Sie sind auf Seite 1von 13

),

theoretical and
applied fracture
mechanics

)" !

II
ELSEVIER

Theoretical and Applied Fracture Mechanics 28 (1997) 81-93

Development and validation of a viscoelastic finite element model


of an L 2 / L 3 motion segment
J.L. Wang a, M. Parnianpour b,., A. Shirazi-Adl c, A.E. Engin d, S. Li e,
A. Patwardhan f
a Department of Orthopaedic Surgery, National Taiwan University Hospital Taipei, Taiwan, ROC
b Department oflndustrial WeMing and System Engineering, The Ohio State University, 210 Baker System, 1971 Neil Ave., Columbus,
OH 43210, USA
c Department of Mechanical Engineering, Ecole Polytechnique, Montreal Canada
d Department of Mechanical Engineering, University of South Alabama, Alabama, USA
e Northwestern University, Sensory Performance Program, Rehabilitation Institute of Chicago, Chicago, IL, USA
f Department of Orthopaedic Surgery, Loyola University Medical Center, Maywood, IL, USA

Abstract
The prediction of the time dependent response of the spine to dynamic loading conditions is essential in understanding
the injury mechanisms leading to occupationally related low back disorders (OLBD). Many previous finite element (FE)
models of the lumbar spine have over-simplified the geometry and the material properties of their elements, yielding results
limited generalizability. This study reports on the development and validation of a nonlinear viscoelastic FE model that can
quantify the mechanical responses of the L2/L3 motion segment to time varying external loads. This model was developed
by consideration of the intrinsic material properties of its individual constituents. A piecewise parameter identification
method was adopted due to the inherent complexity in determining the role and contribution of each element to the overall
behavior of the motion segment. The results of simulation of four loading conditions (quasistatic, constant loading rate,
creep and cyclic relaxation) showed a satisfactory agreement with experimental observations in the literature. The detailed
estimates of the state of stress/strain of this validated FE model can be used to test the role of epidemiological risk factors
such as prolonged awkward posture, speed of lift (strain rate effect) and complex repetitive loading in OLBD. 1997
Elsevier Science Ltd.
Keywords: Lumbar spine; Motion segment; Finite element model; Viscoelasticity; Injury mechanism; Low back pain

1. I n t r o d u c t i o n
As many as 85% of adults experience back pain
that interferes with their work or recreational activity
and up to 25% of the people between the ages of

* Corresponding author. Tel.: + 1-614-2920063; fax: + 1-6142927852; e-mail: pamm@osuergo.eng.ohio-state.edu.

3 0 - 5 0 years report low back symptoms [1]. The


results o f epidemiological studies have associated six
occupational factors with low-back pain symptoms.
These factors are physically heavy work, static work
postures, frequent bending and twisting, lifting and
sudden forceful incidents, repetitive work and exposure to vibration [2]. In a large retrospective survey,
lifting or bending episodes accounted for 33% of all

0167-8442/97/$17.00 1997 Elsevier Science Ltd. All rights reserved.


Pll S 0 1 6 7 - 8 4 4 2 ( 9 7 ) 0 0 0 3 2 - 3

82

J.L. Wang et al. / Theoretical and Applied Fracture Mechanics 28 (1997) 81-93

work-related causes of back pain [3]. The works in


Ref. [4] have identified the combination of lifting
with lateral bending or twisting as a frequent cause
of back injury in the workplace.
Determining the mechanical behavior of the motion segment is important in the study of the etiology
of low back pain. The dynamics of lifting tasks
affect both the active and passive elements of the
spine. However, the response of the spine to dynamic loading is poorly understood. Carried out in
Ref. [5] is a study of the fatiguing dynamic movement of the trunk against a set resistance, showed
that during fatiguing trunk flexion and extension,
there were significant reductions in the velocity,
range of motion and total angular excursion in the
intended (sagittal) plane of motion and a significant
increase in the range of motion and total angular
excursion in the accessory (coronal and transverse)
planes. The presence of more unintended motion in
the accessory planes indicates a loss of coordination
and more injury prone loading conditions for the
spine. Numerous studies have demonstrated that soft
tissues subjected to repetitive loading show creep
and stress relaxation behavior because of their viscoelastic properties [6,7]. Since the internal stability
of the spine is maintained by its passive and active
structures, there is an even greater need for muscular
control in maintaining a given level of spinal stability after repetitive movements. Hence, the presence
of repetitive dynamic trunk exertions increases the
risk by adversely affecting the performance of the
neuromusculoskeletal system (i.e. diminished control
and coordination, reduction in magnitude and rate of
tension generation in the muscles and the reduction
in the passive stiffness of spinal tissues) [5].
The experimental, analytical and computational
simulations have been used to investigate the mechanical behavior of the motion segment. Lumped
parameter models are the most efficient methods of
predicting the gross time dependent behavior of the
motion segments [8]. However, the model parameters
are task dependent and cannot be generalized to
arbitrary loading conditions [9]. The more severe
problem is their inability to provide insight into the
internal stress/strain fields within the structure. The
more detailed structural analysis using a FE model
will provide the states of stress/strain in all constituent elements. Hence, a validated FE model could

