Professional Documents
Culture Documents
ERASMO CARRERA
Aeronautic and Space Engineering Department
Int. J. Appl. Mechanics 2010.02:117-143. Downloaded from www.worldscientific.com
GAETANO GIUNTA
by MONASH UNIVERSITY on 07/17/13. For personal use only.
This paper proposes several axiomatic rened theories for the linear static analysis of
beams made of isotropic materials. A hierarchical scheme is obtained by extending plates
and shells Carreras Unied Formulation (CUF) to beam structures. An N -order approxi-
mation via Mac Laurins polynomials is assumed on the cross-section for the displacement
unknown variables. N is a free parameter of the formulation. Classical beam theories,
such as Euler-Bernoullis and Timoshenkos, are obtained as particular cases. According
to CUF, the governing dierential equations and the boundary conditions are derived
in terms of a fundamental nucleo that does not depend upon the approximation order.
The governing dierential equations are solved via the Navier type, closed form solution.
Rectangular and I-shaped cross-sections are accounted for. Beams undergo bending and
torsional loadings. Several values of the span-to-height ratio are considered. Slender
as well as deep beams are analysed. Comparisons with reference solutions and three-
dimensional FEM models are given. The numerical investigation has shown that the
proposed unied formulation yields the complete three-dimensional displacement and
stress elds for each cross-section as long as the appropriate approximation order is con-
sidered. The accuracy of the solution depends upon the geometrical parameters of the
beam and loading conditions.
1. Introduction
Many primary and secondary structural elements, such as aircraft wings, helicopter
rotor blades, robot arms or structures in civil construction, can be idealised as
beams. An accurate prediction of the mechanics of beams plays a paramount role
Corresponding author
117
March 10, 2010 13:43 WSPC-255-IJAM SPI-J108 S1758825110000500
in their correct, safe and optimal design and utilisation. Engineering elds, such
as Aeronautics, characterised by high levels of technology also require more and
more accurate models. Beam modelling therefore represents an important topic of
research.
Euler-Bernoullis (EB) and Timoshenkos (TB) (see [Timoshenko, 1921], [Timo-
shenko, 1922] and [Timoshenko and Goodier, 1970]) theories represent the classical
one-dimensional models (CMs) for the bending analysis of beam structures made
of isotropic materials. The cross-section is considered to be rigid on its plane. EB
discards the shear deformation while TB accounts for a constant value. Material
stiness coecients should be opportunely reduced. As the beam becomes less and
Int. J. Appl. Mechanics 2010.02:117-143. Downloaded from www.worldscientific.com
less slender, its mechanics is more and more three-dimensional and CMs do not yield
accurate results. The accuracy also depends on the material properties and the cross-
sectional geometry. In the case of I-shaped cross-sections, non-classical deformations
by MONASH UNIVERSITY on 07/17/13. For personal use only.
are important for the global elastic behaviour of the beam. As far as torsional prob-
lems are concerned, Saint-Venants and Prandtls models represent the classical
solutions. Saint-Venant [1855] postulated the displacement eld assuming that the
deformation of a twisted shaft consists of a rigid rotation, as in the case of a circular
cross-section and a warping that is constant along the beam axis. Prandtl [1903]
introduced the membrane analogy for the study of torsion: the problem is reduced to
the study of a homogeneous membrane supported at the edges, with the same out-
line as the beam cross-section and subjected to a uniform tension at the edges and a
uniform transverse bending pressure. Improvements in CMs have been proposed to
account for non-classical eects and non-conventional materials. Some of them are
discussed hereafter. Stephen and Levinson [1979] accounted for a linear variation of
the shear deformation. Within the framework of the nite element method, Kant
and Manjunath [1989] proposed kinematic elds with a second- and a third-order
variation of the longitudinal displacement. The transversal displacement is con-
stant. Beams made of composite and sandwich materials were investigated. These
authors accounted for a linear variation of the transversal displacement in [Kant
and Manjunath, 1990]. Models with quadratic and cubic variation were addressed
in [Manjunath and Kant, 1993]. Transverse stresses were computed via the integra-
tion of the indenite equilibrium equations through three dierent numerical tech-
niques: the direct integration method, the forward and central direct nite dierence
method and the exact curve tting method. Vinayak et al. [1996] and Prathap et al.
[1996] proposed a theory for composite beams in which the longitudinal displace-
ment is a cubic function of the coordinate transversal to the axis of the beam, while
the transversal displacement varies quadratically. Rand [1998] formulated a model
with ve degrees of freedom: three cross-sectional displacements, a twist angle,
which are constant above the cross-section, and a three-dimensional warping func-
tion. Solid and thin-walled beams made of composite materials were investigated.
