Documente Academic
Documente Profesional
Documente Cultură
NUMERICAL MODELING OF
HUMAN FOOT
29
2.1 Finite element model of foot-ankle complex
The FEM is a versatile numerical method which allows stress and strain analyses
foot structure is loaded, stresses are generated in different materials (i.e. tissues).
muscles, ligaments, fascia, and the external environment that arises from foot-
ground interactions. In such a model that mimics the real structures to a certain
In using the FEM, the model of the human foot as a geometrical entity,
source for such complete anatomical structures. The geometrically complex foot
structures, are then discretized into finite number of relatively simple elements
can have its own material properties. The equations for describing the
mechanical behavior of these elements are known. The computer program (e.g.
ABAQUS) can calculate the stiffness matrix of each element, and the stiffness
matrix, of the loads and the boundary conditions allows the response of a model
30
2.1.2 3-D reconstruction of foot geometry
right foot (non-weight-bearing condition) of a male subject (27 years old, height of
169 cm and body weight of 65.1 kg). The CT scan has a resolution of 0.25x0.25
mm2 (pixel size) with a slice spacing of 1 mm. This allows a detailed
including bones, cartilages, and a bulk soft tissue boundary (Fig. 2.1.2.1.). Model
Fig. 2.1.2.1. Segmentation of a human foot from individual coronal CT slices. The bones
were modeled as articulated parts enveloped into a bulk soft tissue
the foot skeletal phase and a lumped soft-tissue phase. The hard foot skeleton
and soft tissue were then reconstructed into a solid model (Fig. 2.1.2.2.). The
31
solid model was finally imported into a pre-processor, PATRAN (MacNeal
surface geometry of various foot bones (tibia, fibula, talus, calcaneus, navicular,
medial, intermediate and lateral cuneiform, and cuboid, the medial, intermediate
and lateral cuneiforms, and the cuboid, metatarsal bones and phalanges) that
Fig. 2.1.2.2. 3-D solid model of foot geometry, including a bulk soft tissue (A) and bones
(B). Note that the foot skeletal was stabilized by actually anatomical ligaments.
plantar fascia, which could not be reconstructed from CT were determined based
Pictures Ltd., London, UK). A total of 134 ligaments (i.e., multiple element
modeling for major ligament bundles) and a fan-shaped plantar fascia structure
32
2.1.3 Model discretization
The reconstructed solid foot model was then used to generate three-dimensional
finite element mesh using various types of structural elements; this process is
known as model discretization. The foot skeleton and the soft-tissue component
integration linear element). Meshes for the bones and the soft-tissue component
share same nodes at the interface except those joint space regions where a
series of contact conditions were defined between adjacent bones (Fig. 2.1.3.1.).
The ligament structures were represented by 3-D truss elements (T3D2) with a
allow these elements to resist tension-producing forces when stabilizing the foot
skeleton.
Fig. 2.1.3.1. Finite element mesh of a human foot with (A) soft tissue and (B) internal
bony structures
33
For the muscular structures, six major extrinsic plantar flexors were
into the posterior extreme of the calcaneus. This facilitates application of G-S
muscle forces through the Achilles tendon-bone junction, and ensure more
realistic muscle load transfer compared to those in previous models, which only
employed nodal points to apply such forces (Gefen et al., 2000, Cheung et al.,
2006). The long tendons of the other five muscles were also inserted into the
model, at their corresponding anatomical attachment sites. This was done using
stringed together to represent the actual tendon trajectory inside the foot.
density used in the FE model was sufficient to reach the converged numerical
results. Mesh refinement process was carried out in a 2-D plane-strain finite
element model based on a sagittal section through the 2nd ray of the foot. The
total strain energy and displacement served as the convergence criteria, with the
testing has revealed that material property of cortical bone may be considered as
34
having ‘nearly’ linearly elastic behavior. Measuring mechanical properties of
cancellous bone tissue is far more difficult than measuring those of cortical bone
this difficulty, the reported cancellous bone tissue modulus ranges from 0.76 to
homogeneous, isotropic, and linear elastic. For the foot bones, Young’s modulus
is taken as 7300 N/mm2, a value that was weighted by Nakamura et al. (1981)
from human cortical and cancellous bone properties and the Poisson ratio is
taken as 0.3. The ligaments, following Cheung et al. (2005), were considered as
In the literature, mechanical properties have been reported for many tendons (e.g.
