Medical Engineering & Physics 20 (1998) 360–373

Finite element modelling of a residual lower-limb in a prosthetic

socket: a survey of the development in the first decade
a,*
Ming Zhang , Arthur F.T. Mak a, V.C. Roberts b

a
Rehabilitation Engineering Centre, The Hong Kong Polytechnic University, Kowloon, Hong Kong, People’s Republic of China
b
Department of Medical Engineering and Physics, King’s College School of Medicine and Dentistry, Denmark Hill, London SE5 9RS, UK

Received 20 January 1997; accepted 5 March 1998

Abstract

A review is presented of the existing finite element models developed from 1987 to 1996 for the biomechanics of lower-limb
prostheses. Finite element analysis can be a useful tool in investigating the mechanical interaction between the residual limb and
its prosthetic socket, and in computer-aided design and computer-aided manufacturing of prosthetic sockets. Various assumptions
and simplifications are made in these models to simplify the actual problem with complex geometry, material properties, boundary
and interfacial conditions, as well as loading situations. The analyses can provide the information on the stress distribution at the
stump/socket interface and within the residual limb tissues. More recently, nonlinear models have been developed taking into
consideration the process of socket rectifications, the slip/friction conditions and material large deformation. The models so far
developed have provided some basic understanding of the biomechanics. Comparison of the predictions of these models with
experimental measurements indicated that the predicted stresses were within the ranges measured, although one-to-one correspon-
dence was difficult to achieve. Further research is still required in order to improve these models to obtain higher precision in the
results taking into account nonlinear and dynamic effects.  1998 IPEM. Published by Elsevier Science Ltd. All rights reserved.

Keywords: Amputation; Finite element model; Prosthesis; Socket design; Stress

1. Introduction to account for variations in the stump shape, variations

in the pressure tolerance of the soft tissues, and special
A lower-limb prosthetic socket provides coupling biomechanical requirements related to support and con-
between the body and the prosthesis. Its aim is to effec- trol. In the current socket design, the principles and tech-
tively integrate the prosthesis as a functional extension niques used by prosthetists in performing such shape
of the body. This goal is limited by the compliance of modifications are largely based on experience. To achi-
soft tissues of the residual limb and their local tolerance eve a satisfactory socket, a subjective trial and error
to externally applied forces. As a matter of fact, the soft approach has to be used until a successful fit is obtained.
tissues around a residual limb are not well-suited for load This kind of subjective knowledge-based design has
bearing and an improper load distribution may cause been used in the current computer-aided design and
pain and skin damage. Therefore, correct shaping of the manufacture (CAD/CAM) of the prosthetic socket. Cur-
socket as a way to distribute the load properly is a criti- rent systems involve three parts: shape sensing to obtain
cal process in lower-limb prosthetic socket design. a digital representation of the stump shape; modifications
The shape of the socket is not an exact replica of the of such a shape into a “socket” shape; and fabrication
residual limb, but includes modifications such that the of the socket by transferring the modified socket shape
load can be transferred more effectively between the to a CAM system [1]. Nowadays, the hardware of a
prosthesis and the residual limb. This modified shape is CAD/CAM system is unlikely to be a main problem.
The primary issue in such a system is how to determine
an appropriate socket shape from the residual limb
* Corresponding author. Tel.: + 852-2766-7664; Fax: + 852-2362- shape, i.e., what kind of socket shape can give an effec-
4365; E-mail: rcmzhang@polyu.edu.hk. tive load transfer and optimal load distribution, and can

1350-4533/98/$19.00  1998 IPEM. Published by Elsevier Science Ltd. All rights reserved.
PII: S 1 3 5 0 - 4 5 3 3 ( 9 8 ) 0 0 0 2 7 - 7
 M. Zhang et al. / Medical Engineering & Physics 20 (1998) 360–373 361

maximise the amputee’s comfort and performance. analysis developed in engineering mechanics. The FE
Owing to the lack of a quantitative assessment method, method is a full-field analysis for determining the state
the modification of the socket is still based substantially of stress and strain in the particular field. This approach
on the subjective experience of the prosthetist and the is well suited for parametric analyses in the process of
user’s feedback. design.
Many design approaches have been developed to FE analysis is implemented by dividing a complex
improve socket shape design. The basic principles vary problem domain, to which there is yet no analytical sol-
from either distributing most of the load over the specific ution or the analytical solution is very difficult to obtain,
load-bearing areas to uniformly distributing the load into a finite number of smaller domains or elements,
over the entire stump. No matter what kind of design each admitting simpler solutions. The solution obtained
principles, designers are interested to know the relation- in this way is approximate, as the model can only allow
ship between the designed socket and the biomechanical a finite number of degrees of freedom. The accuracy of
consequences, including the pressure distribution the solution is largely dependent upon the number of
between the stump and the socket. This will help design- elements of the model, a higher number of elements
ers to evaluate the quality of fitting and to increase the potentially enabling a more accurate solution to be
understanding of the underlying biomechanical rationale. found. However, a penalty must be paid in terms of the
In order to achieve such an aim, a variety of studies amount of computing time required. In practice, a bal-
have been conducted to understand the residual ance is to be achieved between the accuracy of the sol-
limb/socket interfacial biomechanics during the last sev- ution and the computing resources available. Further-
eral decades. Many investigations were undertaken to more, other factors affecting the accuracy of the analysis
measure the pressures and shear stresses applied to the involve model assumptions and parametric inputs of the
surface of the residual limb, as reviewed by Sanders [2] geometry, material properties, boundary conditions and
and Silver-Thorn et al. [3]. Although the pressure and external loads.
shear stress at the socket interface can be assessed at There are two major benefits in using FE analysis.
specific sites by experimental methods, full-field evalu- First, it can be used to determine a particular stress,
ation of the load transfer remains exceedingly difficult strain and motion anywhere within an object that would
to assess experimentally. Evaluation can be made of the theoretically be impossible to obtain by other analytical
stresses applied to the skin surface, but not of stresses or experimental method. Secondly, it can be used as an
within the soft tissues. Errors could result from the inter- effective tool for parametric analysis. This means that
face interference due to the introduction of the trans- structural parameters can be altered and their effects
ducers. Results reported by different groups differed studied very conveniently.
widely. Furthermore, experimental measurements can be Since the finite element methods were introduced to
performed only for an existing socket, but cannot be prosthetic socket design in the late 1980s [5,6], several
used to predict the interface biomechanics before a models have been developed [7–17]. Despite many sim-
socket is fabricated. plifications and assumptions, the model development in
It is anticipated that many of the difficulties associated the first decade showed that FE modelling, if properly
with the experimental measurements can be overcome conducted, can be used efficiently as a useful design tool
by numerical analyses, provided an appropriate compu- for prosthetic sockets [18,19]. In theory, FE analyses
tational model can be developed. As a complement to could predict stress distribution patterns at the interface
clinical measurements, numerical analytical techniques and within the soft tissues and allow parametric evalu-
can be used to facilitate a systematic build-up of the ation, in a systematic manner, of the parametric effects
biomechanical principles behind the design of prosthetic and mechanical response of the residual limb under dif-
sockets. With the emergence of CAD/CAM technology ferent conditions due to socket configurations and mech-
for prosthetic sockets, it is strongly desired that com- anical loading without the need to fit an amputee with
puter models can be used to predict the load transfer a series of sockets. This approach can be used to design
biomechanics of the residual limb/socket interface. It can socket shapes to achieve special load transfer character-
provide quantitative feedback to prosthetists rec- istics before a prosthesis is to be made and put on the
ommending rectification in the CAD software, or even amputee.
better, create new socket shapes with optimal pressure In the current socket design process, as shown in Fig.
distributions. 1, the quality of the socket design is assessed after the
socket is manufactured and fitted. The assessment is
1.1. Finite element analysis done by prosthetist’s observation and the feeling of the
amputee. The FE method itself cannot assess the socket,
A commonly used numerical analysis technique in because the quality of fit depends not only on the mech-
biomechanics is the finite element (FE) technique [4], a anical interactions between the residual limb and the
computational technique for structural stress/strain socket, but also the tolerance of the residual limb tissues
362 M. Zhang et al. / Medical Engineering & Physics 20 (1998) 360–373

