Vous êtes sur la page 1sur 19

A Matlab tool to compute the mechanical

characteristics of any composite section

Rached El Fatmi

Université Tunis Manar, Ecole Nationale d’Ingénieurs de Tunis


Laboratoire de Génie Civil, BP37, Le Belvédère, 1002, Tunis, Tunisie
rached.elfatmi@enit.rnu.tn

ABSTRACT. We present a Matlab program, CSection, devoted to the finite element computation of
the mechanical characteristics of any composite section. These sectional characteristics (SC),
which derive from 3D Saint Venant’s solution, are the full matrix of the structural behavior, the
sectional stress fields and the displacement fields related to the sectional deformations. These
SC, which constitute a relevant set of mechanical information on the elastic behavior of the
composite section, can really help R&D engineers to design and optimize composite section
and to compute composite beams. To illustrate CSection performance, two significant examples
are presented: a homogeneous section and a laminated section made by 8 orthotropic layers.
RÉSUMÉ. On présente un programme Matlab, CSection, dédié au calcul systématique, par élé-
ments finis, des caractéristiques mécaniques d’une section (CMS) composite quelconque. Ces
CMS, qui sont extraites de la solution 3D de Saint Venant, sont l’opérateur de comportement 1D
de la poutre, les champs de contrainte de section et les champs de déplacement qui décrivent la
déformation de la section. Ces CMS, qui constituent autant d’informations sur le comportement
mécanique de la section, offrent à l’ingénieur R&D une reélle aide pour concevoir et optimiser
une section ou calculer une poutre composite. Pour illustrer les performances de CSection deux
exemples significatifs sont présentés: une section homogène et une section stratifiée constituées
de 8 couches orthotropes.
KEYWORDS: beam, Saint Venant, composite, sectional constants, Poisson effect, warping, shear.
MOTS-CLÉS : poutre, Saint Venant, composite, constantes de section, effet Poisson, gauchisse-
ment, cisaillement.

DOI:10.3166/RCMA.22.395-413 c 2012 Lavoisier

Revue des composites et des matériaux avancés – no 3/2012, 395-413


396 RCMA. Volume 22 – no 3/2012

1. Introduction

For homogeneous and isotropic beam, the sectional constants are easy to compute
(closed form results, or numerical computations by common softwares), the elastic
behavior is well controlled, and classical Bernoulli or Timoshenko beam theories are
sufficient for common applications.
In contrast, the behavior of a composite beam is much more complex and not so
easy to understand, to predict and to compute. A major complexity is due to the elastic
couplings between extensional, flexural and torsional deformations that can occur for
an arbitrary composite section (Chandra et al., 1990; Chandra, Chopra, 1991; Rand,
1998; Rappel, Rand, 2000; Volovoi et al., 2001; El Fatmi, Zenzri, 2002; 2004); lami-
nated composite beams are known to exhibit such coupled deformations arising from
the anisotropic nature of the layers and their stacking sequences.
The accuracy of a beam model depends on its aptitude to capture the behavior asso-
ciated with the two small dimensions eliminated in the final (1D) beam analysis. From
the literature surrounding the development of beam theory, three important approaches
can be identified: - beam theories based on kinematic (and static) assumptions using
some engineering experience or intuition (as that of Bernoulli, Timoshenko, Vlasov...
and their extensions); - beam theories based on the asymptotic expansion of the 3D
solution, using small parameter(s) inherent to the beam features (Trabucho, Viano,
1996); - and the one-dimensional beam-like theory which derives from the 3D Saint
Venant (SV) solution of SV’s problem.
Among these different beam approaches, that of SV continues to be considered as
the reference to understand the beam structural behavior. This is due to the status or
the asymptotic nature of 3D SV’s solution: moving away from the end sections, the
3D exact solution tends asymptotically towards 3D SV’s solution. Practically, SV’s
solution, also called central solution (Berdichevsky, 1979; Giavotto et al., 1983), is
known to represent the exact 3D solution in the interior part of a beam, far from the
end-sections where the boundary conditions are applied.
Established first for homogeneous and isotropic section, 3D SV’s solution has
been extended by Iesan (1976) to any composite section wherein each material is
anisotropic. An important property of 3D SV’s solution is that it is linear function of
the cross-sectional stresses (axial force, shear forces, torsional moment and bending
moments). Moreover, (Berdichevsky, 1979) showed that it can be split into a set of
2D-problems on the cross-section and a set of 1D-equations that define a 1D beam-like
theory (denoted by SV-BT in this paper). This splitting has been later used in several
works to compute 3D SV’s solution using the finite element technique (Giavotto et
al., 1983; El Fatmi, Zenzri, 2002; 2004; Sanchez, 2001; Alpdogan et al., 2010). It is
worth noting that the effective computation of SV’s solution provides an important set
of mechanical characteristics that are only dependent of the section nature (shape and
materials):
– Λ, the full (6x6) matrix of the structural behavior that contains (off diagonal)
the sectional constants expressing all the coupling effects;
Characteristics of any composite section 397

