Sie sind auf Seite 1von 15

ARTICLE IN PRESS

International Journal of Industrial Ergonomics 38 (2008) 410424


www.elsevier.com/locate/ergon

Virtual optimisation of car passenger seats: Simulation of static and


dynamic effects on drivers seating comfort
A. Siefert, S. Pankoke, H.-P. Wolfel
Wolfel Beratende Ingenieure GmbH+Co. KG, Max-Planck-Strasse 15, 97204 Hochberg, Bavaria, Germany
Received 30 July 2007; accepted 28 August 2007
Available online 22 January 2008

Abstract
The virtual investigation of static and dynamic effects on seating comfort requires the application of an adequate human model. An
appropriate seat model considering static and dynamic properties of the structure, the foam and the trim is needed to perform an
optimisation for a lower load level on the driver. The evaluation of the seating comfort must be divided into a static and a dynamic part.
For the computation of the relevant physical quantities with the human model CASIMIR and a detailed seat model, the nite-element
solver ABAQUS (ABAQUS Inc., http://www.abaqus.comwww.abaqus.com) is used.
To reect a real driving situation, in the rst step the human model is adapted to the right posture, which is given by the inclination of
the cushion and the backrest. The seating process is then computed by the load due to gravity. The static comfort is mainly evaluated by
the seat pressure distribution. Results such as the H-point and the meat-to-metal value can give additional important informations for
the ergonomic and structural design of the seat. As the model reects the nonlinear properties and the nite-element solver considers the
effects out of nite displacements and contact, a good correlation with measurement is achieved.
The dynamic simulation is carried out by a unit excitation of the seat slides at the clamping points. To consider frequency-dependent
properties of foam, structure and the human body, the computation uses an implicit solver. Therefore the model is linearised after the
nonlinear static seating process.
Dynamic comfort is evaluated by the seat-transfer function. The presented numerical method leads to a good correlation with the
measurements. Superposing the results with real excitation signals enables the estimation of the dynamic loads as muscle or intervertebral
disc forces on the driver.
Altogether this method, in an early state of the development enables the user to optimise a car passenger seat structure due to the static
and dynamic comforts. Considering boundary conditions as higher load amplitudes and accelerations, the advantages of virtual
development can also be applied for construction vehicle seats.
Relevance to industry
The present method allows the evaluation of static and dynamic comforts in a virtual phase of seat development. Besides the reduction
of time and costs, the application of the simulation enables the testing of new materials and ways of construction with low investment.
r 2007 Elsevier B.V. All rights reserved.
Keywords: Human FE model; Static comfort; Dynamic comfort; Seat optimisation; Driver performance

1. Introduction
In many industrial branches the development process of new
products is changing from the experimental testing method to
Corresponding author. Tel.: +49 931 49708 180;
fax: +49 931 49708 650.
E-mail address: siefert@woelfel.de (A. Siefert).

0169-8141/$ - see front matter r 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.ergon.2007.08.016

the digital prototyping method. This procedure can be


observed especially in the automotive industry. There the
market always disposes the OEM to reduce the time and cost
invested on a new product. Corresponding approaches can be
observed in the construction and the commercial vehicle
branches accordingly. There the effort is furthermore increased
by the almost new EU standard 2002/44/EG, which requires
the manufacturers to limit vibration exposure to occupants.

ARTICLE IN PRESS
A. Siefert et al. / International Journal of Industrial Ergonomics 38 (2008) 410424

This process of virtual development started more than a


decade ago. As the evolution of the numerical simulation
tools was happening simultaneously, the solution of
complex problems is possible today. Computations with
nonlinearities due to material, geometry and contact can
nowadays be carried out with a high quality of prognosis.
Correspondingly, the challenge for all complex simulations consists in modelling the structure and its properties.
Compared with requirements of mechanical systems, a
seating comfort simulation is overlaid with the denition
and the mapping of the human body and its behaviour.
In the eld of mapping the human body and its
properties with a nite-element (FE) model, different
approaches have been published for seat comfort simulation within the last years. Thereby, only parts of the human
body or the whole body have been modelled. Particular
works to build the human thigh with pelvis and femur have
been published by Moes and Horvath (2002), Verver (2004)
and Mergl (2006). The presented models have been used for
static comfort simulations.
For whole-body human FE models, publications of Choi
et al. (2006) and Pankoke (2003) exist. Pankoke presented
the human model CASIMIR, which is used in the
simulation method of this study.
Compared to experimental testing, the application of a
simulation method with a validated human model provides
the possibility of investigating the loads on the driver and
evaluating his performance.
Based on the set-up of the complete model, the simulation
of seating comfort can be divided into three steps:





Seat structure
Non-occupied seat
Occupied seat




Non-occupied seat:
Occupant:

Consequently, for computation with a high prognosis a


highly sophisticated model of the occupant is necessary.
The FE model CASIMIR meets these requirements.
Additionally, the evaluation of the workers performance is possible as the dynamic loads in the lumbar
spine and in the muscles in the abdominal cavity can be
calculated.
2. Seat modelling
2.1. Seat structure
The task always starts with the FE modelling of the seat
structure. A seat structure of an upper-class automobile is
shown here (Fig. 1).
The model must always contain all components with a
strong inuence on the resulting values. Therefore an
expansion of the structure by the armrest or operation
elements such as the steering wheel and gear lever is
required.
The structural model is generated and validated within
the scope of the usual development process. Especially in
the automotive area the application of improved crash
models is economically efcient. The set-up of the
structural model can be subdivided into the following
parts:






These steps correspond to the procedure of the experimental testing. In the rst step, the seat structure itself is
modelled and validated to its nonlinear behaviour in joints,
adjusting mechanism and damping behaviour.
The non-occupied seat model thus is an expansion of the
foam cushions and the seat cover and includes the
interaction of these parts to the seat structure.
Finally, the occupied seat denes the whole system.
Because of the inuence on resulting quantities, e.g. the
seat pressure distribution, the interaction denition between the human model and foam cushions is very
important. Within this nal step, the signicance of the
occupant for the vibrational behaviour of the seat becomes
apparent only when considering the acting masses:
25 kg
75 kg

Additionally, the fact that the human body represents a


dynamic system with natural frequencies appearing in the
relevant frequency domain (until approx. 30 Hz) points out
that a decoupled simulation is not reasonable.

411

Structural components
Coupling elements there between (e.g. joints)
Attached assemblies (e.g. airbag, height adjustment
motor)
Compliant interfaces for foam cushions (suspension
systems)

The supporting components as well as the supplements


are usually modelled based on CAD data with shell
elements, beams and springs. Thereby the mass of attached
components, e.g. the adjustment engine, is modelled via
rigid bodies and point masses.
The coupling between structural parts is dened by
connectors that represent the kinematics of the seat.

Suspension system
backrest
(springs+beams)
Structural frame
(shells)
Lordosis support
(shells)
Attached component: height
adjustment motor (rigid body)

Fig. 1. Model of seat structure (DaimlerChrysler, Pankoke et al., 2005).

ARTICLE IN PRESS
412

A. Siefert et al. / International Journal of Industrial Ergonomics 38 (2008) 410424

Seat rail adjustment


(connector element)

Hinge backrest inclination


(connector element)

Fig. 2. Possible variations of seat settings.

Besides the inuence on dynamic behaviour, the connector


elements enable different seat settings, which dene the
practical operation mode (Fig. 2).
If the seat simulation is carried out after a hardware
prototype has been built, the structural model should be
validated by the results of an experimental modal analysis
(EMA). The evaluation is carried out by natural frequencies, mode shapes and damping ratios. The differences
between simulation and experimental testing should be
only a few per cent (Pankoke et al., 2005).
According to the simulation of the static and dynamic
seating comforts, the impact of the suspension systems
(rear and base) is very high. This is based on their
compliance compared to other parts and the direct
interaction with the cushions. Furthermore, the variation
possibilities, e.g. the lordosis support (Fig. 3), must be
considered as their effect on static comfort and on the
drivers performance is very strong.
2.2. Foam cushions and seat cover
2.2.1. Modelling
The modelling of the foam cushions can be done on the
basis of the CAD geometry of the out-of-tool parts.
Alternatively, the STRAK data can be used to represent
the A-surface after the covering process. Within the
modelling of the foam cushion (Fig. 4), the option of
adjusting the thickness should be considered. Besides the
material choice, this provides a second possibility to
optimise the seating comfort.
The foam cushions are modelled with solid elements.
Investigations of different element types have shown that
the application of second-order tetrahedron elements meets
the requirements of computation. This has the advantage
of an automatic meshing being possible and the needed
modelling time is dramatically reduced. However, in order
to guarantee the right contact nodal forces, and especially