The author addressed the importance of the warping function in [Rand, 2000]. Sinus
nite elements were proposed by Ganapathi et al. [1999] to study sandwich beams
March 10, 2010 13:43 WSPC-255-IJAM SPI-J108 S1758825110000500
that account for transverse shear and warping. Starting from that work, Vidal and
Polit [2008] developed a family of nite elements considering a Heaviside function
or the double superposition hypothesis by Li and Liu [1997]. Laminated rectan-
gular beams were analysed. Gjelsvik [1981] presented an isotropic beam theory in
which the beam is split into plate segments. The continuity of the displacements is
imposed via geometric considerations, while global and local equilibrium is obtained
through the principle of virtual work. Wagner and Gruttmann [2002] formulated
a prismatic beam model to study exural shear stresses in thin-walled isotropic
composite beams undergoing torsionless bending. Estivalezes and Barrau [1998],
using a homogenisation method, formulated a simplied calculation method appli-
Int. J. Appl. Mechanics 2010.02:117-143. Downloaded from www.worldscientific.com
functions need to be assumed. Classical models, such as EB and TB, can be retrieved
as particular cases. Naviers, closed form solution is adopted to solve the governing
dierential equations. Slender and deep beams are investigated. The cross-section
is considered to be rectangular or I-shaped. Beams are subjected to bending or tor-
sional loadings. The proposed models are validated towards reference exact solutions
or three-dimensional FEM models.
2. Preliminaries
A beam is a structure whose axial extension, l, is predominant compared to any
Int. J. Appl. Mechanics 2010.02:117-143. Downloaded from www.worldscientific.com
N k
= k , (2.1)
k=1
with
k = {(x, y) : xk1 x xk2 ; y1k y y2k }. (2.2)
are considered, {(xki , yjk )
: i, j = 1, 2} being the coordinates of the corner points
of each k sub-domain. Through the paper, superscript k represents the cross-
section sub-domain index, while as subscript it stands for summation over the range
(1, Nk ). Figures 1(a) and 1(b) present the geometrical data of a rectangular and an
I-shaped cross-section. The reference system is also shown there: x and y axes are
two orthogonal directions laying on the cross-section. The z coordinate is coincident
(a) (b)
Fig. 1. Reference system and geometry of (a) a rectangular and (b) an I-shaped cross-section.
March 10, 2010 13:43 WSPC-255-IJAM SPI-J108 S1758825110000500
n = yz , n = yz , (2.4)
zz zz
by MONASH UNIVERSITY on 07/17/13. For personal use only.
where Dnp , Dnz , and Dp are the following dierential matrix operators
0 0 x z 0 0 x 0 0
Dnp =
0 0
y , D nz = 0
z 0 , Dp = 0
y 0 .
(2.8)
0 0 0 0 0 z y x 0
In the case of beams made of linear elastic materials, the generalised Hooke law
holds. Its compact vectorial form is
= C. (2.9)
p = Cpp p + Cpn n ,
n = Cnp p + Cnn n . (2.10)
March 10, 2010 13:43 WSPC-255-IJAM SPI-J108 S1758825110000500
N Nu F
0 1 F1 = 1
1 3 F2 = x F3 = y
2 6 F4 = x2 F5 = xy F6 = y 2
3 10 F7 = x3 F8 = x2 y F9 = xy 2 F10 = y 3
... ... ...
F (N 2 +N +2) = xN F (N 2 +N +4) = xN1 y . . .
(N+1)(N+2)
N 2
2 2
ux = ux1 ,
uy = uy1 , (3.4)
uz = uz1 + uz2 x + uz3 y
and EB
ux = ux1 ,
uy = uy1 , (3.5)
uz = uz1 ux1,z x uy1,z y,
was formulated to describe the bending mechanics. TB model accounts for constant
shear stress and strain components. In the case of classical models and rst-order
approximation, the material stiness coecients Cij in Eqs. (2.11) should be cor-
rected in order to contrast a phenomenon known in literature as Poissons locking
(see [Carrera and Brischetto, 2008a and 2008b]). This correction is presented in
Appendix A.
are obtained in terms of the displacement components and their derivatives through
the Principle of Virtual Displacements (PVD)
by MONASH UNIVERSITY on 07/17/13. For personal use only.
Li = Lp + Ll . (4.1)
Li represents the strain energy. Lp and Ll stand for the work due to a surface
loading, pk , and a line loading, lk that act on a k sub-domain. stands for a virtual
variation.
+ (Dnp F I)T Cnp (Dp Fs I) + (Dnp F I)T Cnn (Dnp Fs I) + (Dnp F I)T CnnFs Dnz
+ (Dp F I)T Cpp (Dp Fs I) + (Dp F I)T Cpn (Dnp Fs I) + (Dp F I)T Cpn Fs Dnz ]
T
d us dz + u F [Cnp (Dp Fs I) + Cnn (Dnp Fs I) + Cnn Fs Dnz ]
z=l
d us |z=0 . (4.3)
I is the unit matrix. The virtual variation of the strain energy in a compact vectorial
form is
Li = uT K s us dz + uT s us |z=0
z=l
. (4.4)
l
March 10, 2010 13:43 WSPC-255-IJAM SPI-J108 S1758825110000500
s 2
Kxx = C11 E,x s,x + C66 E,y s,y C55 E s ,
z 2
s 2
Kyy = C22 E,y s,y + C66 E,x s,x C44 E s ,
z 2
s 2
Kzz = C44 E,y s,y + C55 E,x s,x C33 E s ,
z 2
s s
Kxy = C12 E,x s,y + C66 E,y s,x , Kyx = C12 E,y s,x + C66 E,x s,y ,
s s
Int. J. Appl. Mechanics 2010.02:117-143. Downloaded from www.worldscientific.com
Kxz = (C13 E,x s C55 E s,x ) , Kzx = (C13 E s,x C55 E,x s ) ,
z z
s s
Kyz = (C23 E,y s C44 E s,y ) , Kzy = (C23 E s,y C44 E,y s ) .