Fig. 1. Flow chart for current socket design and the design with FEA.

to external loads. However, FE analyses can provide teristics of the problem must be captured and
information on the biomechanical interaction between simplifications must be justified relative to the definition
the residual limb and socket and the stresses within the of the problem.
soft tissues. Assessment of the design can therefore be Today, most of the FE work can be facilitated by read-
done before the socket is manufactured by evaluating the ily available commercial FE computer software [21].
FE results, such as the maps of pressure distribution, The art in FE analysis now is really concerned with the
shear stress and interfacial slippage. The design can be development of the FE model, rather than the actual per-
revised until satisfactory results are achieved. Similarly, formance of the calculation. FE packages can give very
a desired pressure pattern can be input into the FE colourful results literally. It is very easy to produce
model, analyses can then be carried out to provide the impressive presentations with beautiful graphics to con-
required socket rectifications needed to produce that vince non-experts who are generally not trained to criti-
pressure pattern. cally review an FE analysis for a particular problem. The
There are two main aims in the development of an FE results may even be misleading, if the model is not a
analysis. First, FE analyses facilitate our understanding valid representation of the real situation. For this reason,
of the interfacial biomechanics. The most powerful the development of the FE model for the stump/socket
ability of FE analysis, compared with experimental structure must be carried out very carefully, assessed
measurements, is the speed with which complicated situ- very critically and validated experimentally.
ations may be analysed parametrically. It is desirable to Silver-Thorn et al. [3] recently presented a review of
investigate systematically the effects of many parameters prosthetic interface stress investigations and concen-
without the need to build prostheses. Secondly, FE trated on the methods of experimental measurements and
analysis may become a useful tool in CAD/CAM socket computational model establishment. Zachariah and San-
systems. The FE method has been identified as a poten- ders [22] reviewed the interface mechanics in lower-limb
tial tool for the theoretical investigation of the stress dis- external prosthetics and discussed the issue of FE model
tribution at the residual limb/prosthetic socket interface. sensitivity analysis and experimental comparison. In that
Combined with the development of CAD/CAM in pros- paper, most FE models on below-knee (BK) prostheses
thetic sockets, FE analysis as a part of CAD process can and one above-knee (AK) model were reviewed. The
provide the prosthetists with quantitative feedback rec- objectives of this paper are to provide a summary of the
ommending rectification, or even better, create a new work, including seven BK and seven AK models, carried
socket shape to obtain a desired pressure distribution out during the first decade between 1987 and 1996 and
[20]. to suggest further research in development of FE model-
It is noted that the results of an FE analysis depend ling of the residual limb/socket system.
very much on the model constructed to represent the
structure. The accuracy of the modelling relies largely
on accurate constitutive modelling of the material 2. Overview of the problems
properties, geometric data, loading characteristics, and
boundary and interfacial conditions. Models are abstrac- The fitting is realised by a socket, which makes direct
tions of reality and are used to simplify the actual prob- contact with the residual limb and forms a critical inter-
lem. The essence of modelling is that the salient charac- face. Since direct skeletal fixation of the prosthesis is
 M. Zhang et al. / Medical Engineering & Physics 20 (1998) 360–373 363

not yet a practical solution, loads must be transmitted mation, i.e. the relative movement between the bone and
through the intervening soft tissues. Thus, a heavy the socket [7,11].
biomechanical demand is put on the residual limb soft
tissues, although they are not suited for such load bear- 2.1. Complexity of the problem
ing. To achieve satisfactory load transmission, mainte-
nance of stability of the prosthesis on the residual limb First, the residual limb has a three-dimensional com-
and sufficient control over the prosthesis, the socket plicated and subject-dependent geometry. Simplification
shape is currently designed by modifying the residual to a two-dimensional or symmetric problem in an
limb geometry. The socket is often fitted with a soft liner attempt to simulate the problem accurately is difficult,
and sock. The components in a typical stump/socket sys- although such a simplification can be useful in prelimi-
tem involve, from outside to the centre, the hard socket, nary studies.
soft liner, socks, soft tissues and bone (Fig. 2), although Secondly, the problem is nonlinear. In common static
the soft liner is sometimes not used. problems, nonlinearity results primarily from three
The basic requirement for FE analyses is to character- sources: (1) geometric nonlinearity, due to large defor-
ise the mechanical interaction between the residual limb mation of materials; (2) boundary nonlinearity, such as
and socket for a given socket fitting. The most important slip/friction contact problem; and (3) material nonlin-
characteristic of the interaction is the stress distribution earity. In a stump/socket system, the soft tissues of the
(including pressure and shear stress) over the entire limb are often subjected to a large deformation during
residual limb surface. The stress distribution within the donning and walking [11]. Slippage exists at the inter-
soft tissues is also important information in socket fit- face [11], although excessive slip is to be avoided in a
ting. Other factors which may determine the fitting qual- good fit. Biological soft tissues obviously exhibit nonlin-
ity are the slippage of the socket over the residual limb ear properties.
and the interfacial connecting stiffness. Here, connecting Furthermore, biological soft tissues are not only non-
stiffness means the ability of the interface to resist defor- linear, but are also viscoelastic, time dependent, aniso-
tropic and are actively being remodelled [23]. Residual
limb tissues consist of skin, subcutaneous fat and the
underlying muscles. The properties of the muscles are
more or less affected by muscle contraction [23]. The
significance of force generation on and the interactive
nature of the active muscles with the passive soft tissue
components have not been well established.
The problem is dynamic, involving not only variable
externally applied loads, but also the inertia loads of the
body segments and artificial components and the time-
dependent properties of materials.
In principle, all these details of the structure and the
procedure should be taken into account in the finite
element model by using sufficiently small elements and
nonlinear iteration solution processes. In practice, this is
not feasible because of limited computer capacity, even
with the largest supercomputer. For such a complicated
system, the FE model has to be established under certain
assumptions, such as idealised or simplified geometry,
material properties and boundary conditions, to reduce
the problem size possibly with an acceptable accuracy.

3. Basic features of previous models

During the last decade, a number of models have been

created for BK and AK residual limb/socket systems
based on different assumptions. Many attempts have
been conducted on the feasibility of the methods. Fig. 3
displays the basic features of these models. Table 1 gives
the size of the model, FE code and computer used in
Fig. 2. Components of residual limb/prosthetic socket system. these studies.
364 M. Zhang et al. / Medical Engineering & Physics 20 (1998) 360–373

Fig. 3. Previous FE models for residual limb and prosthetic socket.

Table 1
Previous model features

Ref. Model size Code Computer CPU time Validation

Elements Nodes

BK Steege [5,24,26] 1017 + 340 1283 SAP IV Cyber 845 n/s prs
Steege [24,25] 1578 1976 GIFT IBM AT 5–11 h prs
Reynolds [7,28] 972 1458 PAFEC Micro VAX II 108 min prs
Sanders [8,29] 840 795 ANSYS VAX 16 min prs + shr
Silver-Thorn [9,30,31] 1688 2221 MARC n/s n/s prs
Quesada [10,32] 655 636 SAP 80 n/s n/s no
Zhang [11,33–36] 1854 + 450 2421 ABAQUS Convex C3840 15 min/step prs + shr

AK Krouskop [6] n/s n/s ANSYS Cyber 175 ⬍ 25 min no

Seguchi [12] 694 683 ANSYS PC n/s no
Brennan [13] 2672 3485 MARC n/s n/s prs
Torres-Moreno [14,37] 2628 1962 ABAQUS Silicon Gr 4D/35 n/s prs
Mak [15] 960 n/s NISA II IBM PC n/s no
Lee [16,20] 2096 2611 ABAQUS Sun Sparc n/s prs
Zhang [17,39,40] 144 + 30 263 ABAQUS Sun Sparc 20 40 s/step no

Key: n/s, not stated; prs, pressure; shr, shear.

