Journal of Biomechanics 34 (2001) 277}283

Technical note
2D calculation method based on composite beam theory for the
determination of local homogenised sti!nesses of long bones
ED rik Estivalezes*, BeH atrice Couteau, Robert Darmana
INSERM U518, Laboratoire de Biome& canique, Amphitheatre Laporte, BP 3103, 31026 Toulouse Cedex 3, France
Accepted 20 August 2000

Abstract

A calculation method using the "nite element technique is presented. Its main objective was to determine strains, stresses and more
particularly sti!nesses in any cross section of a tibia, thus enabling the localisation of tibial torsion in vivo. Each tibial cross section
was considered to be a non-uniform cross section of a composite beam with arbitrary orientation of "bres. The determination of
stresses, strains and sti!nesses within a composite beam cross section has been de"ned by solving a variational problem. The
validation of this method was performed on a tibial diaphysis of which each cross section was assumed to be the cross section of
a composite beam made of orthotropic materials with orthotropic axes of any orientation with respect to the principal axis of the
bone. The comparison of the results, from our model and that of a three-dimensional one, was performed on each nodal value (strains,
stresses) of the meshed cross section as it was impossible to obtain local sti!nesses by experimentation. The good agreement between
the results has validated our "nite element program. Actually, this method has enabled to treat directly 2D geometric reconstructions
from CT scan images with a good accuracy to determine locally the homogenised mechanical characteristics of human tibia in vivo,
and particularly to quantify torsional tibial abnormalities of children without approximation of the shape of the cross section and by
calculating the real moment of inertia J. The importance of the "bre orientation with regards to the sti!ness values has been
emphasised. This 2D method has also allowed to reduce CPU time of the 3D modelling and calculation.  2001 Elsevier Science
Ltd. All rights reserved.

Keywords: Finite element calculation; Mechanical properties; Composite beam; Torsion; Tibia

1. Introduction a one-dimensional model by using integral parameters or

Many investigations have been performed on the anal-
ysis of long bones. On a macroscopic point of view,
Piziali et al. (1976) presented a computer model having
a wider applicability and greater accuracy than previous
models but limited to isotropic material properties. Ken-
nedy and Carter (1985), and Kennedy et al. (1985) studied
the e!ects of heterogeneity, anisotropy and geometric
irregularity on long bone undergoing torsion loading,
and presented a comparative 2D study between simple
models (circular and elliptical annulus models) and "nite
element models of the bone geometry. By analogy with
the classical theory of homogeneous isotropic beams,
some works have been done on composite beams by
reducing the study of a three-dimensional model to
* Corresponding author. Tel.: #33-5-61-49-79-44; fax: #33-5-61-
49-67-45.
E-mail address: erik.estivalezes@purpan.inserm.fr (ED rik Estivalezes).
the homogenisation method in the "eld of mechanics
(Barrau et al., 1985; Gay, 1987).
Composite beams have already been treated in various
works for example, in Nixon (1987); Nouri (1993); Pus-
pita et al. (1993); Whitney and Kurtz (1993); WoK rndle
(1982), but all these studies assumed that an orthotropic
axis of beam constituents was aligned with the longitudi-
nal axis of the beam. However, optimisation of beam
behaviour in mechanics, or bone con"guration in bi-
omechanics led to study composite materials whose or-
thotropic axes were no longer orthogonal to the cross
section (as it could be the case for the lower part of a tibia
because of the geometrical torsion of the section). Conse-
quently, previous theories were no longer valid due to
coupling terms.
This paper presents a method that has been developed
to study tibial rotation abnormalities. Actually, torsional
tibial abnormalities of children are a frequent phenom-
enon which may have important consequences on gait

