Materials Science & Engineering C 117 (2020) 111276

Contents lists available at ScienceDirect

Materials Science & Engineering C

journal homepage: www.elsevier.com/locate/msec

Experimental characterization and micromechanical modeling of the elastic T

response of the human humerus under bending impact
⁎
J. Rahmouna, , H. Naceura, H. Morvana, P. Drazetica, C. Fontaineb, P.E. Mazeranc
a
Univ. Polytechnique Hauts-de-France, CNRS, INSA Hauts-de-France, UMR 8201-LAMIH, F-59313 Valenciennes, France
b
Univ. Lille 2, Faculté de medecine, Laboratoire d'Anatomie, CHRU de Lille, F-59045 Lille, France
c
Univ. de Technologie de Compiègne, CNRS, FRE 2012-ROBERVAL, F-60205 Compiègne, France

A R T I C LE I N FO A B S T R A C T

Keywords: This paper investigates the characterization and numerical modeling of the elastic behavior of the human hu-
Human humerus bone merus bone using a recently developed micromechanical approach coupled to nanoindentation measurements.
Micromechanics At first, standard three-point bending experiments were conducted under low static loading, using several hu-
Nanoindentation merus diaphysis in order to identify the apparent elastic modulus of the bone in static regime. Then, a drop tower
Drop tower impact
impact experiment was used on the same set of humerus diaphysis specimens, in order to assess the elastic
Finite Element
modulus in dynamic regime. These measurements will be used as reference bases for comparison purpose. The
originality of this work, lies in the coupling between a two-phase micromechanical approach based on Mori-
Tanaka homogenization scheme for cylindrical voids and nanoindentation measurements of the elastic modulus
of the bone matrix phase. This model has been implemented using a user defined material subroutine VMAT in
ABAQUS© Explicit code. The bone mechanical response prediction using the proposed methodology was vali-
dated against previous standard experimental data. Finally, it was shown that the numerical predictions are
consistent with the physical measurements obtained on human humerus via the good estimation of the ultimate
impact load.

1. Introduction For the determination of the humerus mechanical properties and its
integration in numerical models, characterization tests of humerus bone
With the growing deployment of airbag protection systems by cars are needed. Moreover, global tests performed on an humerus are also
manufacturers to increase passengers safety, it has been noticed an required for the validation of bone models with respect to the experi-
increased emphasis on the study of the upper extremity injuries in the mental data. Concerning the studies investigating the material proper-
automotive environment [17]. These research studies include the in- ties of the humerus bone, we can cite the study of Singh et al. [33] who
vestigations on injuries of the upper extremities, in particular humeri, supplied the needed mechanical parameters given by different global
from deploying airbags. The humerus is the second most common long tests (three-point bending, fracture toughness, and axial compression).
bone diaphyses known. Its bone functional adaptation to strain is Hoffler et al. [15] performed nanoindentation tests on lamellae level
achieved by the addition and redistribution of tissue [32]. Moreover, and found that the elastic modulus and hardness of human bone tissue
the humeral bone is an hierarchical material whose architecture differs depend strongly on tissue type, anatomical location and individual.
at each level of hierarchy and whose mechanical properties can vary Nevertheless, most of the existing studies concerning global tests on
considerably, even on the same specimen, due to the bone hetero- humerus are mainly done in quasi-static conditions [5,16]. Impact
geneity. Different approaches have been performed to account for the testing provides information about the response of a material to a dy-
multi-scale nature of bone. Homogenization techniques have been de- namic load. This knowledge is essential for materials whose properties
veloped to model its elasticity and fracture risk [21]. Interestingly, are time-dependent, e.g. exhibit viscoelasticity or poroelasticity. Most
continuum micromechanical based approaches allow the modeling of biological materials have such properties and yet their responses at
the bone anisotropic mechanical properties [12,27]. The predictive loading rates that may be found during rapid physiological activities, or
potential of the so-called micromechanical formulation has been shown when subject to traumatic forces such as during falls or accidents, are
using physically and statistically independent sets of experiments. still not well understood. Only in few studies, the humerus has been

⁎
Corresponding author.
E-mail address: [email protected] (J. Rahmoun).

https://doi.org/10.1016/j.msec.2020.111276
Received 15 January 2019; Received in revised form 15 June 2020; Accepted 5 July 2020
Available online 11 July 2020
0928-4931/ © 2020 Elsevier B.V. All rights reserved.
J. Rahmoun, et al. Materials Science & Engineering C 117 (2020) 111276

