Introduction
Understanding of the interacting forces among the components of the human knee during daily physical
activities is of prime importance to prevent injuries on this articulation. In addition, the knowledge of such
forces is fundamental in the proposal of prosthesis design, rehabilitation, and muscular strengthening.
Besides the experimental studies, mathematical and computational models of the knee joint have been
developed to predict the approximated biomechanical behavior of individual fibrous tissue and of the entire
articulation. Some outstanding works, due to their close approximations to real models are those by BaoWillems [A6] (1999), which achieve a mathematical approach to determine the movements of the leg and
thigh. Studies by Li and Gil [A1] (1999) and Li, Lopez and Rubash [A2] (2001) use models of the main
bones and the cartilaginous part and made simplifications about the menisci (using equivalent-resistance
springs) and ligaments. Morales and Ayala [A5] (2001) which deals with a 2D structural analysis where the
modeling of the bones is according to their bony regions.
While the mathematical models of the knee joint can be useful to predict forces and stresses in individual
structures of the articulation, as well as to estimate its kinematics, validation of such models is a
challenging task. It is complicated to make a computational model of the knee that accurately predicts the
response and movements of the articulation as it is resembled by experimental methods.
With the help of digitized geometry and improved capabilities of ANSYS 7.0 in handling contact pairs, the
main objective of this study is to develop a three-dimensional finite element model to predict stresses in
individual components of the knee joint (femur, tibia, menisci, and patella) due to interacting forces among
them when steady state loading is considered. The model developed is a geometric non-linear model that
includes contact path-dependent elements (with friction). Results are compared with the results from linear
modeling of the individual structures forming the articulation.
Procedure
Materials
Mechanical properties of bones vary, in general, according to age, weight, gender, and feeding habits of the
subject under analysis. Properties values given in Tables 4 to 10 (see appendix) were obtained from a dead
body, male 30 years old, 1.70 m tall, and average weight estimated of 70 kg, Haut, Hull, Rashid and Jacobs
[A4]. For analysis purposes a value of 1400 N, which is around twice the estimated weight, was used to
load the knee joint (this loading was also used by Morales and Ayala [A5]). In addition, prescribed
displacements are alternatively used to load the computational model because graphical progression of
convergence is cleaner and easier to follow. Estimated displacement values of 10 mm are applied to the
models inasmuch as these values correspond to the maximum deformation that the meniscus can withstand.
Values presented are considered standard values.
Model
The analysis was carried out according to the geometric models obtained from The standardized femur home
page [W1]. The analysis includes studying the combined behavior of the meniscus, femur, tibia, and
patella, and assuming only an external load applied to the femur, restricting movements the knee joint
instead of applying loads to simulate ligaments and tendons loading (The model constraints used is only
useful for the specified configuration. The applied constraints simulated the duties of ligaments and tendons
in a real knee joint). Due to the intricate model, the 3D model of the articulation was considered to be
extended (0 degrees flexion), formed by only one type of bony material by bone, which is porous and
orthotropic. For comparison purposes, studies were done on independent models (each bone considered
separately) with linear analyses. The cartilaginous parts of the bones were excluded in all the analyses,
linear and non-linear, even though their geometry is known. The reason for this is that the mechanical
properties of these parts are not experimentally characterized yet. In this study the interaction of each part
of the joint with its bony surroundings is taken into account.
Because the present analysis required of three-dimensional nonlinear processing, a solid element was
chosen, SOLID92, which is a structural solid tetrahedral element of 10 nodes, for a space of threedimensional work and that has three degrees of freedom (UX, UY and UZ) by node. Another alternative
was to use an element SOLID95, which offers 20 nodes and also supports geometric non-linearities,
nevertheless, it increases the number of nodes to analyze.
For the components potentially in contact, high order elements CONTA174 and TARGE170 were used.
Element CONTA174 is a three-dimensional 8-node element that allows surface-to-surface contact and
sliding. Element TARGE170 is compatible with element CONTA174 and allows, also, surface-to-surface
contact.
Figure 2. Linearized on
CorelDraw! 8.0
Figure 3. Linearized on
AutoCAD 2002
Figure 4. Greer-Wang
Model
The complete joint, Fig 4, that contains femur, tibia, patella and menisci, was assembled in IronCAD 5.0.
In order to have the total certainty that the obtained model fulfilled the geometry to analyze two revisions
were made: one on the basis of visual appreciation and another one according to FEA.
Analysis
Models
In this study four models were considered, from which progressive data had to be obtained to refine the
final model. The analyzed models were: individual models, simplified model, reduced model and final
model. In the individual models each component was used separately as if all the knee joint worked like a
rigid body, that is to say, passing the forces completely to the following component. The second model,
the simplified model, was used to obtain a first approach to the behavior of the parameters of the contact
pairs. The third model is a reduced model of the final model and it allows us to vary parameters in a fast
way. The final model does not include reductions or simplifications nor an optimized mesh. The four
models are explained next.
The tibia was restricted from its bottom, applying cero displacements along the three directions, UX = UY
= UZ = 0. For each meniscus, all the nodes located on its peripheral surfaces were restricted along the X
and Y directions, keeping free the displacements along the Z direction.