0021-9290/01/$ - see front matter  2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 2 1 - 9 2 9 0 ( 0 0 ) 0 0 1 7 7 - 9
278 E. Estivale% zes et al. / Journal of Biomechanics 34 (2001) 277}283
and joints. The method developed was used as an objec-
tive assessment of bone rigidities for analysing tibial
disorders such as torsional abnormalities of varying se-
verity. In previous works cited in the paper, approximate
calculation of the moment of inertia J has been done, and
only geometrical considerations were taken into account.
Our method is calculating the `reala J in torsion with
account of the real shape of the cross section as well as
the heterogeneity of the spongious bone and the orthot-
ropy of the cortical bone. Consequently, it enables to
determine the localisation of tibial torsion in vivo.
2. Method
In order to directly calculate the homogenised (or
equivalent) sti!nesses of each cross section of a tibia
reconstructed from CT images, some hypotheses have
been done (composite structures assimilated to beams
can be homogenised, meaning that their study is analog-
ous to the study of classical homogeneous beam of the
literature).
Firstly, it is supposed that the studied cross section of
the tibia is one of a theoretical composite beam where
geometrical and mechanical properties are constant with
respect to the coordinate of the longitudinal axis (X ) of
the beam (Fig. 1). In fact, when working in a cross section,
its thickness is supposed to be very small and, conse-
quently, variations of the properties within the thickness
are supposed to be zero.
Fig. 1. De"nition of the theoretical beam from CT images and theoret-
ical beam characteristics for the problem formulation.
Secondly, to avoid any boundary e!ects (Saint-Venant
principle), the considered cross section is located at the
middle of a 200 mm length beam (far from both extremi-
ties of the beam).
Thirdly, this theoretical beam is embedded with free
warping at one extremity while the other extremity
undergoes the di!erent loads. These hypotheses are pure-
ly mathematical and under no circumstances they are
supposed to model a speci"c biomechanical situation,
they represent the classical assumptions of composite
beam theory.
The mechanical formulation of the problem has been
made in a general context by setting a system of theoret-
ical equations.
3. Theory
3.1. General formulation
Fig. 1 shows a typical beam made of N orthotropic
phases whose orthotropic axes are not parallel to the
longitudinal axis of the beam. The beam of span L was
clamped at one end S (x"0) and was free to warp

(regarding the de"nition of the displacement "eld, it is
equivalent to say that the average displacements of the
root cross section equal to zero). At S (x"), the beam

was subjected to di!erent loads like tensile force N ,
V
bending moment M and torque M .
D R
For each loading, a calculation model was determined
by using the principle of virtual work and depending on
the cases, the Lagrange multiplier technique. For brevity
sake, many details of the derivation have been omitted,
readers with further interest are referred to Estivalezes et
al. (1993a,b, 1995).
3.2. Basic assumptions used in the analysis
The classical beam theory is applied (p p , q is
WW XX WX
negligible with respect to the stresses p , q , q , Tim-
VV VW VX
oshenko and Goodier, 1987), which is reinforced by the
fact that the di!erent materials considered were assumed
to be perfectly stuck together with similar Poisson's ratio.
E In any cross section, beam characteristics (geometry,
materials) were constants.
E It was presumed that a cross section S between S and

S underwent a translation, a rotation and a warping.

3.3. Displacement xeld
For each constant loading case N , M and M , since
V D R
these forces or moments were not depending on x,
and according to previous hypotheses, the chosen
 E. Estivale% zes et al. / Journal of Biomechanics 34 (2001) 277}283 279

displacement "eld was the loads N , M and M ;

V W X

u"u (x)!y.h (x)#z.h (x)#g(y, z),

 X W
v"v (x)!z.h (x)#f (y, z),
 V
(1)
;"#j
  p (N ) ) u(x"0) dS2
VV V
1
w"w (x)#y.h (x)#h (y, z).
 V

g(y, z) was the seeking out-of-plane warping function,

#b
 p (M )u(x"0) dS#d
VV W  p (M )u(x"0)
VV X
f (y, z) and h(y, z) were two in-plane warping functions. 1 1
dS!<  , (5)
After derivation of the displacement "eld and con- $
sidering the boundary conditions, the strain "eld where < CVR was the potential energy due to the applied
became $
forces.
By setting the global vector of the unknowns as
e "a!yd#ze, +q,2"+a, b, c, d, e, f, j, b, c, g , g 2 g , where the g are
VV    L G
the nodal warping, the minimisation of U with respect to
*g(y, z) the unknowns leads to the following linear system:
c "b#yf# , (2)
VX *z [K]+q,"[F ]and the solution is given by