– and for each cross-sectional stresses Fi (axial force, shear forces, torsional mo-
ment, bending moments):
- U i , the displacement field related to the sectional deformations that con-
tains the Poisson’s effects (called in-plane warping in this paper) and the out of-plane
warping of the section;
- σ i the sectional stress field.
These sectional characteristics (SC) {Λ, U i , σ i }, are fundamental so that they
can be computed once for all and used later, for a given beam problem, to immedi-
ately obtain the detailed 3D SV-solution. These SC constitute then a relevant set of
mechanical information on the behavior of the composite section that can really help
R&D engineers to design or optimize composite section and to compute composite
beams.
Furthermore, for higher order beam theories which are based on displacement
models, the characteristics U i which describe the sectional deformations can be used
as pertinent modes to enrich the displacement model in order to refine beam theories
(El Fatmi, 2007a; 2007b; El Fatmi, Ghazouani, 2011a; 2011b) and to catch a part of
the end effects.
The economic importance of composite materials is established nowadays: aero-
space construction, competition sport equipments, automotive industry, reinforcements
in civil infrastructures... However, to the author’s knowledge, the systematic compu-
tation of the sectional characteristics of a composite beam does not exist in standard
finite elements codes. Programs or softwares have certainly been developed in re-
search centers of some high-tech industries to compute such characteristics, but these
tools are generally oriented for specific applications and are confidential.
The present work proposes a program, named CSection, implemented on Matlab1
platform, which provides, for a given composite section (shape and materials), all its
characteristics. CSection is developed conforming to the numerical method proposed
by El Fatmi and Zenzri (2002). This method consists in solving, by 2D finite elements,
a set of elastic problems on the section from which {Λ, [U i ]; σ i } are deduced. Mat-
lab is largely used by scientists and engineers. Simple to use, the Matlab version2 of
CSection can be considered as a tool to compute any composite section. However,
such a tool can be implemented on any finite elements code.
To illustrate the efficiency of CSection, two significant examples of section are
computed: a simple rectangular homogeneous and isotropic section and a symmetric
laminated section [0,90,-45,60]s made by 8 orthotropic layers.
In this paper, the next section recalls the essential of 3D-1D SV’s theory for a
composite beam and specify the mechanical characteristics of the section. After a

1. Matlab is developed by MathWorks Inc.


2. A free version of CSection is available for tests and evaluations; one can receive this version upon request
by e-mail.
398 RCMA. Volume 22 – no 3/2012

quick description of CSection, an important section is devoted to the detailed analysis


of the illustrative examples. Besides, the essential of the numerical method used by
CSection is recalled in appendix.

2. 3D/1D Saint Venant’s theory and sectional characteristics

Figure 1. Saint Venant problem

2.1. Saint Venant’s problem extended to composite beam

Extended SV’s problem is a 3D equilibrium elastic problem (Figure 1). The com-
posite beam is along the z axis and is occupying a prismatic domain Ω of a constant
cross section S and length L. Slat is the lateral surface and S0 and SL are the end
cross sections. P = G + GP denotes a point of Ω where G is the center section.The
beam is submitted only to the tractions H 0 and H L acting on S0 and SL , respectively
(vectors and tensors are in boldface character). The materials of the cross section are
anisotropic and perfectly bonded together. The equations of the linearized equilibrium
are:
div σ = 0

in Ω 
1


ε(ξ) = (∇t ξ + ∇ξ) in Ω

2 (1)
σ = K : ε(ξ) in Ω  

σ·n = 0 on Slat


σ · (−z) = H 0 on S0
(2)
σ·z = H L on SL
where ε(ξ) is the strain tensor associated to the displacement field ξ; ∇, (·)t and (:)
denote the gradient, the transpose and the double contraction operators, respectively;
K denotes the elasticity tensor, σ is the stress tensor and n is the unit vector that is
normal and external to Slat .