loaded at level of velocity close to those to which the humerus bone upper face midway between the contact point and the end of the useful
could be subjected during a car accident. We can cite the work of Le- length of the bone (gages 1 and 2) and three other on the lower face
barbé et al. [19] who investigates the human shoulder response to respectively facing the two first gauges and the point of contact as
lateral impact at intermediate velocities. The obtained shoulder injury shown in Fig. 1b. Data recorded using a digital oscilloscope and filtered
mechanism appeared to be similar to those observed in the automotive using CFC 1000 low pass filer to remove high frequency component.
field. After instrumentation, the system humerus-brackets was scanned using
The aim of the present paper is to propose a micromechanical based a 3D scanner VIVID 910 KONICA MINOLTA VI910 (0.1 mm of resolu-
approach for the accurate characterization and modeling of the elastic tion) in order to obtain detailed description of boundary conditions as
behavior of the human humerus under bending impact. The present well as the precise locations of the gages in order to perform compar-
micromechanical model [30,31,36] is based on the coupling between ison with the Finite Element model.
the Mori-Tanaka homogenization scheme [24] for the estimation of the Three-points bending experiments were carried out using a drop
elastic properties of the humeral cortical bone, and a Finite Element tower impact apparatus of 2.30 m height. A drop mass fixed to a car-
model at the macro-level for the prediction of the global humerus riage was allowed to fall freely while guided on vertical rails. The
diaphysis elastic response. The proposed procedure consists in two carriage height can be adjusted to adapt the impact velocity (see
stages: at first, local measurements of the bone material properties were Fig. 1a). A laser sensor M7L was fixed to the drop tower bed in order to
performed thanks to nanoindentation tests using small specimens ex- record the successive positions of the impactor. The humerus in-
tracted directly from the same humerus diaphysis. Once the bone ma- strumentation together with high speed camera (3000 fps) provides a
trix material parameters (Young's modulus, porosity) were determined, precise recording of the evolution of the impact load as well as the
a bending impact test was performed using a solid-shell Finite Element displacements/strain curves function of time.
model [8,26] to validate the global response of the modeled humain
humerus diaphysis. 2.1.2. Static bending experiments
First, static three-points bending experiments were performed on
2. Experimental characterization of the mechanical response of humeri diaphysis using several loads (carriage mass 1 kg, 3 kg, 5 kg).
the human humerus diaphysis under bending loading The measured parameters are the local strains given by the five gages
for each prescribed weight. A least squares minimization technique was
2.1. Three points bending tests of the humerus performed in order to determine the elastic modulus by means of an
inverse analysis which consists in fitting computed data to experimental
2.1.1. Sample preparation and instrumentation measurements which leads to optimal set of constitutive parameters
Five fresh left humeri were obtained from human cadavers with an (longitudinal Young's modulus in our case).
average age of 68.6 ± 7.4 years and an average body height of In Table 1 are listed the geometrical parameters and the mean es-
180 ± 7.8 cm. The humerus bones were supplied by the Department of timates of the apparent longitudinal Young modulus for each humerus.
Anatomy at Lille University School of Medicine in accordance with As it can be observed, the average longitudinal elastic modulus is
French regulations. The average length of these specimens was 19089.8 MPa with a standard deviation of 3082.78 MPa. This estima-
345 ± 14.1 mm with an average mass of 226 ± 56.8 g. Each spe- tion fits within the average values given by [39,25,33]. We can also
cimen has been submitted to a virological and serological analysis to observe that the humerus 582 exhibits the lowest value of the elastic
ensure that they do not exhibit any bone disease on radiographic ex- modulus. This is probably due to the low mass of this humerus.
amination. All specimens were scanned using a medical scanner. These Fig. 2 represents the stress-strain curves of the five humeri based on
acquisitions were performed to extract the digital geometry data and the estimated Young modulus by optimization procedure and the local
saved using the stereolithography (STL) standard format then exported strain measurement of all three tensile gages Within this load range, the
to a Computer Aided Design (CAD) workstation. Finally, the specimens static response of each humerus was always linear.
were embalmed using a solution composed of methanol (35%), phenol
(10%), glycerin (20%) and distilled water (35%) and stored at 4 °C until 2.1.3. Dynamic bending experiments
the beginning of experiments. Several studies [34,35,11] have in- The dynamic impact experiment on the humerus was chosen to hit
vestigated the effects of embalming bones regarding their mechanical in the middle of the diaphysis with a projectile of 8.6 kg mass and
properties. It be founded no or little effect after short to intermediate 20 mm diameter. A drop height of 750 mm was used to generate an
storage periods on the elastic properties of the human bones measured initial velocity of 1.63 m/s at the impact instant. The total impact en-
by experimental tests (see [28,34]). ergy of the present experiment was adjusted to be approximately 10 J.
Each humerus was prepared 24 h before the starting of experiments This value was shown from previous works [7], to be necessary to reach
in order to fix it rigidly inside two aluminum prismatic brackets (see the human humerus ultimate strength. The impact load was measured
Fig. 1b) which were positioned onto a flat support to ensure their using a 4 components load cell sensor (model KISLTER 9327A). A laser
parallelism. One of the metallic brackets was maintained, while the sensor was used to measure the displacement. The humerus in-
other was free to slide in order to adapt it to the desired length of the strumentation together with two high-speed video cameras provide a
bone. For each humerus, four positioning screws (Fig. 1b) were used to precise recording of the evolution of the impact load as well as the
help maintaining fixed the epiphyses and the lateral epicondyle of the displacements/strain curves function of time.
humerus to the brackets. The system humerus-brackets was first posi- The comparison between all humerus responses shows that only one
tioned vertically and a FASTCAST© Polyurethane resin (type F14-F15 humerus which did not break under the impacting load at the imposed
from AXSON Company) was poured into the lower bracket. Three hours initial velocity of 1.67 m/s.
later (recommended resin drying duration), the humerus-brackets Table 2 summarizes data obtained from the dynamic bending ex-
system was flipped and the same operation was done for the other periments. It can be noticed that the average bending stiffness (defined
bracket side. The screws are then removed and the system humerus- as the load divided by the deflection) of all specimens was about 465.6
bracket was separated from the straight support and placed in a cold N/mm with a standard deviation of 132.8 N/mm.
room. Regarding the humerus ultimate impact load, its average value was
Two hours before the starting of experiments, five strain gages (Y 1299.6 N with a standard deviation of 274.24 N. This value is consistent
series 1-LY41-3/350 from HBM) were glued on the humerus diaphysis with the results presented by Duprey et al. [7] who found a mean
to provide maximum tensile and compressive strains. Since the impact maximum force value of 1273 N with a standard deviation of 448 N
will be applied at the diaphysis middle, two gauges are placed on the obtained using dynamic bending tests on humerus with an average

2
J. Rahmoun, et al. Materials Science & Engineering C 117 (2020) 111276

Fig. 1. Experimental device used for the bending impact on the human humerus.