*g(y, z) +q,"[K]\[F ]. By de"nition, the sti!ness of a struc-

c "c!zf# , ture submitted to a loading is the ratio of this applied
VW *y
load by the resulting displacement. Consequently, each
where a}f are constants. sti!ness associated with a loading was obtained by divid-
Supplementary hypothesis for M : Warping did not do ing the applied e!ort by the corresponding global dis-
R placement created according to the behaviour
work on the initial cross section (S ). This meant that
 relationships. Thus, the di!erent sti!nesses could be ex-
the work due to the tensile force and the bending mo-
ments were zero for a displacement resulting from the pressed as:
Tensile sti!ness: 1ES2"N /a; #exural sti!ness with
torsion loading. Then the following relations were V
respect to the y-axis: 1EI 2"M /e #exural sti!ness with
derived: W W
respect to the z-axis: 1EI 2"M /d; torsional sti!ness:
X X
1GJ2"M /f. Comparing to classical sti!ness of iso-
 p (N ) u(x"0) dS"0, 
VV V tropic structures subjected (to tensile: ES, to #exion: EI
W
1 and EI , and to torsion: GJ), here the sti!nesses in
 p (M ) u(x"0) dS"0, (3) X
VV W pointed brackets are homogenised sti!nesses.
1
 p (M ) u(x"0) dS"0,
VV X
1
4. Calculation
where p (N ), p (M ), p (M ) were the normal stresses
VV V VV W VV X A programme written in FORTRAN 77 has been
found for the each previous loading N , M and M .
V W X created. This programme is a modular tool and can be
According to the beam theory, the stress "eld in a ma-
terial `ia could be written as used with any post processing "nite element software. It
means that this program just need an input ASCII data
"le from any meshing program (containing node coordi-
p "Q e #Q c #Q c ,
VVG G VV G VX G VW nate, element connectivities and material properties of
q "Q e #Q c #Q c , (4) each element). This code, which is run for each type of
VXG G VV G VX G VW
q "Q e #Q c #Q c . load, allows the calculation of strains and stresses at each
VWG G VV G VX G VW node of the meshing and gives the local sti!ness in ten-
sion, #exion and torsion of the studied cross section. The
3.4. Numerical resolution method of resolution consists in solving numerically the
linear system [K]+q,"[F ] using the classical method

For each loading case, the solution was obtained by of elimination of Gauss.
the writing of an equivalent functional allowing the cal-
culation of the warping function g(y, z) on the meshed
cross section using the "nite element method. 5. Validation
The following functional was chosen, taking into ac-
count the supplementary hypothesis by means of the It seems very di$cult to determine experimentally the
Lagrange multipliers (j,b,d), these multipliers are zero for local sti!ness of a beam with any cross section and which
280 E. Estivale% zes et al. / Journal of Biomechanics 34 (2001) 277}283
Fig. 2. Meshing of the cross section (3D model (left) and 2D model (right)).
constituents have orthotropic axes of any orientation.
Especially, when such a beam is subjected to torsional
loading, because of coupling e!ects combined to
section warping. So validation has been performed by
comparison of the values given by the presented
2D model and values extracted from the middle cross
section of a 3D "nite element model of the studied beam
(Fig. 2).
The 3D uniform beam of 200 mm length was created
with a 3D "nite element software (Patran7.5, MSC
Corp.). The creation of solid elements was de"ned by the
extrusion of the cross section obtained geometrically
from a CT scan of a tibial diaphysis. The 3D model was
composed of 600 hexaedric elements with a total number
of 945 nodes. One extremity of the beam was constrained,
the other end being subjected to an equivalent torsional
loading. The values of nodal stresses and strains have
been extracted from the middle cross section to avoid
boundary e!ects due to the embedding and to the points
where the loads were applied. The principal orthotropic
axis of the material of the beam (axis No. 1 of the cortical
material) was oriented arbitrarily by an angle of 453 with
respect to the longitudinal axis of the beam (axis x) in the
plane (x,y) as described in Fig. 1. There is no restriction
on the number and type of the di!erent materials con-
stituting the composite beam.
The 2D FEM model corresponding to the tibial dia-
physis cross section was meshed with 3 node triangular
elements. The 2D model was composed of 60 elements
with a total number of 45 nodes.
For both models, the material properties of the cortical
bone, in its orthotropic coordinate system, was taken
from the literature (Hobatho et al., 1991), the elastic
values expressed in the orthotropic coordinate system
de"ned in Fig. 1 are
E "17600 MPa, G "4400 MPa, t "0.461,
  