increase our insight to delineate injury mechanisms


using alternative injury criteria [6]. In addition, numerical simulations can be used to compute biomechanical parameters (i.e. strain energy) that are not
amenable to measurements. Findings from the experiments tend to vary greatly due to the inter-species
variations, state of saturation, age, gender and degeneration states of the disc as well as the differences in
measurement techniques [19,20,27,30,33,41]. Biomechanical evaluations of each risk factor can best be
performed by systematic numerical simulations using
a validated FE model.
The finite element (FE) method has been used for
simulating the mechanical behavior of the spinal
motion segments for over two decades. The following will briefly review the historical development
and a more detailed review can be found in recent
publications [10,11]. The first simplified FE model
was a 2D axisymmetric elastic model of a body-disc
unit [12]. Due to the progress in numerical schemes
for modeling viscoelastic [13-16] and poroelastic
materials [17-20], the better mesh generation and
more extensive experiments to determine the material properties, more realistic simulations can now be
performed [11]. A viscoelastic model of a lumbar
motion segment was developed using a simplified
2D body-disc model [15]. The anulus fibrosus (AF)
was simulated as a homogeneous viscoelastic solid.
The material constants were obtained by matching
the FE output with the gross response of a creep
experiment [21]. Constructed in Ref. [13] is a 2D
body-disc model to simulate the viscoelastic creep
response of rhesus monkey motion segments. In this
study, the AF and nucleus pulposus (NP) were considered to be made of the same material. Since the
geometry and the material properties were over-simplified, the generalization of the results were limited.
In another body-disc axisymmetric viscoelastic model
[14], the uniaxial measurement of the viscoelastic
properties of soft tissue were obtained from the
literature and then adjusted by comparing the creep
response of disc-body unit with experiments. However, only the AF was simulated with the viscoelastic
model.
The primary goal of this work is to develop a
nonlinear viscoelastic FE model that can quantify the
mechanical response (states of stress and strain) of a
functional unit of the spine (L2-L3 motion segment)

J.L. Wang et al. / Theoretical and Applied Fracture Mechanics 28 (1997) 81-93

to time varying extemal loads. Our hypothesis is that


a detailed viscoelastic model incorporating the intrinsic viscoelastic properties of individual constituents
could be used to improve the prediction of the
overall behavior of motion segment under various
loading conditions. The first task was to identify the
parameters of the time dependent (viscoelastic) material models for each element of the L 2 - L 3 motion
segment. Second, the overall response of the motion
segment was evaluated using the experimental resuits in the literature. The foregoing interactive process was repeated across different loading conditions

83

to fine tune all material properties using least squares


optimization. The simulation results of the response
of final model with the optimized set of parameters
were further validated with experimental results not
used .in the previous parameter identification stage.

2. M e t h o d
The geometry of the FE model is based on the
direct measurement of an L 2 - L 3 motion segment
[22]. The functional unit of spine, also called a

Vertebral body (L2)


Spinous proc

Inferior articular process


End plate (L2)

0, c

Anulus fibrosus
nucleus pulposus

End plate (L3)

Vertebral body (L3)

Fig. 1. The finite element mesh of the L2-L3 motion segment. Only the vertebral bodies, disc and spinous processes are shown here. The
isolated eriss-cross sa'uctured anulus fibers and ligaments are not plotted for clarity. The anulus fiber, anulus matrix, nucleus pulposus and
ligaments are simulated as viscoelastic materials where the rest of the elements are elactic.

84

J.L. Wang et a l . / Theoretical and Applied Fracture Mechanics 28 (1997) 81-93

motion segment, consists of two vertebrae and the


disc in between. A mesh of the full motion segment
including the vertebrae, AF, NP and ligaments has
been constructed (Fig. 1). Seven distinct structural
regions were simulated in the current model, namely
AF, NP, cartilaginous endplate, cancellous bone, cortical bone, posterior bony elements and various ligaments. The current model was implemented using
ABAQUS v5.3 [23]. The mechanical models used in
the simulation were the linear elastic solid, viscoelastic solid, nonlinear spring and dashpot and contact
interface (Table 1). The endplate was modeled by
cartilaginous bone at the center and cortical bone at
the periphery. The articular facet joint was meshed
by anatomically realistic geometry and modeled using nonlinear frictionless soft contact surfaces to
simulate the nonlinear interaction of the superior and
inferior joints. The depth of initial contact between

joints was set to 0.4 mm and the pressure at zero


depth of contact was set to 1.6 MPa. The geometry
and the material properties of the elastic elements
(i.e. cortical, cancellous bones, endplate) were purposefully kept identical to those values of the thoroughly validated elasto-static model in Refs. [24,25].
This allows a more systematic evaluation of the
effects of viscoelastic augmentation of the present
model.
The viscoelastic material parameters of individual
components (AF, NP, ligaments) and the motion
segment were determined from the available experimental studies. A piecewise parameter extraction
method was adopted due to the complexity in determining the role of each element and its contribution
to the overall behavior of the motion segment. The
first step was to study the experimental data of the
individual components: such as the stiffness of iso-

Table 1
The optimized material properties and element types of the L 2 / L 3 motion segment

Elastic material properties


component

element type

Y o u n g ' s modulus (MPa)

Poisson ratio

Cortical bone [ 4 9 - 5 2 ]
Cancellous bone [53,54]
Inferior prOcess [24]
Superior process [24]
Endplate [24]
Facet articulation [24]

3-D
3-D
3-D
3-D
3-D
3-D

12000
100
3500
7000
24

0.30
0.20
0.25
0.25
0.40

20-node elastic solid


20-node elastic solid
27-node elastic solid
27-node elastic solid
20-node elastic solid
elastic slide interface

