJ - 4"
R, i - 5.
6
(18x18) (6.283)
It is clear that the forces PORA depend on the values of PN; this is an im-
portant point since the nonlinearity arises from this dependence. Equation
(6.282) estimates the nodal cartesian moments PORA due to rotation in-
crement PORA and yields an appropriate expression for the local rigid body
rotational geometrical stiffness kGR, namely
216 Composite plate and shell element
(3x3)
- [A-OR
k G R-- t R(4)
6
-t
aNN
P
(3",18)(18:>:18)(18",3)(3:>:1)
A- OR
t R(5)
6
-t
aNN
P
(3:>:18)(18:>:18)(18"'3)(3"'1)
A- OR
t R(6)
6
-t
aNN
P J.284)
(3"'18)(18"'18)(18"'3)(3"'~
kGR = (6.285)
(3x3)
. ]
~
~+~+
I", 1{3 I"{ -(~+~+~) I", I~ I"{
( ~+~+~)
I", 1{3 I"{
P"'2~
I",
+P{32"'1{3 $ +P,,{;;
I"{
.
. ~+~+~
The cartesian nodal forces due to three rigid body rotations PORL1 are
- -t
PGL1 = aOR P ORL1
(lSxl) (lSx3) (3xl)
1
kNG= 12r. . Pa . P{3 . P'"( .J.
(6.287)
KE+ALKG=o. (6.289)
1
PNA = "2 [aN{K b) + aN{Ke)]rA, (6.290)
We n,
ref = 0.00560 Pl 2 centra11oad .
6.17 Computational experiments 219
Table 6.1: Central displacement error for a fully clamped isotropic plate
under uniform pressure and central load
I mesh unknowns ~c{p~essure) . Mx{pressure) We {central load) I
3x3 16 -7.68% 16.41 1.03%
5x5 80 -1.58% 21.00 0.89%
7x7 192 -0.38% 21.96 0.80%
9x9 352 0.04% 22.33 0.80%
exact 23.10
2 error %
1
0 I 01"\ ::I .f. unknowns
: 16 ov
352
-1 192
-2
-3
-4 ----ft- w uniform pressure
-5
-6
--e-- w central load
-7
-8
Table 6.2: Central displacement and moment for a thick (i;=8.14) sandwich
plate
mesh unknowns - - We - M~I
3x3 8 31.08{-1.19%) 3058
5x5 104 31.22{-0.74%) 3646
7x7 228 31.22{-0.74%) 3749
9x9 400 31.22{-0.74%) 3767
exact 31.4537 3668.3
error %
2
L5
-e-- wc f
0.5 unknowns
18 104 228 400 f
0
-0.5
-I
-1.5
Ef =6.8e9 Pa E.,= 4.8 e8 Pa
-2 analytical sol. we = 31.4537 mm
Gf = 2.615 e9 Pa =
Gw 38 e6 Pa
Me = 3668.3
Vf =0.3 V., = 0.3
",=0.001 m hw=0.0I5 m
pinched cylinder
E=3.0x 106
v =0.3
h=3.0
L= 600.0
R =300.0
p= 1.0
rigid diaphragm
E= 3.42 x 106
v=O
h =0.25
q = 90.0/ unit area
L=50.0
R=25.0
9= 80
Internal pressure p
z
=
2.0372 psi
y
1<1 t>i
L == 20 in
E
xx
= 2.0 x 106 psi E == 7.5 x 106 psi
yy
Gxy = 1.25 x 106 psi Gxy = 0.625 x 106 psi vxy = 0.25
Table 6.5: Maximum displacement, central hoop stress and central moment
for a glass-epoxy pressurized cylinder (R/h = 20)
I nodes unknowns w~ax x 10 4 O'~Mc I
3x3 22 2.44 79( -33.30%) 16.12 2.080
5x5 94 3.4076(-7.15%) 29.93 4.427
7x7 214 3.6300(-1.09%) 35.15 6.041
9x9 382 3.6558(-0.39%) 37.74 6.98
exact 3.67xl0 [72] 36.70 7.92
224 Composite plate and shell element
w/w
1.1 ~ ref
21 25 29
1
nodes x
0.9
0.8
w pressurized shell
0.5
0.4"
0.3
- 6.825010 '
v - OJ
R - 10.0
h O,04
,
F - I
1
a
;" "J
E - 2,9 10'
v _ 0.22
h -0.32
~ u - 12.0
b- U
F, m 1
F _ I
~
N
l
M
II)
N
~
~
'<I;
;t
~
~
~
::!:
~
Ii
en
w
N~
a:
~
Q.
Z
o
~
::!:
a:
12w
Cl
Figure 6.21: Undeformed and deformed plots for the pinched hemispherical
shell
6.17 Computational experiments 227
Table 6.7: In-plane normalized tip displacement for the twisted beam ;
comparison with 4-node quadrilateral elements; tip displacement x 10- 2
I nodes RSDS[68] QUAD[73] URI[68] URI [74] Present
, 13x3 (0.993) (1.009) (1.004) 0.5347(0.986)
15x3 0.5466(1.007)
25x5 (1.411) (1.001) (0.999)
Exact 0.5424
element are presented in Table 6.7 where it is also compared with other
available solutions. Very good agreement is again obtained. Figure 6.22
provides us with plots of the undeformed and deformed beams.
El = 40, G12 = 0.5, G23 = 0.2, 1I12 = 0.5, G 13 = G12, 1I13 = 1I12
E2 E2 E2 (6.293)
- 104 ,
W = WE2h3.
qo
.A
El = 25 X 106 E2 = 106
G 12 = 0.5 X 106 G 13 = G 23 = 0.2 X 106 (6.294)
1/12 = 1/13 = 0.25
and of the core were (transverse isotropy with respect to the z axis was
assumed)
230 Composite plate and shell element
s = l/h - (a b 1) - (a b 1)
U xx 2' 2' 2 U yy 2' 2' 2
The thickness of each face sheet was h/10. Normalized stresses at selected
through the thickness locations as obtained from the present analysis and
the three dimensional solution of [76] are presented in Tables 6.9, 6.10.
El = 25 X 106 E2 = 106
G 12 = 0.5 X 106 G 13 = G23 = 0.2 X 106 (6.296)
1I12 = 1I13 = 0.25
~ol ~I/.
dof wlwexact error (%) 112
24 0.9164 -8.35
54 0.9653 -3.47
96 0.9834 -1.66
c III
150 0.9917 -0.83
216 0.9962 -0.38 0.q (/ 12,0)
IIh=\o
zlh
zlh
-0.5 -0.5 , - - - - elasticity
-0.4 -0.4167 \?",
V shell element
-0.333 'r7 '
V
....
..........
-0.3
-0.25 V ...... - ........................
-0.2 ...............
sbell element -0.167 V "" .... ,
-0.1 elasticity -0.0832 V \
o o V \
0.1 0.0832 V ./
0.167 V .... .... '"
------
........
0.2
0.25 V
0.3 ....