3.1. BK models meshed into 3D solid elements with linear elastic proper-
ties. The liner was modelled as either solid elements or
Steege et al. [5,24–27] established the first FE model as elastic springs.
of the BK residual limb and prosthetic socket to discuss Reynolds [7] and Reynolds and Lord [28] established
whether interfacial pressures could be predicted. The an axisymmetric model with an idealised geometry to
internal and external geometry of the residual limb was predict the effects of friction on the skin pressures by
based on transverse computer tomography (CT) scans. comparing the pressures predicted under two extreme
The socket was assumed to be rigid and to be fully con- assumptions: either no slip or frictionless at the
nected with the stump. The bone and soft tissue were skin/socket interface. The model was also used to inves-
 M. Zhang et al. / Medical Engineering & Physics 20 (1998) 360–373 365

tigate the effect of distal-end load bearing on the press- the use of interface elements to simulate the slip/friction
ure distribution on the residual limb surface by model- conditions at the skin/liner interface. The analyses were
ling no distal-end loading and full distal-end loading. performed in two steps corresponding to the two stages
Their 3D model was established based on a real limb of deforming the residual limb: from the free limb state
geometry obtained from digitising the loose plaster wrap to that being donned into the rectified socket shape with-
of a residual limb and the bone shape obtained from out any additional external load and from the donned
digitising a set of biplanar X-rays. The soft tissues were unloaded state to the fully loaded case.
meshed into solid elements with different Young’s mod-
uli assigned to the elements in different regions, simulat- 3.2. AK models
ing the site-dependent stiffness of the tissues around the
residual limb. The bone and socket were modelled as Krouskop et al. [6] established the first FE model for
rigid boundaries without distal-end loading, while socket the AK socket. The stump shape was measured using
rectification was performed by giving a radial displace- two diametrically opposed contacting/retracting probes.
ment to the nodes on the external surface of the limb. The bone position was set using information gathered
The soft liner was not considered. with the ultrasonic apparatus. The elastic moduli meas-
Sanders [8] and Sanders and Daly [29] established a ured by ultrasonic method were assigned to different 3D
3D linear FE model with assumptions of no slip at the solid elements. A “desired interfacial pressure pattern”
interface and no distal-end loading without consideration was applied to the residual limb and then the displace-
of socket rectifications. The bone was taken to be a rigid ments of the nodes on the skin surface required to achi-
boundary. The soft tissues were modelled as solid eve that pressure distribution were predicted. These sur-
elements with two layers, skin plus fat, and muscle. The face nodes displacements would be relevant to the
liner was meshed into linear solid elements. The socket socket rectification.
was meshed into quadrilateral shell elements, while a Seguchi et al. [12] modelled the acrylic socket without
portion of the shank was meshed into beam elements. considering the characteristics of bulk soft tissue. A
The load was applied at the distal end using data meas- socket with a uniform thickness was meshed into quadri-
ured by a pylon transducer fitted in the shank. Varying lateral thin shell elements. It was not clearly mentioned
loads were applied to predict the transient stress patterns how the load and boundary conditions were applied and
during walking. how the soft tissue interacted with the socket.
Silver-Thorn [9] and Silver-Thorn and Childress Brennan and Childress [13] created a 3D AK model
[30,31] created a 3D model with an approximate generic using the geometry obtained from digitising the wrap
limb geometry. The internal bone was modelled as a cast and from CT scans. The bone and the soft tissue
rigid boundary. The soft tissues, soft liner and socket were meshed into solid elements. The socket was mod-
were meshed into solid elements. Socket rectifications elled as a rigid boundary. Socket rectification was
and an external load were applied to the socket. Para- applied to the skin surface nodes. A load of half a body
metric analyses were conducted for an unrectified and a weight was applied to the proximal side of the hip. The
PTB rectified socket. model was used to investigate the pressure distributions
Quesada and Skinner [10,32] established a 3D model with three kinds of sockets, namely the Ischial–Gluteal
of an exoskeletal patellar tendon bearing below-knee Bearing socket, Partial IRC socket and Total IRC socket.
socket to predict the stresses in relation to some para- Torres-Moreno [14] and Torres-Moreno et al. [37]
metric variations. In this model, the bone was a rigid established an FE model utilising the 3D geometric con-
boundary, and the whole prosthesis without fitting a soft figuration of the internal tissues obtained from nuclear
liner was meshed. The soft tissues were assumed to be magnetic resonance (NMR). The bone and soft tissues
a set of elastic springs. It was assumed that there was were modelled by triangular prisms and hexahedrons,
no slip at the skin/socket interface. The load was applied respectively. The model was divided into two parts to
at the heel of the foot. conform to tissues both within and distal to the region
Zhang [11] and Zhang et al. [33–36] created a 3D near the brim. Prescribed nodal displacements and inter-
nonlinear model, using the actual limb geometry and tak- face pressures were applied for proximal and distal
ing into consideration the socket rectification and the parts, respectively.
slip/friction conditions between the skin and liner. In this Mak et al. [15] established axisymmetric and 3D AK
model, the bone, soft tissue and liner were meshed into models based on the mid-part of the residual limb
solid elements. The socket was taken to be a rigid shapes, obtained from CATSCAN images. The soft
boundary. Different Young’s moduli were assigned to tissues were meshed into hexahedron and wedge
the soft tissues at different regions. External load was elements. The bone was rigidly fixed. Radial displace-
applied to the proximal side of the bone. Socket rectifi- ments were prescribed at the nodes on the skin surface
cations were applied to the external surface of the liner. to simulate the socket geometry.
One of the main differences from the other models lies in Lee et al. [16,20,38] created a 3D AK model by mesh-
366 M. Zhang et al. / Medical Engineering & Physics 20 (1998) 360–373