Achilles tendon and anterior cruciate ligament). Modulus values are generally in
the range of 500~1,850 MPa (Yamamoto et al., 1992, Danto and Woo, 1993).
For the Achilles tendon, uni-axial tension tests have been conducted by Wren et
al. (2001) at lower strain rates of 1%/s and relatively higher strain rate of 10%/s.
The mean moduli found were 816 MPa (± 218) at 1%/s rate and 822 MPa (±211)
35
at 10%/s rate, respectively. Considering the quasi-static loading system used in
the current study, as will be discussed in the following sections, the elastic
modus obtained at lower strain rate is chosen and a common Poisson’s ratio of
0.3 was used as the material properties for the Achilles tendon in the current foot
FE model.
For the flexor tendons, the literature contains the least amount of
stiffness values of flexor hallucis longus (FHL) and peroneus brevis (PB) were
43.3 N/mm (± 14.1) and 43.6 N/mm (± 18.9) (Maffulli et al., 2008), respectively.
Using the relation, a flexor tendon of the foot with a typical cross-sectional area
of 12.5 mm2 was calculated to have a Young’s modulus of 450 MPa (Garcia-
Gonzalez et al., 2009). Thus, the Achilles and other flexor tendons, were
idealized as isotropic linear elastic materials with different Young’s moduli of 816
MPa (Wren et al., 2001) and 450 MPa (Garcia-Gonzalez et al., 2009), and a
36
Table 2.2.2.1. Summary of FE model listing element type and material properties for
different model entities.
Element Cross-
Entity type E (MPa), v section Reference
(ABAQUS) (mm2)
4-node Cortical and cancellous
Bone tetrahedral 7300, 0.3 bone properties weighted
continuum by Nakamura et al. (1981)
4-node Indentation test conducted
Cartilage tetrahedral 1.01, 0.4 on 1st MTP joint by
continuum Athanasiou et al. (1998)
Properties obtained from
2-node
collateral ligaments of the
Ligament tension-only 260, 0.4 18.4
human ankle joint by
truss
Siegler et al. (1988)
4-node Tensile modulus obtained
Achilles
tetrahedral 816, 0.3 at lower strain rate (Wren et
tendon
continuum al., 2001)
Calculation based on the
Other 2-node stiffness values of flexor
flexor tension-only 450, 0.3 12.5 hallucis longus and
tendons cable peroneus brevis (Garcia-
Gonzalez et al., 2009)
2-node
Plantar Wright and Rennels,
tension-only 350, 0.4 290.7
fascia (1964))
truss
4-node Stress-strain curve
Plantar
tetrahedral Hyperelastic determined experimentally
soft-tissue
continuum by Chen et al. (2011)
37
2.2.3 Constitutive model for plantar soft tissue
The plantar soft-tissue is a typical load-bearing soft tissue that undergoes very
strain behavior when subject to high loads during gait. Furthermore, the plantar
soft-tissue in normal foot is often rich in fluid content, and its material behavior
be derived from a ‘strain-energy density function’. This function defines the strain
energy stored in the material per unit of reference volume (volume in the initial
configuration) as a function of the strain at that point in the material. For this
materials, as is the case for the plantar soft tissue. And various hyperelastic
material models have been successfully used to model plantar soft tissue
behavior in finite element simulation (Lemmon et al., 1997, Cheung et al., 2005,
form, the Marlow form, the Mooney-Rivlin form, the neo-Hookean form, the
Ogden form, the polynomial form, the reduced polynomial form, the Yeoh form,
and the Van der Waals form (Hibbert and Karssonn, 2006). This study focuses
on the Ogden-form hyperelastic constitutive model that are widely used in the
38
literature to model the plantar soft tissue material behavior, because of its
U U 1 , 2 , 3 2.1
U
i i p , i = 1, 2, 3 2.2
i
where p is the hydrostatic part of stress, whereas the first term is relative to the
deviatoric part. For incompressible materials (i.e. total volume remains constant
incompressibility:
1 , 2 3 1 / 2 2.4
U U 1 , 2 , 3 U , 1 / 2 2.5
39
U
2.6