Fig. 3. Possible variations of seat lumbar support.

for contact problems, a modied element approach must


be applied.
The seat cover can be modelled with a shell layer, dened
by the contact faces of the foam cushions. As the nodes of
the foam and the cover are merged a shear force
transmission between these parts is possible, which does
not denitely reect the real behaviour. This approach can
represent the stiffening effects of the trimming process,
though simplied. Compared to an alternative modelling
method, including the trimming process (Cakmak et al.,
2006), this procedure is very time efcient.
2.2.2. Material properties
Usually, several different PUR-foam materials are
applied in a modern passenger car seat. For example, in

ARTICLE IN PRESS
A. Siefert et al. / International Journal of Industrial Ergonomics 38 (2008) 410424

413

Fig. 5. Experimental set-up for foam material tests.

Fig. 4. Model of cushions with tetrahedron elements (Pankoke et al.,


2005).

the cushion and backrest ank areas, the application of a


stiffer mixture is preferred in order to guarantee more
lateral hold during rolling turn.
The static material properties under compressive
loadwhich is the dominating real load caseare
determined according to the corresponding test standard
(ISO 3386-1, 1986). The tests are done as uni-axial
compression tests on a hydraulic test stand, with measurement of the forces carried out along the deformation path
(stroke). In order to estimate the inuence of the spread for
standard factory models, three specimens each should be
investigated.
As a material approach for the description of foams, a
hyper-elastic law is used in ABAQUS. The elasticity of the
material is described via the potential energy U of elastic
deformation. The applied potential function for foams
(Eq. (1)) shows a deviation of the approach according to
Ogden, considering the nearly full compressibility of PURfoams.
Strain energy potential of compressive foams:


N
X
2mi ^ ai ^ ai ^ ai
1
el ai bi
U
l l2 l3  3 J
 1 .
(1)
bi
a2 1
i1 i
The potential energy U is dened by the following
parameters: mi are the coefcients of initial shear modulus,
l13 the principal stretches, ai the standard material
parameter, bi the coefcients for degree of compressibility
and Jel the elastic volume ratio.
The free material parameters mi, ai and bi are determined
within an iterative identication, where the average value
out of loading and unloading is the objective.
The dynamic properties of the materials are described by
their complex stiffness, which must be dened depending

on loading frequency and static pre-deformation. The


dynamic tests performed in this regard are thus carried out
with varying static pre-displacement levels. The specimens
are excited with harmonic signal in the domain of 130 Hz.
Fig. 5 shows the experimental set-up for the static and
dynamic tests of the foam material.
The denition of the dynamic behaviour of the foams is
based on the linearisation around the respective operating
point (static pre-strain). The dynamic effects for the
analysis in the frequency domain are dened by indication
of frequency-dependent terms for modulus of storage and
loss. The comparison of the simulation results for the
dynamic stiffness with the executed tests, performed for
validation, shows excellent correlation.
Fig. 6 shows the comparison for the static stressstrain
curve of a standard seat-cushion-foam under uni-axial load.
Fig. 7 displays the relation of the experiment and the
computation for the complex dynamic Youngs modulus in
the frequency domain from 1 to 25 Hz.
Besides the uni-axial investigations, more realistic tests
with a volumetric load by a spherical punching tool are
carried out. Thereby the static pre-stress is because of the
weight of the indenter tool. The dynamic load is then
applied by a base point excitation. The evaluation is carried
out by the transfer function between the amplitude at
the foam surface and the excitation amplitude. Fig. 8 shows
the results of the measurement and the simulation.
Although the foam material parameters are only determined by a uni-axial compression load case, the volumetric
test with shear effects at the lateral anks shows a very
good correlation.
3. Human FE model CASIMIR
3.1. Requirements for occupant models
Today, in many elds various human models are used in
distinct applications (Digital Human Modeling Conference,
2002; Wolfel et al., 2004). Most common is the use of CADbased human models such as RAMSIS (Human Solutions)
or UGS Classical Jack, which are applied for the design of
the interior package. Within the application of these models,

ARTICLE IN PRESS
414

A. Siefert et al. / International Journal of Industrial Ergonomics 38 (2008) 410424

Fig. 6. Comparison of measurement and simulation for static load.