2.2. 3D Saint Venant’s solution and sectional characteristics

3D SV’s solution is the unique (z-polynomial) solution that exactly satisfies Equa-
tion (1) and satisfies the boundary conditions (2) only in terms of resultant:
Z Z
( σ · (−z), σ · z) dS= (H 0 , H L ) dS
Z S ZS
(GP ∧ σ · (−z), GP ∧ σ · z) dS= (GP ∧ H 0 , GP ∧ H L ) dS
S S
Characteristics of any composite section 399

The expression of 3D SV’s solution (Ladevèze, Simmonds, 1998; El Fatmi, Gha-


zouani, 2011a) is given by:
6
X
ξ sv (x, y, z) = u(z) + ω(z) ∧ GP + Fi (z)U i (x, y) (3)
i=1
6
X
σ sv (x, y, z) = Fi (z) σ i (x, y) (4)
i=1

where Fi ∈ {Tx , Ty , N, Mx , My , Mt } are the components3 of the classical sectional


stresses [R, M ] defined by
Z Z
σ · z dS = R = [Tx , Ty , N ] GP ∧ σ · z dS = M = [Mx , My , Mt ]
S S

and which verify the 1D equilibrium equations

R0 = 0 M0 + z ∧ R = 0 (5)

where (·)0 denotes the derivative with respect to z. The sectional displacements [u, ω]
are related to [R, M ] by the structural (1D) behavior given by:
   0   
γ u +z∧ω R
= =Λ (6)
χ ω0 M

where Λ is the 1D structural compliance (6×6) operator of the beam. An important


particularity of 3D SV’s solution is that the set {Λ, U i , σ i } depends only on the
section nature (shape and materials). Note that U i =[Πi , ψ i ] contain Πi , the in-plane
warpings (Poisson’s effects) mode, and ψ i , the out-of-plane warpings. The sectional
characteristics {Λ, [Πi , ψ i ], σ i } constitute a set of mechanical information on the
elastic behavior of the section.

2.3. Saint Venant (1D) beam-like theory

Let us come back to the 3D SV-problem. As soon as the sectional characteristics


{Λ, [Πi , ψ i ], σ i } are known, one can get the 3D SV’s solution in two steps:
1- Λ is used to solve the 1D problem defined by the following 1D-equations

R0 = 0
   
γ R
= Λ [R, M ]0,L = [F , C]0,L (7)
M0 + z ∧ R = 0 χ M

These 1D-equations define the SV’s beam-like theory (SV-BT).

3. The axial force, the shear forces, bending moments and torsional moment, respectively.
400 RCMA. Volume 22 – no 3/2012

2- the 1D solution {u(z), ω(z), R(z), M (z)} of (7) and {[Πi , ψ i ], σ i } are then
used to generate the 3D SV’s solution, conforming to (3,4). Note that this step is
straightforward and does not need any further computation.
R EMARK 1. — Conforming to SV’s principle, 3D SV’s solution gives a good descrip-
tion of the exact solution far from the end sections, in the interior part of the beam.
Besides, SV’s solution corresponds to SV’s problem where the end sections are free
to warp. However, in practice, SV-BT may be used to treat other boundary conditions.
For instance, a built in section is modeled by the 1D conditions [u=0, ω=0]; in that
case, the built-in effect can make the 3D SV’s solution generated by Equations (3-4)
acceptable in a shorter interior part of the beam. 

3. Quick description of Matlab version of CSection

Figure 2. Examples of homogeneous and composite sections

CSection is a program implemented on Matlab platform, devoted to the compu-


tation of the mechanical characteristics of any composite section and developed con-
forming to the numerical method given by El Fatmi and Zenzri (2002). The section
(Figure 2) may be solid or walled, open or closed, symmetric or not, and made of
elastic and anisotropic materials perfectly bonded together. To compute a section, two
steps are needed:
1- Matlab provides4 a tool (pdetool) allowing to define the 2D geometry of the
section (with sub-domains) and to mesh it in triangles; then, user defines the sub-
domain materials: for each material, the elastic constants are given with respect to the
material axes, and these material axes have then to be specified with respect to the
beam axes;
2- CSection uses these section data (meshing+materials) to systematically com-
pute the sectional characteristics {Λ, [Πi , ψ i ], σ i }. The results can be printed, visu-
alized or exported to other use.

