Medical Engineering and Physics 43 (2017) 86–91

Contents lists available at ScienceDirect

Medical Engineering and Physics

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

Technical note

Above-knee prosthesis design based on fatigue life using finite

element method and design of experiment
Suwattanarwong Phanphet a,c, Surangsee Dechjarern b,∗, Sermkiat Jomjanyong a
a
Industrial Engineering Department, Chiang Mai University, Chiang Mai 50200, Thailand
b
Production Engineering Department, King Mongkut’s University of Technology North Bangkok, Bangkok 10800, Thailand
c
Department of Industrial Technology Faculty of Science and Technology Chiang Mai Rajabhat University, Chiang Mai 50300, Thailand

a r t i c l e i n f o a b s t r a c t

Article history: The main objective of this work is to improve the standard of the existing design of knee prosthesis de-
Received 29 October 2015 veloped by Thailand’s Prostheses Foundation of Her Royal Highness The Princess Mother. The experimen-
Revised 30 August 2016
tal structural tests, based on the ISO 10328, of the existing design showed that a few components failed
Accepted 1 January 2017
due to fatigue under normal cyclic loading below the required number of cycles. The finite element (FE)
simulations of structural tests on the knee prosthesis were carried out. Fatigue life predictions of knee
Keywords: component materials were modeled based on the Morrow’s approach. The fatigue life prediction based
Knee prosthesis on the FE model result was validated with the corresponding structural test and the results agreed well.
ISO 10328 The new designs of the failed components were studied using the design of experimental approach and
Finite element
finite element analysis of the ISO 10328 structural test of knee prostheses under two separated loading
Design of experiment
cases. Under ultimate loading, knee prosthesis peak von Mises stress must be less than the yield strength
Fatigue
of knee component’s material and the total knee deflection must be lower than 2.5 mm. The fatigue life
prediction of all knee components must be higher than 3,0 0 0,0 0 0 cycles under normal cyclic loading. The
design parameters are the thickness of joint bars, the diameter of lower connector and the thickness of
absorber-stopper. The optimized knee prosthesis design meeting all the requirements was recommended.
Experimental ISO 10328 structural test of the fabricated knee prosthesis based on the optimized design
confirmed the finite element prediction.
© 2017 IPEM. Published by Elsevier Ltd. All rights reserved.

1. Introduction The type of prosthetic knee used depends on the amputee’s

strength and activity level. There are three types of knee mecha-
Currently, there are more than 50,0 0 0 amputees in Thailand. nisms, which are side bars with joints, single-axis knee prosthe-
The main causes are 45% road-related accidents, 25% landmine ac- sis and polycentric knee prosthesis [2,3]. Polycentric knee pros-
cidents, 20% chronic diabetes and 10% of other causes. The cost of theses are more stable and easier to control during swing and
total leg prosthesis system for Thai amputee is high due to the stance phase, but are more expensive. A polycentric knee mech-
high cost of knee prosthesis. The Prostheses Foundation of Her anism is a two-six bar linkage joint. Several researches of poly-
Royal Highness the Princess Mother has been producing and giv- centric knee prostheses involve the study of their dynamic stabil-
ing prostheses to disabled people, in Thailand and other South- ity [4–6]. The optimization of knee prosthesis based on structural
East Asian countries, in need of replacement limbs since 1992. The strength by varying design parameters has not been explored. Fi-
study of Thailand’s Prostheses Foundation OPD cards showed that nite element (FE) analyses are commonly used to investigate the
the knee prostheses often need repairing after 2 years of using. The fatigue strength and wear resistance of knee implants since FE
main cause was fatigue and wear of the knee component parts. modeling allows easy parametric study without the need to man-
Moreover, the polycentric knee prosthesis developed by the Foun- ufacture knee implants [7–9]. An FE model-based topology opti-
dation does not meet the ISO-10328 [1] structural standard test for mization of knee prosthesis is reported to lower its weight while
a knee prosthesis. successfully increase its strength [10]. The full factorial design of
experiment or Taguchi method is suitable for parametric study of
complex mechanisms such as monolimb [11,12] since there is in-
∗
Corresponding author. teraction between the design parameters. This work aims to study
E-mail addresses: [email protected] (S. Phanphet), the failure of knee prosthesis under the structural test and to de-
[email protected], surdec20 0 [email protected] (S. Dechjarern),
velop an optimum design of knee prosthesis to meet the require-
[email protected] (S. Jomjanyong).