ing bone and the soft tissues into eight-node hexahedron either at the distal end or proximal end. Sometimes
elements, based on the limb geometry obtained from interface pressures are directly prescribed.
magnetic resonance imaging (MRI). The socket was
modelled as a rigid boundary. The load was applied to
the proximal end of the femur.
Zhang and Mak [17,39,40] established a 2D nonlinear 4. Geometry
model based on a sagittal plane profile of an AK stump
and socket with consideration of socket rectifications and To create an FE model, first the geometry of the mod-
the slip/friction interface conditions. In the model, the elled objects needs to be obtained, including the shapes
bone, soft tissues and socket were meshed into quadrilat- of the free residual limb, internal bone, the socket and
eral elements. Socket rectifications were applied to the the liner if involved. Although some models were based
external surface of the socket by prescribing the radial on ideal and simplified geometry, an accurate description
displacements. The external force was applied to the should be based on an actual geometry.
Traditional CAD/CAM socket systems usually
proximal end of the femur. The ABAQUS interface
involve a shape digitiser or laser scanner [41–45], which
elements were used to connect the skin and the socket.
can get the shape of the stump and socket, but only the
Various coefficients of friction were given to investigate
surface information. Some techniques, such as X-ray [7],
the effect of friction on the load transfer. The effects of
CT [46], MRI [47] and ultrasound [48–53], are also used
three distal-end boundary conditions, namely no distal-
to measure the geometry of the stump surface and the
end contact/loading, full distal-end contact and air-
internal bones. However, there are limitations with each
cushion loading at the distal-end, were investigated. The
method, as reviewed by Silver-Thorn et al. [3], for
air-cushion loading was modelled by using hydraulic example, X-ray and CT techniques expose the subject to
fluid elements to simulate a sealed air pocket at the dis- ionising radiation.
tal end.
In summary, the following features have been
included in the existing models.
5. Material properties
1. Stump sock, a thin soft layer, is not often considered
in the FE modes. No rationale is offered for this sim- Table 2 presents the material properties used in the
previous models. All the FE models developed so far
plifying assumption beyond the necessity to reduce
assumed the bone, soft tissues, liner and socket to be
the scope of the problem.
solid-like continua. They were also assumed to be iso-
2. The main objects analysed are the soft tissues and the
tropic, uniform or locally uniform, and linearly elastic,
liner, if fitted, or the socket, which form the contact
although Steege and Childress [25] had attempted to use
interface. The soft liner was not fitted in the existing
nonlinear properties, a Mooney–Rivlin model, for the
AK models.
soft tissues. It seems not to be a problem to obtain the
3. Soft tissues consist of skin, muscle and fat. However,
mechanical properties of prosthetic components and
they are normally considered as one object with sim- bone. However, the problem remains to obtain the in
plified mechanical properties, such as isotropic, uni- vivo mechanical properties of soft tissues.
form and linearly elastic, but location-dependent, Soft tissues are structures of irregular geometry and
considering the different stiffness over different sites. complex components, with complex material properties
The soft tissues were either meshed into solid such as non-homogeneity, anisotropicity, viscoelasticity
elements, or modelled as an elastic spring support. and time dependency. Many models treat the soft tissues
4. The hard socket and the bone, which are much stiffer as fibre-reinforced composites [54], poroelastic and
than the soft tissues and liner, are often simplified as multi-phasic elements [55–57]. These complex models
rigid boundaries. However, for special requirements, are particularly used in developing an accurate descrip-
they, or one, is often meshed. tion of the mechanical properties of soft tissues. To com-
5. The friction/slip contact conditions between the plete an FE model, however, we are much more inter-
stump and socket have been considered in some mod- ested in the mechanical properties of bulk soft tissues of
els. Most other models assumed either no-slip or fric- the residual limb, especially measured in vivo. Although
tionless interface. A more precise treatment should the soft tissues can be divided into skin, muscles and
involve a frictional contact problem, using some spe- fat, etc., most models considered them as a whole. The
cial elements, such as interface element. bulk properties are required in the FE model. For an
6. Socket rectifications are considered in some models isotropic elastic material, two material constants, viz.
by giving a radial displacement to the nodes on the Young’s modulus and Poisson’s ratio, are required.
external surface of the limb or the liner or the socket. To obtain the material constants of the soft tissues,
7. The external loads are applied normally at a point the majority of studies reported in the literature use
 M. Zhang et al. / Medical Engineering & Physics 20 (1998) 360–373 367

Table 2
Material properties used in previous models

Ref. Bone Soft tissue Soft liner Hard socket

E (GPa) ␯ E (kPa) ␯ E (kPa) ␯ E (GPa) ␯

BK Steege [5,24,26] 15.5 0.28 60 0.49 380 – – –

Steege [24,25] 15.5 0.28 60 0.49 380 0.49 – –
Reynolds [7,28] – – 50–145 0.45 – – – –
Sanders [8,29] – – 3.5–131 0.49 1800 0.39 1 0.35
Silver-Thorn [9,30,31] – – 60 0.45 380 0.49 1 0.3
Quesada [10,32] – – 247–2490 – – – 14 0.13
Zhang [11,33–36] 0.01 0.49 160–260 0.49 380 0.49 – –
AK Krouskop [6] – – n/s n/s – – – –
Seguchi [12] – – – – – – n/s n/s
Brennan [13] 15.5 0.28 60 0.49 – – – –
Torres-Moreno [14,37] 17.6 0.3 24.6 0.4999 – – – –
Mak [15] – – 60 0.49 – – – –
Lee [16,20] 15.5 0.3 35–120 0.4999 – – – –
Zhang [17,39,40] 15 0.3 150 0.45 – – 15 0.3

Key: n/s, not stated.

indentation tests [7,9,58–62]. Some electromechanical developed for investigation of articular cartilage proper-
devices were developed to perform such tests and to ties, were conducted with the assumption that the tissue
obtain the force–displacement relationship. Recently an was a semi-infinite elastic medium. The formula [68]
ultrasonic indentation system for assessment of mechan- often used for a rigid, plane-ended and cylindrical
ical properties of soft tissues in vivo was developed by indenter was E = p/(2.677w0a), where E is the Young’s
Zheng et al. [63,64], which has a potential for clinical modulus; p the load applied by the indenter; w0 the depth
use. A pen-size hand-held probe was composed of an of indentation; and a the radius of the indenter. This
ultrasound transducer and a load cell. The thickness and solution has not considered the effect of the finite thick-
deformation of the soft tissue layer during indentation ness. Hayes et al. [69] obtained a theoretical solution to
can be determined from the ultrasonic echo. The force the problem associated with axisymmetric indentation,
response can be measured by the low profile load cell. where the elastic layer was bonded to a rigid foundation.
Other techniques reported for measuring the soft For a rigid, plane-ended, cylindrical, frictionless
tissue mechanical properties include other use of ultra- indenter, the Young’s modulus could be obtained by
sound [65,66] and MRI [67]. A non-invasive in-vivo E = p(1 − ␯2)/[2aw0k(a/h,␯)], where E is the Young’s
method has been reported to measure the mechanical modulus; p the load applied by the indenter; w0 the depth
properties of the bulk soft tissues by a pulsed ultrasonic of indentation; a the radius of the indenter; h the thick-
Doppler system [65,66]. An ultrasonic transducer was ness of the measured layer; ␯ the Poisson’s ratio; and k
used to measure internal displacements resulting from a geometry and material dependent factor. A table of
external acoustical perturbations. factor k was obtained for several values of ratios of
More attention should be paid to how to extract the indenter radius to tissue thickness a/h and the Poisson’s
Young’s modulus from the force–displacement relation- ratio ␯. Hayes’ solution could be used to consider the
ship [11]. General indentation tests give a force–dis- layered effect and the lateral expansion of the tissue dur-
placement curve, the slope of which is not the Young’s ing indentation. However, Hayes’ solution is based on
modulus. Two methods have been reported so far to the assumptions of infinitesimal deformation and fric-
derive the effective Young’s moduli from the load–dis- tionless indenter interface. Recently, Zhang et al. [70]
placement curve. investigated the influence of friction and large defor-
One method reported by Reynolds [7] was by match- mations during indentation tests of a layered material
ing the experimental force–displacement curves with the bonded to a rigid foundation using FE methods. Modi-
curves predicted by linear FE modelling of indentation fied k values were presented for the calculation of the
into a layer of tissue with idealised mechanical proper- effective Young’s modulus in the same manner as
ties. The modulus which could give a curve close to the Hayes’ solution, except that the results presented have
experimental one was assumed to be the equivalent taken into consideration large deformations and friction
modulus. at the skin and indenter surface.
The other method is a mathematical formula transfer. Poisson’s ratio is another material parameter often
Early theoretical interpretations of indentation data, needed to input for the elastic properties. In general, soft
368 M. Zhang et al. / Medical Engineering & Physics 20 (1998) 360–373