4. The user can also import a meshed section from other programs.
Characteristics of any composite section 401

The 2D finite elements are the classical 6-nodes triangles. The time cost (cpu time)
to compute all the sectional characteristics of a given section is just of a few seconds
(Table 1) on a common personal computer.

4. Two illustrative examples

To illustrate CSection performance, two examples (Figure 3) of section (denoted


by S1 and S2) are presented:

Figure 3. Homogeneous section and laminated section

S1: a homogeneous rectangular section [h,b=h/2] made by an isotropic material


[E=1000, ν=0.3];
S2: a symmetric laminated section [0,90,-45,60]s made by 8 transversely isotropic
layers of the same thickness: the geometry is [b,h=0.8b,t=0.1b] and the mate-
rial is given by [EX =137.6 GPa, EY =EZ =14.448 GPa, νXY =νXZ =νY Z =0.21,
GXY =GXZ =GY Z =5.48 Gpa, where (X, Y , Z) are the principal material axes].

For each section, [ Λ, U i , σ i ] are computed by CSection and presented; Table 1


gives the cpu5 times corresponding to both sections. However, for more details, the
following critical points are specified:
S1: the variation of the shear coefficients6 related to the shear forces, with respect
to the aspect ratio h/b, and with respect to Poisson’s ratio ν;
S2: for a shear force (Ty ), the y-variation, through the layers, of the shear ; and
for an axial force (N), the x-variation of the 0/90-interlaminar stresses which may
induce the delamination of the layers (see the free edge effect in (Pipes, Pagano, 1970;
El Fatmi, Zenzri, 2004)).

5. cpu times obtained on a personal computer using Intel Duo processor 2.53MHz and 4Go Ram.
6. These shear coefficients, that appear naturally in SV-BT, correspond to the so-called (Timoshenko) shear
correction factors.
402 RCMA. Volume 22 – no 3/2012

Table 1. Size problem and cpu time


number of elements number of nodes degrees of freedom cpu times
S1 664 1389 4171 4
S2 2304 4781 14347 12

4.1. Homogeneous rectangular section

Structural behavior. Let Γ=Λ−1 , the structural rigidity operator. For S1, which is
isotropic and x-y-symmetric, Γ is diagonal and given for h=1 by:

 
150.8542 0.0000 0.0000 0.0000 0.0000 0.0000

 0.0000 160.1812 0.0000 0.0000 0.0000 0.0000 

 0.0000 0.0000 500.0000 0.0000 0.0000 0.0000 
Γ=  (8)

 0.0000 0.0000 0.0000 41.6667 0.0000 0.0000 

 0.0000 0.0000 0.0000 0.0000 10.4167 0.0000 
0.0000 0.0000 0.0000 0.0000 0.0000 10.9945

Sectional deformations. Poisson’s effects (Figure 4) are only due to the axial force
(N ) and the bending moments (Mx , My ). Out of plane warpings (Figure 5) are only
due to the shear forces (Tx , Ty ) and the torsional moment (Mt ).
Sectional stresses. The (canonical) sectional 3D stress fields σ i are those obtained
for unit cross-sectional stresses Fi =1. The axial 3D stresses (Figure 6, left) reduced to
σzz are only due to the axial force (N ) and the bending moments (Mx , My ). The shear
stresses (Figure 6, right) reduced to (σxz , σyz ) are due to the shear forces (Tx , Ty ) and
the torsional moment (Mt ).
All these results are expected and classical. However, the next result is more criti-
cal and more significant.
The shear coefficients. One can deduce7 from the computations of Γ for different
h/b ratios and different Poisson’s ratios ν, the values of the shear coefficients (kx , ky );
the results are given in Table 2. The common value used for rectangular cross-section
is, for all h/b ratio, kx =ky =5/6=0.833; this value is correct only if ν=0. For a fixed
value of h/b, the shear coefficients are slightly influenced by ν, but for a fixed ν 6= 0,
they strongly depend on h/b ratio. These results are conform to those presented in
(Lowe, 1971; Renton, 1991) where SV’s problem limited to shear force has been
studied (for a rectangular section) by Fourier series.