http://dx.doi.org/10.1016/j.medengphy.2017.01.001
1350-4533/© 2017 IPEM. Published by Elsevier Ltd. All rights reserved.
 S. Phanphet et al. / Medical Engineering and Physics 43 (2017) 86–91 87

a b c

Fig. 1. (a, b) alignment setting conditions according to ISO 10328 [1], (c) The structural testing machine.
Table 1
Mechanical properties of knee prosthesis materials [14–17].

Material /Properties Density (g/cc) Young’s modulus (GPa) Yield stress (MPa) Poisson’s ratio Ultimate stress (MPa)

1. Aluminum ADC 12 2.823 71.0 144.35 0.33 205.31

2. SUS 304 8.00 200 250 0.29 505
3. SUS 630 (H900) 7.80 196 1034 0.272 1241
4. Ester Polyurethane (PU) 1.2 0.0689 56 0.4 56

ments of ISO 10328:2006 under static and cyclic tests. This paper phase control as shown in Fig. 2(a). The knee prosthesis compo-
applies finite element analysis and design of experiment to study nents consist of cover joint bar, knee joint upper, stopper, joint
and optimize the design parameters of a knee prosthesis in order bar, roller pin A upper and lower, roller pin B, knee joint lower
to increase its fatigue life. and friction pin C. Total weight of a knee joint is 500 g. Materi-
als of knee component are stainless Steel SUS 304 and SUS 630,
2. Methods Aluminum ADC 12 and Polyurethane. Two bearings were used in
roller pin assembly of the knee prosthesis and the other connec-
2.1. Structural testing of knee prosthesis tions were shirk-fitted and glued together.
The CAD geometry of the knee prosthesis is shown in Fig. 2(b).
The ISO 10328:2006 structural testing of the knee prosthesis A 3D finite element mesh of the knee prosthesis and testing fix-
specifies forces for 3 amputee’s weight levels at normal walking tures was created using ABAQUS auto-mesh option. All simple
phase (cyclic loading) and severe ultimate load during heel-strike parts were meshed with 3D-C3D8R elements, while all complex
(condition I) and toe-off (condition II). The alignment is defined in parts were meshed with 3D-C3D4 elements [13]. The testing fix-
the ISO 10328 by the distances F, O and U in X, Y and Z directions tures were assumed rigid. The number of elements range from
at three reference planes on the total limb replacement, i.e. hip 1640 elements for roller pin to 53,268 elements for upper knee
level, knee level, foot level, as shown in Fig. 1. This illustrates the joint part. The elastic-plastic material model was applied for all
walking motion during heel-strike (condition I) and toe-off (con- materials. The Young’s modulus, Poisson’s ratio, density, yield and
dition II) [10]. The ISO 10328 requires knee prostheses to undergo ultimate strength of all materials are shown in Table 1. Contact
3,0 0 0,0 0 0 cyclic loading and static ultimate loading without fail- pairs between pins, upper, lower and joint bars, where bearings
ure to any knee components and a combined knee deflection of were used and sliding allowed, were modeled with friction con-
less than 5 mm. The alignment and forces used in this knee pros- tact with the friction coefficient of 0.1. Tie constraints were applied
thesis structural testing are according to P4 (80 kg) condition I. The to glued and shirk-fitted assembly pairs. All connections between
structural testing took place at the Sirindhorn National Medical Re- knee prosthesis and testing fixture were assumed tied together and
habilitation Centre, Thailand, using a 5 kN servo-pneumatic fatigue no movements were allowed. Cyclic loading of 1230 N was applied
testing machine by Si-Plan Electronics Research Ltd (Fig. 1(c)). The at the top loading fixture. The point of loading at the bottom plate
cyclic loading was 1230 N at speed of 1 Hz. The knee testing is set was fixed. The alignment of knee testing model resembled that of
up by using two Aluminum tubes and top and bottom plate fix- the ISO 10328 experimental knee test under P4 and condition I as
tures to tie all components together (Fig. 1). The lengths of fixture shown in Fig. 2(c).
tubes are 150 mm and 500 mm for top and bottom parts, respec-
tively. At the top plate, FT is 89 mm. and OT is −74 mm. At knee
level, Fk is 56 mm. and Ok is −48 mm. Fb and Ob is −52 mm. and 2.3. Fatigue life prediction
39 mm. respectively at the lower (foot) plane.
Fatigue life prediction of knee prosthesis components are based
2.2. Finite element analysis on the Morrow’s Approach [18].The equation is as follows:

The polycentric knee prothesis,designed by the Prostheses σa   b

= 2N f (1)
Foundation, is a four-bar linkage knee joint with a hydraulic swing σf
88 S. Phanphet et al. / Medical Engineering and Physics 43 (2017) 86–91

a b c
FT

Stopper

Knee joint
Roller pin B upper

Roller pin A
Upper
Joint bar
Cover
Friction pin C

Roller pin A
Knee joint Lower
Lower
FB

Fig. 2. (a) The knee prosthesis of the Prostheses Foundation, (b) CAD geometry of the prosthesis, (c) the FE model of knee prosthesis and testing fixtures according to ISO
10328 [10].

a b c d

Fig. 3. (a) Fatigue failure of knee prosthesis, (b) Contour plot of displacement of a knee prosthesis and fixture (c) von Mises stress of knee prosthesis and fixtures, (d) von
Mises stress of knee prosthesis.

where σ a , σ ’f , b and Nf are the stress amplitude, mean stress, fa- Table 2
S-N curve equations for fatigue life predictions
tigue strength coefficient, fatigue strength exponent and number of
for all materials used.
life cycles, respectively. Rearranging Eq. (1) to obtain the equation
for the calculation of the number of cycles to failure, Nf : Value b σf
  1b SUS630 H900 − 4.49 × 10 − 2 1380
1 σa SUS304 − 7.20 × 10 − 2 586
Nf = (2) 7.75 × 10 − 2
2 σf Aluminum ADC 12 − 378
Ester Polyurethane − 9.85 × 10 − 2 108
The S-N curves of each material [14−17] were obtained and
converted to log-log curves to calculate the values of σ f and b. The
values of σ f and b for knee prosthesis material are as shown in study by adapting their sizes and shapes at 3 levels as shown in
Table 2. Table 5. Applying a full factorial DOE [19], 27 FE models of the
structural test of knee prosthesis were studied. Knee prosthesis
2.5. Designing of experiments (DOE) was subjected to 2 separated loading cases; a normal cyclic load-
ing of 1230 N and a static ultimate loading of 4130 N. Three design
The study of the Prostheses Foundation OPD cards revealed that criteria conditions were identified as follows:
the failure of knee prosthesis were found in 3 components, namely,
the stopper, the pyramid lower knee joint and the joint bar. In or- - Under ultimate loading, the total deflection of knee prosthesis
der to design and optimize fatigue strength of knee prostheses, the must be less 2.5 mm and the maximum stress of components
three failed components were selected as design parameters in this must be less than its yield strength.
 S. Phanphet et al. / Medical Engineering and Physics 43 (2017) 86–91 89

Table 3
Design Parameters for DOE of FE models of the knee prosthesis testing.

Parameters Levels

1 2 3
Thickness of joint bar (A) 7.5 mm 8.3 mm 9.3 mm

Cross-section length and shape of pyramid lower knee joint (B) ∅12.5 mm ∅12.5 mm ∅13.5 mm

Diameter of stopper (C) ∅8 mm ∅10 mm ∅12 mm

Table 4
Validation of FEM fatigue lifetime prediction at 95% confidence interval (2σ ).