by MONASH UNIVERSITY on 07/17/13. For personal use only.
z z
(4.5)
pkT = {pk
xx pk
yy pk
xy pk
yx pk
xz pk
yz }. (4.8)
They act as shown in Fig. 2. The lateral surfaces {Sk : = x, y} of the beam are
dened on the basis of the normal versor {nk : = x, y}. A normal versor with
March 10, 2010 13:43 WSPC-255-IJAM SPI-J108 S1758825110000500
Fig. 2. Components of a surface loading lateral surfaces and normal vectors of the beam.
the same orientation as x or y axis identies a positive lateral surface. The external
virtual work due to p is
Lp = Lkp + Lkp + Lkp + Lkp + Lkp + Lkp k . (4.9)
xx yy xz yz xy yx
where
xk
2
+
(Ekx , Ekx ) = (F (x, y2k ), F (x, y1k )) dx,
xk
1
(4.11)
y2k
+
(Eky , Eky ) = (F (xk2 , y), F (xk1 , y)) dy.
y1k
March 10, 2010 13:43 WSPC-255-IJAM SPI-J108 S1758825110000500
lkT = {lxx
k k
lyy k
lxy k
lyx k
lxz k }.
lyz (4.12)
= [pk ky
xx E + pk kx
yx E
k
+ lxx F (xkl , ylk ) + lyx
k
F (xkl , ylk )]k ,
xx xx yx yx
uy : (C12 E,y s,x + C66 E,x s,y )uxs + (C22 E,y s,y + C66 E,x s,x )uys
by MONASH UNIVERSITY on 07/17/13. For personal use only.
uz : (C13 E s,x C55 E,x s )uxs,z (C23 E s,y C44 E,y s )uys,z
+ (C44 E,y s,y + C55 E,x s,x )uzs C33 E s uzs,zz
= [pk ky
xz E + pk kx
yz E
k
+ lxz F (xkl , ylk ) + lyz
k
F (xkl , ylk )]k .
xz xz yz yz
(4.15)
For a xed approximation order, the nucleo has to be expanded versus the indexes
and s in order to obtain the governing equations and the boundary conditions
that concern the desired model.
upon the assumption that the external loadings vary towards z in the following
manner
k k
Pxx sin(z)
Lxx sin(z)
k
k
P yy sin(z)
L yy sin(z)
P sin(z)
k
L sin(z)
k
k xy k xy
p = , l = . (5.2)
P k
sin(z)
L k
sin(z)
yx
yx
P k cos(z)
Lk cos(z)
xz
xz
k
k
Pyz cos(z) Lyz cos(z)
Int. J. Appl. Mechanics 2010.02:117-143. Downloaded from www.worldscientific.com
This last assumption does not represent a loss in generality, since a generic loading
can be approximated via its Fouriers series expansion (see [Carrera and Giunta,
by MONASH UNIVERSITY on 07/17/13. For personal use only.
Uy : (C12 E,y s,x + C66 E,x s,y )Uxs + (C22 E,y s,y + C66 E,x s,x
+ 2 C44 E s )Uys (C23 E,y s C44 E s,y )Uzs
k kx k ky
= [Pyy E + Pxy E + Lk k k k k k
yy F (xl , yl ) + Lxy F (xl , yl )]k ,
yy yy xy xy
Uz : (C13 E s,x C55 E,x s )Uxs (C23 E s,y C44 E,y s )Uys
+ (C44 E,y s,y + C55 E,x s,x + 2 C33 E s )Uzs
k ky k kx
= [Pxz E + Pyz E + Lk k k k k k
xz F (xl , yl ) + Lyz F (xl , yl )]k .
xz xz yz xz
(5.5)
March 10, 2010 13:43 WSPC-255-IJAM SPI-J108 S1758825110000500
For a xed approximation order, the algebraic system has to be assembled according
to the summation indexes and s. Its solution yields the maximal displacement
amplitudes. The strains are retrieved by the geometric relations, Eqs. (2.10), and
the stresses via the generalised Hooke law, Eqs. (2.7).
ings. The half-wave number is equal to the unit. The results are validated towards
elasticity solutions or three-dimensional FEM models.
by MONASH UNIVERSITY on 07/17/13. For personal use only.
2 a2 zz xx 2 a xz
zz = 1+ , xx = 1+ , xz = 1+ . (6.1)
6 l2 Pxx Pxx 3 l Pxx
They are all evaluated at y = 0. In the case of zz and xx , z equals l/2, while for
xz , z = 0. Several values of the x coordinate are considered. Tables 2 and 3 present
stresses for l/a = 100 and l/a = 50. Since bending mechanics is predominant,
the proposed theories all accurately predict the bending stress zz . The rst-order
approximation yields a constant stress component xx . The third-order model accu-
rately describes the transverse shear stress component, but it does not satisfy the
by MONASH UNIVERSITY on 07/17/13. For personal use only.