A very common form of strain energy function is that proposed by Ogden (1972):
i 1 i
N
U 1 , 2 , 3 2 i 3 i / i 2.7
i 1
parameters (i.e. material constants) that describe the behavior of this rubber-like
material model. These material constants, and αi, can be positive or negative,
1 N
i i k
2 i 1
2.8
where is the ground state (i.e. initial) shear modulus. Introducing Eq. (2.7) into
hyperelastic Ogden formulation with stress () and stretch () relation given as:
2
1 ( / 2) 1 2.9
For 1st-order Ogden model, we only have two unknown material constants,
40
In the literature, the stress-strain curves of the plantar soft tissue are often
obtained based on tests of the fat pad under the heel bone (i.e. heel pad)
(Erdemir et al., 2005, Cheung et al., 2005, Lemmon et al., 1997). Since the
primary focus of the current model is the sub-MTH region, the material behavior
has to be determined for soft tissues specific to that location. For this purpose, an
the forefoot plantar soft tissue. The indentation force-displacement curve was
directly measured from soft tissue under the metatarsal heads, and was used for
experiments on human plantar soft tissue under the 2nd MTH, based on this
customized tissue tester. The parameters were chosen such that the reaction
forces upon indentation optimally fit the experimental observations; this yielded
= 3.75 x10-2 MPa and α = 5.5. (See chapter 3, section 3.1.4.3, for detailed
metatarsal heads (MTHs) (Cavanagh et al., 2000, Lott et al., 2007, Bus et al.,
41
2008). Those insoles are often fabricated using solid foams that are made up of
polyhedral cells that pack in three dimensions. The foam cells can be either open
(e.g. plastazote) or closed (e.g. Ethylene-vinyl acetate (EVA) foam). Soft insoles
made from such foam materials have excellent energy absorption properties: the
stiff elastic materials for a certain stress level (Nigg et al., 1988). Typical
stressstrain (σε) curve of four types of foam padding material that are
commonly used for fabrication of insoles in therapeutic footwear are shown in Fig.
2.2.4.1. Values were calculated from the force (F) and displacement (D) data
4F D
, 2.10
s 2 L
Fig. 2.2.4.1. Stress-strain curve obtained from uni-axial compression test of deformable
foam padding materials
42
The samples prepared were 30 mm in diameter (s) and 13.6 mm thick (L).
rate of 4 mm/s. From the foam stress-strain curve, three stages can be
1. At small strains (5%) the foam deforms in a linear elastic manner due to cell
wall bending.
by the elastic buckling of the columns or plates that make up the cell edges or
walls. In closed cells the enclosed gas pressure and membrane stretching
3. Finally, a region of densification occurs, where the cell walls crush together,
In the present model, the foot-supporting surface that the foot makes contact with
as rigid body, and a 6.8 mm thick, flat, foam pad. Due to the highly-deformable
2 i i
2
1 2 i 3 i 3 J el i i 1
1
U
i 1 i
2 i 2.11
i
Where, i 2.12
1 2 i
43
Values of these elastomeric foam parameters of the hyperfoam model (i;αi; βi)
ABAQUS (Hibbert and Karssonn, 2006). Table 2.2.4.2. shows the calculated
hyperelastic constants that fitted to the set of stressstrain (σε) curve data of
Table 2.2.4.2. The material constants of the foam pad used as the foot-supporting
interface from the hyperfoam strain energy function
α1 (MPa) β1 1 α2 (MPa) β2 2
0.467 15.5 0.021 -0.11 3.62 -1.05
Cheung et al., 2005), ligament failure (Wu, 2007), plantar fascia release (Gefen,
2002), and internal tissue stresses (Gefen, 2003, Chen et al., 2010b), most
existing models primarily focus on static standing posture, and modeling of the
muscles has been largely ignored. Thus, compared to the majority of existing foot
models which focus on standing (Gefen, 2002, Cheung et al., 2005, Chen et al.,
during heel rise, and forefoot sub-MTH stress/strain responses is the focus.