0.333 V./"' ....
./
0.4 (j (/12,0) 0.417 ~./ iiy/O,V2)
.q ./
0.5''''''''''''- , 0.5'
..~-.l.--...L.-.......- - . l . - -......- .....- - - '
o 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
(0190/0)
simply supported
z
K ~/ t::>-x
Table 6.11: Central normalized displacement ill and through the thickness
stresses (jxx, (jyy, for a (0/90/0) rectangular laminate for various aspect
ratios s. CPT is an abbreviation for classical plate theory
s = l/h ill - (a b 1) - (a b 1)
f7 xx 2' 2' 2 f7 yy 2' 2' 6"
2 present 9.8156
elasticity 8.1700
4 present 3.0320 0.6000 0.0972
elasticity 2.8200 0.7260 0.1190
10 present 0.9982 0.6012 0.0420
elasticity 0.9190 0.7250 0.0435
20 present 0.6384 0.6012 0.0285
elasticity 0.6100 0.6500 0.0299
50 present 0.5191 0.6020 0.0233
elasticity 0.5200 0.6280 0.0259
100 present 0.5009 0.6021 0.0224
elasticity 0.5080 0.6240 0.0253
CPT 0.5030 0.6230 0.0252
Table 6.12: Normalized through the thickness stresses O'xy, O'xz, O'yz, for a
(0/90/0) rectangular laminate for various aspect ratios s
0.5 Z ~ TRIC
0.4 0 3-D
\
Ll CPT
0.2 [ Ll
oI
0.1 Ll _ _
I I I I I I
~tyz (a/2,0,zJ
I
o 0.005 0.01 0.015 0.02 0.025 0.03 0.0 0.04
-0.1 I- Ll
o
-0.2 ~ Ll 0
-0.3
Ll
-0.4
-0.5
Table 6.13: Normalized buckling loads (. ~:i~) for a (0/90)8 square laminate
(l/h = 10). CPT, HSDT, are abbreviations for classical plate theory and
higher-order shear deformation theory, respectively
400mm
I<::l l>-I
clamped edge
clamped edge
w(~m~m~)~~~~~=j____~/:::: __________~
21~
I.S point A
o experiment
1.6
point B
1.4
-------
1.2
/"
//
// _-...0-
O.S // //...0-
// //0
0.6
/ ..-0"/
/~/
~'l/
~'/'
0.4
/d'
O.H- I //
//
. // P (kg)
Olk/
o-t-T~-1-
2 4 6 S 10 12 14 16 IS 20
...
co
~ lii'"
a>
co
~ ~
~
"!
...
'" a '"
CD
O!
'"
~
'"
~
<D
~
III
co
~
l!iJ
~
;;;
~
...'"
'"
o
M
~
i.,
(/)
~
II>
N
~
~
~
ci:.
z
~
~
~
::1:
~
W
N~
II:
'"
<If
()
...J
b
...J
0..
Table 6.14: Normalized buckling loads (. :;i!) for a (45/ -45/0/90)8 square
laminate (l/h = 10) for various aspect ratios s
10 27.0966 26.799
20 37.4311 37.115
50 42.1436 41.877
100 42.9401 42.819
z.w
IV
y. v
u=w=, ='11 = 0
x,u
. - ~90
2 '
T 45
w=9=O
441 nodes
-45
2398 unknowns
EI = 150GPa
E2 = 10 GPa
G I2 =6GPa=GI3 hL =1.35 x 104 m
G23 =4GPa Rlh = 92.59
VI2=VI3=v23=O.25
a l = 2.5 x 10.8 C I
a2 = 30 x 10-6 C l
Figure 6.28: Composite cylindrical panel under edge load and temperature
increase
four buckling modes due to applied edge load are shown in Figs. 6.30-6.33,
while the first four buckling modes due to temperature are depicted in Figs.
6.34-6.37.
'v
+
Figure 6.29: Finite element mesh for the {45/ -45/0/90)8 cylindrical panel
242 Composite plate and shell element
c;:?
f'-
M
~
6.
Ql
en
~
10
N
~
~
'<l:
ri:
z
~
~
U
en
::2;
'v
+
Ii
en
w
a:
N
II
o
...J
b
...J
Il..
Z
o
~
::2;
a:
ou..
W
o
Figure 6.30: First elastic buckling mode (Per = 65914Pa) for the composite
cylindrical panel
6.17 Computational experiments 243
~
I"-
M
i
~
II)
N
~
~
~
~
Ii:
~
!(
D..
<3
en
>v
::::E
~
;g +
w
II:
"!
~
9
D..
Z
o
~
~
f(
w
o
g
t.:.
M
~
~CD
en
~
Il)
N
~
~
.q;
d:
I
en
::i!
'v
Ii +
en
w
II:
'"
N
g"
b
--'
D..
Z
o
~
::i!
II:
ou.
W
Cl
Figure 6.32: Third elastic buckling mode (PeT = 71700 Pa) for the compos-
ite cylindrical panel
6.17 Computational experiments 245
It)
<0
M
tOJ
(J)
~
It)
N
~
~
~
d:
z
~
~
(J)
::!!
+
~w
a:
~
~
b
'\/
....I
a.
z
o
~
::!!
a:
1(
w
c
:;:
co
c.;
~
~
til-
C/l
.,j.
l/)
N
~
'<I:
~
z
~
C3
C/l
:i:
'v
~
Ii +
faII:
en
~
Q.
Z
o
~
:i:
II:
f2w
o
Figure 6.34: First thermal buckling mode (Tcr = 168.74C) for the com-
posite cylindrical panel
6.17 Computational experiments 247
8
OJ
M
~
5r
~
Ii)
N
~
~
'<I:
r:i:
z
~
~
~
~
Ii
en
+
'\/
w
a:
o
N
~
Il..
Z
o
~
~
a:
~
w
Cl
~
Oi
(0)
~
10
N
~
~
~
ci:
~
~
::;:
~
'v
Ii
en +
w
a:
CJi
g
ga.
z
o
~
a:
2
w
c
Figure 6.36: Third thermal buckling mode (Tcr = 294.71 DC) for the com-
posite cylindrical panel
6.17 Computational experiments 249
;:Ii
0;
~
c.;
~
.,
6.
en
..j.
~
on
N
~
~
~
~
ri:
z
~
~
U
en
::E
'v
Ii
en +
w
II:
N
~
a.
z
o
~
::E
II:
f(
W
C
--
c:::;:::::::.. \-- --
-- -- c:::;:::::::..
--
--
c:::;:::::::..
c::::::>
-- -- --
-- ~w
--
--
(45/-45/0/90)s 735 nodes
3954 unknowns
EI =20GPa
R = 250 mm
E2 =! GPa
h=20mm
G I2 =0.6 GPa =G I3
1= 1400mm
G23 =0.5 GPa
Rlh= 12.5
V 12 =V 13 =V 2J =0.25
Figure 6.40: First buckling mode for the composite pressurized vessel
l
6.17 Computational experiments
CS:Cl>:l>L 96-fin'l-9Z S"Z N'II::U'Id n -l:l N'Il:ll'ldr.>S~ L" L=S3l:l L " z=~:n .10ld NOll"'1~Il:lOlj3IJI
CS:Cl>:l>L 96-00'l-9Z S"ZN'Il:l.1'1d n-l:l N'Il:l.1'1d/:>Sr-I (')'1r-1~3NL"L=S3l:l L"Z=:>1 .10ld
FR NGE
I
t,;)
c;,
NGE
FRINGE PLOT LC=2.2 RES=1.1(VEC-MAG) MSC/PATRAN R-l.4 PATRAN 2.5 26-Aug-96 14:45:57 >I:>--
EFORMATION PLOT LC=2.2 RES=1.1 MSC/PATRAN R-1.4 PATRAN 2.5 26-Aug-9614:45:57
"'rj
......
aq oo
~
(1) .