boundary condition for the case of xx . In this last case, the fourth-order is required
in order to match the reference solution. Figures 4 (a) to 4(c) present the variation
in the stress components versus x/a for l/a 100. The theories all match the exact
0.5 0.5
0.4 TGB 0.4
EB
0.3 TB 0.3
0.2 N=1 0.2
0.1 N=2 0.1
N=3
x/a
x/a
0 N=4 0
-0.1 -0.1 TGB
-0.2 -0.2 N=1
-0.3 -0.3 N=2
N=3
-0.4 -0.4 N=4
-0.5 -0.5
-1 -0.5 0 0.5 1 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4
zz
*
*xx
(a) (b)
0.5
0.4
0.3
0.2 TGB
0.1 TB
0 N=1
x/a
N=2
-0.1 N=3
-0.2 N=4
-0.3
-0.4
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
*xz
(c)
Fig. 4. (a) ,
zz (b)
xx and (c)
xz versus x/a for l/a 100, bending loading and rectangular
cross-section.
March 10, 2010 13:43 WSPC-255-IJAM SPI-J108 S1758825110000500
solution in the case of zz . The third-order model yields a boundary value of xx
by MONASH UNIVERSITY on 07/17/13. For personal use only.
that is worse than that computed via the second-order one. Table 4 presents the
results regarding the case in which l/a is equal to ten. zz is no longer symmetric
versus the x axis. An expansion order as high as ve is required to match the exact
solution up to the fth signicant digit. The fourth-order model is also very accu-
rate. The results for l/a equal to ve are presented in Table 5. An accuracy of the
rst ve signicant digits calls for the sixth-order approximation. The fourth-order
model is also accurate. The results regarding the case in which l/a equals two are
presented in Table 6. The higher order eects become relevant since the problem is
governed by three-dimensional mechanics. Figures 5 (a) to 5(c) present the variation
of the stress components above the cross-section. The stress component zz is no
longer linear.
In the second case, a torsional moment is applied to the beam by means of two
line loadings of equal maximal amplitude L3+ 1
xx = Lxx = Lxx , as shown in Fig. 6.
Classical models, such as EB and TB, yield displacement, strain and stress elds
equal to zero for any case in which the resultant of the external loadings is equal
to zero, even though the resultant moment is not. This is due to the fact that the
0.5 0.5
0.5
0.4 0.4
0.3 0.3
0.2 0.4 -0.4 0.2
0.1 0.1
x/a
x/a
(a) (b)
0.5
0.4
0.3
0.2 TGB
0.1 TB
0 N=1
x/a
N=2
-0.1 N=3
-0.2 N=5
-0.3
-0.4
-0.5
-0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
*
xz
(c)
Fig. 5. (a) ,
zz (b)
xx and (c)
xz versus x/a for l/a = 2, bending loading and rectangular
cross-section.
March 10, 2010 13:43 WSPC-255-IJAM SPI-J108 S1758825110000500
loadings coherent to these models are all applied to the section centroid. The rst-
order approximation does not yield the trivial solution, although the results may
not be accurate. According to this model, the cross-section is too sti on its plane.
The following non-dimensionalised shear stress component yz is accounted for
ba2 a2 yz
yz = yz = , (6.2)
Mt (z = 0) l Lxx
where Mt (z = 0) is the torsion moment at z = 0. The shear stress component is
evaluated at 2x/a = 1 and y = z = 0. The results are assessed towards the solution
present in [Timoshenko and Goodier, 1970]. This solution is addressed as TGT and
it was obtained through Prandtls membrane analogy. For a beam with a square
cross-section, the reference solution is 4.807. This value does not depend upon the
span-to-height ratio. The results are presented in Table 7 for an approximation order
as high as 12 and several values of l/a. The twelfth-order approximation matches
the reference solution for l/a as low as 50. The two solutions dier by about 2.2%,
in the case of l/a = 2. In Table 8, the twelfth-order theory is compared to the
reference solution. The aspect ratio is assumed as a free parameter. The span-to-
height ratio ranges between 100 and two. The one-dimensional model matches the
exact solution for l/a as low as 50, the error being about 0.4%. For l/a = 5 and
b/a = 3, the error is about 19%. In the case of l/a = 2, it is equal to about 35%. It
should be noticed that, in these cases, l/b is 4/3 and one. The result for the cases
l/a 50 10 5 2
N = 12 4.807 4.804 4.796 4.700
N =9 4.782 4.774 4.746 4.514
N =7 4.858 4.857 4.853 4.784
N =5 4.934 4.937 4.946 4.959
N =3 3.000 2.996 2.985 2.877
N =1 3.000 2.988 2.951 2.720
March 10, 2010 13:43 WSPC-255-IJAM SPI-J108 S1758825110000500
0.5 0.5
Int. J. Appl. Mechanics 2010.02:117-143. Downloaded from www.worldscientific.com
0.375 0.375
0.25 0.25
0.125 0.125
by MONASH UNIVERSITY on 07/17/13. For personal use only.
x/a
x/a
0 0
-0.125 -0.125
N=3 N=3
-0.25 N=4 -0.25 N=4
-0.375 N=7 -0.375 N=7
N=9 N=9
-0.5 -0.5
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
*yz (y=0, z=0) *yz (y=0, z=0)
(a) (b)
Fig. 7.
yz versus x/a for (a) l/a 50 and (b) l/a = 2 in the case of square cross-section under
torsional loadings.
l/a = 2 and b/a = 3 is not reported since l/b becomes lower than unity and the
structure is, therefore, no longer a beam. Figures 7(a) and 7(b) present the variation
of yz towards x/a for l/a 50 and l/a = 2. A square cross-section is considered.