E "8800 MPa, G "5120 MPa, t "0.4,
  
E "9600 MPa, G "2960 MPa, t "0.464.
  
6. Results
The results were plotted at each node, as the coordi-
nate and number of node were the same in the two
models (Fig. 2). The horizontal axis represents the num-
ber of each node of the meshing of the cross section, and
the vertical axis represents strain or stress values.
The comparison of the results was performed for each
loading case, but for brevity sake only results corre-
sponding to the torsion loading case Mt is presented, as it
required the calculation of the three others : N , M and
V W
M (Eq. 3).
X
For the torsion case, only strains and stresses in the
plane (x, z) and (x, y) were non-negligible, and as they
were proportional to the strains, only stresses are illus-
trated in Fig. 3. The comparison consisted in the calcu-
lation of the percentage of error between the 2D and the
3D model for each node; Error (%)"(100"(A!B)/A")
where A is the value from the 3D Model and B is the
value from the 2D Model. Due to its de"nition, the
parameter error is only reliable for values that are not
closed to zero (if A is close to zero, then the ratio
"(A!B)/A" tends to high values of percentage).
As a matter of fact, when comparing stress values
taken from the 3D and the 2D models, the maximum of
error committed was for values closed to zero. This is
illustrated in the Fig. 3 for nodes number 22, 23 and 24
and for nodes number 10, 11, 12 and 34, 35, 36. The
global average error was 21.5% in the plane (x,z) and
33.29% in the plane (x,y). When neglecting these values
closed to zero, the global average error was reduced to
5% in the plane (x,z) and 8.8% in the plane (x,y). It
demonstrated the accuracy of the 2D method, globally
less than 9%, with respect to a 3D "nite element model.
Concerning the in#uence of the principal orthotropic
axis orientation with respect to the di!erent sti!nesses in
tension, bending and torsion, for each variation of the
orthotropic axis orientation, from 0 to 90 with a pitch of
153, all sti!nesses were quickly reduced. It was noted that
the decrease of tension and bending sti!nesses was be-
 E. Estivale% zes et al. / Journal of Biomechanics 34 (2001) 277}283 281

Fig. 3. Comparison between proposed model and MSC/Patran for stress q and for stress q in MPa.
VX VW