~
~
......
~ ....
...... (1)
en '"C
(1) [
C")
o (1)
::l
0-
c"" 0-
00
~ ~
(1)
-
Jg'
.2971
[
g. .23n a
.1783
.1188
j-' .05942
6.17 Computational experiments 255
These values are the highest of all previously obtained. Of course, a more
optimum orientation of fibers should be obtained by employing formal op-
timization procedures.
256 Composite plate and shell element
Cf 0>
diaphragm
EI = 150GPa
c;~~. E2 = 10 GPa
GI2 =6GPa=G I3
G23 =4GPa
fc::!
0.250 mcf::J 1.40 m
t>j<I
0.30 m
t>j v 12 =v 13 =v23 =0.25
a l = 2.5 x 10-8 C- I
a =30 x 1O.6 c- 1
2
-1 10-4 m
-V (751-151010), hL ' 1.35 X
8
RI h = 231.48
7 V (601-60)4
R= 0.250m
6 iii (301-30)4
5 -V (45101-45190),
4 II (45/-4510190),
3 v (0I901019j I),
2 v (90)g
" (O)g
t1-
8'l..- -::::::J (751-15I0I0
lL
7 J (601-60)4
b::
6 - . (3 01-30)4
~
5
C:::
'-- -:II '=:J (45101-45190)
4 --. (451-45I0I90) ,
3 -::J (019010190),
IJ.
2 -:1 (90)g
J,..
... (0),
o 5 10 15 20 25 30 35
critical temperature (0 C)
Figure &.42: Rocket-like shell; critical pressures and temperatures for eight
lamination schemes
6.17 Computational experiments 257
Figure 6.43: First thermal buckling mode for the (75/ -15/0/0)8 composite
shell at Tcr = 29.4193C
258 Composite plate and shell element
6.18 Problems
1. A strain rosette is placed on a plane so that it forms the following
three angles with respect to the x axis: a) 30, b) -30, c) 90. The
strains determined by the rosette during a static test are El = 800,
E2 = 200, E3 = -50.
express the stress resultants in the three nodal local cartesian coor-
dinate systems.
9. On the edges of a triangular side are acting cartesian moments M 1 ,
M 2 , respectively. What are the edge symmetrical and antisymmetri-
cal moments and rotations?
10. Provide a definition for the shear correction factor. Under which
displacement or strain assumptions is the shear correction factor not
required?
11. A triangular element of area n is subjected to a uniform pressure
load of intensity q. What is the local and global vectors of kinemati-
cally equivalent loads? How would the presence of concentrated nodal
forces and moments affect the formation of this vector?
12. Derive explicitly the expressions for the natural thermal load vector
if the temperature interpolation relations (6.205) are used. Proceed
with the evaluation of all integrals.
13. How does the geometrical stiffness of the shell element relate to the
geometrical stiffness of the beam element?
14. Prove expression (5.261).
15. For the pinched cylinder and the Scordelis-Lo roof shown in Fig. 6.18
provide the symmetry boundary conditions
16. Explain the rational for the creation of a weighted antisymmetrical
shear stiffness for the triangular element TRIC.
17. Why are rigid body rotations important for the formation of the ge-
ometrical stiffness?
18. Comment on the possibility of computing the through the thickness
stresses at the triangular nodes
19. Why are the natural modes and forces additive during a geometrically
nonlinear analysis?
20. In which cases do you expect the shear correction factor to signifi-
cantly affect the deformation?
21. What is the physical meaning of matrices a and A? Starting from
aA = 118, provide cross relations for the sub matrices of a and A
260 Composite plate and shell element
Computational statistics
The above CPU time list shows that most CPU time is primarily consumed
on algebraic operations such as matrix multiplications, and secondarily on
the core finite element computational procedures such as formulation of
stiffness matrices, assembly procedures, solution of the linear system of
equations, etc. This is in direct contrast with classical finite element meth-
ods in which most of the computing time is consumed on the computation
of the stiffness matrices, the assembly procedure, the solution of the linear
system of equations etc. More specifically the following observations are
made:
is sufficiently shaped and tailored for the computer. One important note:
the model problem which we have chosen here represents a medium size
problem. That is, with the increase of the unknowns it is expected that
procedures such as solution of the linear system of equations will consume
more computing time. This, however, represents another modern research
area in computational mechnics which deals with efficient solution algo-
rithms in association with fast finite element technology.
Chapter 8
Nonlinear analysis of
anisotropic shells
method, refined to work with TRIC, shows many advantages over classical
formulations: analytic and elegant expression for all elemental matrices; a
series of vector and matrix multiplications that can be easily optimized for
maximum speed; material generality; a choice of Newton-Raphson or modi-
fied Newton-Raphson solution algorithms; accurate location of bifurcation,
limit and displacement points; computational efficiency and economy; and
full potential for vectorization/parallelization. The arc-length method and
its variations have been extensively used by many authors for the study of
nonlinear shell behavior -we quote here the works by Riks [82], Crisfield
[83], [81] and Papadrakakis [84]. However, most of the studies are conducted
using classical finite element methods and confined to isotropic panels of
simple geometry. Thus we aim at developing a computational procedure for
large-scale nonlinear analysis of isotropic and composite shells of arbitrary
geometry by blending our finite element methodology with the arc-length
method. The latter is refined to work in concert with the natural mode
method.
KTU=P, (8.1)
where K T is the global tangent stiffness matrix. We have seen that the
tangent stiffness matrix is given by the expression
KT ~ [T:6[ii~[ ~ JiiN]T06]
natural stiff. (I2xI2)
,
'"
local siff. k
(18x18)
global stiff. K E
(18x18) (8.2)
We note the first advantage of our formulation: The local elastic stiffness
matrix a~K NaN along with k-G and a~K NGaN can be expressed analyt-
ically. These elemental matrices are all added together to form the local
tangent stiffness matrix K LT in an element level. Thus only matrix mul-
tiplications Tb6 K LTT06, with T06 being a hyper diagonal matrix, suffice
for the formation of the global tangent stiffness. This action will signifi-
cantly speed up the computation process. Naturally, vector and parallel
computers can perform these operations very fast.
Now during nonlinear analysis with automatic load adjustments we will
use two measures to guide us in identifying stable, neutral, strongly and
weakly unstable regions. These measures include the energy of the struc-
ture (or stiffness parameter) using a predicted displacement vector, and
the nature of eigenvalues of the tangential stiffness matrix including posi-
tive eigenvalues, zero eigenvalues, number of negative eigenvalues etc. The
eigenvalue information is used to determine the sign of the determinant of
the tangent stiffness matrix. Symbolically, these parameters are expressed
as
1. When both the energy and first eigenvalue are positive, the motion
of the structure is said to be stable.
2. When the energy is positive but the first eigenvalue is zero a bifurca-
tion point has been encountered. Depending on the number of zero
eigenvalues, the possibility exists for other equilibrium branches be-
sides the primary path. Various bifurcations may include the follow-
ing: stable symmetric bifurcation; unstable symmetric bifurcation;
and unsymmetric bifurcation.