Table 1 Table 2
Mechanical properties of human humeral bone obtained from static loading. Mechanical properties of human humerus bone for impact experiments.
Humerus ID Length Mass Elastic modulus Humerus ID Velocity Energy Bending stiffness Ultimate impact load

(mm) (g) (MPa) (m/s) (J) (N/mm) (N)

535 325 250 18,954 535 1.67 11.99 541 1592

514 361 354 19,614 514 1.63 11.42 642 1342
517 355 305 22,445 517 1.67 11.99 473 1239
485 331 262 14,099 485 1.65 11.71 344 1456
582 351 184 20,337 582 1.72 12.72 328 869

2.1.4. Densitometric measurements

Quantifying regional heterogeneities in bone characteristics such as
the density might improve our understanding of the injuries mechanism
of the extremity fractures. Few studies investigate the regional variation
of the density of cadaver long bone. We can cite the densitometric
measurements of Boos et al. [4] in 14 cadaver tibias who found that the
mean bone densities varied between 123.79 mmg/mm3 and 267.99
mmg/mm3. For the humerus bone, Boos et al. [4,6] showed that the
mean bone mineral density of the metaphysis region was 0.91 ± 0.224
g/cm3, and the one of the epiphysis was 0.559 ± 0.139 g/cm3.
In this study, densitometric measurements were performed after
each dynamic test at the useful length of the impacted humerus in order
to determine the porosity f of the diaphyseal cortical bone of the hu-
merus. After removing brackets and strain gages, the humeri were
cleaned from the bone marrow, cut into 8 zones (z1 to z8) and weighted
using a precision weighing balance from OHAUS company (Fig. 3).
Then, the true density dt were calculated from specimens weights in air
and inside water by means of the Archimedes principle.
In Table 3 are listed true densities of each zones of the humerus
bones. It can be observed that the mean true densities dt varied between
Fig. 2. Experimental response of the humerus bones under static bending.
1.69 g/cm3 and 1.89 g/cm3 showing the existence of interindividual

velocity of 1.77 m/s. One can also cite the work of Mukherjee et al. [25]
who performed dynamic bending tests on the shoulder bones using
three different impact velocities. The mean ultimate impact load of
humerus bones corresponding to the lower impact velocity considered
was about 1275 N with a standard deviation of 671.75 N. This ultimate
impact load increased naturally with the increasing of the impact ve-
locity.
Fig. 3. Specimens localization on the humerus.

3
J. Rahmoun, et al. Materials Science & Engineering C 117 (2020) 111276

Table 3
True densities of human humerus bone (g/cm2).
Humerus ID z1 z2 z3 z4 z5 z6 z7 z8 Mean SD

535 1.79 1.84 1.87 1.75 1.71 1.90 2.17 1.79 1.85 0.14
514 1.87 1.92 1.81 1.79 1.86 1.95 1.86 1.51 1.82 0.13
517 1.66 1.93 1.88 1.75 1.87 2.35 1.90 1.80 1.89 0.20
485 1.56 1.76 2.36 1.67 1.85 1.81 1.78 1.70 1.81 0.23
582 1.83 2.13 1.84 1.80 1.98 1.89 1.53 1.84 1.86 0.17

differences. Moreover, regarding the same humerus, there is also in-

traindividual variances with respect to the eight zones.
Finally, the porosity of cortical bone was obtained using the relation
f=1 − dr where dr is the relative density defined as the ratio of the
apparent density da by the true density dt of the bone. The average
porosity value of the diaphyseal cortical bone of the humerus was found
Fig. 5. Load as function of the indentation depth.
to be equal approximately to 0.09.