No. element
48
176
32
30
40
38

Viscoelastic material properties


Zener model
Ligaments [24]
Anulus fiber [26]

nonlinear
nonlinear
nonlinear
nonlinear

spring
dashpot
spring
dashpot

100
50
960
480

Prony Series
relaxation of

shear

bulk

time (s)

Anulus matrix: E = 8.0 MPa, v = 0.45

gj = 0.3991

k I = 0.3991

r I = 3.45

g2
g3
g4
gl
g2
g3
g4
g5

k 2 = 0.300
k 3 = 0.149
k 4 = 0.150
kI = 0
k2 = 0
k3 = 0
k4 = 0
k5 =0.8

~'2 =
r3=
74 =
rI=
z2 =
"r3 =
z4 =
r5 =

Nucleus pulposus: E = 2.0 MPa, v = 0.49

Total No. of elements


Total No. of nodes
Total No. of variables

=
=
=
=
=
=
=
=

0.0000
0.3605
0.1082
0.6375
0.1558
0.1202
0.0383
0

100
1000
5000
0.141
2.21
39.9
266
500

60

108

2122
4349
11166

J.L. Wang et al. / Theoretical and Applied Fracture Mechanics28 (1997) 81-93
Table 2
The experimentsused for determining the viscoelasticmaterial properties
Materials
Loading
Specimen
Anulus fiber
CSR
RTT
Anulus matrix
CSR
MS
creep
AM
Nucleus pulposus
shear
NP
creep
MS (IDP)
Ligaments
CSR
ALL

Viscoelastic model
Zener model
Prony, short-term
Prony, long-term
Prony, short-term(shear)
Prony, long-term(bulk)
Zener model

85

Exp. data
[26,24]
[32]
[33]
[34]
[35]
[36,37]

CSR: constant strain rate, RTr: rat tail tendon, MS: motion segment, AM: anulus matrix, NP: nucleus pulposus, IDP: intmdiscal pressure.
ALL: anterior longitudinal ligament.

lated fibers, long term stiffness modulus of the anulus matrix (AM), shear modulus of the NP, and
temporal response of spinal ligaments. Since the
material properties were not available for all individual components, the overall behavior of the motion
segment (body-disc-body unit) was also used in this
process. An iterative method was used to obtain an
'optimized set' of material properties that satisfy the
viscoelastic behavior of the motion segment across
the available (and often incomplete) experimental
data. The overall optimized material properties are
listed in Table 1 and the experiments used for obtaining the material properties are listed in Table 2.

The ratios of stiffness were 1.66 for anterior to


posterior, 1.79 for outer to inner at the anterior
region and 1.39 for outer to inner at the posterior
region [31 ].

2.1. Material parameters for anulus fibrosus

where GR(t) is the normalized relaxation modulus.


g; is the weighing factor ranged from 0 to 1. r i is
the relaxation time constant and G O is the instantaneous stiffness at t = 0.
It is known that the viscoelastic behavior of the
soft tissues may be dominated by both the short and
long time scales [6]. It is reasonable to find the
short-term constants first before extracting the longterm constants. Experiments of constant loading rates
[32] were used to determine the short-term constants
(Fig. 2a). If the Poisson ratio is assumed to be time
invariant at short time scales (within a few seconds),
the normalized relaxation moduli for Young's modulus can also be applied for shear and bulk modulus.
The reported regional biphasic material properties
[33] were transformed to the equivalent viscoelastic
material properties with FE simulation of confined
compression experiments (Fig. 2b) to find the longterm constants. Based on our simulations, a good
curve fit cannot be obtained by the variation of bulk
modulus alone. With the short term shear modulus

The anulus fibrosus (AF) was considered to be a


nonlinear, nonhomogeneous, criss-cross structured,
viscoelastic composite of collagen anulus fibers
(AFib) embedded in a viscoelastic AM of ground
substance.
2.1.1. Anulus fiber
The temporal response of rat tail tendon (RTT) at
different strain rates [26] was used to obtain the
stress-strain curve of AFib by using a Zener model,
since RTT is primarily composed of parallel collagen
fibers. The detailed development of material properties AFib using Zener Model can be found in Ref.
[6]. Eight layers of AFib were represented in the AF.
In each layer, the angle of the fiber was about + 30
with the horizontal mid-plane of the disc. The overall
volume fraction of collagen in the matrix was kept at
16% [27-30]. The regional variations were accounted for by augmenting the stiffness of fibers.

2.1,2. Anulus matrix


The Prony series was used to simulate the viscoelastic behavior of the AM in the current model.
The equation of stiffness for the Prony series can be
described as:
GR(t)=

=1Go

&
2~gi(1-e
i- l

-t/',)

(1)

J.L. Wung et uI./

86

(a)

Theorsticul

md Applied

Fracture

Mechanics

-~

0.25

I
0.2
I
,. 1_

28 (1997) 81-93

3%lsec.
O.B%lsec.
+
0
O.O3%/sec.
------Optimized
Curves

Strain

(%)

I
20 1
;:
b
E
!I
;;

r331
- - FEM Optimized

15t

(r-sq=99.92%)