3. When energy and first eigenvalue are both zero the structure is at
neutral equilibrium; other equilibrium branches may exist. In general,
the existence of zero eigenvalues indicate the presence of bifurcations
and the possibility of secondary paths.
4. When one or more negative eigenvalues of the tangent stiffness are
found, the structure operates in an unstable region:
270 Nonlinear analysis of anisotropic shells
where F"t is the vector of internal loads. If we denote the unbalanced force
vector by
Xi = OAi+IXf + xi (8.7)
P
Xi = K-1p
Ti , (8.8)
Xi
C
= - K-1
Tigi (8.9)
8.2 The incremental/iterative scheme 271
where
P)t P
a1 = ( xi xi'
a2 = 2 [~Ui + xi]t xf, (8.14)
a3 = [2~Ui + xi]t xi.
For the first increment, 8>'1 is given as an input data and is used to de-
termine the arc-length radius ~l which remains constant throughout the
computational procedure. For subsequent increments, the arc-length is de-
termined by the roots of the quadratic equation.
Now before initiating the incremental/iterative procedure the following
computing steps are undertaken:
E= 3105 N/mm2
v=O.3
L=504mm
R=2540mm
t= 12.7mm
9 = 5.6773 0
Rlt= 200
Pc (N)
2500
2000
1500
1000
500
wc(mm)
o
o 5 10 15 20 25 30
Figure 8.1: Hinged isotropic shell; comparison with the reference solution
276 Nonlinear analysis of anisotropic shells
Pc (N)
2500
2000
1500
1000
500
0
1/ iteration NO
0 5 10 15 20 25
wc(mm)
30
25
20
15
10
0
0 5 10 15 20 25
60
50
40
30
20
10
iteration NO
o --:--~
o 5 10 15 20 25
P
c 9 x 9 nodes in quarter (TRIC)
free
E= 3105 N/mm2
v=0.3
L=504mm
R=2540mm
t= 6.35 mm
9= 5.6773 0
Rlt=400
500
300
100
-100
-300
wc(mm)
-500
o 5 10 15 20 25 30
PeeN)
600
400
200
-200
P 4 8
-400
30 r We (mm)
25
20
15
10
5
iteration NO
4 8 12 16 20 24 28 32
70
60
". ............
50
40
30
20
10
4
- 8 12 16 20 24 28 32
~
0
5
p..
p..
C1l
C'D
~CI.l
_00
.......
f-'
t-;)
t-;)
00
00 I~
I~
280 Nonlinear analysis of anisotropic shells
*************************************************************
TRIANGULAR COMPOSITE ELEMENT T RIC
*************************************************************
Increment Iteration Time/Tangent Stiff. Time/increment
1 4 1.39 4.24
2 4 1.38 4.26
3 4 1.39 4.28
4 4 1.40 4.28
5 4 1.40 4.26
6 3 1.41 3.69
7 3 1.38 3.65
8 3 1.40 3.66
9 3 1.39 3.65
10 3 1.39 3.65
11 4 1.39 4.24
12 4 1.39 4.23
13 6 1.40 5.43
14 6 1.39 5.45
15 6 1.39 5.46
16 6 1.39 5.42
17 6 1.39 5.41
18 6 1.40 5.44
19 8 1.38 6.59
20 6 1.39 5.43
21 8 1.40 6.69
22 8 1.41 6.68
23 10 1.39 7.78
24 12 1.39 8.98
25 11 1.39 8.35
26 47 1.39 29.62
27 28 1.39 18.39
28 9 1.39 7.18
29 5 1.39 4.82
30 4 1.38 4.25
31 5 1.39 4.84
32 7 1.39 6.02
33 4 1.39 4.26
34 3 1.40 3.71
286 Nonlinear analysis of anisotropic shells
32 3 19.99 30.61
33 3 19.90 30.51
34 2 19.87 27.55
peared to be the case in all examples considered here. We stress that these
are only preliminary observations and that much more research is needed
before more definitive statements can be made. The deformation of the
{452/ - 452)68 panel at various load levels is illustrated in Figs. 8.14-8.17.
The plots reveal a double snap-through and snap-back.
We conclude the numerical examples with a postbuckling analysis of
a 4-layer (45/ - 45/ - 45/45) laminated composite cylinder shown in Fig.
8.18. One edge of the cylinder is clamped while the free edge is subjected
to a compressive force P which is distributed to the nodal points. An initial
imperfection in the form of the first buckling mode scaled down to 5% of
the shell thickness is given before the nonlinear analysis is initiated. For
points A, B, we provide the load-displacement curves in Figs. 8.18, 8.19.
Deformation plots at two distinct loads are shown in Figs. in Figs. 8.20,
8.21. We observe that following buckling, the composite cylinder reaches
a value after which only local edge movements occur while the cylinder
body close to the clamped support seems to be relatively unaffected. From
this numerical example we deduce the intricate load and energy absorption
mechanisms of composite cylinders that makes them attractive for crash
prone vehicles.
8.3 Numerical examples 289
EI = 150GPa
E2 = 10 GPa
L=504mm GI2 =6GPa=G I3
8 = 5.6773 0 a.
I
=2.5 x 10-8 C-I
Rlh=400 a. = 30 x 10-6 C-I
2
hL =1.35 X 10-4 m
(0/90/0/906)s
PeeN)
15000
10000
5000
-5000
We (mm)
-10000
5 10 15 20 25 30
10
2
iteration NO
o
o 5 10 15 20 25 30 35 40 45 50 55 60
60
50
40
30
20
10
o
--
... -...
o 5 10 15 20 25 30 35 40 45 50 55 60
o r wA
-0.5
-I
-1.5
-2
-2.5
-3
-3.5
-4
-4.5
==r: ,vA
-5
-0.012 -0.01 -0.008 -0.006 -0.004 -0.002 o 0.002 0.004
12000 f Pc (N)
10000
8000
6000
4000
2000
0
-6000 ~ ~ wc(mm)
-8000
0 5 10 15 20 25 30
6000
4000
2000
-2000 r
(45/-45 2)6S
,
\. /'
wc(mm)
-4000 ' . I ,
0 5 10 15 20 25 30
12000 f Pc (N)
10000
8000
6000
4000
2000
=t
0
(451-4510/90/60/-30)3S
~ wc(mm)
-6000
0 2 4 6 8 10 12 14 16 18 20
20 t I (45l4VO/ 90 2)3S
lO
0
0 20 40 60 80 100 120 140 160 180
100
.'
90
80
70
60
50 ..... ' .. / ........... .. - .... _-" ... -.... -.
40
II " ....... -
30
20 WI (45/-45 2)6S
lO
0
0 20 40 60 80 100 120 140
80
70
60 t. .---. .....
:"
# - ..
- ..........
50
40
30
20
lO ~
iteration NO
I === I 1 _I
20 40 60 80 100 120 140 160 180 200 220 240
Figure 8.13: Decomposition of the strain energies for the three composite
panels
8.3 Numerical examples 293
25000
20000 (451-451-45/45)
p
15000
R =0.250m
10000
a =1.40m
5000
0 uA (m)
0 0.00005 0.0001 0.00015 0.0002 0.00025 0.0003
30000
25000
20000
15000
10000
5000
vim)
01/
0 0.0005 0.001 0.0015 0.002 0.0025
30000
25000
20000
15000
10000
5000
II
wim)
0
0 0.0005 0.001 0.0015 0.002 0.0025
25000
20000
15000
10000
5000
uim)
0
0 0.00005 0.0001 0.00015 0.0002 0.00025 0.0003
30000
25000
20000
15000
10000
5000
0
0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012
30000
25000
20000
15000
10000
5000
0
-0.00054 -0.00045 -0.00036 -0.00027 -0.00018 -0.00009 0
lU~OO '
'"'
-....