conducted with pyramidal indenters, a quadratic dependence between

2.2. Micromechanical characterization of the elastic properties load and displacement can be found. The relation between load F and
indentation depth h is given by:
The heterogeneous and hierarchical nature of the diaphyseal cor- 2
tical bone of the humerus makes it unlikely for a single measurement to F= Er h2
π (1)
fully characterize its structural integrity or accurately predict fracture
risk. Hence, this study aims to understand the contributions of archi- Er is the reduced modulus which relied to the bone elastic properties
tecture and material properties on the elastic behavior by means of (E and ν fixed to 0.3) and the indenter elastic properties (Ei=1141 GPa
nanoindentation experiments. The well-known principle of na- and νi=0.07 for diamond material) by the relation:
noindentation lies in bringing a very small tip to the surface of the bone 1 1 − ν2 1 − νi2
to make an imprint. Elastic modulus and hardness of the material are = +
Er E Ei (2)
extracted from experimental readings of indenter load and depth of
penetration. The load versus indentation depth curve were fitted in the Fig. 5
Three small specimens were dissected from the humerus diaphysis using the quadratic rheological model performed by Mazeran et al.
and were milled down to a prismatic shape of 8 × 5 × 2 mm. The [22,23] for the determination of bone mechanical properties.
sampling area is located in the humerus distal part and is considered The results of this experimental campaign have provide mean values
quite far from the rupture zone and has no apparent damage (Fig. 4). of the cortical bone elastic modulus of 19.50 ± 0.28 GPa in the
Two of samples are intended for measures in the longitudinal direction longitudinal direction and of 18.6 ± 2.8 GPa in the transverse direc-
of the humerus. The third one was intended for measures in the radial tion [41]. Moreover, the third specimen was tested on two localizations
direction. These specimens were mechanically polished by an auto- (the osteons and the interstitial tissues). The results showed mean va-
matic grinding machine having off-centered rotating movements in lues of the cortical bone elastic modulus of 18.2 ± 2.4 GPa for in the
order to achieve smooth and flat surface with substantially smaller osteons tissue and of 21.1 ± 2.98 GPa in the interstitial tissue.
roughness compared to indentation depths. The average of the micromechanical elastic modulus denoted by Em
At first an abrasive polishing is applied with hydration using ultra- for infinite strain rate derived from nanoindentation tests on the spe-
pure water and discs of different grain sizes 800, 1200, 2400 and 4000. cimens of the humerus diaphyseal cortical bone has been found to be
Every polishing stage was done is 120 s with a rotary speed of 100 rpm. equal to 19.2 ± 3.1 GPa which fits within the average values available
A final polishing with a felt disk coupled to a suspension of alumina in the literature (see [41]).
particles of size 0.02 μm was then performed. This last polishing stage It has to be noticed that the micromechanical Poisson ratio denoted
was applied with a fixed rotary speed of 100 rpm during 360 s. At the by νm was not studied in the present investigation, its value was fixed to
end an ultrasound rinsing bath was applied using ultra-pure water 0.3 as commonly proposed for long bones in the literature (see [3,29]).
during 5 mn.
After this preparatory step, nanoindentation tests were performed 3. Theoretical formulation of the micromechanics based model of
with a nanoindenter G200 from AGILENT using a Berkovich tip and the the diaphyseal humerus bone
method proposed by Mazeran et al. [22]. This method is based on four
stages protocol (loading, first 300 s hold plateau, partial unloading and In this section, the micromechanical modeling of the elastic beha-
second 300 s hold plateau) in order to exhibit the time-dependent re- vior of the diaphyseal cortical bone of the humain humerus will be
versible and irreversible behavior. For each specimen, 40 indents were presented.
carried out in order to reduce the error associated with the measure- The diaphyseal cortical bone of the human humerus, in the present
ment of the standard deviation. As nanonidentation experiments are investigation, was considered as a porous media with a matrix phase

Fig. 4. Bone specimens holder for polishing stage

and nanoindentation experiments.

4
J. Rahmoun, et al. Materials Science & Engineering C 117 (2020) 111276

Fig. 6. Description of the representative elementary volume.

weakened by micro-pores and exhibits transversely isotropic symmetry. 9κ mμm (1 − f )

En =
The representative elementary volume REV of the bone is composed of 3κ m + μm (7)
an isotropic linear elastic solid matrix with stiffness ℂm and of a set of
Moreover, if we consider that the aspect ratio is α=1 (the case of
cylindrical shaped pores having an elasticity tensor ℂp and volume
circular cylindrical voids), the transversal Young's modulus reads as:
fraction f. The pores were taken to be identical in shape and orientation
which characterized by a normal unit vector denoted by n (see Fig. 6). 18κ mμm (1 − f )(3κ m + μm )
Ep =
Taking advantage of the solution of the Eshelby equivalent inclusion 9fκ m (3κ m + 4μm ) + 2(3κ m + μm )2 (8)
problem [9,18], the Mori-Tanaka homogenization scheme [24] pro-
Making the same assumption of circular cylindrical voids, the
vides the following expression of the homogenized elastic stiffness
components of the homogenized elastic stiffness tensor ℂhom have been
tensor for the porous media (see [27,30,31] for more detail):
computed by means of the Eq. (3) and reported in Appendix B.
hom Finally, the obtained micromechanical model is implemented in a
VMAT within the explicit dynamics code ABAQUS© for the prediction
= m + f [(p − m)−1 + ]−1 : [(1 − f )  + f [ + : (p − m)]−1 ]−1
of the mechanical response of the humerus bone under low velocity
(3) impact. This model needs three materials parameters: the matrix elastic
1 modulus Em obtained by the nanoindentation tests, the diaphyseal
where , Iijkl = + δil δkj ) is the fourth-order symmetric identity
(δ δ
2 ik jl
tensor and δij is the Kronecker delta symbol. ℙ is the Hill tensor [13] cortical bone porosity f deduced from the densitometric measurements
which depends on the shape and orientation of the pores (considered of the humerus bones and the matrix Poisson ratio νm given by the
here as cylindrical voids of elastic stiffness tensor ℂp=0) and on the literature.In order to limit the number of unknown material parameters
elastic stiffness of the solid matrix ℂm. in this investigation, the aspect ratio is taken α=1, which corresponds
We model the void as an infinite cylinder aligned along the e3 = n to the assumption of cylindrical voids having a circular cross section.
(rectangular cartesian coordinate) axis, and of elliptical cross-section of
b 4. Multi-scale Finite Element model of a human humerus bone
aspect ratio α = a geometrically prescribed by:
under impact
x12 x2
+ 22 ≤ 1, − ∞ < x3 < + ∞
a2 b (4) 4.1. 3D mesh generation of the humerus