tissues are assumed to be incompressible, that is the require master surfaces to be smooth. Currently, smoo-
Poisson’s ratio is 0.5. Since it is difficult to ensure the thing a 3D deformable surface is not available.
convergence of the computation, the Poisson’s ratio is
sometimes assumed to be 0.4 to 0.499. Some previous 6.2. Socket rectification
models in which the soft tissues were modelled as elastic
spring supports [10,26] dealt with the soft tissues as fully The main aim of socket rectifications is to distribute
compressible without taking into consideration the lat- the load over the desired regions of the residual limb.
eral expansion of the tissues. Vannah and Childress [71] So the simulation which aims to predict the load distri-
found the force–displacement relationship of soft tissues bution must take it into consideration. Failure to impose
held in a narrow space, such as the residual limb tissues the required socket rectification will make the model
between the bone and socket, was sensitive to the Pois- unrealistic. In practice, the stump is donned into the
son’s ratio. socket with a rectified shape. Owing to the difficulties
involved in modelling such a donning procedure with a
large slippage, previous models impose the socket recti-
6. Boundary conditions fication via displacement boundary conditions
[7,9,11,17,26]. Usually the differences between the
6.1. Friction/slip conditions between the residual limb socket and stump (amount of the rectification) shapes
and the socket were applied to the nodes on the external surface by
applying the radial displacement boundaries.
Some models assumed there is no slip between the
skin and the socket. However, work done by Quesada
and Skinner [10] revealed that there was a large ratio of 7. Load states
shear stress to normal stress, which was much larger than
the coefficient of friction between the skin and socket. An FE analysis requires a numerical description of all
The necessity for consideration of the friction/slip inter- external loads applied to the structure. These loads are
face was further indicated by Zhang et al. [11,36,72]. In usually variable and are not always precisely known.
their model, interface elements were used to simulate the The challenge in FE analysis is often how to simulate
friction/slip condition between the skin and the liner for the loading situation appropriately.
BK model, and between the skin and socket for AK In previous models, the load applied to the model has
models. The analysis showed that small local slippage normally been simplified to one point, applied to the
always exists, even though large global slippage can be interface either at the proximal site by applying the load
avoided. If the slip was to be modelled as a tissue defor- to the bone, or at the distal end applied to the end of
mation in the non-slip model, it will overestimate the the prosthesis. The loads have often been obtained either
shear stresses, while underestimating the maximum experimentally measurements or have been assumed
pressure by at least 50%. A non-friction model ignores loads relating to the body weight. Sanders [8] measured
the shear action, but overestimates the maximum press- the shank forces using strain-gauge instrumentation on
ure by about 100%. the shank pylon. The other method for obtaining the
When surfaces are in contact they usually transmit external force involves the measurement of the ground
shear as well as normal force. Sliding will occur if the reaction forces and the joint angles.
magnitude of the shear stress reaches a certain value. To In modelling the donning procedure, particular com-
model the friction/slip contact problems, some special plications are encountered in experiencing a large slipp-
methods have been developed in some commercial FE age between the structures. Torres-Moreno [14] divided
packages. Contact problems can be divided into infini- the donning event into two distinct stages, proximal fit-
tesimal sliding or finite sliding, between the contacting ting and distal fitting. The first stage was fitting of the
surfaces over finite areas, or between discrete points. The proximal brim portion by prescribing the relevant nodal
contact bodies can be deformable or rigid. To model the displacements and solving for the pressure at the surface
specialised cases, some special FE elements have been of the residual limb. The second stage was to describe
developed, such as gap element, stress/displacement the load transfer characteristics of the residual limb by
slide plane or slide line element, tube to tube contact imposing a series of prescribed pressure patterns. Nor-
element, stress/displacement interface element and mal pressure was applied to the external surface located
stress/displacement rigid surface. Because infinitesimal- in regions distal to the brim. Zhang [11] and Zhang and
sliding formulation does not require monitoring of the Mak [17] performed the analyses in two steps corre-
contact between a slave node and the entire master sur- sponding to the two stages of deformation of the limb
face, it is computationally less expensive than finite-slid- tissues: from the free limb state to that being donned into
ing contact. The cost savings are most dramatic in 3D the rectified socket shape, and from the donned shape to
contact problems. In addition, this formulation does not the fully loaded case. In the first step, pre-stresses were
 M. Zhang et al. / Medical Engineering & Physics 20 (1998) 360–373 369

obtained by giving displacement boundary conditions to tures. The predicted resultant shear stresses were less
the external surface of the liner or the socket. In the than the experimental values at all measured sites. Best
second step, the analysis was executed by maintaining matches were achieved at postero-distal and antero-
the rectified shape and the pre-stresses, then applying proximal sites. Consistent mismatches were seen. Zhang
external forces due to weight bearing during gait. [11] and Zhang et al. [73] measured the interfacial press-
ure up to 283 kPa, and shear stress up to 44 kPa, at the
7.1. Iterative solution for large deformations critical regions of BK sockets using triaxial transducers.
The FE analytical pressures up to 226 kPa were, on aver-
To study the effect of large deformations, an FE age, 30% lower than those measured. The difference
model is normally analysed in an iterative manner, might result from the fact that the load applied to the
where the displacements and loads are applied in a series prosthetic limb in the FE model was static and equal to
of small increments. Iteration with each increment the body weight, whereas the actual load during walking
ensures that nonlinear equilibrium (convergent solution) was dynamic and larger than the body weight. The com-
is achieved. On each increment, iterated as an individual parison suggests that the dynamic analysis is necessary
problem, the geometry of the model is updated based on in future FE modelling studies. Owing to the application
the deformed geometry of the previous increment. The of the interface elements to simulate the slip/friction con-
stiffness matrix required for the solution of the model is ditions at the stump/socket interface, the shear stresses
dependent on the spatial orientation and location of the have better agreement, difference in the magnitudes
elements within the model. Owing to the continual being 10%, on average, and the directions of the result-
change of these terms, it gives certain geometric nonlin- ant shear stresses are identical to the experimental
earity to the model. To do this using a commercial FE results.
package, a nonlinear geometry option is normally In general, comparison of the stresses predicted by FE
required. analysis and those measured suggested that the results
can be in the same range. However, one-to-one corre-
spondence has not been achieved. The validation is lim-
8. Validations of the FE models ited to a number of the points where stresses can be mea-
sured.
So far, the major method to validate the modelling is
to compare the stresses predicted by the FE model with
those measured experimentally. Owing to the limitation 9. Main results
of the clinical measurement, the comparison can only be
done at a limited number of sites. A comparison of the The results predicted depend largely on the accuracy
FE models with their corresponding clinical experi- of model establishment and the input parameters. How-
mental measurements is presented below. ever, the parametric analyses still give useful infor-
Steege et al. [24] predicted normal stresses on a BK mation for the clinical socket design.
stump between 0 and 120 kPa, and the experimental
measurements were between 0 and 128 kPa. Silver- 9.1. Socket rectification
Thorn [9] predicted that the pressures on a BK socket
varied from 0 to 275 kPa, whereas those measured were Socket rectifications applied to the model largely
0 to 205 kPa. Brennan and Childress [11] predicted affect the distribution of the interfacial stresses. For the
pressures for AK sockets ranging from 0 to 150 kPa, BK socket, UCL socket rectification patterns were often
whereas their measurements on AK subjects varied from used in the modelling [7,9,11]. As expected, the socket
0 to 40 kPa. Although the subject and sockets used for rectifications cause an increase in pressures over the
experimental and FE analytical studies were the same, (load-tolerant) areas with negative rectification, and
the significant peak areas predicted by FE models were decrease in pressure over the (load-sensitive) areas with
not verified experimentally. Lee et al. [47] reported that positive rectification and the maximum pressure often
the pressures predicted on AK subjects ranged from 1 occurs at the heaviest rectified area—the patellar tendon.
to 60 kPa, whereas measured pressures ranged from 11 Silver-Thorn’s models [31] show that modification of
to 77 kPa. Sanders [8] predicted the normal stress for two primary rectification areas, the patellar tendon and
a BK socket ranging from 0 to 159 kPa, whereas the popliteal regions, may significantly influence the press-
measurements during walking were from 0 to 169 kPa. ure experienced at other areas of the residual limb. In
Sanders predicted the shear stress to be between 0 and Reynolds’ model [7], the “heavily rectified socket”
41 kPa, but the experimental results showed a range resulted in increased pressures over the entire stump,
from 0 to 9 kPa. Qualitatively, shapes of analytic and more obviously over the positive rectified areas, com-
experimental interface stress waveforms were similar in pared with a lightly rectified socket. In Zhang’s models
that they were double-peaked with some distinct fea- [11,36], different degrees of rectification were analysed.
370 M. Zhang et al. / Medical Engineering & Physics 20 (1998) 360–373

The results showed that more heavy rectification will by the distal end if there is no gap between the stump
cause an increase in pressures over the load bearing end and socket. This value could be larger than the toler-
areas, such as patellar tendon, popliteal and lateral and ant ability of the stump-end. The real situation will be
medial aspects of the tibia, and a decrease in pressures between the two extreme cases: no distal-end loading
over the load sensitive areas, such as distal anterior tibia, and full distal-end loading, depending on the gap
tibial crest, fibular head and end of stump. In Zhang’s remaining between the stump and the socket. The sealed
models, it was also found that the PTB socket, following air at the distal-end will produce a positive pressure
the UCL PTB rectification map, seems to generate over- which will help to support roughly 10% of the vertical
concentrated pressures in some areas. The pressures load.
applied to 56% of the interface contact area are lower
than 40 kPa. A small area suffers pressures of more
9.4. Interfacial friction
than 160 kPa.