Q) Q)
...r::: "'0
00
.::
.... Co)
0..
ESLlOO'
Co)
0
.....'"' E9SlQO'
..........
tI)
....
.::ce
0
00
I>LOEOO'
l!";)
-.:::I'
I
...... ...........
0
....
I>'I>I>EOO' l!";)
-.:::I'
00 ..........
-
00
Q)
SLSEOO' ...r:::
ce .....
.::
ce ......
9OC1>OO' 0
~
.....::
-....
Q) 0
.:: 9ELI>OO'
ii3
0
.:: S
Z L9~SOO'
'"'
.
Q)
L6SSOO' 0
..-i
c-:I
9l09OO'
00
Q)
9SI>9OO' '"'
....be
;::l
DEFORMAnON PLOT
9O:0S:I>~ L6-lnr-8~ S'l NV\:ilVd I>' L-I:I NVI:I1Vd/OSr-l L'L=S3I:1l'l=01 lOld
FRINGE
0 9O:0S:I>L L6-lnr-8L S'G NV\:ilVd n-I:I NVI:I1Vd/OSV\I (~VV\I-03"h ' L=S3I:1l' l=01 lOld
300
~
0
M
Chapter 9
Programming aspects
* paracb.h
integer znn,zne,znde,zndf,zu,zf
integer zmaxa,zlb,znl,znne,znmt,znumna,
znnped2,znna,znumk,
zndh,znumh,zheat
parameter(znn=500,zne=1000,znne=2,zndf=6,
znmt=zne,znde=zne,znnped2=znne*zndf,znna=znnped2-6,
znumk=znnped2*znnped2,znumna=znna*znna,
zndh=2,znumh=znne*zndh,zheat=znumh*znumh,
zu=1000000,zf=znn*zndf,znl=30,
zmaxa=zf+1,zlb=zf)
znne: maximum number of node per element (2 for the beam element)
The user must modify these parameters according to his problem and
available core memory. For example, he can set the maximum number of
nodes equal to 200 (znn=200).
idel{it): identification number for element it. For the beam element,
idel{it} =2
idel(iel): identification number for element iel. For the beam element
idel{iel} =2
node(iel,k): connectivity array for element iel with nodes k=l and
k=2. If node{iel,1}=2, then the first node of element iel connects to
global node 2. The same convention holds for the second node
theta(i,imt): fiber angle for layer i and material type imt formed by
the local cartesian axis x' and fiber angle 1
press: Pressure magnitude with sign opposite to all local z' axes.
306 Programming aspects
iprint: printout
iprint=. true.: extensive printout
iprint=.Jalse.: limited printout
The output information is written on file outb.dat
9.2 Input data 307
ipost
ipost=.true.: postprocessing is conducted with the computation of
forces, moments, natural energies for all elements and the estimation
of the through the thickness stresses for a selected element
All this information is written on output file nmq.dat
ipatran
ipatran=.true.: a neutral interface to Patran [87] is provided. The
finite element mesh is written on file beam.dat, while the nodal
displacements on file beam.dis
iwhichel: element for which the through the thickness stresses are
required
The input data for the problem of Fig. 5.20 comprising 6 nodes, 5 beam
elements and 30 unknowns is given as
6, 5, 30, 1
1,0.,0., O.
2,5.,0., O.
3,10.,0., O.
4, 15., 0., O.
5, 20., 0., O.
6, 25., 0., O.
1, 0, 0, 0, 0, 0, 0
2, 1, 2, 3, 4, 5, 6
3,7,8,9, 10, 11, 12
4, 13, 14, 15, 16, 17, 18
5, 19, 20, 21, 22, 23, 24
6, 25, 26, 27, 28, 29, 30
1, 2, 2, 1, 1, 2.5, 10., 0., 5.
2, 2, 2, 2, 1, 7.5, 10., 0., 5.
3, 2, 2, 3, 1, 12.5, 10., 0., 5.
4, 2, 2, 4, 1, 17.5, 10.,0., 5.
5, 2, 2, 5, 1, 22.5, 10., 0., 5.
1, 2, 1,2,0.,0.,0., O.
2,2,2,3,0.,0.,0., O.
3,2,3,4,0.,0.,0., O.
4,2,4,5,0.,0.,0., O.
5,2,5,6,0.,0.,0., O.
308 Programming aspects
4
45.,90. ,0.25,3.e7,3.e6,3.e6,1.e6,1.e6,1.e6
0.3,0.3,0.3,0.10e-6,0.10e-6,1.,1.,0.,1.,1.,3000.
-45.,90.,0.25,3.e7,3.e6,3.e6,1.e6,1.e6,1.e6
0.3,0.3,0.3,0.10e-6,0.10e-6,1.,1.,0.,1.,1.,3000.
-45.,90.,0.25,3.e7,3.e6,3.e6,1.e6,1.e6,1.e6
0.3,0.3,0.3,0.10e-6,0.10e-6,1.,1.,0.,1.,1.,3000.
45.,90.,0.25,3.e7,3.e6,3.e6,1.e6,1.e6,1.e6
0.3,0.3,0.3,0.10e-6,0.10e-6,1.,1.,0.,1.,1.,3000.
'unifp' , -0 .
1, O. , O. , 0., 0., O. ,0. ,0. ,0. ,0. ,0. ,0. ,0.
2, O. , O. , O. , O. , O. ,0. ,0. ,0. ,0. ,0. ,0. ,0.
3, O. , O. , O. , 0., O. ,0. ,0. ,0. ,0. ,0. ,0. ,0.
4, O. , O. , 0.,0.,0.,0.,0.,0.,0.,0.,0.,0.
5, O. , O. , 0., 0., O. ,0. ,0. ,0. ,0. ,0. ,0. ,0.
6, 0.1, 0.1, -0.1, -0.1, O. ,0. ,0. ,0. ,0. ,0. ,0. ,0.
'static','isimpgs',.true.,true.,.true.,tr~e.,l
Following the reading of the input data, the routine constructs the
global load vector, computes the total thickness for each material type as
well as the z coordinates for all plies, estimates the length of each element,
and depending on user requests prints out all parameters and prepares the
neutral Patran file. Some of these tasks are accomplished with the Fortran
modules
enddo
enddo
c ..... CONVERT DEGREES TO RADS
do imt=l,nmt
do j=l,nl(imt)
theta(j,imt)=theta(j,imt)*acos(-1.)/180.
th12(j,imt) =th12(j,imt)*acos(-1.)/180.
enddo
enddo
c ..... Z-COORDINATES
do imt=l,nmt
do i=O,nl(imt)
if(i.eq.O) hz(i+l,imt)=thic(imt)/2.
if(i.ne.O) hz(i+l,imt)=hz(i,imt)-hl(i,imt)
enddo
enddo
c ... Compute beam length
do k=l,ne
xl=xn(node(k,l
yl=yn(node(k,l
zl=zn(node(k,l
x2=xn (node (k, 2
y2=yn(node(k,2
z2=zn(node(k,2
bl(k)=sqrtx2-xl)**2+(y2-yl)**2+(z2-z1)**2)
enddo
Routine findsky defines the skyline storage, the positions of the diag-
onal terms and computes the total number of elements under the skyline.