The determination of the Hill tensor ℙ, which is based on the work The selected human humerus to be analyzed was the Humerus 517
of [13] (see also [18,37]), consists in evaluating the integral: which did not fail under the impact of an energy of 10 J. Therefore, it is
ξi [K m ( ξ )]−jp1 ξq |(ij)(pq) assumed that this humerus exhibited an elastic behavior without da-
α
Pijpq =
2π
∫| ξ|=1 ξ12 + α 2ξ22
ds ( ξ ) mage nor failure and therefore it could be modeled using the present
(5) micromechanical elastic model.
The first step of the elaboration of the Finite Element (FE) model
for which the integration is on the unit circle centered at the origin of
was the generation of the 3D CAD geometry of the humerus diaphysis
space (ξ1, ξ2), defined by | ξ | = 1. K ipm ( ξ ) = Cijpq
m
ξ j ξq is called the acoustic
by starting from the digitalized STL standard format. As shown in
tensor. The notation ∣(ij)(pq) indicates the symmetrization with respect to
Fig. 5a, the digitalized STL mesh was split into two different parts
the couples (i,j) and (p,q).
which represent the inner and outer surfaces of the humerus. Each STL
The analytical evaluation of the tensor ℙ is not an easy task. For
surface was, partitioned along the transverse direction into 10 small
transversely isotropic materials, it is interesting for standard notations
parts (Fig. 7a) using CATIA© CAD software. Then, small local surfaces
and the associated tensorial representations to introduce the Walpole's
were built up using their edges generated previously by CATIA© CAD
base [38] which is constituted of a set of direction-related fourth-order
facilities. Once the total inner and outer surfaces of the humerus dia-
tensors i , i = 1, …6. By using this algebra and for cylindrical pores
physis were reconstituted, a first mapped FE shell mesh model was
oriented with respect to the e3 = n axis, Walpole's components of the
generated using only quads. Then, a 3D mesh of the humerus volume
Hill tensor ℙ are given by [18,38]:
was extruded using 7920 hexahedrons, thanks to the 3D meshing fa-
 = (2κp , n, 2μp , 2μn , l, l) (6) cilities available in HYPERMESH© software [1] (Fig. 7b).

where κp, l, n, μp and μn the plane-strain bulk modulus, cross modulus, 4.2. Modeling of the drop tower impact on the humerus
axial modulus under uni-axial strain, axial-shear modulus, and trans-
verse-shear modulus, respectively. These components may be in- The final stage of this investigation was the elaboration of the FE
tegrated analytically and are given in Appendix A. model including boundary conditions, material parameters, contact and
Thus, for cylindrical voids of elliptical cross-section in an isotropic loading conditions, in order to simulate the impact bending of the hu-
medium, we obtain the following expression of the longitudinal Young's merus and try to validate the material model. To this purpose, two
modulus which it is independent of the aspect ratio α. brackets with concentrated mass elements of 1080 g have been modeled

5
J. Rahmoun, et al. Materials Science & Engineering C 117 (2020) 111276

Fig. 7. Humerus Finite Element model.

Fig. 8. Drop tower modeling using Finite Element method.

using hexahedral elements with the assumption of a rigid elastic ma- fixed (see Fig. 9(d)). For the next of study, the material parameters used
terial. The center of mass of each rigid bracket was allowed to move in FE simulations of the humerus 517 are Em=19.2 MPa, νm=0.3 and
along the X-axis and to rotate around the Y-axis as shown in Fig. 8. The f=0.09.
humerus epiphyses were embedded into the brackets using a prismatic
shaped solid of 71 × 71 × 55 mm and having the resin material 5. Discussion
parameters (ρ=1500 kg/m3, E=400 MPa, ν=0.3). For the sake of
biofidelity, the inner part of the humerus diaphysis was filled by a solid This study was performed for the mechanical characterization of the
representing the marrow with elastic properties E=1 MPa, ν=0.3 and elastic behavior of the human humerus under impact using a newly
ρ=1000 kg/m3 (see [14]). The system brackets-humerus was posi- developed multi-scale model. The originality of the proposed approach
tioned on a plate of 400 × 150 × 10 mm size and of 10, 77 kg mass consists in the identification of only few material parameters to de-
which has been modeled as an analytical rigid body. scribe the whole complex anisotropic behavior of the humerus bone.
The impactor was considered as a cylindrical shell of 20 mm dia- To investigate the effect of the impact velocity on the elastic re-
meter and 8.6 kg of mass. It has been modeled using S4R shells of sponse of the humerus under bending impact, three different velocities
ABAQUS© elements and a rigid elastic material (see Fig. 8). All contact (1.5, 1.77 and 1.9 m/s) were applied. As we can show in Fig. 10(a) and
interfaces were modeled using “General Contact” conditions without (b), the impact velocity affected clearly the elastic properties of the
friction. The FE simulation was carried out using a total duration of humerus. An increasing of the impact velocity leads naturally to an
10 ms to cover the entire impact process. The 3D mesh of the whole FE increasing of bending stiffness and strain.
system includes 7920 C3D8R solid elements and analytical solid. The Fig. 11 shows the evolution of humerus response under the impact
FE simulation was carried out using the Explicit Dynamics algorithm of load. One can observe globally an overall good agreement in the esti-
ABAQUS© software. mation of ultimate load in comparing the predicted numerical solution
to the various experimental responses. From Table 3, one can notice
4.3. Analysis of the material parameters influence that the micromechanical model gives an ultimate impact load of
1201 N with a minimum error of 7% [33].
In order to analyze the influence of materials parameters on the FE The comparison between experimental measurements and the nu-
simulation of the humerus elastic response under bending impact, dif- merical predictions of the impact load shows that the predicted ulti-
ferent set of parameters are considered. Fig. 9 represents the load-dis- mate load is somehow under-estimated, this is probably due to the local
placement curves of the modeled humerus 517 under three values of bone material identification. Indeed, it has been reported [20] that
porosity, matrix modulus, longitudinal and transverse elastic moduli. nano-indentation experiments often induce damage into the bone prior
As can bee seen in Fig. 9(a), the increased cortical porosity f rendered to the physical measure, therefore the resulting elastic characteristics
lower structural stiffness as well as lowered peak force along with an are often underestimated even if the embalming solution has a small
increased ability to absorb energy before failure. On the other hand, we effect on Young's modulus and other properties measured in static tests.
have studied in Fig. 9(b) the variation of the matrix elastic modulus Em However, some authors [10,40] found that impact energy absorption
around its mean value determined by nanoindentation experiments. We seems to be reduced by quite a large amount, and that results of high
notice that an increase in matrix elastic modulus is correlated with an strain rate tests may be more affected than the results of static tests.
increase of bone elastic properties namely the bending stiffness. The Nevertheless the remaining minor deviations could be simply attributed
same phenomenon occurs for the longitudinal elastic modulus En at to the natural differences in geometry (size and morphology) between
fixed transverse one Ep (Fig. 9(c)). However, there was no detectable all tested humeral bones, which is common for most of biological ma-
effect of the variation of the transverse elastic modulus on the elastic terials [2].
response of the humerus when the longitudinal elastic modulus was However, despite the aforementioned experimental uncertainties,