7. kx = 2(1 + ν) Γ
Γ
11
ky = 2(1 + ν) Γ
Γ
22
33 33
Characteristics of any composite section 403

Table 2. Shear coefficients with respect to ν (with h=2b) and to h/b (with ν=0.3)
ν 0.00 0.15 0.30 0.45 h/b 1 2 5 10
kx 0.833 0.817 0.784 0.749 kx 0.828 0.784 0.478 0.179
ky 0.833 0.833 0.833 0.833 ky 0.828 0.833 0.833 0.833

Figure 4. Poisson’s effects for the homogeneous section S1

Figure 5. Out-of-plane warpings for the homogeneous section S1

Figure 6. Stress fields for the homogeneous and isotropic section S1


404 RCMA. Volume 22 – no 3/2012

4.2. Laminated section

Structural behavior. For S2, Γ=Λ−1 is not diagonal and given by:
 
6.1657 0.0000 −1.8607 0.0000 0.0000 0.0000

 0.0000 4.0000 0.0000 0.0000 0.0000 0.0000 

−1.8607 0.0000 38.3708 0.0000 0.0000 0.0000
Γ = 109 × 
 
 (9)

 0.0000 0.0000 0.0000 3.7226 0.0000 0.0645 

 0.0000 0.0000 0.0000 0.0000 3.0925 0.0000 
0.0000 0.0000 0.0000 0.0645 0.0000 0.5722

The constants Γ13 and Γ46 rise from the the extension-bending and the bending-
twist couplings, respectively.
Sectional deformations. In contrast with the case of homogeneous and isotropic
section, each cross-sectional stress (Tx , Ty , N, Mx , My , Mt ) contributes to the Pois-
son’s effect deformation (Figure 7) and to the out-of-plane warping (Figure 8) of the
section (this is due to the composite nature of the section).

Figure 7. Poisson’s effects for section S2

Sectional stress fields. Important: for S2 the materials are homogeneous by sub-
domain. Right to an interface between two different layers (1 and 2), only the stress
vector ( σ · y) has theoretically to be continuous ( σ 1 · y= σ 2 · y). In CSection the
stress fields are displayed by subdomain! Therefore, it is normal to see discontinuities
for (σxx , σxz , σzz )-fields right to the interfaces. In contrast, when (σxy , σyy , σyz )-
fields coincide right to an interface, that is significant of the numerical quality of the
results; see for example in Figure 9, the discontinuities for σzz , and how σyz coincide.
Characteristics of any composite section 405

Figure 8. Out-of-plane warpings for section S2

Figures 9-10-11 depict the fields of each component of the 3D stresses (only the
main components are presented). For the shear force Tx and the torsional moment Mt
(Figure 11 ), in contrast with the case of homogeneous and isotropic section, σxx and
σzz are not zero, and note that they are of the same magnitude with σxz and σyz . The

Figure 9. The main components of (Ty )-stress fields for section S2

Figure 10. The main component of (Mx , My , N )-stress fields for section S2
406 RCMA. Volume 22 – no 3/2012

Figure 11. The main components of (Tx , Mt )-stress fields for section S2

shear through the layers. Figure 12 depicts, for a shear force Ty , the y-variation of
the shear σyz . Note that, as expected, the shear appears continuous through the layers
and vanishes at the extremities. The free edge effect. For an extension (N ), Figure 13
gives the x-variation of the inter-laminar stresses σyy on 0/90-interface. Note that two
curves are given: a gray curve for the 90-layer and a black curve for the 0-layer, and
how they coincide. The concentration of σyy (traction) explains the delamination that
Characteristics of any composite section 407

Figure 12. Variation, through the layers, of the shear σyz for a shear force Ty

starts from the free edge of the laminated beam; this phenomenon, known as the free
edge effect, is contained in SV’s solution.

Figure 13. Variation of the interlaminar stresses σyy on 0/90-interface for a tension

4.3. Comments

The results presented above as critical points on the shear coefficient, the shears
through the layers and the inter-laminar stresses have been given to emphasyze the
quality of the mechanical and numerical results that can be provided by the computa-
tion of the sectional characteristics that derive from SV’s theory. Similar results can be
obtained for any composite section, in a systematic way by CSection. However, one
can find other examples of homogeneous and composite sections in (Sanchez, 2001;
408 RCMA. Volume 22 – no 3/2012

El Fatmi, Zenzri, 2002; 2004), where SV-results are compared with results from liter-
ature, and where more details are presented; such as shear centers, torsion constants,
elastic couplings...