Testing average lifetime (Cycle) FEM fatigue lifetime prediction (Cycle) 95% confidence interval (2σ ) (Cycle) Error (%)
1
(893,614 + 912,811)/2 = 903,213 Nf = 1
2
( 208
586
) −7.20×10−2 =887,0 0 0 (903,213–887,0 0 0) = 16,213 < 32,425 pass err = (9,03,213 ,−88,7,0,00),
9,032,13
∗ 100 = 3.59

a b c

Fig. 4. Contour plots of von Mises stress of knee prosthesis (a) case 3, (b) case 14, (c) case 25.

- Under cyclic loading, the fatigue life prediction must be over Applying Eq. (2) and the values of σ f and b in Table 2 for SUS304,
3,0 0 0,0 0 0 cycles. the predicted number of cycles to fatigue failure of knee joint bar
under cyclic loading is 887,0 0 0. For the statistic reliability at 95%
3. Results (2σ ), the difference between the experimental testing results and
the FEM prediction must be ±32,425 cycles. It can be seen from
3.1. Validation of fatigue life prediction Table 4 that the difference between the FE fatigue prediction and
the knee structural test is 16,213 and it is within 95% confidential
The results of experimental structural test showed that failures interval. The fatigue prediction based on FEM error is only 3.59%
of the knee prostheses were caused by fatigue failure at joint bars from the results of experimental structural tests.
after the average number of cycles of 903,213. Fig. 3 shows the
locations of fatigue failure of knee joint bar under cyclic loading. 3.2. Design of knee prosthesis under static ultimate and cyclic
Fig. 3(b) shows an FEM predicted contour plot of the displacement loadings
of the knee prosthesis and fixtures. It is clear that the maximum
displacement occurs at the knee prosthesis as expected. The FE Fig. 4 shows the contour plots of von Mises stress distribution
model of knee prosthesis structural test result shows that the max- of three knee prosthesis design cases, which are case 3, 14 and 25.
imum stress values of 208 MPa is at the joint bar (Fig. 3 (c−d)). It is clear that as the thickness of the joint bar increased, its max-
90 S. Phanphet et al. / Medical Engineering and Physics 43 (2017) 86–91

Table 5
DOE results of FE models of structural tests of knee prosthesis under cyclic and static ultimate loading (P4 - condition I).

Case Joint bar Lower Stopper Deflection ultimate (mm.) Ultimate load fail/pass

Stress (MPa) Nf (Cycles) Stress (MPa) Nf (Cycles) Stress (MPa) Nf (Cycles)