Although the third-order approximation accounts for a cubic variation versus x/a,
only the linear term is present for l/a 50. The next relevant contribution is due
to the fourth-order term. Fourth-, seventh- and ninth-order models can barely be
distinguished. In the case of l/a = 2, the third-order term becomes relevant. In
both cases, the main contribution to the solution is given by the fourth-order term.
Figures 8(a) and 8(b) show the out-of-plane warping of the cross-section at z = 0
for b/a = 3/2 and l/a = 100 and two. The results are computed via the ninth-order
approximation. The transverse displacement is non-dimensionalised with respect to
the maximum value of the warping of an elliptic section of equal area and aspect
ratio (see [Timoshenko and Goodier, 1970])
a2 b E
w = 2 2
w. (6.3)
Lxx l(b a ) 1 +
Warping in the case of a rectangular cross-section is higher than that of the equiv-
alent elliptic cross-section. As noted by Saint-Venant [1855], in the case of singly
connected sections and for a given cross-sectional area, the lower the polar moment
March 10, 2010 13:43 WSPC-255-IJAM SPI-J108 S1758825110000500
w* w*
3 3
2 2
1 1
0 0
-1 1 -1 1
-2 0 2y/b -2 0 2y/b
-1 2x/a 0 -1
-1 2x/a 0 -1
1 1
(a) (b)
Fig. 8. Transverse displacement uz
versus x/a and y/b for (a) l/a = 100 and (b) l/a = 2 at z = 0
Int. J. Appl. Mechanics 2010.02:117-143. Downloaded from www.worldscientific.com
via the ninth-order approximation, rectangular cross-section (b/a = 3/2) under torsional loadings.
of inertia, the higher the torsional rigidity. Circular cross-sections yield the largest
by MONASH UNIVERSITY on 07/17/13. For personal use only.
(a) (b)
Fig. 9. Loading conguration in the case of (a) bending or (b) torsional loadings, I-shaped
cross-section.
March 10, 2010 13:43 WSPC-255-IJAM SPI-J108 S1758825110000500
(a) (b)
Fig. 10. FEM models of the beams subjected to the surface loading. Due to the symmetry, only
half of the structure is modelled.
Int. J. Appl. Mechanics 2010.02:117-143. Downloaded from www.worldscientific.com
Absolute values are considered for the case in which a displacement component
by MONASH UNIVERSITY on 07/17/13. For personal use only.
presents an anti-symmetrical variation across the section. The solution to the prob-
lem via the proposed hierarchical models is compared to the three-dimensional FEM
solution obtained via the MSC-Nastran commercial code. The eight-node HEXA8
element is assumed. Due to problem symmetry versus the z = l/2 plane, half of the
structure is analysed in the FEM model (see Figs. 10(a) and 10(b)). The mesh is
such that ux,max and uz,max converged up to the rst four digits and doubling the
elements yields a lower change in uy,max than 0.5%. In the case of the slender beam,
40000 solid elements are assumed, while for l/a equal to ten, the number of elements
is 115200. The results are presented in Table 9. The maximum approximation order
is considered to be 18. In the case of slender beams, the maximum value of ux is
at x = 0 and y = s2 . The mechanics is governed by pure bending. The displace-
ment component uy is anti-symmetrical towards the x-axis. The maximum of the
Table 9. Maximum displacements for the I-shaped cross-section subjected to bending loading.
10 ux,max 105 |uy,max | 103 uz,max 10 ux,max 103 |uy,max | 102 uz,max
l/a = 100 l/a = 10
FEM 3D 3.640a 2.699b 8.570c 4.059d 3.102b 8.748c
N = 18 3.640 2.695 8.570 4.051 3.040 8.745
N = 10 3.640 2.695 8.570 4.038 2.973 8.742
N =9 3.640 2.694 8.570 4.038 2.967 8.742
N =8 3.640 2.694 8.570 4.035 2.958 8.740
N =7 3.640 2.694 8.570 4.034 2.946 8.740
N =6 3.640 2.694 8.570 4.026 2.941 8.738
N =5 3.640 2.694 8.570 4.025 2.929 8.738
N =4 3.640 2.693 8.570 4.014 2.867 8.743
N =3 3.640 2.693 8.570 4.014 2.814 8.743
N =2 3.639 2.693 8.568 3.878 2.805 8.593
N =1 3.639e 0.001e 8.569e 3.886e 0.058e 8.669e
TB 3.639f 8.568 3.881f 8.568
EB 3.636f 8.568 3.636f 8.568
Maximum value at: (a) (0, s2 /2, l/2); (b) (a/2 + s1 , b/2, l/2); (c) (a/2 s1 , 0, 0);
(d) (a/2, b/2, l/2). (e) Maximum position dierent from the reference FEM solution. (f) The
cross-section is rigid on its plane.
March 10, 2010 13:43 WSPC-255-IJAM SPI-J108 S1758825110000500
absolute value occurs at x = a/2 + s1 and y = b/2. This is due to Poisson eect.