5. CSection: usefulness of such a tool

5.1. Computation of composite section and composite beam

For beam, in standard finite elements codes, only the classical Bernoulli or Timo-
shenko (1D) finite elements are generally proposed to the user. Further, for a given
elastic 1D beam problem:
1- the sectional constants have to be given by the user;
2- after computation of the 1D solution, the 3D stresses, right to a cross-section,
have to be deduced by the user.
In contrast, if the computation of the sectional characteristics is possible, the user
can compute and store a set of k sectional characteristics [Λk , σ ik ] to be used later to
solve (homogeneous or composite) beam problems:
1- Λk , for the 1D finite element computations;
2- σ ik , to systematically deduce the 3D stress8 field that corresponds to 3D SV’s
solution (or the 3D central solution).
In that way, except the design of the sections, all the computations are left to the
computer.
The behavior of a composite beams is much more complex than homogeneous and
isotropic ones, especially when anisotropic materials are of consideration. Such a tool
can really help R&D engineers to simulate9 and understand the behavior of composite
beams, and later to design and optimize composite sections. The present version of
CSection is developed as a tool on Matlab plateform, but it can be implemented in any
finite element codes. The essential of the numerical method used by CSection is given
in appendix.

8. 3D SV’s stress field.


9. Note that, even when it comes to plates or composite shells, the classical mechanical tests are usually
carried out on samples whose appearance is that of a composite beam; and such a tool can be used to analyze
composite tests.
Characteristics of any composite section 409

5.2. Higher order beam theories development

To improve the prediction of the classical beam theories which assume a rigid
body motion of the cross-section, the so-called higher order beam theories allow some
deformation of the section; they are based on the following displacement models:

rigid motion of the section


z }| n
{ X
ξ(u, ω, ηi , ...) = u(z) + ω(z) ∧ GP + ηi (z)Ψi (x, y) (10)
i=1

where ηi are control parameters of the sectional displacement modes Ψi which are
supposed to be known. The characteristics [Πi , ψ i ] which describe the sectional in-
and out of-plane deformation modes can be used as pertinent modes to enrich the
displacement model in order to derive such theories; using this way higher order beam
theories have been recently proposed for homogeneous section in (El Fatmi, 2007a;
2007b) and for composite sections in (El Fatmi, Ghazouani, 2011a; 2011b).

References

Alpdogan C., Dong S., Taciroglu E. (2010). A method of analysis for end and transitional
effects in anisotropic cylinders. International Journal of Solids and Structures, Vol. 47,
pp. 947-956.
Berdichevsky V. L. (1979). Variational asymptotic method of shell theory construction. PMM,
Vol. 32, pp. 553-576.
Chandra R., Chopra I. (1991). Experimental and theoritical analysis of composite i-beams
with-elastic couplings. AIAA Journal, Vol. 29, No. 12, pp. 2197-2206.
Chandra R., Stemple A. D., Chopra I. (1990). Thin-walled composite beams under bending,
torsional and extensional loads. J. Aircraft, Vol. 27, No. 7, pp. 619-626.
El Fatmi R. (2007a). Non-uniform warping including the effects of torsion and shear forces.
part-I: A general beam theory. International Journal of Solids and Structures, Vol. 44,
pp. 5912-5929.
El Fatmi R. (2007b). Non-uniform warping including the effects of torsion and shear forces.
part-II: Analytical and numerical applications. International Journal of Solids and Struc-
tures, Vol. 44, pp. 5930-5952.
El Fatmi R., Ghazouani N. (2011a). A higher order composite beam theory built on 3d saint
venant’ solution, part-1: theoretical developments. Composite Structures, Vol. 93, pp. 557-
566.
El Fatmi R., Ghazouani N. (2011b). A higher order composite beam theory built on 3d saint
venant’ solution, part-2: built-in effect influence on the behavior of end-loaded cantilever
beams. Composite Structures, Vol. 93, pp. 567-581.
El Fatmi R., Zenzri H. (2002). On the structural behavior and the saint venant solution in the
exact beam theory. Application to laminated composite beams. Computers and Structures,
Vol. 80, pp. 1441-1456.
410 RCMA. Volume 22 – no 3/2012

El Fatmi R., Zenzri H. (2004). A numerical method for the exact elastic beam theory. Applica-
tions to homogeneous and composite beams. International Journal of Solids and Structures,
Vol. 41, pp. 2521-2537.