44
loading transfer needs to be considered. In this study, relative articulating
movements of the bony joints were simulated in the entire foot (see below). This
which have joints that were over-constrained by “fusing” them (Jacob, 2001b,
Gefen, 2002, Wu, 2007, Gefen et al., 2000, Gefen, 2003) or using kinematic-
Neither of these two reflect the anatomical joint constraints that are actually
governed by the bony congruence facets and passive stabilizers (i.e. ligaments
and fascia) (Leardini et al., 2000). The modeling strategy utilized here enabled
the external loading boundary conditions. Moreover, a model that includes joint
articulation could be potentially validated via joint kinematics and contact forces
Relative articular joint movements were included in the foot model. These are
located at the ankle (i.e., joint between articular surface of the tibia and fibula and
medial, intermediate and lateral cuneiform, and cuboid), tarsometatarsal (i.e., the
medial, intermediate and lateral cuneiforms, and the cuboid, articulating with the
bases of the metatarsal bones) and metatarsophalangeal (MTP) joints, with the
45
elements (Master/Slave approach, see Fig. 2.3.1.1.) were created for the
joints. Nodes on one surface (the slave surface) contact the discretized
segments on the other surface (the master surface). This allows the bones to
Fig. 2.3.1.1. Master/Slave approach to model contact interactions. The contact behavior
strictly follows the kinematic implications that slave nodes cannot penetrate master
surface segments (Hibbert and Karssonn, 2006).
interaction, 9 for tarsus joint interactions (including 1 for subtalar joint), 5 for
geometry of individual articular cartilage of these bony joints was not accounted
for; instead, their mechanical properties that contribute to normal joint loading
transfer were modeled. Thus, the governing stiffness during contact was chosen
46
to be 1.01 MPa, based on the compressive material properties of the foot joint
cartilage obtained from the 1st MTP joints (Athanasiou et al., 1998).
The functional roles of the extrinsic plantar flexor muscles are to provide stability
of the foot during the stance phase (Sutherland et al., 1980, O'Connor and Hamill,
tendon structures, which allow them to affect multiple joints inside the foot. This
major bony joints, including the ankle, the subtalar, and indirectly the
tendon is the most dominant extrinsic plantar flexor. Various studies using
among the Achilles tendon, plantar fascia and MTP joints, both statically (Carlson
et al., 2000) and dynamically (Erdemir et al., 2004), early described by (Hicks,
1955) as the Windlass mechanism. With the foot stabilized by flexor muscles
upon heel-rise, substantial dynamic ground reaction forces (GRF) are imposed
underneath the metatarsal heads (MTHs) (Chen et al., 2010b). Such potentially
47
‘detrimental’ stresses may only last for a short-dwelling time in a normal foot,
provided that the coordinative muscles involved activate the right amount of force
at the exact right time during gait (Perry, 1992, Hayafune et al., 1999).
As the areas under the MTHs are of particular clinical interest, the instant
the forefoot force reaches a peak (i.e. second peak GRF probably corresponding
replicating the foot pose relative to the ground. Specifically, the second
reflect mean forefoot orientation in the sagittal plane (i.e. the main loading plane)
at push-off (Fauth, 2002). The tibia, fibula and the superior surfaces of the soft
simulated via force vectors in axial alignment with the tendons attached (Fig.
2.3.2.2.).
During push-off, the ground reaction force (GRF) vector, acting at its
center of pressure at the forefoot segment, causes a bending moment along the
ankle joint. This depends on the joint positions and is balanced by the plantar
posterior (TIBP), Flexor hallucis longus (FHL), Flexor digitorum longus (FDL),
Peroneus brevis (PB) and Peroneus longus (PL). Note that the G-S complex is
dominant, since the other five extrinsic flexors have only small lever arms.