tween two and three times higher than for torsion sti!-
ness. A variation of 153 with respect to the principal axis
(oriented at 03) lead to a 4% decreasing of the sti!ness in
tension and bending and a 2% decreasing of the torsional
sti!ness. So it was deduced that a variation of the ori-
entation of the principal orthotropic axis was less a!ect-
ing the torsional sti!ness.
This programme has been used to calculate the in#u-
ence of the principal orthotropic axis orientation with
respect to the di!erent equivalent (or homogenised) sti!-
nesses in tension, bending and torsion. The Fig. 4 illus-
trates the values of the di!erent sti!nesses when the
orientation of the principal orthotropic axis (axis n31 of
Fig. 1) varies from 0 to 903 with a pitch of 153 in the plane
(x,y).
7. Discussion
The presented model is able to treat any structure
comparable to a beam made of orthotropic materials
where the orthotropic principal orientation is not neces-
sarily coincident with the beam longitudinal axis. Actual-
ly, in the lower part of the tibia, the cross section is
twisted of about 153 with respect to the principal axis of
the tibia, and this o!-axis angular variation is quite
signi"cant on the values of the di!erent equivalent rigid-
ities. For a beam made of one material with a spiral
distribution of the orthotropic axes, as the orientation
varies with respect to the localisation in the cross section
of the considered "nite element, the calculation can be
performed by considering the material of each element
282 E. Estivale% zes et al. / Journal of Biomechanics 34 (2001) 277}283
Fig. 4. Variation of the tensile sti!ness 1ES2, of the #exural sti!ness 1EI 2, of the #exural sti!ness 1EI 2 and of the torsional sti!ness 1GJ2 in MPa
W 
with respect to the orientation of the orthotropic axis.
apparently di!erent. It can be done by expressing the
element sti!ness matrix of the orthotropic coordinate
frame in the global coordinate frame of the cross section.
Actually, the material is the same for the entire cross
section, but the change of orientation results in a di!erent
sti!ness matrix in each element when it is expressed in the
global coordinate system.
The only limitation in using this methodology is the
`cuttinga of the tibia to slices and to consider that each
slice corresponds to the cross section of a composite
beam made of di!erent materials. Nevertheless, it seems
to be the only way to determine the locally equivalent
sti!nesses in any point of the tibia, between distal and
proximal regions to "nally determine the localisation of
torsion tibial in vivo. Furthermore, the spatially vari-
ations of the geometry and material properties in the
x-direction are indirectly taken into account as each
cross section, reconstructed from CT scan images,
can have a di!erent shape and can be made of any
material (especially, in the epiphyses made of spongious
and cortical bone). Of course, the use of a classical
3-D approach will allow to determine all components
of stress but it will only give the global rigidity of the
entire tibia and there will be no way to obtain the local
ones.
This programme has also been used to study tibial
abnormalities in children (Limbert et al., 1998) and the
results have shown the importance of considering simul-
taneously the geometrical and mechanical properties in
the assessment of child bone rigidities and "nally demon-
strated that the bone sti!ness of children with torsional
deformities was not altered.
To conclude, this calculation model has enabled to
study composite beams and particularly long bones of
any cross section subjected to di!erent loads. The use of
the principle of virtual work combined to the Lagrange
multiplication technique was well adapted to solve this
kind of problem. The formulated hypotheses as well as
the method used have been demonstrated and numer-
ically validated. The coupling e!ects as well as warping
were considered.
The comparison between the obtained results and
those from a three-dimensional "nite element calculation
software enabled to validate our calculation method on
the determination of local sti!nesses of a human tibial
diaphysis. The method presented has given a close ap-
proximation (inferior to 9%) of the di!erent sti!nesses in
any tibial diaphysis cross section when compared to
a classical 3D FEM software. It has also demonstrated
the e!ects of the orientation of the principal orthotropic