1 208 8.87 × 105 150 7.47 × 104 38 1.94 × 104 2.82 Fail
2 91 8.23 × 1010 112 3.24 × 106 69 4.55 × 101 3.48 Fail
3 256 4.96 × 104 83 1.55 × 108 58 2.65 × 102 3.21 Fail
4 128 7.53 × 108 94 3.11 × 107 62 1.35 × 102 3.43 Fail
5 133 4.42 × 108 85 1.14 × 108 69 4.55 × 101 3.28 Fail
6 140.8 2.00 × 108 122 1.07 × 106 61 1.59 × 102 3.25 Fail
7 151 7.58 × 107 99 1.59 × 107 38 1.94 × 104 3.18 Fail
8 101.3 1.94 × 1010 78 9.79 × 107 51 9.79 × 102 3.15 Fail
9 124.3 1.13 × 109 102 1.08 × 107 48 1.81 × 103 3.12 Fail
10 434 3.25 × 101 155 4.89 × 104 56 3.79 × 102 3.05 Fail
11 176 8.70 × 106 78 3.45 × 108 64 9.76 × 10 2.86 Fail
12 168 1.77 × 107 76 4.82 × 108 22 4.99 × 106 1.42 Fail
13 182 5.71 × 106 95 2.71 × 107 67 6.13 × 10 3.07 Fail
14 202 1.30 × 106 73 8.11 × 108 53 6.62 × 102 2.87 Fail
15 429 3.88 × 10 109 4.60 × 106 46 2.79 × 103 2.37 Fail
16 152 6.73 × 107 91 4.72 × 107 22 4.99 × 106 3.07 Fail
17 229 2.33 × 105 86 9.79 × 107 43 5.53 × 103 2.88 Fail
18 273 2.03 × 104 88 7.27 × 107 38 1.94 × 104 2.47 Fail
19 144 1.42 × 108 81 2.12 × 108 59 2.23 × 102 3.09 Fail
20 129 6.76 × 108 80 2.49 × 108 57 3.16 × 102 2.82 Fail
21 163 2.62 × 107 69 1.68 × 109 46 2.79 × 103 2.54 Fail
22 198 1.76 × 106 94 3.11 × 107 58 2.65 × 102 3.14 Fail
23 202 1.33 × 106 62 6.67 × 109 51 9.79 × 102 2.82 Fail
24 214 6.18 × 105 95 2.71 × 107 21 8.00 × 106 2.37 Fail
25 173 1.16 × 107 73 8.11 × 108 23 3.17 × 106 1.52 Pass
26 155 5.37 × 107 62 6.67 × 109 22 4.99 × 106 1.5 Pass
27 204 1.17 × 106 91 4.72 × 107 53 6.62 × 102 2.44 Fail

Table 6
The optimized knee prosthesis, stress values and lifetime prediction under cycling loading

Thickness of joint bar 9.3 mm Square-shaped pyramid lower knee ∅13.5 mm Stopper diameter ∅ 10 mm

Stress 154.8 MPa Stress 62 MPa Stress 22 MPa

Nf = 5.37 × 107 Cycle Nf = 6.67 × 109 Cycle Nf = 4.99 × 106 Cycle

Fig. 5. Main effect plot of design parameters on (a) von-Mises stress under ultimate loading, (b) deflection under ultimate loading, (c) Fatigue life cycle under cyclic loading.

imum stress decreased while the maximum stresses of knee cover fatigue failure were calculated for each component as shown in
and friction pin increased because higher force was transferred to Table 5. The fatigue life of the knee prosthesis is the shortest life
other knee components. Maximum von Mises of three components prediction of its components. Under ultimate load, the maximum
were used for fatigue life prediction. Applying Eq. (2) and the val- stress of each knee component was compared with its material
ues of σ f and b in Table 3, the predicted number of cycles to yield stress, resulted in pass or fail.
 S. Phanphet et al. / Medical Engineering and Physics 43 (2017) 86–91 91