6
J. Rahmoun, et al. Materials Science & Engineering C 117 (2020) 111276

Fig. 9. Influence of humerus material parameters on the load-displacement curves: (a) porosity, (b) matrix modulus, (c) longitudinal modulus, (d) transverse
modulus.

one can remark that the obtained ultimate impact load shows a rela- Finite Element model in capturing the true humerus bone response for
tively limited error (7%) compared to the mean experimental measured the three-point bending tests.
load. The same remark can be made regarding the instant necessary to These results seem promising for future applications such as eva-
reach the ultimate force (1.35 ms), which is very close to the mean luation of injury risk in lateral car crashes [19]. As the proposed model
experimental one (1.40 ms) (Table 4). can utilize detailed numerical representations of the human humerus
Moreover, the strain response of strain-gauge-acquired values at the with accurate geometric and material properties, a perspective is to
mid-diaphysis of human humeri was investigated. At this location, the realize a geometric personalization of the model. This would allow the
maximum strain values ranged from 0.25% to 0.97% with an average assessment of the influence of a model personalization. In addition, this
experimental value of 0.66 ± 0.17. The error of the mean measured humerus model can be integrated in a whole-body model for applica-
maximum strain from the FE-calculated maximum strain was found to tions such as shoulder lateral impacts. It can be subjected to a given
be equal to 3%. This minimum error assesses the performance of the loading condition, producing response characteristics intended to

Fig. 10. Effect of the impact velocity: (a) load versus displacement curve, (b) strain versus time curve.

7
J. Rahmoun, et al. Materials Science & Engineering C 117 (2020) 111276

mortem human subject testing. These models remove the issue of re-
peatability, allowing for the implementation of parametric studies often
used by vehicle designers to improve automotive safety.

6. Conclusions

In conclusion, the framework of homogenization was proposed for

the modeling of elastic behavior for diaphyseal cortical bone. The
elastic properties of the human humerus were investigated by coupling
the Mori-Tanaka scheme and the micromechanical measurements of the
cortical bone properties using nanoindentation experiments.
The proposed model predicts adequately the experimental results
via the estimation of the ultimate impact load that the humerus may
encounter before fracture. Further improvements will be performed to
the present model to be able to simulate the bone fracture by taking into
account the nonlinear behavior and including the strain rate effects and
effective fracture criteria.

Fig. 11. Mechanical response of the humerus under impact loading (experiment CRediT authorship contribution statement
vs numerical).
J. Rahmoun: Conceptualization, Investigation, Methodology,
Writing - original draft. H. Naceur: Conceptualization, Investigation,
Table 4
Comparison of the ultimate impact load and maximum strain of the modeled Methodology, Software, Writing - review & editing. H. Morvan:
humerus (Numerical vs. Experiment). Data curation. P. Drazetic: Supervision. C. Fontaine: Supervision,
Validation. P.E. Mazeran: Data curation.
Load Time Maximum strain

(N) (ms) (%) Declaration of competing interest

FE model 1201 1.35 0.64
Experiment (mean) 1299 1.40 0.66 The authors declare that they have no known competing financial
Error model 7% 3% interests or personal relationships that could have appeared to influ-
ence the work reported in this paper.
mimic the human body in a real world scenario. The primary advantage
of human body models, however, is that they provide insight into the Acknowledgments
onset of injury and resultant local response characteristics thereafter
through the use of stress and strain data measured throughout the en- The authors gratefully acknowledge the financial support from the
tire model. Such information is unobtainable during volunteer or post French Ministry of Research.

Appendix A. Hill tensor for elliptic cylindrical inclusions

The Hill tensor for cylindrical inclusions of elliptical cross-section with the aspect ratio α embedded in an isotropic solid matrix, can be written,
by using Eq. (6), in the form:
α 3 α 3κ m + μm (3α + 4) 1 1
2kp = m m
, n = 0, 2μp = , 2μn = , l=0
1 + α (3κ + 4μ ) 2
(1 + α ) μm (3κ m + 4μm ) 1 + α 2μm (A.1)
where κ and μ are the bulk and shear moduli of the bone matrix.
m m

Appendix B. Elastic stiffness tensor for circular cylindrical inclusions

The Walpole's components of the homogenized stiffness tensor of the humeral bone are:

(
hom = 2κphom , nhom, 2μphom , 2μnhom , l hom, l hom )
where:

C1111 = κphom + μphom , C1122 = κphom − μphom , C2323 = μnhom , C3333 = nhom , C1133 = l hom

The non-zero components of the homogenized elastic stiffness tensor for cylindrical inclusions of circular cross-section (the case of aspect ratio
α=1) and of volume fraction f in an isotropic medium can be calculated analytically, by using Eqs. (3) and (A.1) and are written as:

8
J. Rahmoun, et al. Materials Science & Engineering C 117 (2020) 111276

⎧C μm (1 − f )(3κ m + μm )(2f + 1)(3κ m + 4μm )

= C2222 =
⎪ 1111 (3μm + fμm + 3fκ m)(3κ m + μm + 3fκ m + 7fμm )
⎪
⎪C μ (1 − f )(3κ m + μm )(3κ m − 2μm + 6fμm )
m
=
⎪ 1122 (3μm + fμm + 3fκ m)(3κ m + μm + 3fκ m + 7fμm )
⎪
⎪ μm (3κ m − 3fκ m − 2μm + 2fμm )
⎪C1133 = C2233 = 3μm + fμm + 3fκ m
⎨ μm (−3κ m − 6fκ m − 4μm + 4fμm + 9f 2 κ m)
⎪C3333 = −
⎪ 3μm + fμm + 3fκ m
⎪ 2μm (1 − f )
⎪C2323 = C1313 =
⎪ 1+f
⎪ μm (−3κ m − 6fκ m − 4μm + 4fμm + 9f 2 κ m)
⎪C1212 = −
⎩ 3κ m + μm + 3fκ m + 7fμm (B.1)

References technique combined with meshless methods, Advances in Biomechanics and Tissue
Regeneration (2019) 201–213.
[22] P.-E. Mazeran, M. Beyaoui, M. Bigerelle, M. Guigon, Determination of mechanical
[1] Altair Engineering, HyperMesh© 10.0 User’s Guide, Altair Engineering Inc., World properties by nanoindentation in the case of viscoelastic materials, International
Headquarters 1820 E. Big Beaver Rd. Troy MI 48083-2031 USA, 2010. Journal of Material Research 103 (2012) 715–722.
[2] M.S.R. Aziz, B. Nicayenzi, M.C. Crookshank, H. Bougherara, E.H. Schemitsch, [23] S. Jaramillo Isaza, P.-E. Mazeran, K. El Kirat, M.-C. Ho Ba Tho, Time-dependent
R. Zdero, Biomechanical measurements of stiffness and strength for five types of mechanical properties of rat femoral cortical bone by nanoindentation: an age re-
whole human and artificial humeri, J. Biomech. Eng. 136 (5) (2014) 1–10. lated study, Journal of Material Research 29 (2014) 1135–1143.
[3] V. Baca, Z. Horak, P. Mikulenka, V. Dzupa, Comparison of an inhomogeneous or- [24] T. Mori, K. Tanaka, Averages stress in matrix and average elastic energy of materials
thotropic and isotropic material models used for FE analyses, Med. Eng. Phys. 30 with misfitting inclusions, Acta Metallia 21 (1973) 571–574.
(2008) 924–930. [25] S. Mukherjee, A. Chawla, S. Borouah, D. Sahoo, M.W.J. Arun, Dynamic Properties of
[4] C. Boosa, K. Fink, P. Stomberg, W. Köller, B.W. Igl, M. Russlies, The intraindividual the Shoulder Complex Bones, Society of Automotive Engineers, 2011, pp. 11–17.
agreement of the bone density of the human proximal tibia, Ann. Anat. 189 (2007) [26] H. Naceur, S. Shiri, D. Coutellier, J.L. Batoz, On the modeling and design of com-
515–519. posite multilayered structures using solid-shell finite element model, Finite Elem.
[5] E.F. Carrera, F.A. Nicolao, N. Archetti Netto, R.L. Carvalho, F.B. dos Reis, Anal. Des. 70 (2013) 1–14.
E.J. Giordani, A mechanical comparison between conventional and modified an- [27] H. Naceur, J. Rahmoun, J. Halgrin, P. Chabrand, Multiscale finite element modeling
gular plates for proximal humeral fractures, Journal of Shoulder and Elbow. of ductile damage behavior of the human femur under dynamic loading,
Surgery Board of Trustees (2008) 631–636. International Journal of Damage Mechanics 24 (2015) 418–445.
[6] P. Clavert, R.-M. Javier, J.L. Charrissoux, L. Obert, L. Pidhorz, F. Sirveaux, [28] C. Ohman, E. Dall’Ara, M. Baleani, S. Van Sint Jan, M. Viceconti, The effects of
P. Mansat, T. Fabre, How to determine the bone mineral density of the distal hu- embalming using a 4% formalin solution on the compressive mechanical properties
merus with radiographic tools? Surg. Radiol. Anat. 38 (2016) 389–393. of human cortical bone, Clin. Biomech. 23 (2008) 1294–1298.
[7] S. Duprey, K. Bruyere, J.P. Verriest, Experimental and simulated flexion tests of [29] L. Peng, J. Bai, X. Zeng, Y. Zhou, Comparison of isotropic and orthotropic material
humerus, International Journal of Crashworthiness 12 (2) (2007) 153–158. property assignments on femoral finite element models under two loading condi-
[8] X. Elie-Dit-Cosaque, A. Gakwaya, H. Naceur, Smoothed finite element method im- tions, Med. Eng. Phys. 28 (3) (2006) 227–233.
plemented in a resultant eight-node solid-shell element for geometrical linear [30] J. Rahmoun, F. Chaari, E. Markiewicz, P. Drazetic, Micromechanical modeling of
analysis, Comput. Mech. (2014) 1–22. the anisotropy of elastic biological composites, Multiscale Modeling and Simulation
[9] J.D. Eshelby, The determination of the elastic field of an ellipsoidal inclusion and 8 (1) (2009) 326–336.
related problems, Proceedings of the Royal Society Series A 252 (1957) 561–569. [31] J. Rahmoun, A. Auperrin, R. Delille, H. Naceur, P. Drazetic, Characterization and