10 i-

i
5 -

Time (seconds)
Fig. 2. The material parameters of the anulus matrix. (a) The short-term response: A set of experiments [32] using thirty cycles of loading
was conducted at different strain rates (3, 0.3, & O.O03%/second).
Only one ascending cycle of one motion segment are shown here. The
average of the short term constants of eight motion segments are: g, = 0.3991, T, = 3.45 seconds. (b) The long-term response of the anulus
matrix subjected to constant compression load. The experiment 1331 was conducted in compression within a confmed space to find the
resistance of AF. Hence, the result represent the mechanical properties of the AM rather than the AFib. The same boundary and loading
conditions were simulated by FE model and its prediction indicated a very strong fit to the experimental results (Rz = 99.2%).

fixed, the long term shear relaxation was augmented


and varied to provide a good fit.
The modulus of the ground matrix of AM in
elasto-static FE model 1241 was set to 4.2 MPa. In
order to get similar gross responses between the
elasto-static model and the viscoelastic model during
short loading durations often used in the experimental studies (i.e. in the range of 30 s), the initial
modulus of the Ah4 was increased from 4.2 to 8
MPa. In the viscoelastic FE simulation, an initial

stiffness of 8 MPa will be reduced to 4.2 MPa after


30 s with the determined relaxation modulus.
2.2. Material parameters for nucleus pulposus
The nucleus pulposus (NP) has often been considered as an inviscid elastic fluid for most elastic FE
models, However, recent results [34] have shown
that the NP behaves more like a viscoelastic material
in the short-term response, Therefore, the material

87

J.L. Wang et aL / Theoretical and Applied Fracture Mechanics 28 (1997) 81-93


'

L9
=

0.8 ~

-g"5

0.6

o Experimental data [34]


. . . .
.
=
. %)

E
r"

.o

0.4
0.2

0,.

. . . . . . . . . . . . . . . . . .

10-2

10-1

100

101

10 2

10 3

Time (second)
Fig. 3. The best fit of reduced shear relaxation function for the nucleuspulposususingseries based on experimental data [34].

parameter of the short-term normalized relaxation


shear modulus was obtained from the same experimental results [34] (Fig. 3). However, their suggested
shear modulus for NP (7.2 kPa, E = 0.02 MPa with
v - - 0.49) was not considered. With such a low stiffness value, the simulated intradiscal pressure (IDP)
in the full motion segment would be much smaller
than that observed in the experiments (more than a
hundredfold difference). Therefore, a larger stiffness
was considered ( E = 2.0 MPa, v = 0.49) to ensure
matching the experimentally observed IDP.
The short-term relaxation of bulk modulus was
assumed to be zero due to the incompressibility o f
the NP, which conserves the fluid content within a
short time scale. Nevertheless, the fluid loss (here
modeled as volume change) must be considered during prolonged loading such as creep tests. The longterm relaxation of the bulk modulus was estimated
by a loss of 10% in 1DP after 3 h of 1200 N
compressive creep loading on the motion segment
based on Ref. [35]. The relaxation of the shear
modulus was assumed to be zero since most of the
shear stiffness (95%) is already relaxed within the
short-term response.
A 1 h creep simulation with 600 N of compressive load was performed using the current model.
About a 9% decline in the IDP and a 4% volume
loss were observed. They are in agreement with the
poroelastic FE simulations [17]. It is noteworthy that
before incorporating the change in bulk modulus to

represent the volume loss, the FE viscoelastic model


predicted a small rise in the IDP which was contrary
to the experimental observation, but in agreement
with an earlier model which considered only the
AFib to be viscoelastic [14].
2.3. Material parameters f o r ligaments

The ligamentous structures (capsular, anterior and


posterior longitudinal, intertransverse, interspinous,
ligamentum flavum and supraspinous) of the motion
segment were also represented by a Zener model.
The cross sectional area and elastic response were
adopted from an elastic model [24,25], whereas constant strain rate loading experiments of sensitivities
of anterior longitudinal ligament (ALL) [36,37] were
used to estimate parameters of Zener model using
the procedure reported elsewhere [6]. The results of
these simulations for ALL were consistent with experimental results [36,37] (Table 3). Since the variaTable 3
The experimental [36] and simulated stiffness of the anterior
longitudinal ligament (ALL) during three constant strain rate
tensile loadings usingZener model [6]
Loading rate
Exp. data [36]
Zener model
Slow (0.1 mm/s)
81.7 (SD 37.2)
78.8
Medium (2.5 ram/s)
85.2 (SD 32.6)
88.6
Fast (200 rnm/s)
200 (SD 99.6)
188
SD: Standard deviation

88

J.L. Wang et al, / Theoretical and Applied Fracture Mechanics 28 (1997) 81-93

tion in temporal response has not been reported for


the remaining ligaments, the viscoelastic material
constants of all ligaments were assumed to be similar
to those of ALL.

set of loading conditions: (1) application of a pure


moment ranging from 0 to 20 Nm on the full motion
segment (including the ligaments and posterior elements) in flexion, extension, axial rotation and lateral bending within 30 s, (2) pure compression on
full motion segment from 0 to 4000 N within 30 s,
(3) constant strain rate axial loading, (4) 1 h compressive creep loading and (5) cyclic stress relaxation loading at 0.01, 0.1 and 1 Hz for 15 cycles.
The last three experiments used b o d y - d i s c - b o d y units

3. M o d e l v a l i d a t i o n
The behavior of the full motion segment is complex and required careful validation using a diverse

2O

500~

E
Z
r- 10

E
o
=E

2001
5

r!

xO Flexion
Extension
Lateral Bending
Axial Rotation
5
Rotation (deg.)