For more information on this storage scheme see the text of Bathe [79].
The storage scheme is declared in the following loops
c----------------------------------------------------------
c Piecewise integration through the layers.
c----------------------------------------------------------
do i=l,nl(imt)
rknll=rknll+storec(i,l,l)*(hz(i,imt)-hz(i+l,imt))
rkn12=rkn12+storec(i,1,1)*(1./2.)*(hz(i,imt)**2-hz(i+l,imt)**2)
rkn22=rkn22+storec(i,1,1)*(1./3.)*(hz(i,imt)**3-hz(i+l,imt)**3)
rkn44=rkn44+storec(i,1,1)*(1./12.)*(hz(i,imt)-hz(i+l,imt))
rkn55=rkn55+storec(i,1,1)*(1./4.)*(hz(i,imt)-hz(i+l,imt))
rkn661=rkn661+storec(i,3,3)*(1./3.)*(hz(i,imt)**3-hz(i+1,imt)**3)
rkn662=rkn662+shear(i,1,1)*(1./12.)*(hz(i,imt)-hz(i+l,imt))
rlp=rlp+shear(i,l,l)*(hz(i,imt)-hz(i+l,imt))
enddo
c----------------------------------------------------------
c The beams natural stiffness matrix
c----------------------------------------------------------
snb (1 , 1) =rkn11
snb(1,2)=rkn12
snb (1 ,3) =0 .
snb(1 ,4)=0.
snb (1 , 5) =0 .
snb (1 ,6) =0 .
312 Programming aspects
snb(2,2)=rkn22
snb(2,3)=O.
snb(2,4)=O.
snb(2,5)=O.
snb(2,6)=O.
snb(3,3)=rkn33
snb(3,4)=O.
snb(3,5)=O.
snb(3,6)=O.
snb(4,4)=rkn44
snb(4,5)=O.
snb(4,6)=O.
snb(5,5)=rkn55
snb(5,6)=O.
snb(6,6)=rkn661+rkn662
an (1 , 1)=-1.
an (1 ,7)=1.
an(2,5)=-1.
an(2,11)=1.
an(3,3)=2./bl(id)
an(3,5)=-1.
an(3,9)=-2./bl(id)
an(3,11)=-1.
an(4,6)=1.
an(4,12)=-1.
an(5,2)=2./bl(id)
an(5,6)=1.
an(5,8)=-2./bl(id)
an(5,12)=1.
an(6,4)=-1.
an(6,10)=1.
In the above array, bl (id) represents the length of element id. A series of
matrix multilplications then yield the local and global elemental cartesian
stiffnesses. The same computing procedures are also followed in setting
up the natural thermal load and transforming it to the local and global
9.5 Assembly of the global stiffness matrices 313
coordinates. Then, the global thermal load vector is assembled into the
global loading vector via
c
c ASSEMBLE THE THERMAL LOAD VECTOR
c
do 2040 in=l,nne
ip=node {id, in)
do 2060 iu=l,ndf
i=indx{ip,iu)
if(i.eq.O) goto 2060
v(i)=v(i)+sttin-l)*ndf+iu)
2060 continue
2040 continue
At this point, the global load vector including the applied loads and
temperatures is defined and stored into the one dimensional array v{nu),
where nu denotes the total number of unknowns.
c ..... M A I N ASS E M B L Y L 0 0 P
do 140 in = 1,noel(id)
ip = node(id,in)
do 160 iu = 1,ndf
i = indx(ip,iu)
if(i.eq.O) goto 160
do 170 inn = 1,noel(id)
ipp = node(id,inn)
do 180 ju = 1,ndf
j = indx(ipp,ju)
if(i.gt.j) goto 180
ij=maxa(j)+(j-i)
a(ij)=a(ij)+k(iu,in,ju,inn)
180 continue
314 Programming aspects
170 continue
160 continue
140 continue
All elemental cartesian stiffnesses are assembled into the one dimen-
sional array a(nwk), where nwk is the total number of entries in the sky-
line.
9.7 Postprocessing
Following computation of the global nodal unknowns, and if desired by
the user, the program proceeds to postprocessing of the results by calling
routine bepost. The routine first computes the generalized natural forces
and moments and subsequently estimates the cartesian stress resultants as
they refer the local elemental coordinate system. Then, it computes all
natural energies. These operations are performed for all elements via the
commands (follow also the comments before each task)
c-----------------------------------------------------------
c ..... COMPUTE R = a R ELEMENT NATURAL MODES
c N N
c 6x1 6x12 12x1
c-----------------------------------------------------------
9.8 Geometrical stiffness 315
call zerov(rhoen,6*1)
call mxv(dan,6,rhoel,12,rhoen)
c-----------------------------------------------------------
c COMPUTE P = k R ELEMENT NATURAL FORCES
c N NE N
c 6x1 6x6 6x1
c-----------------------------------------------------------
call zerov(fnat,6*1)
call mxv(dd,6,rhoen,6,fnat)
c-----------------------------------------------------------
c GET NATURAL ENERGIES 1/2 Pn rn
c--------------------------------------------------~--------
en1=(1./2.)*fnat(1)*rhoen(1)
en2=(1./2.)*fnat(2)*rhoen(2)
en3=(1./2.)*fnat(3)*rhoen(3)
en4=(1./2.)*fnat(4)*rhoen(4)
en5=(1./2.)*fnat(5)*rhoen(5)
en6=(1./2.)*fnat(6)*rhoen(6)
all=en1+en2+en3+en4+en5+en6
c-----------------------------------------------------------
c ..... COMPUTE R = a R ELEMENT NATURAL MODES
c N N
c 6x1 6x12 12x1
c-----------------------------------------------------------
call zerov(rhoen,6*1)
call mxv(dan,6,rhoel,12,rhoen)
c-----------------------------------------------------------
316 Programming aspects
c
c EFFECTS of PNl
c
if(igeoms.eq.'simpgs'.or.igeoms.eq.'psimpgs') then
call zerov(copnl,12*12)
copnl(2,2)=(pn(1)/bl(id
copnl(3,3)=(pn(1)/bl(id
copnl(8,8)=(pn(1)/bl(id
copnl(9,9)=(pn(1)/bl(id
9.8 Geometrical stiffness 317
copn1(2,8)=-(pn(1)/bl(id))
copn1(3,9)=-(pn(1)/bl(id))
copn1(8,2)=-(pn(1)/bl(id))
copn1(9,3)=-(pn(1)/bl(id))
endif
c
c EFFECTS of PN2
c
if(igeoms.eq.'psimpgs') then
call zerov(copn2,12*12)
copn2(4,2)=-(pn(2)/bl(id))
copn2(4,8)=(pn(2)/bl(id))
copn2(6,4)=-(pn(2)/2.)
copn2(6,10)=-(pn(2)/2.)
copn2(10,2)=(pn(2)/bl(id))
copn2(10,8)=-(pn(2)/bl(id))
copn2(12,4)=(pn(2)/2.)
copn2(12,10)=(pn(2)/2.)