[10] R. Gauthier, H. Follet, M. Langer, S. Meille, J. Chevalier, F. Rongiéras, F. Peyrin, micromechanical modeling of the human cranial bone elastic properties, Mech. Res.
D. Mitton, Strain rate influence on human cortical bone toughness: a comparative Commun. 60 (2014) 7–14.
study of four paired anatomical sites, J. Mech. Behav. Biomed. Mater. 71 (2017) [32] B. Santos, C. Quental, J. Folgado, M. Sarmento, J. Monteiro, Bone remodelling of
223–230. the humerus after a resurfacing and a stemless shoulder arthroplasty, Clin.
[11] E.H. Van Haaren, B.C. Van der Zwaard, A.J. Van der Veen, I.C. Heyligers, Biomech. 59 (2018) 78–84.
P.J. Wuisman, T.H. Smit, Effect of long-term preservation on the mechanical [33] D. Singh, A. Rana, S.K. Jhajhria, B. Garg, P.M. Pandey, D. Kalyanasundaram,
properties of cortical bone in goats, Acta Orthop. 79 (5) (2008) 708–716. Experimental assessment of biomechanical properties in human male elbow bone
[12] C. Hellmich, J. Barthelemy, L. Dormieux, Mineral-collagen interactions in elasticity subjected to bending and compression loads, Journal of Applied Biomaterials and
of bone ultrastructure-a continuum micromechanics approach, Eur. J. Mech. A Functional Materials (2019) 1–13.
Solids 23 (2004) 783–810. [34] U. Stefan, B. Michael, S. Werne, Effects of three different preservation methods on
[13] R. Hill, A self-consistent mechanics of composite materials, Journal of the the mechanical properties of human and bovine cortical bone, Bone 47 (2010)
Mechanics and Physics of Solids 13 (1965) 213–222. 1048–1053.
[14] K.-Y. Ho, J.H. Keya, C.M. Powers, Comparison of patella bone strain between fe- [35] T. Toop, T. Muller, S. Huss, P.H. Kann, E. Weihe, S. Ruchloltz, R. Peter, Embalmed
males with and without patellofemoral pain: a finite element analysis study, J. and fresh frozen human bones in orthopedic cadaveric studies: which bone is au-
Biomech. 47 (1) (2014) 230–236. thentic and feasible? Acta Orthop. 39 (5) (2012) 543–547.
[15] C.E. Hoffler, K.E. Moore, K. Kozloff, P.K. Zysset, M.B. Brown, S.A. Goldstein, [36] F. Vandenbulcke, Caractérisation et modélisation multi-échelle du comportement
Heterogeneity of bone lamellar-level elastic moduli, Bone 26 (6) (2000) 603–609. mécanique à la rupture du membre scapulaire sous sollicitations dynamiques, PhD
[16] S. Hungerer, F. Wipf, G. von Oldenburg, P. Augat, R. Penzkofer, Complex distal Thesis University of Valenciennes, 2015.
humerus fractures—comparison of polyaxial locking and nonlocking screw con- [37] J.R. Willis, Bounds and self-consistent estimates for the overall moduli of aniso-
figurations—a preliminary biomechanical study, J. Orthop. Trauma 28 (3) (2014) tropic composites, Journal of the Mechanics and Physics of Solids 25 (1977)
130–136. 185–202.
[17] R. Kaufman, L. Fraade-Blanar, A. Lipira, J. Friedrich, E. Bulger, Severe soft tissue [38] L.J. Walpole, On the overall elastic moduli of composite materials, Journal of the
injuries of the upper extremity in motor vehicle crashes involving partial ejection: Mechanics and Physics of Solids 12 (1969) 235–251.
the protective role of side curtain airbags, Accid. Anal. Prev. 102 (2017) 144–152. [39] H. Yamada, F.G. Evans (Ed.), Strength of Biological Materials, Williams and
[18] N. Laws, R. McLaughlin, The effect of fiber length on the overall moduli of com- Wilkins, Baltimore, MD, 1970.
posite materials, Journal of the Mechanics and Physics of Solids 27 (1979) 1–13. [40] X. Zhai, J. Gao, Y. Nie, Z. Guo, N. Kedir, B. Claus, T. Sun, K. Fezzaa, X. Xiao,
[19] M. Lebarbé, P. Vezin, F. Rongiéras, D. Lafont, Human shoulder response to high W.W. Chen, Real-time visualization of dynamic fractures in porcine bones and the
velocity lateral impact, Stapp Car Crash Journal 61 (2017) 27–51. loading-rate effect on their fracture toughness, Journal of the Mechanics and
[20] A. Malandrino, A. Fritsch, O. Lahayne, K. Kropik, H. Redl, J. Noailly, D. Lacroix, Physics of Solids 131 (2019) 358–371.
C. Hellmich, Anisotropic tissue elasticity in human lumbar vertebra, by means of a [41] P.K. Zysset, X. Edward Guo, C. Edward Hoffler, K.E. Moore, S.A. Goldstein, Elastic
coupled ultrasound-micromechanics approach, Mater. Lett. 78 (2012) 154–158. modulus and hardness of cortical and trabecular bone lamellae measured by na-
[21] M. Marques, J. Belinha, A.F. Oliveira, R.M. Natal Jorge, Chapter 10: determination noindentation in the human femur, J. Biomech. 32 (10) (1999) 1005–1012.
of the anisotropic mechanical properties of bone tissue using a homogenization