1 0 0 ~

10

(a) Deformation
14 . . . . . . . . . . . . . . . . . . . . . . . . . . .

5
Rotation (deg.)

-~t0

(b) Facet Force

~'2.5
10

=E

eL
u.

4!
2

5
10
Rotation (deg.)
(c) Maximum Anulus Fiber Strain

o ~

0.5

i
2000

4000

Force (N)
(d) Disk Pressure at
Pure Compression

Fig. 4. The responses of the full L 2 - L 3 viscoelastic FE model during pure flexion, extension, lateral bending and rotation moments up to 20
Nm, and a pure compression up to 4000 N. The maximum Ioadings were attained within 30 seconds (to match the experimental procedure).
The responses are well within the results from the previously reported elastic model [39] that have been validated extensively with the
experimental results in the literature.

J.L. Wang et aL / Theoretical and Applied Fracture Mechanics 28 (1997) 81-93

89

Table 4
The comparison of mechanical responses of the elasto-static FE model [38] and the current viscoelastic model (30 s loading duration) during
pure loading of the full L2-L3 motion segment
Loading
condition

Deformation (o)
at 20 Nm

Facet force (N)


at 20 Nm

Flexion [38]
Current model

7.2
8.3

12
87

2.7
2.5

Extension [38]
Current model

6.4
5.9

332
422

3.9
3.7

Lateral bending [38]


Current model

8.3
8.6

25
49

4.2
6.1

Axial torque [38]


Current model

2.6
4.4

215
266

7.0
7.0

[32]; hence, the spinous processes and ligaments of


the F E model were taken out for these simulation.
3.1. Short-term (30 s) loading

to

The full motion segment was subjected


moments 20 N m in pure flexion, extension, lateral
bending and rotation for 30 s. The results o f the
simulation were compared to the elasto-static model
results [38] because both models had the similar
geometry and elastic elements (Fig. 4, Table 4). The
current model shows a similar stiffness during extension and lateral bending but predicted a more compliant behavior during axial rotation and flexion

25

x
+
O
---. 0 1 5 I- - '

,
2 I

0.05 ]
0

Max. strain (%) of AFib


at 5 rotation

which caused higher facet forces. The m a x i m u m


fiber strains are similar for both models. The predicted IDP during 3000 N of pure compressive loading of the current viscoelastic and elastic models are
2.1 MPa (Fig. 4D) and 2.6 MPa [39], respectively.
The IDP o f the current simulation is within the
experimental results which ranged from 1.8 to 2.4
MPa at the same loading condition [40].
3.2. Axial compressive at constant loading rates
The simulations for constant strain rate loading
using both the short- and long-term constants for
fiber matrix, the criss-cross structured A F i b and the

3%lsec.
'
0.3%lsec.
0 03%lsec
Exp. Data [32]
ABAQUS simulation

~'~

_~ ' ~ ~ o -

.
0.5

~- ~x"

-;

.
1.5

S t r a i n (%)
Fig. 5. The comparison of short term responseof the viscoelastic FE model and the averaged experimental results of eight motion segments
[32]. The loading rates arc 3, 0.3, 0.03%/second.

90

J.L. Wang et al. / Theoretical and Applied Fracture Mechanics 28 (1997 ~81-93

4 :3 5j" |

ABAQUSsimulation

Averaged

.,~

Standard

......

results [32]
deviation

"

2.5

0"51 .....
0

500

..........................................................
1000

t 500

2000

2500

3000

3500

Time (sec.)
Fig. 6. The comparison of long term creep response of the averaged experimental results of eight motion segments [32] and the FE model
simulation. The instantaneous strain is subtracted to allow ensemble averaging of creep responses.

viscoelastic NP in the FE model were compared to


the average behavior of the eight motion segments
undergoing uniaxial compression at three rates (3,
0.3 and 0.03%/s, Fig. 5). The model predicts a more
compliant response at all loading rates. Given considerable inter-specimen variation, the predictions
derived from the current FE model are satisfactory
even though no attempt was made to match the
stiffness properties of any single motion segment.
This confirmed the validity of the estimated short
term parameters.

3.3. Compressive creep


The validation of the long term behavior of the
model was confirmed by examining the accuracy in
predicting the average axial strain during 1 h creep
incorporating all the estimated material properties.
The creep axial loading was 426 N and the instantaneous strain was subtracted to allow ensemble averaging of creep responses over the eight specimens
[32]. To further illustrate the variation in behavior
between the specimens, the mean plus/minus one
standard deviation of the experimental measurements
are also shown in Fig. 6. The predicted creep curve
of FE simulation was located within one standard
deviation.

3.4. The responses of cyclic relaxation


The cyclic axial loading experiments of an L 1 - L 2
disc-body-disc lumbar motion segment undergoing

426 N preload [9] was also used for validation. The


loading condition was from 0 to 100 p~m at 0.01, 0.1
and 1 Hz. The results of several predicted mechanical responses (relaxation, stiffness, and hysteresis

Table 5
The comparison of mechanical responses from the experiments
with the predictions of the standard linear solid (SLS) model [9]
and current FE model during cyclic stress relaxation. Eight lumbar
spine motion segment were tested in the experiment
Frequency (Hz)

Experiment

SLS

FEM

Relaxation of loading (dimensionless)


0.01
0.51 (SD 0.07)
0.40 (SD 0.07)
0.10
0.19 (SD 0.05)
0.08 (SD 0.02)
1.00
0.10 (SD 0.04)
0.01 (SD 0.00)

0.590
0.219
0.082

Stiffness (MN/m)
0.01
1.62 (SD 0.31)
0.10
2.23 (SD 0.45)
1.00
2.42 (SD 0.51)

NA
NA
NA

1.21
1.70
1.86

Modulus (MPa)
0.01
0.10
1.00

5.0
NA
5.4

9.74
13.64
14.89

10.5 (SD 3.80)


14.3 (SD 5.03)
15.5 (SD 5.47)

Relative hysteresis energy (dimensionless)


0.01
0.08 (SD 0.05)
0.033
0.10
0.08 (SD 0.04)
NA
1.00
0.11 (SD 0.05)
0.001
NA: not available.