copn2(2,4)=-(pn(2)/bl(id))
copn2(8,4)=(pn(2)/bl(id))
copn2(4,6)=-(pn(2)/2.)
copn2(10,6)=-(pn(2)/2.)
copn2(2,10)=(pn(2)/bl(id))
copn2(8,10)=-(pn(2)/bl(id))
copn2(4,12)=(pn(2)/2.)
copn2(10,12)=(pn(2)/2.)
c
c EFFECTS of PN4
c
call zerov(copn4,12*12)
copn4(4,3)=(pn(4)/bl(id))
copn4(3,4)=(pn(4)/bl(id))
copn4(4,9)=-(pn(4)/bl(id))
copn4(9,4)=-(pn(4)/bl(id))
copn4(5,4)=-(pn(4)/2.)
copn4(4,5)=-(pn(4)/2.)
copn4(5,10)=-(pn(4)/2.)
copn4(10,5)=-(pn(4)/2.)
copn4(10,3)=-(pn(4)/bl(id))
318 Programming aspects
copn4(3,10)=-(pn(4)/bl(id
copn4(10,9)=(pn(4)/bl(id
copn4(9,10)=(pn(4)/bl(id
copn4(11,4)=(pn(4)/2.)
copn4(4,ll)=(pn(4)/2.)
c
c EFFECTS of PN6
c
call zerov(copn6,12*12)
copn6(5,2)=(pn(6)/bl(id
copn6(5,8)=-(pn(6)/bl(id
copn6(6,3)=(pn(6)/bl(id
copn6(6,9)=-(pn(6)/bl(id
copn6(11,2)=-(pn(6)/bl(id
copn6(11,8)=(pn(6)/bl(id
copn6(12,3)=-(pn(6)/bl(id
copn6(12,9)=(pn(6)/bl(id
copn6(2,5)=(pn(6)/bl(id
copn6(8,5)=-(pn(6)/bl(id
copn6(3,6)=(pn(6)/bl(id
copn6(9,6)=-(pn(6)/bl(id
copn6(2,ll)=-(pn(6)/bl(id
copn6(8,ll)=(pn(6)/bl(id
copn6(3,12)=-(pn(6)/bl(id
copn6(9,12)=(pn(6)/bl(id
endif
Finally, it adds all separate contributions and transforms the local geomet-
rical stiffness to the global coordinate system. This is accomplished via
commands
do i=l,12
do j=l,12
kgeom(i,j)=kngl(i,j)+copnl(i,j)+
copn2(i,j)+copn4(i,j)+copn6(i,j)
enddo
enddo
c
c T
c FORM k = T k T
9.9 Assembly of the geometrical stiffness 319
c G 06 G 06
c (12x12) (12x12)(12x12)(12x12)
c
call zerov(multll,12*12)
call zerov(gg,12*12)
call multtt(kgeom,t04,multl1,12,12,12)
call multtt(t04t,multll,gg,12,12,12)
c..... MA I N ASS EM B L Y L0 0P
do 140 in = 1,noel(id)
ip = node(id,in)
do 160 iu = 1,ndf
i = indxCip,iu)
if(i.eq.O) goto 160
do 170 inn = 1,noel(id)
ipp = node(id,inn)
do 180 ju = 1,ndf
j = indx(ipp,ju)
if(i.gt.j) goto 180
ij=maxa(j)+(j-i)
b(ij)=b(ij)+k(iu,in,ju,inn)
180 continue
170 continue
160 continue
140 continue
9.10 Scholium
We emphasize that only the general logic of the computer program is given
in this chapter. As stressed in the beginning, we believe that if the reader
320 Programming aspects
comprehends the theory presented he can easily follow the computer pro-
gram and in addition refine it and expand it according to his own flavor
and technique.
Appendix A
X2 -
[ Y2 -
Xl]
YI = [XI2]
Yl2 , (A.l)
z2 - zl zl2
and the differences of the global coordinates between point P* and node 1
as
Xl]
Y: - YI
X* -
Xle]
[ = [Yle . (A.2)
Z - Zl Zle
In addition
322 Geometry of the beam element in space
Xl -
[ Yl - Y:X*] =
[xel]
Yel . (A.3)
Zl - Z Zel
The entries of the first row in To comprise components of the unit vector
PX' = CXIX]
[ Cx' Y = l1 [X12]
Y12 , (A.4)
CX' Z Z12
where
where 0 is the area of the triangle formed by the two nodes and the extra
point defined as
1
0= -JC12
2
+ C2 2 + C3 2 , (A.7)
in which
Finally, the components of the unit vector Pyl are simply given by
We are now in the position to define the matrix of direction cosines via
cx, X c x' Y zl
c x'
To = [ cy'x Cy'y cy'z . (A.lO)
Cz'x Cz'y Cz'z
Note that matrix To is derived in an identical manner for the 3-node shell
element by simply replacing the extra point P* with node 1, and the beam
nodes 1, 2 with the triangular nodes 2, 3, respectively.
Appendix B
Included with in the floopy disk accompanying the book is a model com-
puter program for static and buckling analyses of isotropic and laminated
composite beams, frames and large three-dimensional beam assemblies.
The name of the program is beam1.f and is written in standard Fortran
77. It can be compiled and run on any computer with Fortran 77. The pro-
gram was originally written and compiled on a Sun Sparcstation 10 using
the command 17 beam1.f.
In addition to the source code, four example problems are provided to
familiarize the user with the input data and the execution of the computer
program. All axamples were described in the text.
The following files are included in the floopy disk:
[11] RW. Clough. The finite element method: A personal view of its
original formulation. in From Finite elements to the Troll platform,
Ivar Holland 70th anniversary, pages 89-97, 1994.
[12] M.J. Turner, RW. Clough, RW. Martin and L. Topp. Stiffness and
deflection analysis of complex structures. J. Aero. Sci., 25:805-823,
1956.
[13] RW. Clough. The finite element method in plane stress analysis.
Proc. 2nd ASCE Conf. on Electronic Computation, Pitsburg, Pa.,
Sept. 1960.
[18] KK Gupta and J.L. Meek. A brief history of the beginning of the
finite element method. Int. J. Numer. Meth. Engrg., 39:3761-3774,
1996.
[19] O.C. Zienkiewicz and Y.K Cheung. The Finite Element Method in
Continuum and Structural Mechanics, volume 1. McGraw-Hill, New
York, 1965.
[21] O.C. Zienkiewicz and Y.K Cheung. Finite elements in the solution of
field problems. The Engineer, 220:507-510, 1965.
BIBLIOGRAPHY 329
[22] O.C. Zienkiewicz and RL. Taylor. The Finite Element Method, vol-
ume 1. McGraw-Hill, fourth edition, 1991.
[23] O.C. Zienkiewicz. Origins, milestones and directions of the finite ele-
ment method - a personal view. Archives of Comput. Meth. Engnrg.,
2:1-48, 1995.
[24] Lord Rayleigh. On the theory of resonance. Trans. Roy. Soc., A161:77-
118, London, 1870.
[25] W. Ritz. Uber einer neue methode zur losung gewissen variations-
probleme der mathematischen physik. J. Reine angew. Math., 135:1-
61, 1909.
[26] B. Siuru and J.D. Busick. Future Flight: The next generation of air-
craft technology. TAB AERO, Blue Ridge Summit, PA, 1994.