4. Discussion and conclusion graduate education scholarship. The authors are grateful to Thai-
land’s Prostheses Foundation of Her Royal Highness The Princess
The interaction plots of design parameters on total deflection Mother for providing the information and knee prostheses for this
under ultimate loading and fatigue life cycle under cyclic loading study. The authors appreciate the assistance of the Sirindhorn Na-
were constructed. The results showed that the dimensions of the tional Medical Rehabilitation Centre, Thailand, in performing the
joint bar, the lower knee joint and the stopper are interrelated ISO 10328 structural tests on knee prostheses for this study.
on their effect on the total deflection of knee prosthesis under
the ultimate force and overall fatigue life of knee prosthesis under References
normal cyclic loading. Fig. 5 shows the main effect plot of design
[1] ISO 10328:2006. Prosthetics-structural testing of lower-limb prostheses − Re-
parameters on von-Mises stress under ultimate loading, deflection quirements and test methods. Switzerland: International Standard; 2006.
under ultimate loading, and fatigue life cycle under cyclic loading. [2] Tang PC, Ravji K, Key JJ, Mahler DB, Blume PA, Sumpio B. Let them walk!
The joint bar size and shape are the most influenced parameters Current prosthesis options for leg and foot amputees. J Am Coll Surg
2008;206(3):548–60.
in case of deflection and maximum stress under severe ultimate [3] Radcliffe CW. Four-bar linkage prosthetic knee mechanisms: kinematics, align-
force. The cross section area of the stopper is the most influenced ment and prescription criteria. Prosthet Orthot Int 1994;18:159–73.
parameter in case of overall fatigue life of knee prostheses. This is [4] Ochsner A, Altenbach H. Optimization of the design of a four bar mechanism
for a lower limb prosthesis using the taboo search algorithm. Adv Biomech
because polyurethane has a low fatigue limit.
Syst Mater Adv Struct Mater 2013;40:107–25.
The best knee prosthesis design was found to be a joint bar [5] Hobson DA, Torfason LE. Computer optimization of polycentric prosthetic knee
with uniform thickness of 9.3 mm, square shaped pyramid with mechanisms. Bull Prosthet Res 1975;10-23:187–201.
[6] Poliakov O, Chepenyuk O, Pashkov Y, Kalinin M, Kramar V. Multicriteria syn-
cross section length of 13.5 mm and a stopper diameter of 10 mm
thesis of a polycentric knee prosthesis for transfemoral amputees. Int J Chem
(see Table 6). It has the fatigue lifetime of 4.99 × 106 cycles as de- Biol Eng 2012;6:257–62.
termined by fatigue life of the stopper and it is not plastically de- [7] Completo A. Relationship of design features of a stemmed tibial knee prosthe-
formed under ultimate force. The optimized design of knee pros- sis with stress shielding and end-of-stem pain. Mater Des 2009;30:1391–7.
[8] Baldwin MA. Dynamic finite element knee simulation for evaluation of knee
theses was fabricated by the Prostheses Foundation. The Sirindhorn replacement mechanics. J Biomech 2011;45(3):474–83.
National Medical Rehabilitation Centre performed the ISO 10328 [9] Chen NZ, Winson C, Lee C, Zhang M. A numerical approach to evaluate the
structural test on them. The test result confirmed the finite ele- fatigue life of monolimb. Med Eng Phys 2006;28:290–6.
[10] Reist TA, Andrysek J, Cleghorn WL. Topology optimization of injection moldable
ment prediction that the knee prosthesis can withstand more than prosthetis knee joint. Comput Aided Des Appl 2010;7(2):247–56.
3,0 0 0,0 0 0 cycles of cyclic loading and does not fail under ultimate [11] Lee WC, Zhang M. Design of monolimb using finite element modelling and
loading. It was found that the optimum prosthesis passed the re- statistic-based Taguchi method. Clin Biomech 2005;20:759–66.
[12] Alauddin M. Prediction of tool life in end milling y response surface method-
quirements of the ISO 10328 structural test. ology. J Mater Process Technol 1997;71:456–65.
[13] ABAQUS. Documentation. Providence, RI, USA: Dassault Systèmes; 2011.
Competing interests [14] Guo F. Fatigue life prediction of SUS 630 (H900) steel under high cycle loading.
Acta Mech Solida Sin 2013;26:584–91.
[15] Okayasu M. A study of the mechanical properties of an Al–Si–Cu alloy (ADC12)
None declared produced by various casting processes. Mater Sci Eng 2012;543:185–92.
[16] Loos MR. Enhancement of fatigue life of polyurethane composites containing
carbon nanotubes. Compos: Part B Eng 2013;44:740–4.
Funding
[17] North American Stainless Long products stainless steel grade sheet. AISI304
June 2007 Available at:http://www.northamericanstainless.com/wp-content/
Partly by the National Research Council of Thailand uploads/2010/10/Grade- 304- 304L.pdf. Accessed 26.
[18] Saroosh MA. High cycle fatigue life prediction of cold forging tools based on
workpiece material property. J Mater Process Technol 2007;191:178–81.
Ethical approval [19] Fisher RA. The design of experiments. 9th ed. New York. USA: Hafner Publish-
ing Company, inc.; 1971. ISBN 0-02-844690-9.
Not required

Acknowledgment

The work described in this paper was partly supported by a

grant from the National Research Council of Thailand for the post-