Giavotto V., Borri M., Mantegazza P., Ghiringhelli G., Carmaschi V., Maffioli G. (1983).
Anisotropic beam theory and applications. Computers and Structures, Vol. 16, pp. 403-
413.

Iesan D. (1976). Saint-venant’s problem for inhomogeneous and anisotropic elastic bodies.
Journal of Elasticity, Vol. 6, Vol. 6, pp. 277-294.

Ladevèze P., Simmonds J. (1998). New concepts for linear beam theory with arbitrary geometry
and loading. European Journal of Mechanics, Vol. 17, No. 3, pp. 377-402.

Lowe P. (1971). Classical theory of structures. Cambridge University Press, U. K.

Pipes R. B., Pagano N. J. (1970). Interlaminar stresses in composite laminates under uniform
axial extension. J. Composite Materials, pp. 538-548.

Rand O. (1998). Interlaminar shear stresses in solid composite beams using a complete out of
plane shear deformation model. Computers and Structures, Vol. 66, No. 6, pp. 713-723.

Rappel O., Rand O. (2000). Analysis of elastically coupled thick-walled composite blades.
International Journal of Solids and Structures, Vol. 37, pp. 1019-1043.

Renton J. D. (1991). Generalized beam theory applied to shear stiffness. International Journal
of Solids and Structures, Vol. 27, No. 15, pp. 1955-1967.

Sanchez P. (2001). Implementation and illustrations of the exact beam theory (mise en oeuvre
et illustrations de la théorie exact des poutres). Unpublished doctoral dissertation, Ecole
Normale Supérieure de Cachan (Paris-VI).

Trabucho L., Viano J. M. (1996). Mathematical modelling of rods, handbook of numerical


analysis, vol. iv. P.G. Ciarlet, P. G. & J.L. Lions, J. L. éditors, Elsevier.

Volovoi V. V., Hodges D. H., Cesnik C. E. S., Popescu B. (2001). Assessment of beam model-
ing methods for rotor blade applications. Mathematical and computer modelling, Vol. 33,
pp. 1099-1112.

Annex A. The numerical method

This appendix recalls the essential of the numerical method used by CSection:
– the mathematical characterization of the sectional characteristics;
– and, some points related to the computation technique.

It is worth noting that the method leads to solve a set of 2D-problems using the
classical 6-nodes triangle as finite element but these problems do not correspond to the
classical plane-stress or plane-strain models. More details can be found in (El Fatmi,
Zenzri, 2002).
Characteristics of any composite section 411

A.1. Mathematical characterization of the sectional characteristics

The problem is to find, for a given section, the sectional characteristics {Λ, U k ,
σ } for (k = 1, ..., 6). These can be deduced from the solutions of 6 problems P k
k

defined on the cross section.


Let λij be the components of Λ and V (x, y) a 3D vector field defined on the
section. We introduce the constant vectors
   ∗ 
λ3k η3  k 
 λ4k  ∗  βx
k ∗
 η k
λ =   η =  ∗  β = βyk 
 4   (11)
λ5k  η5
∗ βzk
λ6k η6

and the displacement field ξ ∗ =ξ ∗ (η ∗ , V ∗ ) with its correspondent strain tensor ε∗ =


ε(ξ∗ )
   ∗ 
εxx Vx,x
 εyy   Vy,y ∗
1 ∗ 2 ∗ ∗
  
2 η5 z − η6 zy + Vx
   ∗ ∗

γxy   Vx,y + Vy,x
ξ∗ =  − 12 η4∗ z 2 + η4∗ zx + Vy∗  ε∗ = 
   
= ∗ ∗  (12)

 γzx   η1 − η6 y + Vz,x 
  
η3∗ z + η4∗ xy − η4∗ zx + Vz∗  γzy   η2∗ + η6 x + Vz,y ∗ 

εzz ∗ ∗ ∗
η3 + η4 y − η5 x

Note that ε(ξ ∗ ) is independent of z.