Both maximum values of ux and uy occur at z = l/2, according to Eqs. 5.1. The
maximum value of uz is at x = (a/2 + s1), y = 0 and z = 0. For z = l, there is the
minimum of uz . This displacement component does not present any symmetrical
variation along the section although the mechanics is governed by bending. Classical
models yield the same value of uz,max as higher order theories do. The section is
considered rigid on its plane. The Poisson eect is neglected. The rst-order approx-
imation accounts for it, but the section is still too sti on its plane. At least the
second-order terms are required. The rst-order approximation does not yield the
correct position of the maximum value of the displacements, although the values of
Int. J. Appl. Mechanics 2010.02:117-143. Downloaded from www.worldscientific.com
ux,max and uz,max are close to those computed via higher order approximations. The
analytical solution matches the FEM one. In the case of the span-to-height ratio
equal to ten, the maximum values of |uy | and uz occur at the same locations as for
by MONASH UNIVERSITY on 07/17/13. For personal use only.
the slender beam. The position of ux,max changes. It is at x = a/2 and y = b/2.
It shifts towards the loading application area. As far as four signicant digits are
concerned, the convergence of uz,max is obtained via the third-order approximation.
Classical models yield a value that diers by about 2%. The maximum values of ux
and uy present a slower convergence than uz . This is more pronounced in the case
of uy , since it is due to the Poisson transverse contraction. The maximum value of
ux , computed via EB, diers by about 10% from the one obtained via the maximum
approximation order. The dierence is about 4% in the case of TB and second-order
approximation. For approximation orders higher than the third-one, the even terms
do not improve the value of ux,max except for the fourth signicant digit.
The results regarding the beam subjected to the torsional loadings are presented
in Table 10. The maximum order of the approximation is considered to be 21. The
Table 10. Maximum displacements for the I-shaped cross-section subjected to torsional loadings.
103 |ux,max | 103 |uy,max | 105 uz,max 10 |ux,max | 10 |uy,max | 102 uz,max
l/a = 100 l/a = 10
FEM 3D 2.481a 5.580a 5.265b 2.408a 5.164a 4.870b
N = 21 2.365 5.318 5.093 2.293 4.939 4.675
N = 20 2.365 5.318 5.093 2.289 4.937 4.673
N = 19 2.331 5.242 5.021 2.261 4.872 4.612
N = 10 2.091 4.703 4.425 2.031 4.412 4.112
N=9 1.941 4.367 4.044 1.899 4.118 3.777
N=8 1.941 4.367 4.044 1.891 4.118 3.779
N=7 1.674 3.766 3.301c 1.655 3.593 3.118c
N=6 1.674 3.766 3.302c 1.644 3.598 3.128c
N=5 1.636 3.679 3.177c 1.609 3.520 3.013c
N=4 1.636 3.679 3.177c 1.598 3.510 3.011c
N=3 0.632 1.422 1.378c 0.656 1.401 1.332c
N=2 0.632 1.422 1.378c 0.629 1.388 1.328c
N=1 0.190 0.428 0.203c 0.209 0.429 0.201c
Maximum value at: (a) ((a/2+s1 ), b/2, l/2); (b) (a/2, b/2, 0). (c) Maximum position dierent
from the reference FEM solution.
March 10, 2010 13:43 WSPC-255-IJAM SPI-J108 S1758825110000500
FEM models have the same mesh as those developed for the solution of the previous
problem. A further renement of the mesh, by doubling the elements, yields a lower
variation of the displacements than 0.5%. The maximum absolute values of the
displacement components ux and uy occur at the points (a/2 + s1 , b/2, 0) for both
values of the span-to-height ratio. The maximum value of uz is at (a/2, b/2) and
z = 0. Its location is not correctly predicted by approximation orders as low as the
seventh one. Due to the loading condition, the convergence is also slow in the case
Int. J. Appl. Mechanics 2010.02:117-143. Downloaded from www.worldscientific.com
of the slender beam. Tenth- and 21st-order approximations at best dier by 11%.
The FEM solution and maximum approximation orders that have been adopted
by MONASH UNIVERSITY on 07/17/13. For personal use only.
dier by about 5% for both values of the span-to-height ratio. The solution above
the whole cross-section may not always be accurately predicted via the information
in a single point as theories based on Mac Laurins polynomial functions are. For
a xed number of unknowns, the assumption of other polynomial approximations,
such as the Lagrange ones, should be investigated. For higher approximation orders
than the fourth one, the odd terms do not improve the results for slender beams.
For l/a equal to ten, only the fourth signicant digit changes. The results may be
improved by assuming higher approximation orders than those considered here. As
far as the computational costs are concerned, the proposed analytical models require
less than a second, regardless the approximation order. The FEM solution based
on the proposed models, not reported here for sake of brevity, is obtained in few
seconds for a very ne mesh. For the reference three-dimensional FEM solutions,
more than two hours are required.