9.2. Socket and liner materials Reynolds [7] established axisymmetric models with
idealised BK geometry under assumptions of frictionless
Several investigators studied the effects of changes in or totally rough interface between the skin and the socket
the mechanical properties of the socket and liner to explore the friction effect on the load transfer. The
materials on the interfacial stresses for below-knee pro- results show that there are dramatic effects of friction
stheses. Quesada and Skinner [10] concluded that on the load distribution and interfacial stiffness. Zhang
interfacial pressures could be reduced by the use of and his colleagues [11,17,35,72] investigated the friction
lower modulus prosthetic materials and thinner fabri- effects by using 3D BK and 2D AK models. In these
cation, thereby providing greater deformation of the models, interface elements were used to connect the skin
prosthesis with less deformation of the soft tissues. How- and socket interface with various coefficients of friction
ever, this may lower the interfacial stiffness and influ- assigned to the interface. The results reveal that the fric-
ence the control of the prosthesis. Zhang [11] investi- tion between the skin and the socket plays a critical role
gated the influence of the liner stiffness on the stress in determining the interfacial load transfer character-
distribution. The results showed that the average press- istics. Adequate friction is useful not only to prevent slip
ure over the entire stump surface was not sensitive to but also to assist in load transfer. The lower the friction,
the liner stiffness, although the pressures in some critical the larger the pressure needed to support the body
areas were changed slightly. Silver-Thorn and Childress weight. This is further supported by experimental evi-
[31] found that the interfacial pressure distribution was dence. Large friction between the skin and the socket
moderately sensitive to both socket and liner material may damage the skin due to a large interfacial shear
properties. Increased liner stiffness may result in stress; while a small coefficient of friction will require
increased pressure in the patellar tendon region and large interfacial pressure to support the body weight.
decreased pressure over the tibial crest. Thus, low frictional interface should always not be taken
as more desirable than higher frictional interface. The
9.3. Distal-end loading investigation suggested that a proper choice of surface
friction may balance these two requirements of good
Reynolds [7] investigated the effect of distal-end load- limb control and minimisation of interfacial hazards.
ing using idealised axisymmetric BK models. The distal-
end loading effect was investigated by comparing the
9.5. Stump shape and soft tissue stiffness
pressures over the stump surface with and without distal-
end contact. The results show obvious large pressures
over the side surface of the stump and low interfacial An increase in residual limb length corresponds to an
stiffness in the no end-bearing model. Zhang and Mak increase in the overall surface area available for load
[17,40] explored the effect of distal-end boundaries on bearing. Quesada and Skinner [10] found that a 2 cm
the load transfer in AK sockets by modelling three dis- decrease in stump length increased distal stresses by
tal-end boundary conditions: (1) no distal-end loading 18%. A change in stump length from 84 to 181 mm
by giving a gap between the skin and the socket to decreased shear stresses at the patellar tendon, lateral
ensure that there is no distal-end contact during loading; and medial tibial flares, and fibular head and decreased
(2) full distal-end boundary without a gap between the the pressures at the critical area of the patellar tendon.
skin and the socket at the distal end area and with the The FE analyses also investigated the influence of the
skin in full contact with the socket; (3) air-cushion taper of the residual limb. Increased global taper resulted
boundary simulating a suction socket by assuming there in changes in the interfacial pressures: decreased press-
is a sealed air cavity over the distal end. It was found ures in the regions of the fibular head, tibial crest, and
that more than 60% of the total vertical load is supported patellar tendon [31].
 M. Zhang et al. / Medical Engineering & Physics 20 (1998) 360–373 371

9.6. External forces and socket alignment material properties and boundary and loading conditions.
The analyses have provided information on the stress
The effects of socket alignment have usually been distribution and slippage at the stump/socket interface
investigated by modifying the direction of the external and parametric effects. These advances have no doubt
force applied to the socket. Reynolds [7] predicted the helped to lay important groundwork for future develop-
pressures for two loading cases: neutral alignment and ments. However, we must realise that the greater part of
bench alignment (5° flexion in the sagittal plane and 5° the efforts in this first decade of the FE analyses in the
abduction in the frontal plane). The resulting alterations prosthetic socket biomechanics were method-oriented
to the pressure distribution were not large; maximum dif- rather than problem-oriented. All we can say at present
ference in local peak pressure was about 13% in the is that the technique is promising and that a considerable
anterolateral region. Sanders [8], however, found sub- amount of work is still required before complete confi-
stantial changes in the pressure distribution as a result dence in the method is obtained.
of alterations to the alignment of the order of 12°. Further development of FE modelling of the stump
and socket system should address two primary objec-
tives. First, FE analyses should continue to increase the
10. Summary and future trends understanding of the interfacial biomechanics and should
investigate systematically the effects of various design
We believe that FE analysis will offer a strong poten- parameters. To achieve this aim, progress should be
tial in making significant contributions to the improve- made towards more precise modelling of the problems
ment of lower-limb prosthetic delivery. The develop- at hand. Secondly, FE modelling will be a tool in
ments achieved in the first decade have shown this CAD/CAM socket systems, to provide prosthetists with
capability. Although the earlier analyses have not solved quantitative feedback recommending socket rectifi-
the problems completely, they have indirectly demon- cations. Applied research should be directed towards
strated the feasibility of using new methodology in integrating together FE analysis and the CAD/CAM sys-
understanding the biomechanics involved. As modelling tem.
continues to be refined, more precise formulations of One of the main problems confronted in FE analysis
material properties, boundary conditions and loading is the modelling of soft tissues, which have complex
characteristics will become issues of high priority. properties and may undergo large deformation. Lack of
Development of FE models can be phased into three an accurate description of the nonlinear properties of the
generations. The first generation of the modelling is lin- soft tissues, such as nonlinearity, time-dependency and
ear analysis established under linear assumptions, that anisotropicity, will prevent the development of a precise
is, linear material properties, linear geometry with model. An approximately biphasic stress–strain or hyp-
infinitesimal deformation, and linear boundary con- erelastic curve may be a more accurate description of
ditions without considering the friction/slip conditions. their mechanical response than a linear model. The
Models in this generation make a relatively small elements developed in commercial FE packages have
demand on CPU time. Most models developed in the limitations, such as a limited aspect ratio and angle. If
first decade are in this stage. The second generation can the deformed element exceeds the ranges, the analysis
be called nonlinear analysis, taking into consideration will be purged. More robust computational schemes for
the nonlinear effects, involving nonlinear material special biomechanical applications need to be developed.
properties, nonlinear geometry, and nonlinear boundary Simulation of the donning procedure is another dif-
conditions. The analysis required an iteration procedure. ficulty. To date, socket rectification is imposed by
While relatively more CPU time is required, more accur- prescribing the displacement boundary condition at the
ate solutions can be obtained. Some nonlinear analyses nodes on the socket or liner. This means that the amputee
have been presented in more recent literature puts the limb into the liner if fitted and an unrectified
[11,17,33,34,36]. The third generation of the analyses socket first, then the socket deforms into the rectified
should be dynamic. This analysis should consider not shape. In a real situation, however, the amputee normally
only variable external loads, but also inertial effects. So puts on the liner first, then dons the rectified socket. It
far, dynamic models for the stump/socket system have is still hard, at this time, to model the large slip during
not been reported, although variable external loads have the donning procedure.
been applied in some analyses. Other developments are required to better simulate the
The models developed in the first decade have taken slip/friction interface conditions between the skin and
into consideration actual limb geometry, socket rectifi- the socket, to describe the geometry of stump and bone
cation, nonlinear interface conditions, large deformations precisely, and to optimise the meshing structure. Experi-
and variable external loads. Various assumptions and mental validation of the modelling is also required.
simplifications are made in the model development to To use FE analyses in CAD/CAM systems, applied
simplify the actual problem with complex geometry, research should be carried out to link the two together.
372 M. Zhang et al. / Medical Engineering & Physics 20 (1998) 360–373