The contact pair between the femur (Contact) and the menisci (Target) is defined as Contact Pair 1, while
that between the menisci and the patella (Contact) and the tibia (Target) is defined as Contact Pair 2. The
contact pairs were realized by means of node components.
Values obtained from the convergent contact pair are shown in Table 10.
Femur (Pa)
Tibia (Pa)
Menisci (Pa)
1.5667E8
4.4563E7
4.1757E7
6.7340E7
4.9649E8
1.4994E8
9.4517E7
3.3975E8
2.1327E7
2.24406E6
3.44056E6
2.41243E7
Von Misses stress showed on last table are graphically displayed in Fig. 14, Fig. 15 and Fig. 16. These
results are Von Misses Stress (averaged), in Pa.
Tibia (Pa)
Patella (Pa)
Menisci (Pa)
SEQV (Avg)
3.0866E8
1.51125E9
5.4797E9
4.0061E7
X (SX)
53558.3
2.8957E6
5.37498E6
9483.01
Y (SY)
89836.8
1.86947E6
5.36128E7
347021
Z (SZ)
543606
3.32007E6
1.3696E7
99587.9
Von Misses stress showed on the last table are graphically displayed in Fig. 17, Fig. 18, Fig. 19 and Fig. 20.
These results are Von Misses Stress (averaged), in Pa.
The obtained results, maximum values, from full analysis model are showed on the next table:
Tibia (Pa)
Patella (Pa)
Menisci (Pa)
SEQV (Avg)
2.65237E8
6.07005E8
1.00215E8
1.11417E7
X (SX)
3.32104E6
4.9664E6
1.25758E6
15854.7
Y (SY)
3.47354E6
7.40549E6
3.11426E6
68197.8
Z (SZ)
2.75184E6
1.77663E7
7.403E6
2497.52
Von Misses stress showed on last table are graphically displayed in Fig. 21, Fig. 22, Fig. 23 and Fig. 24.
These results are Von Misses Stress (averaged) in Pa.
Trying to simplify obtained results and showed a faster comparison between the analysed models, we make
a comparative chart displaying Von Misses Equivalent Stress from Femur among different models. These
are showed on Fig. 25.
Two of the important parts of this work are to distinguish the discrepancy of the values of stress between
linear and non-linear models and the location of the maximum stress.
In the location of the maximum stress, a simple linear analysis of the Femur shows points of maximum
stress in the central part and on the area close to the upper "head", whereas a non-linear analysis displays a
bigger stress on the area in contact with the Patella. For the Tibia similar results are obtained. The linear
analysis exhibits bigger concentration of stress in central areas, while the non-linear analysis shows
maximum stress over areas that interact with the Patella. The Menisci also display different strain and stress
levels between the linear and non-linear analysis; however, it is not possible to identify an unique area of
maximum strain.
The comparative chart in Fig. 25 emphasizes the lack of correspondence of the stress values obtained from
the linear analysis and those from the non-linear analysis.
From these models a logical order of stress levels is expected, being higher for the stresses in the lineal
model, then decreasing for the reduced model and the lowest level of stress is present in the complete
model. This behavior of stresses is inferred from the fact that in the lineal analysis each body part receives
directly the full 1400 N load, obtaining a high level of stress. In the non-lineal models the body part that
takes the load directly is the Femur, and this explains why the obtained values in the graphs are closer. The
other components do not take the complete load since they are deformable bodies and a portion of the load
is taken by them.
Another outstanding aspect of this work is the numeric evaluation of the control parameters of the contact
pair (FKN and FTOLN, mainly). The achieved descriptions are aimed to get a selection criterion for the
parameter values involved. The numeric evaluation of this parameters is important since without them a
complete analysis of the articulation as a whole is not possible.
The complete model meets the initial expectations since it allows to control the four bodies involved by
means of the contact pairs, while achieving the convergence of a non lineal analysis without using
simplifications of the physical model. In addition, in the full model each body can be analysed separately
and also the disparity among lineal and not lineal analysis values can be analysed in graphic form, which
accentuates the importance of this type of analysis.
Conclusion
Numerical structural analysis of human knee joint, one of the most important joint in the human body, has
great importance because this analysis take us closer to a complete knowledge of its mechanical behavior.
Reaching this kind of simulated analysis instead of intrusive procedures or destructive tests are part of the
biomechanical highest goals.
In order to get a full knowledge of human knee joint behaviour it is necessary to consider, in addition to the
analysed model, the cartilaginous parts that cover the bones, the compact bone (bones internal part), and
accurate menisci model as well as ligaments to achieve a closer simulation to reality.
Throughout this paper some of the most important parameters to control the geometry of the movement are
given, and the divergence of the results as well as the progressive convergence approach are discussed. It is
also implicitly shown the model evolution method to find the parameters that lead us to achieve the
objectives of this paper.