P ROPOSITION 2. — k ∈ {3, 4, 5, 6}. For (η k , V k ) minimizing the functional J k de-


fined by Z
J k (η ∗ , V ∗ ) = ε∗ : K : ε∗ dS − ηk∗ (13)
S
we have
λk

= ηk   

−λ2k



Uk = V k + β k +  λ1k  ∧ GP (14)
αk




σk = Kε k 

where the constants (αk , βzk , λ1k , λ2k ) for k ∈ {3, 4, 5, 6} are given by:
Z Z 
βzk = − C 3 .V k dS λ2k = C 4 .V k dS 

ZS SZ (15)
α k
= − C 6 .V k dS λ1k = − C 5 .V k dS 
S S

with C k = σ k · z. The constants (βxk , βyk ) are given by Equation (22) which needs
the solutions of the problems P 1 and P 2 .
412 RCMA. Volume 22 – no 3/2012

For the solution of problems P 1 and P 2 , we introduce the strain tensors defined
by:

   
0 0

 0 


 0 

4
 0  5
 0 
D =  D =  (16)

 −α4 y + Vx4 


 −α5 y + Vx5 

 α4 x + Vy4   α5 x + Vy5 
βz4 − λ14 x − λ24 y + Vz4 βz5 − λ15 x − λ25 y + Vz5
which suppose that the problems P 4 and P 5 are solved.

P ROPOSITION 3. — For (η 1 , V 1 ) and (η 2 , V 2 ) respectively minimizing the function-


als
Z Z
1
J 1 (η ∗ , V ∗ )) = (ε∗ : K : ε∗ ) dS- (ε∗ K : D 5 -V ∗ .C 5 ) dS (17)
2 S S
Z Z
∗ 1
2 ∗
J (η , V )) = (ε : K : ε ) dS- (−ε∗ K : D 4 +V ∗ .C 4 ) dS (18)
∗ ∗
2 S S

we have

λ1

= η1 

−(λ21 − βy5 )
  


1 1
U = β +  (λ11 − βx5 )  ∧ GP + V 1 (19)
α1




5
σ1 = K : (ε1 − D )

λ2

= η2 

−(λ22 + βy4 )
  


2 2
U = β + (λ12 + βx4 )  ∧ GP + V 2 (20)
α2




4
σ2 = 2
K : (ε + D )

where, with C 1 = σ 1 · z and C 2 = σ 2 · z. All the constants that remain are given
by
Z Z 
βz1 = − C 3 .V 1 dS βz2 = − C 3 .V 2 dS 


S S
Z Z 



λ21 − βy5 = C 4 .V 1 dS λ22 + βy4 = 4
C .V dS2 


SZ SZ (21)
λ11 − βx5 = − C 5 .V 1 dS λ12 + βx4 = − C 5 .V 2 dS  


ZS ZS 


α1 = − C 6 .V 1 dS α2 6
= − C .V dS  2 

S S
Characteristics of any composite section 413

and Z 
βxk = − C 1 .V k dS 

for (k = 1, ..., 6) ZS (22)
βyk = − C 2 .V k dS 

S

A.2. Computation technique

The propositions 2 and 3 identify six minimization problems allowing the determi-
nation of the sectional characteristics {Λ, U k , σ k }. In each minimization problem,
the unknowns are the constant vector, η k , and the 3D vector field V k , which is only
function of the sectional coordinates (x, y). Using the finite element method to deter-
mine such unknowns, only the cross-section S has to be discretized.
For a given composite cross-section (geometry and materials), the program achieves
successively the numerical minimization of the quadratic functionals J 3 , J 4 , J 5 , J 6
and then J 1 , J 2 . The cross-section is discretized into triangular or quadrilateral ele-
ments, and for every functional J k , the numerical minimization consists of solving a
linear system in which the unknowns are the four components of the constant vector
η k and the degrees of freedom of the three components of the vector field V k .
Using the standard techniques of the finite element method, the stiffness matrix and
the second member of the linear system are derived respectively from the quadratic
term and the linear term of the functional J k . Note that in the Propositions 1 and 2,
the symmetry of Λ has been ignored in order to get the same quadratic term for the
functionals J k involved in the six problems; therefore, the stiffness matrix is the same
for the six problems and only the second members differ (which is very economic).

Vous aimerez peut-être aussi