0.18
0.16
0.048

J.L. Wang et a L / Theoretical and Applied Fracture Mechanics 28 (1997) 81-93

energy) obtained from the FE simulations were compared with the experimental observations including
the results of the optimized lumped parameter model
(using standard linear solid, SLS) [9] in Table 5.
Assuming the difference in the mechanical responses
between levels of L 1 / L 2 and L 2 / L 3 motion segments can be neglected, the results of the FE model
provided a better match with the experimental data
than the optimized lumped model. The relative hysteresis energy loss showed the least agreement between the experiments and model predictions.

4. Discussion
Several assumptions and limitations are applicable
to the current model. First, it should be noted that the
ideal test for obtaining the short-term constants of
AM should be the loading of the AM alone, not the
body-disc-body motion segment (Fig. 2(a)). Since
the duration of loading was short and the range of
axial displacement was quite small (0.1 mm), the
resistance of the L 1 - L 2 specimens to deformation
could be largely attributed to the AM. The contribution of the AFib to resisting the deformations is
considered negligible due to the small deformation
of motion segment and small disc bulging. The
contribution of NP was also neglected due to its
smaller stiffness; therefore a larger stress will be
resisted by the AM than by the NP at the same
loading displacement.
Second, it is known that the mechanical behavior
of the disc during creep is greatly influenced by the
fluid movement in the AF and NP; hence the poroelastic model incorporating osmotic and swelling
pressures may be more appropriate to predict the
response [20]. However, there is the intrinsic viscoelastic behavior observed for the disc [41]. Therefore, the fluid flow may not be the only mechanism
to describe the creep and recovery behaviors [42].
Numerous studies have also demonstrated that soft
tissues subjected to repetitive loading showed creep
and stress relaxation behaviors because of their viscoelastic properties [6,7]. The present work simulated the loss of fluid in the disc by reducing its bulk
modulus. More comparative studies are needed to
delineate the full benefits and disadvantages of these

91

alternative modeling strategies (i.e. viscoelastic,


poroelastic and poroviscoelastic models). It should,
however, be iterated that no other time-dependent FE
model of spine has been validated by such diverse
sets of loading conditions.
Third, the validation of the viscoelastic material
properties is a continuous process. Future simulations are intended to model the recently reported
time dependent response of a large number of motion segments given in Ref. [43]. They also [44,45]
indicated that the viscoelastic properties of ligaments
and the disc increase their resistance to flexion by
12% if the duration of movement is reduced from 10
to 1 second. The sustained flexion was also shown to
reduce resistance by 42% in just 5 min and by 67%
in 1 h [46]. The drop at 5 min is concluded to be
dominated by rapid stress relaxation of ligaments
[46,47] where the disc slower relaxation is due to
slower process of fluid flow [48].
The lumped parameter models have long been
used to delineate the viscoelastic behavior of living
tissues. In the comparison of SLS model and FE
simulation, we found the FE analysis to be better
than the SLS model in prediction of various mechanical variables. The hysteresis energy is the least
predictable variable for both models. Nevertheless, it
must be noted that none of the cyclic relaxation
experiments were used for the estimation of material
properties. Obtained in [9] are the material constants
from creep response for their SLS model which was
then used to predict the stress relaxation behaviors. It
can be concluded that the material constants of the
SLS model can be used to predict the responses to
the loading conditions that were used in its development. Less accurate predictions are, however, expected when one considered new loading conditions.
Our results implied that: with material properties
obtained from individual elements, the response to
different loading conditions can, in general, be predicted more accurately without continuously modifying the model parameters.
This validation study, conf'uaning the generalizability of the current model, allows one to investigate
the injury mechanisms caused by the occupational
risk factors using series of systematic numerical
simulations based on :different failure criteria. The
results of our ongoing research to perform these
investigations are forthcoming.