First, experimental techniques need to be streamlined, so for analyses of stresses within above-knee stumps. Proceedings
as to directly obtain the input data, such as limb and of 7th World Congress of ISPO, Chicago, 1992:147.
[16] Lee VSP, Solomonidis SE, Spence WD. Biomechanical model-
socket geometry, material properties and the external ling of the interface between residual limb/prosthesis interface
loading. Auto-meshing techniques should be for transfemoral amputees using finite element analysis. Proceed-
implemented to establish the model quickly and prop- ings of the Eighth International Conference on Biomedical Engin-
erly. Proper post-processing tools should also be eering, Singapore, 1994:333–5.
developed to present the results in a clinically relevant [17] Zhang M, Mak AFT. A finite element analysis of the load transfer
between a residual limb and its prosthetic socket—Roles of
manner, such as a map of interfacial pressure and shear interfacial friction and distal-end boundary conditions. IEEE
stress distributions as well as internal stress/strain distri- Trans Rehabil Eng 1996;4(4):337–346.
butions. These can be easily displayed on the screen, to [18] Childress DS, Steege JW. Computer-aided analysis of below-
give designer a direct and quantitative impressions of the knee socket pressure. J Rehabil Res Dev Progr Rep 1987 vol 24:
mechanical interaction. 22–4.
[19] Lord M, Turner-Smith A. Orthopaedic prosthetics and orthotics
in rehabilitation. In: Bronzino JD, editor. The biomedical engin-
eering handbook. Boca Raton, Florida: CRC Press, 1995:2055–
References 70.
[20] Lee VSP, Solomonidis SE, Spence WD, Paul JP. A study of the
[1] Houston VL, Burgess EM, Childress DS, Lehneis HR, Mason biomechanics of the residual limb/prosthesis interface in trans-
CP, Garbarini MA, LaBlanc KP, Boone DA, Chan RB, Harlan femoral amputees. Proceedings of Eighth World Congress of
JH, Brncick MD. Automated fabrication of mobility aids ISPO, Melbourne, Australia, 1994:79.
(AFMA): Below-knee CASD/CAM testing and evaluation pro- [21] Lee AJC. The place of numerical and experimental methods in
gram results. J Rehab Res Dev 1992;29(4):78–124. biomedical engineering. In: Middleton J, Pande GN, Williams
[2] Sanders JE. Interface mechanics in external prosthetics: review of KR, editors. Recent advances in computer methods in biomechan-
interface stress measurement techniques. Med Biol Eng Comput ics and biomedical engineering. Swansea: Books and Journals
1995;33:509–16. International Ltd., 1992:1–7.
[3] Silver-Thorn MB, Steege JW, Childress DS. A review of pros- [22] Zachariah SG, Sanders JE. Interface mechanics in lower-limb
thetic interface stress investigations. J Rehabil Res Dev external prosthetics: a review of finite element models. IEEE
1996;33(3):253–66. Trans Rehabil Eng 1996;4(4):288–302.
[4] Husain R, Chao EYS. A survey of finite element analysis in [23] Fung YCB. Mechanics of soft tissues. In: Skalak R, Chien S,
orthopaedics biomechanics: the first decade. J Biomech editors. Handbook of bioengineering. New York: McGraw-Hill,
1983;16:385–409. 1987: Chapter 1.
[5] Steege JW, Schnur DS, Vorhis RL, Rovick JS. Finite element [24] Steege JW, Schnur DS, Childress DS. Prediction of pressure at
analysis as a method of pressure prediction at the below-knee the below-knee socket interface by finite element analysis. ASME
socket interface. Proceedings of RESNA 10th Annual Confer- Symposium on the Biomechanics of Normal and Pathological
ence, California, 1987:814–6. Gait, Boston, 1987: BED-vol 4:39–43.
[6] Krouskop TA, Muilenberg AL, Doughtery DR, Winningham DJ. [25] Steege JW, Childress DS. Finite element modelling of the below-
Computer-aided design of a prosthetic socket for an above-knee knee socket and limb: phase II, modelling and control issues in
amputee. J Rehabil Res Dev 1987;24(2):31–8. Biomechanical System Symposium. ASME WAM, BED, 1988:
[7] Reynolds DP. Shape design and interface load analysis for below- vol 11:121–9.
knee prosthetic sockets. Ph.D. Dissertation. University College, [26] Steege JW, Childress DS. Finite element prediction of pressure
University of London, 1988. at the below-knee socket interface. Report of ISPO Workshop on
[8] Sanders JE. Ambulation with a prosthetic limb: mechanical CAD/CAM in Prosthetics and Orthotics, Alexandria. Va: ISPO
stresses in amputated limb tissues. Ph.D. Dissertation. University 1988:71–82.
of Washington, 1991. [27] Steege JW, Silver-Thorn MB, Childress DS. Design of prosthetic
[9] Silver-Thorn MB. Prediction and experimental verification of socket using finite element analysis. Proceedings of 7th World
residual limb/prosthetic socket interface pressure for below-knee Congress of ISPO, Chicago, IL, 1992:273.
amputees. Ph.D. Dissertation. Northwestern University, 1991. [28] Reynolds DP, Lord M. Interface load analysis for computer-aided
[10] Quesada P, Skinner HB. Analysis of a below-knee patellar ten- design of below-knee prosthetic sockets. Med Biol Eng Comput
don-bearing prosthesis: a finite element study. J Rehab Res Dev 1992;30:419–26.
1991;3:1–12. [29] Sanders JE, Daly CH. Normal and shear stresses on a residual
[11] Zhang M. Biomechanics of the residual limb and prosthetic limb in a prosthetic socket during ambulation: comparison of
socket interface in below-knee amputees. Ph.D. Dissertation. Uni- finite element results with experimental measurements. J Rehabil
versity of London, 1995. Res Dev 1993;30(2):191–204.
[12] Seguchi Y, Tanaka M, Nakagawa A, Kitayama I. Finite element [30] Silver-Thorn MB, Childress DS. Use of generic, geometric finite
analysis and load identification of above-knee prosthesis socket. element model of the below-knee residual limb and prosthetic
Proceedings of the Fourth International ANSYS Conference and socket to predict interface pressures. Proceedings of Seventh
Exhibition, Pittsburg, 1989:12.31–12.44. World Congress of ISPO, 1992:272.
[13] Brennan JM, Childress DS. Finite element and experimental [31] Silver-Thorn MB, Childress DC. Parametric analysis using the
investigation of above-knee amputee limb/prosthesis systems: a finite element method to investigate prosthetic interface stresses
comparative study. ASME Advances in Bioengineering for persons with trans-tibia amputation. J Rehabil Res Dev
1991;BED,20:547–50. 1996;33(3):227–38.
[14] Torres-Moreno R. Biomechanical analysis of the interaction [32] Quesada P, Skinner HB. Finite element analysis of theprosthesis
between the above-knee residual limb and the prosthetic socket. model alterations on socket/stump interface stresses. Proceedings
Ph.D. Dissertation. University of Strathclyde, 1991. of Seventh World Congress of ISPO, 1992:275.
[15] Mak AFT, Yu YM, Hong LM, Chan C. Finite element models [33] Zhang M, Roberts VC. Development of a nonlinear finite element
 M. Zhang et al. / Medical Engineering & Physics 20 (1998) 360–373 373