7. Conclusions
A unied formulation, named CUF, has been proposed for beam structures. Higher
order models that account for shear deformations and in- and out-of-plane warp-
ing can be formulated straightforwardly. Within the proposed formulation, classi-
cal models, such as Euler-Bernoullis and Timoshenkos, are regarded as particular
cases. Closed form, Navier type solution for beams subjected to bending and tor-
sional loadings has been addressed. Rectangular and I-shaped cross-sections have
been considered. The following conclusions can be drawn:
(i) As far as bending loadings are concerned, the proposed formulation allows
obtaining results as accurate as desired through an appropriate choice of the
approximation order. The results of one-dimensional CUF models match the
three-dimensional mechanics.
(ii) Classical models are accurate where the mechanics is governed by global bend-
ing, that is, for slender beams.
March 10, 2010 13:43 WSPC-255-IJAM SPI-J108 S1758825110000500
(iii) The rst-order model can be regarded as a transition between classical models
and rened theories since it accounts for non-classical eects although not
accurately in the case of deep beams.
(iv) In the case of a deep beam with a rectangular cross-section, the sixth-order
model matches the exact reference solution.
(v) Mechanics due to torsional loadings is more dicult to describe accurately.
Higher approximation order than the bending case are required. The accuracy
depends on the type of the cross-section, on the span-to-height ratio and on
the aspect ratio.
(vi) Within CUF, the description of warping does not require any specic warping-
Int. J. Appl. Mechanics 2010.02:117-143. Downloaded from www.worldscientific.com
The assumption of Lagranges polynomials for the approximation above the cross-
section represents an attractive solution to capture non classical, local phenomena
on the cross-section. This will be the subject of further investigation together with
the investigation of cross-section of aeronautical interest, such as wings and rotor
blades proles.
Acknowledgments
Second author is supported by the Ministere de la Culture, de lEnseignement
Superieuer et de la Recherche of Luxembourg under AFR-Postdoc grant TR-PDR
BFR07-136.
Appendix
A. Correction of Poissons Locking
The Poisson ratio couples the normal deformations along the spatial directions x,
y, and z. Because of this, constant and linear approximations of the displacements
do not yield accurate results, even in the case of slender beams. This phenomenon
is known in literature as Poissons Locking (PL). A modied version of materials
constitutive equations, in which the stiness coecients are opportunely modied,
should therefore be adopted. A correction of the PL is obtained imposing xx and
yy equal to zero in Hookes law. An algebraic linear system is obtained in xx and
yy . By substituting its solution in Hookes equation regarding zz , the reduced
stiness coecient Q33 is obtained
C23 C12 C13 C22 C11 C23 C12 C13
Q33 = C33 C13 2 C C + C23 2 C C (A.1)
C12 11 22 C12 11 22
March 10, 2010 13:43 WSPC-255-IJAM SPI-J108 S1758825110000500
Q33 = E (A.2)
k i
E, = [(xk )i+1 (xk1 )i+1 ]
by MONASH UNIVERSITY on 07/17/13. For personal use only.
x s,x
i+1 2
1
[(y k )j++1 (y1k )j++1 ]. (B.1)
j++1 2
k 1
E, y s,y
= [(xk )i++1 (xk1 )i++1 ]
i++1 2
j
[(y k )j+1 (y1k )j+1 ]. (B.2)
j+1 2
1
Eks = [(xk )i++1 (xk1 )i++1 ]
i++1 2
1
[(y k )j++1 (y1k )j++1 ]. (B.3)
j++1 2
k j
E, x s,y
= [(xk )i+ (xk1 )i+ ] [(y k )j+ (y1k )j+ ]. (B.4)
i+ 2 j+ 2
k i
E, y s,x
= [(xk )i+ (xk1 )i+ ] [(y k )j+ (y1k )j+ ]. (B.5)
i+ 2 j+ 2
k i 1
E, xs
= [(xk2 )i+ (xk1 )i+ ] [(y k )j++1 (y1k )j++1 ]. (B.6)
i+ j++1 2
1
Eks,x = [(xk )i+ (xk1 )i+ ] [(y k )j++1 (y1k )j++1 ]. (B.7)
i+ 2 j++1 2
k 1 j
E, ys
= [(xk )i++1 (xk1 )i++1 ] [(y k )j+ (y1k )j+ ]. (B.8)
i++1 2 j+ 2
1
Eks,y = [(xk2 )i++1 (xk1 )i++1 ] [(y k )j+ (y1k )j+ ]. (B.9)
i++1 j+ 2
March 10, 2010 13:43 WSPC-255-IJAM SPI-J108 S1758825110000500
References
Berdichevzy, V. L. [1976] Equations of the theory of anisotropic inhomogeneous rods,
Doklady Akademii Nauk SSR 228, 558561.
Carrera, E. [2003] Theories and nite elements for multilayered plates and shells: A
unied compact formulation with numerical assessment and benchmarking, Archives
Int. J. Appl. Mechanics 2010.02:117-143. Downloaded from www.worldscientific.com
1037.
Carrera, E. and Brischetto, S. [2008a] Analysis of thickness locking in classical, rened
and mixed multilayered plate theories, Composite Structures 82(4), 549562.
Carrera, E. and Brischetto, S. [2008b] Analysis of thickness locking in classical, rened
and mixed theories for layered shells, Composite Structures 85(1), 8390.
Carrera, E. and Giunta, G. [2008] Hierarchical models for failure analysis of plates bent
by distributed and localized transverse loadings, Journal of Zhejiang University SCI-
ENCE A 9(5), 600613.