All assumptions used in other similar studies were made in order are to save processing time and to avoid
convergence problems. The assumptions made in this work are only for data or parameters not defined
completely. The main contribution of the present work is: to be able to simulate the whole articulation
using three-dimensional solid models, including friction among the bodies.
As a future work, a validation of obtained results can be sought based on the reactions that a live subject
under study can undergo and then compare them with the reactions obtained in this simulation or other
simulations.
A possible method of validation it would be to simulate this system with ADAMS and then make a
comparison of stresses and force transmission among models.
Finally, its emphasized the difference between results obtained from linear analyses and non-linear
analyses and the associated complexity of the latter (geometric non-linearities, contacts non-linearities,
multiples contact pairs and highly deformable materials).
[L1] Cailliet R., Sndromes dolorosos, Rodilla, Editorial Manual Moderno, Segunda edicin,
Mxico, Pag.: 8-15
[L2] Gupta , K. K., Meek J. L., Finite Element Multidisciplinary Analysis, AIAA Education Series
(American Institute of Aeronautics and Astronautics, Inc., Pag.: 1-20.
[L3] Moaveni, S. , Finite Element Analysis, Theory and application with ANSYS, Prentice Hall,
Pg: 1-6, 216-261
[L4] Cook, R.D., Finite element modeling for stress analysis, John Wiley & Sons, Pag: 275-295
[L5] Reddy, J.N., An introduction to the finite element method, McGraw Hill Publising Company
[L6] ANSYS Training Manual 6.0, Ansys Inc. Basic non-linearities, Advanced non-linearities,
Contact and Bolt pretensin.
[A3] S. J. Piazza, S.L. Delp, Three-dimensional dynamic simulation of total knee replacement
motion during a step-up task, December 2001, Journal of Biomechanical Engineering, ASME.
[A4] T.L. Haut, M.L. Hull, M. M. Rashid, C. R. Jacobs, A finite element model of the human
knee joint for the study of tibio-femoral contact, June 2002, Journal of Biomechanical
Engineering, ASME.
[A5] L. Morales Acosta, A. Ayala Ruz, Knee Joint Biomechanics, October 2001, SOMIM
Congress 2001, Celaya, Gto., Mxico.
[A6] H. Bao, P.Y. Willems, On the kinematic modeling and the parameter estimation of the
human knee joint, August 1999, Journal of Biomechanical Engineering, ASME.
http://www.cineca.it/hosted/LTM-IOR/back2net/ISB_mesh /form_fem.html
http://www.ibv.org/iinformacion/index.html
http://www.per.ualberta.ca/biomechanics/bwwnofrm.htm
http://www.biomech.com/
Appendix
Table 4. Mechanical properties (Material 1)
Porous
Mu
0.2
Compact
Ex
1.2 E10
6.1 E9
Ey
1.34 E10
8.692 E9
Ez
2 E10
1.2793 E10
PRxy
0.38
0.38*
PRyz
0.24
0.24*
PRxz
0.22
0.22*
Gxy
4.55 E9
2.3061 E9**
Gyz
6.23 E9
4.0411 E9**
Gxz
5.61 E9
3.588 E9**
Compact
Ex
1.85 E10
6.9 E9
Ey
2.275 E10**
8.5 E9
Ez
4.9262 E10**
1.84 E10
PRxy
0.49*
0.49
PRyz
0.14*
0.14
PRxz
0.12*
0.12
Gxy
6.4255 E9**
2.4 E9
Gyz
1.3119 E 10**
4.9 E9
Gxz
9.382 E9**
3.6 E9
2 E7
Ey
1.4 E8
Ez
2 E7
PRxy
0.2
PRyz
0.3
PRxz
0.2
Gxy
6.9565 E6**
Gyz
8.0706 E7**
Gxz
3.913 E6**
Compact
Ex
1.85 E10
6.91 E9
PRxy
0.3
0.3*
2 E7
Ey
1.4 E8
Ez
2 E7
PRxy
0.2
PRyz
0.3
PRxz
0.2
Gxy
6.9565 E6**
Gyz
8.0706 E7**
Gxz
3.913 E6**
Contact pair 2
FKN
1.0
1.0
FTOLN
0.01
0.01
Material
Contact
Status
Exclude
everything /
Red.
penetration
Include
everything /
Close gap
Contact
pair 2
Contact
pair 3
Contact
pair 4
FKN
0.8
0.8
0.25
0.1
FTOLN
8.5
0.1
0.01
Material
Exclude
Exclude
Exclude
Exclude
Initial
Penetration everything everything everything everything
Close Gap Close Gap Close Gap
Auto offset Reduce
penetration
Contact
pair 1
Contact
pair 2
Contact
pair 3
Contact
pair 4
FKN
0.8
0.8
0.25
0.1
FTOLN
8.5
0.1
0.01
Material
Initial
Penetration
Exclude
everything
Exclude
everything
Exclude
everything
Exclude
everything
Auto offset
Reduce
Close Gap
penetration
Close Gap
Close Gap
* These values were taken considering that the Poisson ratio is constant for the spongy bone as well as for
the compact one. There are no reported values in the literature.
** These values were approached by a linear interpolation, since they are not reported in the literature.