axis on the calculation of the sti!nesses. Furthermore,
calculation has been achieved with a rather small number
of element, as the meshing was only done on the cross
section (45 nodes and 60 elements instead of 945 nodes
and 600 elements in the 3D model), and the solution was
obtained quickly.
This program has enabled a better understanding of
tibial rotational abnormalities by bringing a quantitative
part (taking into account the mechanical and geometrical
properties of each cross section from epiphysis to diaphy-
sis) to the qualitative one obtained by the use of imagery
technique like radiography, MRI, or CT scans.
Actually, the use of this method has been a
major element for the determination of homogenised
 E. Estivale% zes et al. / Journal of Biomechanics 34 (2001) 277}283
mechanical characteristics of human tibia (Limbert
et al., 1997) and for the quanti"cation of rotational
abnormalities in children (Limbert et al., 1998)
by the determination of the localisation of tibial torsion
in vivo.
References
Barrau, J-J., Chambard, O., Gay, D., Nuc, M., 1985. HomogeH neH isation
en torsion d'une poutre composite. In: Proceedings of Troisieme
Colloque Tendances Actuelles en Calcul de Structures. Bastia, Plu-
ralis, pp. 283}296.
Estivalezes, E., 1995. Etude et deH veloppement d'un eH leH ment de poutre
composite pour un reve( tement deH seH quilibreH . PhD. Thesis, I.N.S.A.,
Toulouse, France.
Estivalezes, E., Barrau, J-J., Ramahefarison, E., 1993a. DeH termination
des contraintes normales et de cisaillement dans la section
droite d'une poutre composite. In: Proceedings of 113 Congres
Franc7 ais de MeH canique. Lille Villeneuve d'Ascq, France, pp.
201}204.
Estivalezes, E., Barrau, J-J., Ramahefarison, E., 1993b. Determination
of homogenised characteristics for composite beams. In: Proceed-
ings of the sixth European Conference on Composite Material.
Whoodhead Publishing, Bordeaux, France, pp. 211}216.
Gay, D., 1987. Les poutres composites; #exion, torsion. In: Hermes
(Ed.), MateH riaux Composites. Hermes, Paris, pp. 283}313.
Hobatho, M.C., Ashman, R.B., Darmana, R., Morucci, J.P., 1991.
Assessment of the elastic properties of human tibial cortical bone by
ultrasonic measurements. In: Van Der Perre, G., Lowet, G., Bor-
gwardt Christensen, A. (Eds.), In vivo Assessment of Bone Quality
by Vibration and Wave Propagation Technique. Durham, UK, pp.
45}56.
283
Kennedy, J.G., Carter, D.R., 1985. Long bone torsion: I E!ects of
heterogeneity, anisotropy and geometric irregularity. Journal of
Biomechanical Engineering 107, 183}188.
Kennedy, J.G., Carter, D.R., Caler, W.E., 1985. Long bone torsion: II.
A combined experimental and computational method for determin-
ing an e!ective shear modulus. Journal of Biomechanical Engineer-
ing 107, 189}191.
Limbert, G., Estivalezes, E., Couteau, B., Hobatho, M.C., 1997. In vivo
structural analysis of human tibia: determination of homogenised
mechanical characteristics. Innovation in Technology in Biology
and Medicine 18, 187}200.
Limbert, G., Estivalezes, E., Hobatho, M.C., Baunin, C., Cahuzac, J.P.,
1998. In vivo determination of homogenised mechanical character-
istics of human tibia. Application to the study of tibial torsion in
vivo. Clinical Biomechanics 13, 473}479.
Nixon, M. W., 1987. Extension-twist coupling of composite circular
tubes with application to tilt rotor blade design. American Institute
of Aeronautics and Astronautics 87}0772.
Nouri, T., 1993. HomogeH neH isation et calcul des contraintes de cisaille-
ment dans les poutres composites a sections quelconques et a con-
stituants orthotropes. Ph.D. Thesis, UniversiteH Paul Sabatier,
Toulouse, France.
Piziali, R.L., Hight, T.K., Nagel, D.A., 1976. An extended structural
analysis of long bones-Application to the human tibia. Journal of
Biomechanics 9, 695}701.
Puspita, G., Barrau, J-J., Gay, D., 1993. Computation of #exural and
torsional homogeneous properties and stresses in composite beams
with orthotropic phases. Composite Structure 24, 43}49.
Timoshenko, S.P., Goodier, J.N., 1987. Theory of Elasticity, 3rd ed.
McGraw-Hill, New York.
Whitney, J.M., Kurtz, R.D., 1993. Analysis of orthotropic laminated
plates subjected to torsional loading. Computational Engineering 3,
83}97.
WoK rndle, R., 1982. Calculation of the cross section properties and the
shear stresses of composite rotor blades. Vertica 6, 111}129.