Fig. 7. Comparison of measurement and simulation for dynamic Youngs modulus.

the mechanical properties of the package and the human,


beside the kinematics, are not considered. Accordingly,
distinctions from the real conditions can occur. Especially
the inuence of compliant materials as foam of the seat and
soft tissue of the human body, which cause great deformations under static loads, is not considered. The procedure to
support these geometric methods with measure quantities,
such as the H-Point, is not possible within a complete virtual
development process.

Consequently, for this procedure the application of


human models, considering the mechanical behaviour,
represents an improvement. This effect comes along with
the advantage of carrying out simulations to evaluate static
and dynamic seating comforts.
Within the group of models including the mechanical
properties, one must differentiate between phenomenological
and anatomical set-ups. Phenomenological models serve
to reproduce one single characteristic of man precisely

ARTICLE IN PRESS
A. Siefert et al. / International Journal of Industrial Ergonomics 38 (2008) 410424

415

Fig. 8. Comparison of measurement and simulation for a dynamic test with a spherical punch tool.

m1

qK

q1
Knoblauch, 1992

Fairley &
Griffin, 1989

q3

m3

m2

q2
DIN 45676, 2003

qS

q4

m4

q5

m5

qS

m0

q1

m0

qS

m1

q2

m2

q3

m3

m0

Fig. 9. Phenomenological models of a man in the sitting posture.

described (a phenomenon). The advantage of a simple


model conguration is limited to the fact that these
models are not applicable for the simulation of properties
beyond this phenomenon. Most of these phenomenological
models represent the whole body in the standing or
in a sitting posture. Fig. 9 shows models from Fairley
and Grifn (2003), Knoblauch (1992) and DIN 45676
(2003).
On the contrary, anatomical models represent man on
the basis of his physiological characteristics. Depending on
the desired results, these models can describe the whole
body or a part of it. Fig. 10 displays models of Moes and
Horvath (2002), Verver (2004) and Mergl (2006), which
represent a compliant thigh including pelvis and femur and
have been applied for seat comfort simulations.
If appropriately modelled and validated, anatomical
models have the capability of predicting the mechanical
behaviour of physical quantities that are not part of the
input data for set-up generation.

Fig. 10. Anatomical models of the compliant thigh with pelvis and femur.

Thus, these models provide the possibility of evaluating


especially the dynamic seating comfort via vibration parameters sensed by man. Furthermore, the results can be

ARTICLE IN PRESS
416

A. Siefert et al. / International Journal of Industrial Ergonomics 38 (2008) 410424

taken into account with subjective assessment (e.g. forces


in the lumbar spine, accelerations of different abdominal
parts).
The following preconditions must be fullled by an
anatomical model to allow a numerical evaluation of static
and dynamic seating comforts according to whole-body
behaviour:







Model geometry derived from human anatomy


Masses, stiffness and damping properties dened via
physiological data
Detailed model of the lumbar spine
Consideration of static muscle activation for the static
equilibrium of the upper part of the body under gravity
load
Consideration of dynamic muscle effect representing
frequency-dependent stiffness and damping properties

The requirement for a detailed model of the lumbar


spine is based on the fact that this part of the body is the
most important for the whole-body vibrations of man.
After the model generation, the set-up and its properties
must be validated with test person measurements. The
most important parameters for static and dynamic seating
comforts are as follows:
Fig. 11. CASIMIR model for man of the 50th percentile.