92

J.L. Wang et al. / Theoretical and Applied Fracture Mechanics 28 (1997) 81-93

Acknowledgements
Support was provided by the Bureau of Workers
Compensation, the Industrial Commission of Ohio,
Vestibular Development Fund and NIDRR, RERC
grant #H133E30009. The invaluable assistance of
Patrick Sparto, Professor Kamran Barin, Professor
William S. Marras and Professor Hans-Joachim
Wilke is greatly appreciated.
References
[1] J.W. Frymoyer, Magnitude of the problem, in: J.N. Weinstein, S.W. Weisel (Eds.), The Lumbar Spine, W.B. Saunders, Philadelphia, 1990.
[2] J.W. Frymoyer, M.H. Pope, J.H. Clements, D.G. Wilder, B.
MacPherson, T. Ashikaga, Risk factors in low back pain: An
epidemiological survey, J. Bone Jt. Surg. Ser. A 65 (2)
(1983) 213-218.
[3] D.K. Damkot, M.H. Pope, J. Lord, J.W. Frymoyer, The
relationship between work history, work environment and
low-back pain in men, Spine 9 (4) (1984) 395-399.
[4] J.D.G. Troup, J.W. Martin, D.C.T. Lloyd, Back pain in
industry: A prospective survey, Spine 6 (1981) 61-69.
[5] M. Parnianponr, M. Nordin, N. Kahanovitz, V. Frankel, 1988
Volvo award in biomechanics. The triaxial coupling of torque
generation of trunk muscles during isometric exertions and
the effect of fatiguing isoinertial movements on the motor
output and movement patterns, Spine 13 (9) (1988) 982-992.
[6] J.L. Wang, M. Parnianponr, A. Shirazi-Adl, A.E. Engin,
Failure criterion of collagen fiber: Viscoelastic behavior simulated by using load control data, Theor. Appl. Fract. Mech.
27 (1997) 1-12.
[7] S.A. Goldstein, T.J. Armstrong, D.B. Chaffm, L.S. Matthews,
Analysis of cumulative strain in tendons and tendon sheaths,
J. Biomech. 20 (1) (1987) 1-6.
[8] J.L. Wang, M. Parnianpour, A. Shirazi-Adl, A.E. Engin, The
review and evaluation of viscoelastic models for collagen
fiber during constant strain rate loading, Biomed. Eng. Appl.
Basis Commun. 9 (1) (1997) 5-19.
[9] S. Li, A. Patwardhan, F. Amiriouche, R. Havey, K. Meade,
Limitations of the standard linear solid model of intervertebrai discs subject to prolonged loading and low-frequency
vibration in axial compression, J. Biomech. 28 (1995) 779790.
[10] A. Shirazi-Adl, M. Pamianpour, Finite element model studies
in lumbar spine biomechanics, in: C.T. Leondes (Ed.),
Biomechanics Systems Techniques and Application, Gordon
and Breach Publishers, 1997, in press.
[11] L.G. Gilbertson, V.K. Goel, W.Z. Kong, J.D. Clausen, Finite
element methods in spine biomechanics research, Crit. Rev.
Biomed. Eng. 23 (5-6) (1995) 411-473.
[12] T. Belytschko, R. Kulak, A finite element method for a solid
enclosing a inviscid incompressible fluid, J. Appl. Mech. 40
(2) (1973) 609-610.

[13] D.R. Furlong, A.N. Palazotto, A Finite element analysis of


the influence of surgical herniation on the viscoelastic properties of the intervertebral disc, J. Biomech. 16 (10) (1983)
785-795.
[14] A. Lamer, Shirazi-Adl, G. Drouin, G. Mclntyre, Viscoelastic
finite element stress analysis of a lumbar disc-body unit in
compression, 33rd Annual Meeting Orthopaedic Research
Society, 1987, p. 430.
[15] Y.K. Liu, G. Ray, Systems identification scheme for the
estimation of the linear viscoelastic properties of the intervertebral disc, Aviat. Space Environ. Med. (1978) 175-177.
[16] Y.M. Lu, V.M. Gharpuray, W.C. Hutton, The effect of
posture, loading rate and diurnal fluid changes on the propensity to disc prolapse, Spine (1997), in press.
[17] M. Argoubi, A. Shirazi-Adl, Poroelastic creep response analysis of a lumbar motion segment in compression, J. Biomech.
29 (10) (1996) 1331-1339.
[18] B.R. Simon, J.S.S. Wu, M.W. Carlton, J.H. Evans, L.E.
Kazarian, Structural models for human spinal motion segments based on a poroelastic view of the intervertebral disk,
J. Biomech. Eng. 107 (1985) 327-335.
[19] B.R. Simon, J.S.S. Wu, M.W. Carlton, L.E. Kazarian, E.P.
France, J.H. Evans, O.C. Zienkiewicz, Poroelastic dynamic
structural models of rhesus spinal motion segments, Spine 10
(6) (1985) 494-507.
[20] J.P. Laible, D.S. Pilaster, B.R. Simon, M.H. Krag, M. Pope,
L.D. Haugh, A dynamical material parameter estimation
procedure for soft tissue using a poroelastic f'mite element
model, J. Biomech. Eng. 116 (1994) 19-29.
[21] L.E. Kazarian, Creep characteristic of the human spinal
column, Orthop. Clin. North Am. 6 (1975) 1.
[22] A. Shirazi-Adl, Stress analysis of the lumbar disc-body unit
in compression: A three dimensional nonlinear finite element
study, Spine 9 (2) (1984) 120-134.
[23] Hibbitt, Karlsson, Sorensen, ABAQUS User's Manual, version 5.3., 1993.
[24] A. Shirazi-Adl, A.M. Ahmed, S.C. Shrivastava, Mechanical
response of a lumbar motion segment in axial torque alone
and combined with compression, Spine 11 (9) (1986) 914927.
[25] A. Shirazi-Adl, G. Drouin, Load-bearing role of facets in a
lumbar segment under sagittal plane Ioadings, J. Biomech. 20
(6) (1987) 601-613.
[26] R.C. Haut, R.W. Little, A constitutive equation for collagen
fibers, J. Biomech. 5 (5-A) (1972) 423-430.
[27] D. Brickley-Parsons, M.J. Glimcher, ls the chemistry of
collagen in intervertebral discs an expression of Wolff's law?
A study of human lumbar spine, Spine 9 (2) (1984) 148-163.
[28] D.R. Eyre, Biochemistry of the intervertebral disc, Int. Rev.
Connect Tissue Res. 8 (1979) 227-287.
[29] J.O. Galante, Tensile properties of the human lumbar annulus
fibrosus, Acta Orthop. Scand. Suppl. 100 (1967).
[30] R.L. Stevens, R. Ryvar, W.R. Robertson, J.P. O'Brien, H.K.
Beard, Biological changes in the anulus fibrosus in patients
with low-back pain, Spine 7 (3) (1982) 223-233.
[31] D.L. Skaggs, M. Weidenbanm, J.C. Iatridas, A. Ratcliffe,
V.C. Mow, Regional variation in tensile properties and bio-