model for analysis of stump/socket interface stresses in below- [52] Faukner V, Walsh N, Gall N. Ultrasound as an aid to prosthetic
knee amputee. In: Held KD, Brebbia CA, Ciskowski RD, Power socket design. J Rehabil Res Dev Progr Rep 1986:7.
H, editors. Computational biomedicine. Southampton: Compu- [53] Rogers B, Morimoto AK, Faulkner V, Kozlowski DM, Walsh
tational Mechanics Publisher, 1993:209–214. NE, Vick DD. Evaluation of a 3 D ultrasound imaging system
[34] Zhang M, Turner-Smith AR, Roberts VC. A non-linear finite for BK prosthesis fabrication. Proceedings of 8th World Congress
element analysis of the stump/socket interface. Second World of ISPO, 1995:126.
Congress on Biomechanics, Amsterdam, The Netherlands, [54] Decraemer WF, Maes MA, Vanhuyse VJ. A non-linear viscoelas-
1994:385. tic constitutive equation for soft biological tissues, based on a
[35] Zhang M, Turner-Smith AR, Roberts VC. Friction, slip and structural mode. J Biomech 1980;13:559–64.
stresses at residual limb/prosthetic socket interfaces—a 3-D non- [55] Mow VC, Kuei SC, Lai SC, Armstrong GG. Biphasic creep and
linear finite element analysis. World Congress On Medical Phys- stress relaxation of articular cartilage in compression: theory and
ics and Biomedical Engineering, Rio de Janeiro, Brazil, experiments. J. Biomech. Eng., Trans. ASME 1987;102:73–84.
1994:861. [56] Mak AFT, Lai WM, Mow VC. Biphasic indentation of articular
[36] Zhang M, Lord M, Turner-Smith AR, Roberts VC. Development cartilage, part I: theory analysis. J Biomech 1987;20:703–14.
of a nonlinear finite element modelling of the below-knee pros- [57] Oomens CWJ, Campen DH, Grootenboer HJ. A mixture
thetic socket interface. Med Eng Phys 1995;17(8):559–66. approach to the mechanics of skin. J Biomech 1987;20:877–85.
[37] Torres-Moreno R, Solomonidis SE, Jones D. Three-dimensional [58] Liu GHW, Mak AFT, Lee SY. Indentation properties of the soft
finite element analysis of the above-knee residual limb. Proceed- tissues around below-knee stumps. Proceedings of 7th World
ings of the Seventh World Congress of ISPO, 1992:274. Congress of ISPO, Chicago, 1992:150.
[38] Lee VSP, Solomonidis Se, Paul JP. Mechanical behaviour of [59] Mak AFT, Liu GHW, Lee SY. Biomechanical assessment of
amputee stump/socket using finite element analysis. Proceedings below-knee residual limb tissue. J Rehabil Res Dev
of International Society of Biomechanics, XIVth Congress, Paris, 1994;31(3):188–98.
1993:774–5. [60] Torres-Moreno R, Solomonidis SE, Jones D. Geometrical and
[39] Zhang M, Mak AFT. Effect of friction on the load transfer mechanical characteristics of the above-knee residual limb. Pro-
between an above-knee residual limb and prosthetic socket. Pro- ceedings of 7th World Congress of ISPO, 1992:149.
ceedings of Second Medical Engineering Conference, Taiwan, [61] Vannah WM, Childress DS. Indentor tests and finite element
May 1996. modelling of bulk muscular tissue in vivo. J Rahibil Res Dev
[40] Zhang M, Mak AFT. Finite element analyses of the stump end 1996;33(3):239–52.
loadbearing on above-knee prosthetic socket. Proceedings of an [62] Bader DL, Bowker P. Mechanical properties of skin and underly-
International Conference on Biomedical Engineering, Hong ing tissues in vivo. Biomaterials 1983;4:305–8.
Kong, 1996:94–7. [63] Zheng YP, Mak AFT. An ultrasound indentation system for
[41] Saunders CG, Bannon M, Sabiston RM, Panych L, Jenks SL, biomechanical properties assessment of soft tissues in-vivo. IEEE
Wood IR, Raschke S. The CANFIT system: shape management Trans Biomed Eng 1996;43(9):912–8.
technology for prosthetic and orthotic application. J Prosthet [64] Mak AFT, Zheng YP. Biomechanical assessment of residual limb
Orthot 1989;1(3):122–30. tissue using a force transducer in series with an ultrasound thick-
[42] Oberg K, Kofman J, Karisson A, Linstrom B, Sigblad G. The ness gage. Proceedings of Eighth World Congress of ISPO,
Capod system: a Scandinavian system for prosthetic sockets. J 1995:127.
Prosthet Orthot 1989;1(3):139–48. [65] Krouskop TA, Dougherty DR, Vinson FS. A pulsed Doppler
[43] Boone DB, Burgess EM. Automatic fabrication of mobility aids: ultrasonic system for making noninvasive measurements of the
clinical demonstration of the UCL computer aided socket design mechanical properties of soft tissue. J Rehabil Res Dev
system. J Prosthet Orthot 1989;1(3):187–90. 1987;2:1–8.
[44] Walsh NE, Lancaster JL, Faulkner VW, Rogers WE. A com- [66] Malinauskas M, Krouskop TA, Barry PA. Noninvasive measure-
puterized system to manufacture prostheses for amputees in ment of the stiffness of tissue in the above-knee amputation limb.
developing countries. J Prosthet Orthot 1989;1(3):165–81. J Rehabil Res Dev 1989;3:45–52.
[45] Schreiner RE, Sanders JE. A silhouetting shape sensor for the [67] Roger SI, McGovern TF, Chung KC. Biomechanics of tissue dis-
residual limb of a below-knee amputee. IEEE Trans Rehabil Eng tortion and stiffness by magnetic resonance imaging. In: Bader
1995;3(3):242–53. DL, editor. Pressure sores—clinical practice and scientific
[46] Smith KE, Vannier MW, Commean PK. Spiral CT volumetry of approach. London: MacMillan Press, 1990:177–90.
below-knee residua. IEEE Trans Rehabil Eng 1995;3:235–41. [68] Sokoloff L. Elasticity of articular cartilage. Fed Proc Am Soc
[47] Lee VS, Solomonidis SE, Spence WD. Magnetic imaging of the Exp Biol 1966;25:1089–95.
trans-femoral residual limb. Proceedings of the Eighth World [69] Hayes WC, Keer LM, Herrmann G, Mockros LF. A mathematical
Congress of ISPO, 1995:127. analysis for indentation tests of articular cartilage. J Biomech
[48] He P, Xue K, Bismar HA, Murka P. Application of ultrasonic 1972;5:541–51.
imaging technique to prosthetic socket design. Proceedings of [70] Zhang M, Zheng YP, Mak AFT. Estimating the effective Young’s
Annual International Conference of IEEE Engineering in Medi- modulus of soft tissues from indentation tests. Med Eng Phys
cine and Biology Society, vol. 3, 1991:1849–1850. 1997;19(6):512–7.
[49] He P, Xue K, Murka P, O’beness F. A PC-based ultrasonic data [71] Vannah WM, Childress DS. Modelling the mechanics of nar-
acquisition system for computer aided socket design. Proceedings rowly contained soft tissues: the effects of specification of Pois-
of 19th Annual Meeting and Scientific Symposium of American son’s ratio. J Rehabil Res Dev 1993;30(2):205–9.
Academy of Orthotists and Prosthetists, 1993:22. [72] Zhang M, Turner-Smith AR, Tanner A, Roberts VC. Frictional
[50] He P, Xue K, Chen Q, Murka P, Schall S. A PC-based ultrasonic action at residual limb/prosthetic socket interface. Med Eng Phys
data acquisition system for computer aided socket design. IEEE 1996;18(3):207–14.
Trans Rehabil Eng 1996;4(2):114–9. [73] Zhang M, Turner-Smith AR, Roberts VC, Tanner A. Clinical
[51] Faukner V, Walsh N, Gall N. A computerized ultrasound shape investigation of the pressure and shear stress on the below-knee
sensing mechanism. Orthot Prosthet 1988;41(4):57–65. stump with a prosthesis. Med Eng Phys 1998 (in press).