[30] S.P. Timoshenko and J.N. Goodier. Theory of Elasticity. 3rd Edn.,
McGraw-Hill, New York, 1970.
[31] S. Timoshenko. Strength of materials, Part II. 2nd Edn., Van Nos-
trand, Amsterdam, 1974.
[32] F.P. Beer and E.R Johnston Jr. Mechanics of Materials. Mc Graw-
Hill, 1981.
[33] A. Venkatesh and K.P. Rao. A laminated anisotropic curved beam and
shells stiffening element. Comput. Struct., 15:197-200, 1982.
[34] F. Yuan and RE. Miller. A new finite element for laminated composite
beams. Comput. Struct., 31:737-745, 1989.
[36] A.T. Chen and T.Y. Yang. Static and dynamic formulation of a sym-
metrically laminated beam finite element for a microcomputer. J.
Compos. Mater., 19:459-475, 1985.
[46] J.H. Argyris and L. Tenek. A natural triangular layered element for
bending analysis of isotropic, sandwich, laminated composite and hy-
bride plates. Comput. Methods Appl. Meeh Engrg, 109:197-218, 1993.
[47] J.H. Argyris and L. Tenek. Buckling of multilayered composite plates
by natural shear deformation matrix theory. Comput. Methods Appl.
Meeh Engrg, 111:37-59, 1994.
BIBLIOGRAPHY 331
[48] J.H. Argyris and L. Tenek. Linear and geometrically nonlinear bending
of isotropic and multilayered composite plates by the natural mode
method. Comput. Meth. Appl. Meeh Engrg., 113:207-251, 1994.
[49] J.H. Argyris and L. Tenek. Nonlinear and chaotic oscillations of com-
posite plates and shells under periodic heat load. Comput. Methods
Appl. Meeh Engrg, 122:351-377, 1995.
[51] J.H. Argyris and L. Tenek. An efficient and locking-free flat anisotropic
plate and shell triangular element. Comput. Meth. Appl. Meeh. Engrg.,
118:63-119, 1994.
[60] D.J. Allman. Evaluation of the constant strain triangle with drilling
rotations. Int. J. Num. Meth. Eng., 26:2645-2655, 1988.
[63] J.H Argyris and D.W. Scharpf. Large deflection analysis of prestressed
networks. ASCE J. Struet. Div., 98:633-654, 1972.
[64] J.H Argyris and P.C. Dunne. A simple theory of geometrical stiffness
with applications to beam and shell problems. lSD-Report No 183,
Stuttgart, 1975.
[67] T.J.R. Hughes and W.K Liu. Nonlinear finite lement analysis of shells
part ii: two-dimensional shells. Comput. Meth. Appl. Meeh. Engrg.,
27:167-182, 1981.
[69] J.C. Simo and D.D. Fox. On a stress resultant geometrically exact
shell model. part ii: the linear theory; computational aspects. Comput.
Meth. Appl. Meeh. Engrg., 73:53-92, 1989.
[71] KJ. Bathe and E.N. Dvorkin. A formulation of general shell elements-
the use of mixed interpolation of tensorial components. Int. J. Numer.
Meth. Eng., 22:697-722, 1986.
BIBLIOGRAPHY 333
[72] K.P. Rao. A rectangular laminated anisotropic shallow thin shell finite
element. Comput. Meth. Appl. Mech. Engrg., 15:13-33, 1978.
[73] R.H. MacNeal and R.L. Harder. A proposed standard set of problems
to test finite element accuracy. Finite Elements Anal. Des., 1:3-20,
1985.
[75] N.D. Phan and J.N. Reddy. Analysis of laminated composite plates
using a higher-order shear deformation theory. Int. J. Numer. Meth.
Eng., 21:2201-2219, 1985.
[78] A.K Noor. Stability of multilayered composite plates. Fibre Sci. Tech-
nol., 8:81-89, 1975.
[85] E. Hinton and D.R.J. Owen. Finite Element Software for Plates and
Shells. Pineridge Press, Swansea, U.K., 1984.
[86] User's guide. FORTRAN 3.0.1. Sun Microsystems, Mountain View,
California, 1994.
[87] User's guide. PATRAN - Vol. 1. PDA Engineering, Santa Ana, CA,
1984.
Index
warping, 224
weight sensitive structures, 38
Mechanics
SOUD MECHANICS AND ITS APPLICATIONS
Series Editor: G.M.L. Gladwell
44. D.A. Hills, P.A. Kelly, D.N. Dai and A.M. Korsunsky: Solution of Crack Problems. The
Distributed Dislocation Technique. 1996 ISBN 0-7923-3848-0
45. V.A. Squire, R.J. Hosking, A.D. Kerr and P.J. Langhorne: Moving Loads on Ice Plates. 1996
ISBN 0-7923-3953-3
46. A. Pineau and A. Zaoui (eds.): IUTAM Symposium on Micromechanics of Plasticity and
Damage of Multiphase Materials. Proceedings of the IUTAM Symposium held in Sevres,
Paris, France. 1996 ISBN 0-7923-4188-0
47. A. Naess and S. Krenk (eds.): IUTAM Symposium on Advances in Nonlinear Stochastic
Mechanics. Proceedings of the IUTAM Symposium held in Trondheim, Norway. 1996
ISBN 0-7923-4193-7
48. D. Iean and A. Scalia: Thermoelastic Deformations. 1996 ISBN 0-7923-4230-5
49. J. R. Willis (ed.): IUTAM Symposium on Nonlinear Analysis of Fracture. Proceedings of the
IUTAM Symposium held in Cambridge, UK 1997 ISBN 0-7923-4378-6
50. A. Preumont: Vibration Control ofActive Structures. An Introduction. 1997
ISBN 0-7923-4392-1
51. G.P. Cherepanov: Methods of Fracture Mechanics: Solid Matter Physics. 1997
ISBN 0-7923-4408-1
52. D.H. van Campen (ed.): IUTAM Symposium on Interaction between Dynamics and Control in
Advanced Mechanical Systems. Proceedings of the IUTAM Symposium held in Eindhoven,
The Netherlands. 1997 ISBN 0-7923-4429-4
53. N.A. Fleck and A.C.F. Cocks (eds.): IUTAM Symposium on Mechanics of Granular and
Porous Materials. Proceedings of the IUTAM Symposium held in Cambridge, U.K. 1997
ISBN 0-7923-4553-3
54. J. Roorda and N.K. Srivastava (eds.): Trends in Structural Mechanics. Theory, Practice,
Education. 1997 ISBN 0-7923-4603-3
55. Yu. A. Mitropolskii and N. Van Dao: Applied Asymptotic Methods in Nonlinear Oscillations.
1997 ISBN 0-7923-4605-X
56. C. Guedes Soares (ed.): Probabilistic Methodsfor Structural Design. 1997
ISBN 0-7923-4670-X
57. D. Franyois, A. Pineau and A. Zaoui: Mechanical Behaviour of Materials. Volume I:
Elasticity and Plasticity. 1998 ISBN 0-7923-4894-X
58. D. Franyois, A. Pineau and A. Zaoui: Mechanical Behaviour of Materials. Volume II:
Viscoplasticity, Damage, Fracture and Contact Mechanics. 1998 ISBN 0-7923-4895-8
59. L. T. Tenek and J. Argyris: Finite Element Analysis for Composite Structures. 1998
ISBN 0-7923-4899-0