J.L. Wang et al./ Theoretical and Applied Fracture Mechanics 28 (1997) 81-93

[32]

[33]

[34]

[35]

[36]

[37]

chemical composition of the human lumbar anulus fibrosus,


Spine 19 (12) (1994) 1310-1319.
S. Li, Response of human intervertebral disc to prolonged
axial loading and low-frequency vibration, Doctoral Dissertation, Department of Mechanical Engineering, University of
Illinois at Chicago, 1994.
B.A. Best, F. Guilak, L.A. Setton, W. Zhu, F. Saed-Nejad,
A. Ratcliffe, M. Weidenbaum, V.C. Mow, Compressive mechanical properties of the human annulus fibrosus and their
relationship to biochemical composition, Spine 19 (2) (1994)
212-221.
J.C. latridas, L.A. Setton, D.C. Blood, M. Weidenbanm,
V.C. Mow, Mechanical behavior of the human nucleus pulposus in shear, in: 41st Annual Meeting, vol. 20 (1), Orthopaedic Research Society, Orlando, Florida, 1995.
D.S. McNally, M.A. Adams, Internal intervertebral disc mechanics as revealed by stress profllometry, Spine 17 (l)
(1992) 66-73.
P. Neumann, T.S. Keller, L. Ekstrom, T. Hansson, Effect of
strain rate and bone mineral on the structural properties of
the human anterior longitudinal ligament, Spine 19 (2) (1994)
205-211.
P. Neumann, T.S. Keller, L. Ekstrom, L. Perry, T. Hansson,
D.M. Spengler, Mechanical properties of the human lumbar
anterior longitudinal ligament, J. Biomech. 25 (10) (1992)
1185-1194.

[38] A. Shirazi-Adl, personal communications, 1996.


[39] A. Shirazi-Adl, G. Drouin, Nonlinear gross response analysis
of a lumbar motion segment in combined sagittal loadings, J.
Biomech. Eng. 110 (1988) 216-222.
[40] H.S. Ranu, A.I. Denton, R.A. King, Pressure distribution
under an intervertebral disc: An experimental study, J.
Biomech. 12 (1979) 807-812.
[41] W. Koeller, W. Meier, F. Hartmann, Biomechanical propertics of human intervertebral discs subjected to axial dynamic
compression: A comparison of lumbar and thoracic discs, J.
Biomech. 9 (7) (1984) 725-733.

93

[42] A.F. Mak, The apparent viscoelastic behavior of articular


cartilage: The contributions from the intrinsic matrix viscoelasticity and interstitial fluid flows, J. Biomech. Eng. 108
(1986) 123-130.
[43] M.A. Adams, D.W. McWillan, T.P. Green, P. Dolan, Sustained loading generates stress concentrations in lumbar intervertebral discs, Spine 21 (4) (1996) 434438.
[44] M.A. Adams, P. Dolan, Recent advances in lumbar spinal
mechanics and their clinical significance, Clin. Biomech. l0
(l) (1995) 3-19.
[45] M.A. Adams, P. Dolan, Time-dependent changes in the
lumbar spine's resistance to bending, Clin. Biomech. 11 (4)
(1996) 194-200.
[46] R.J. Hindle, M.J. Pearcy, A. Cross, Mechanical function of
the human lumbar interspinous and supraspinous ligaments,
J. Biomed. Eng. 12 (1990) 340-344.
[47] L.H. Yahia, J. Audet, G. Drouin, Rheological properties of
the human lumbar spine ligaments, J. Biomed. Eng. 13
(1991) 399-406.
[48] M.A. Adams, W.C. Hutton, The effect of fatigue on the
lumbar intervertebral disc, J. Bone Jr. Surg. Set. B 65 (2)
(1983) 199-203.
[49] N. Brown, C. Saputa, J. Black, Young's modulus of living
human bone, Trans. 27th Annual Meeting of the Orthopaedic
Research Society, vol. 6, 1981, p. 41.
[50] A.H. Burstein, D.T. Reilly, M. Martens, Aging of bone
tissues: Mechanical properties, J. Bone Jt. Surg. Set. A 58
(1976) 82-86.
[51] D.R. Carter, W.E. Caler, Uniaxial fatigue of human cortical
bone. The influence of tissue physical characteristics, J.
Biomech. 14 0981) 461-470.
[52] F.G. Evans, Mechanical Properties of Bone, Springfield, IL,
1973.
[53] D. Lindahl, Mechanical properties of dried defatted spongy
hone, Acta Orthop. Scand. 47 (1975) I 1-19.
[54] H. Yamada, Strength of Biological Materials, Williams and
Wilkins, Baltimore, 1970.

Das könnte Ihnen auch gefallen