48
Analysis commenced with this baseline model, driven by musculoskeletal loading.
A maximum vertical GRF of 623.1 N, approximating the second GRF peak during
walking (measured for this particular subject who provides the foot geometry
using Kistler force plates during gait analysis), was generated by contracting the
(1995), which simulated the activity of the flexor tendons by the use of linear
actuators.
Fig. 2.3.2.2. Cut through the element mesh of the finite element model of muscular foot
and ankle complex, incorporating internal soft tissue, skeletal structures, ligaments,
plantar fascia, and musculotendinous units for push-off simulation. Application of muscle
forces were simulated by force vectors align with the tendons attached. FAT =
Gastrocnemius-soleus complex, FTIBP = Tibialis posterior, FFHL= Flexor hallucis longus,
FFDL = Flexor digitorum longus, FPB = Peroneus brevis, and FPL = Peroneus longus.
49
The initial configurations of the metatarsals were affected slightly because
joints. Solution of the model converged with the 2nd metatarsal shaft oriented at
Table 2.3.2.2. The input forces in the muscles applied through the nodes connected to
tendon elements to drive the foot finite element model
th
i muscle Muscle forces No. of nodes
#1 Gastrocnemius Soleus FAT 1620N 30
The corresponding total muscle force (TMF) was computed and the forces
TMF i 1,2,3,4,5,6
PCSAi
MFi 2.13
6
PCSA j
j 1
This method was similar to the one used by Salathe and Arangio, (2002) in an
analytical foot structure model for estimation of foot flexor muscle forces. The
forces calculated in the present study are listed in Table 2.3.2.2. This
50
(1983). Additional analysis was also undertaken, whereby activation of toe
flexors was not considered (i.e. forces applied to FHL & FDL approach zero), in
order to determine the significance of toe flexor muscle forces in realistic toe
load-bearing. The models were solved using the general purpose FE analysis
general frictional contact conditions were utilized. Sliding contact algorithm was
Fig. 2.3.3.1. Relative sliding of points with contact constraint. Note the possible evolution
of contact between node 101 and its master surface, BSURF, involving sliding contact
conditions with friction.
51
Using this modeling approach, the contact tractions, including shear interaction,
over the whole plantar surface segments can be computed. Moreover, this
algorithm also accounts for the stick, slip or separation of contacting surfaces
classical Laws of Friction. Thus, Coulomb's law follows the three rules of
According to Coulomb friction, the relationship between friction force (Fy) and
Fy
Fy Fz , Tan( ) 2.14
Fz
Coulomb friction is used to model the frictional contact between the foot
the local contact pressure would be applied to the surface of the deforming
52
The classical laws of friction offer a paradigm in which the resistance to relative
physical laws but rather “abstractions” that have been widely taught and adopted.
Classical friction does not adequately describe the complex interactions between
compliant, resilient, non-uniform surfaces such as those found in typical foot and
accurate 3-D contact stresses in the numerical modeling, the coefficient of friction
or shear traction ratio, µ, has to be measured experimentally and input into the
foot model. This is, in fact, one of the primary sources of complexity that is
studies have largely relied on those empirical values of µ. This is probably one of
interface are exclusively focused on plantar pressure prediction, while only few
53
calculate the actual shear traction ratio, µ, during foot-ground interaction. Details
With full muscular loads applied, this baseline foot model was successfully
solved in a typical geometry at heel rise, with the ankle and metatarsophalangeal
2.3.4.1). The ABAQUS CAE/post-processor was used to report results from finite
and von Mises stresses (VMS) in bony structures and internal tissue strains in
ligaments, plantar fascia and plantar soft tissue, and the model’s experimental
54
Fig. 2.3.4.1. The finite element predicted stress response of a whole human foot
subjected to comprehensive musculoskeletal loading corresponding to a heel-rise
posture. Outcome measures of primary interest in foot biomechanics, including (A)
stress distributions of bulk soft tissue, (B) metatarsal bones, (C) ligaments and plantar
fascia, can be obtained.
55