Cowper, G. R. [1966] The shear co-ecient in Timoshenko beam theory, Journal of
Applied Mechanics 33(10), 335340.
Estivalezes, E. and Barrau, J. J. [1998] Analytical theory for an approach calculation of
composite box beams subjected to tension and bending, Composites Part B 29(4),
371376.
Ganapathi, M., Patel, B. P., Polit, O. and Touratier, M. [1999] A C1 nite element includ-
ing transverse shear and torsion warping for rectangular sandwich beams, Interna-
tional Journal for Numerical Methods in Engineering 45(1), 4775.
Gjelsvik, A. [1981] The Theory of Thin-Walled Bars (John Wiley and Sons, Inc., New
York).
Hodges, D. H. [1990] A review of composite rotor blade modeling, AIAA Journal 28(3),
561565.
Hodges, D. H. and Yu, W. [2007] A rigorous, engineer-friendly approach for modelling
realistic, composite rotor blades, Wind Energy 10, 179193.
Kant, T. and Manjunath, B. S. [1989] Rened theories for composite and sandwich beams
with C0 nite elements, Computers and Structures 33(3), 755764.
Kant, T. and Manjunath, B. S. [1990] Higher-order theories for symmetric and unsym-
metric ber reinforced composite beams with C0 nite elements, Finite Elements in
Analysis and Design 6(4), 303320.
Kapania, R. K. and Raciti, S. [1989a] Recent Advances in Analysis of Laminated Beams
and Plates, Part I: Shear Eects and Buckling, AIAA Journal 27(7), 923934.
Kapania, R. K. and Raciti, S. [1989b] Recent Advances in Analysis of Laminated Beams
and Plates, Part II: Vibrations and Wave Propagation, AIAA Journal 27(7), 935946.
Lee, J. and Lee, S. [2004] Flexural-torsional behavior of thin-walled composite beams,
Thin-Walled Structures 42(9), 12931305.
Lee, J. [2005] Flexural analysis of thin-walled composite beams using shear-deformable
beam theory, Composite Structures 70(2), 212222.
March 10, 2010 13:43 WSPC-255-IJAM SPI-J108 S1758825110000500
Li, X. and Liu, D. [1997] Generalized laminate theories based on double superposition
hypothesis, International Journal for Numerical Methods in Engineering 40(7), 1197
1212.
Liu, G. R. and Zhang, G. Y. [2009] A novel scheme of strain-constructed point interpo-
lation method for static and dynamic mechanics problems, International Journal of
Applied Mechanics 1(1), 233258.
Manjunath, B. S. and Kant, T. [1993] Dierent numerical techniques for the estimation
of multiaxial stresses in symmetric/unsymmetric composite and sandwich beams with
rened theories, Journal of Reinforced Plastics and Composites 12(1), 237.
Murty, A. V. K. [1970] Analysis of short beams, AIAA Journal 8(11), 20982100.
Prandtl, L. [1903] Zur Torsion von Prismatischen Staben, Zeitschrift Fur Physik 4,
758770.
Int. J. Appl. Mechanics 2010.02:117-143. Downloaded from www.worldscientific.com
Rand, O. [1998] Fundamental closed-form solutions for solid and thin-walled composite
beams including a complete out-of-plane warping model, International Journal of
Solids and Structures 35(21), 27752793.
Rand, O. [2000] On the importance of cross-sectional warping in solid composite beams,
Composite Structures 49(4), 393397.
Reddy, J.-N. [2004] Mechanics of laminated composite plates and shells. Theory and Anal-
ysis (CRC Press).
Saint-Venant, B. [1855] Memoire Sur la Torsion des Prismes, Memoires des Savants
Etrangers 14, 233560.
Stephen, N. G. and Levinson, M. [1979] A second order beam theory, Journal of Sound
and Vibration 67(3), 293305.
Timoshenko, S. P. [1921] On the corrections for shear of the dierential equation for
transverse vibrations of prismatic bars, Philosophical Magazine 41, 744746.
Timoshenko, S. P. [1922] On the transverse vibrations of bars of uniform cross section,
Philosophical Magazine 43, 125131.
Timoshenko, S. P. and Goodier, J.-N. [1970] Theory of elasticity (McGraw-Hill, New York).
Tsai, S. W. [1988] Composites Design (Dayton, Think Composites).
Vidal, P. and Polit O. [2008] A family of sinus nite elements for the analysis of rectangular
laminated beams, Composite Structures 84(1), 5672.
Vinayak, R. U. and Prathap, G. and Naganarayana, B. P. [1996] Beam elements based on
a higher order theory - I. Formulation and analysis of performance, Computers and
Structures 58(4), 775789.
Yu, W. and Hodges, D. H. and Volovoi, V. and Cesnik, C. E. S. [2002] On Timoshenko-
like modeling of initially curved and twisted composites beams, International Journal
of Solids and Structures 39(19), 51015121.
Yu, W. and Hodges, D. H. [2004] Elasticity solution versus asymptotic sectional analysis
of homogeneous, isotropic, prismatic beams, Journal of Applied Mechanics 71(1),
1523.
Wagner, W. and Gruttmann, F. [2002] A displacement method for the analysis of exural
shear stresses in thin-walled isotropic composite beams, Computers Structures 80(24),
18431851.