Static seating pressure distribution


Dynamic mass with excitation of the seating man at the
buttocks
Transfer functions of the excitation from the buttocks to
the head
Transfer functions of the excitation from the buttocks to
further parts, particularly to measuring points at
shoulder and lumbar spine

3.2. Occupant model CASIMIR


CASIMIR represents a dynamic, anatomical predictive
FE model of a man in a sitting posture (Siefert et al., 2006).
Currently, it is available for the FE code ABAQUS. In
the course of its application for seating comfort, the
following components have been developed recently:







Detailed model of the lumbar spine including frequencydependent damping properties of the intervertebral discs
Detailed model of the relevant abdominal and dorsal
musculature
Dynamic model of the abdominal cavity
Detailed skeletal model of pelvis, femur, tibia, cervical
spine, head and arms
Model of the body soft tissue in the relevant contact
regions to the seat, including static and dynamic tissue
properties

Whereas Fig. 12 shows the detailed parts of the


abdominal musculature and the lumbar spine with intervertebral discs, Fig. 11 displays a view of the complete

model. The presented 50th percentile reects a man with


averaged anthropometric values.
Within the detailed model of the abdominal musculature, all relevant muscles are considered with nonlinear and
frequency-dependent spring and dashpot elements. The
muscle activation for the posture under gravity is
computed by an optimisation routine following the
principle of minimising the required energy. The determined values correspond to electromyographic measurements.
The material behaviour of the different components of
the lumbar spine as annulus and nucleus are dened over
literature values. All bone components are modelled as
rigid bodies, as their stiffness, compared to the compliant
parts as muscles and tissue, is very high. The set-up of the
whole model and its components are described in detail by
Buck (1997) and Pankoke (2003).
3.3. Adaptability: percentile and posture
For the widespread application of a human model in the
evaluation of the static and dynamic seating comforts, a
modulation of the set-up must be possible. Thereby,
percentile, body height and mass on the one hand, which
have been described statistically, and the posture on the
other hand must be considered. Both parameters have
signicant inuence on the interaction with the seat and
the static pressure distribution hereupon as well as on the
dynamic whole-body-vibration behaviour.

ARTICLE IN PRESS
A. Siefert et al. / International Journal of Industrial Ergonomics 38 (2008) 410424

The range of required percentiles depends on the desired


customer clientele. While for the seating comfort of luxury
cars the overall spectrumstarting with the 5%-female
(percentile f05) via the 50%-male (m50) until the 95%-male
(m95)is signicant, the range decreases for small cars to
the lower percentiles.
According to such specications and by means of an
individualisation algorithm, covering a choice of seven
anthropometric values, as described in Pankoke et al.
(1998), CASIMIR can be adapted to the value combination
of percentiles or even individuals. Fig. 13 shows the

417

CASIMIR family with the average denition m50 and


the limits f05 and m95.
While the decision for the percentile is related to the
customer, the global adaptation of the posture is primarily
dened by the seat and its possible adjustment ranges,
which are given by the package design. The adjustment of
the CASIMIR posture is effected via the denition of
relative joint angles and a related adaptation of the model
parameters (Pankoke, 2003).
Additionally, it is possible to dene posture variations
that depend on the driver himself or that can vary during
long driving distances. CASIMIR has the potential to
consider a lordotic or kyphotic bearing of the lumbar
spine, see Fig. 14. Thereby, it is possible to model a
concentrated or a more relaxed posture of the driver on the
one hand and investigate the inuence of common lumbar
support systems of seats on the other.
4. Static seating and static comfort
4.1. Interaction between occupant and seat model

Fig. 12. CASIMIR: Detailed model of abdominal muscles (left) and


lumbar spine (right).

For a realistic reproduction of static seating under


gravity loading, an appropriate contact denition between
the seat and the driver is essential, whereby the effects of
friction must be considered.
The interaction between the CASIMIR and the foam
surface is dened by two independent contact pairs,
backrest and back on the one hand and cushion and
buttocks on the other hand. The chosen algorithm is a socalled masterslave contact, where internal variables are
created for each node of the slave surface. The applied
Lagrange multipliers are discontinuous, as a hard

Fig. 13. CASIMIR familyf05m50m95 (left to right).

ARTICLE IN PRESS
418

A. Siefert et al. / International Journal of Industrial Ergonomics 38 (2008) 410424

shape, the thickness and the material properties of the


foam cushions and on the compliance of the suspension
system.
Here the simulation results for a luxury passenger car are
shown




Fig. 14. CASIMIR: Posture variation of lumbar spine: lordotic (left) and
kyphotic (right).

contact formulation is chosen. The computation of the


contact problem is carried out by considering the friction
by a coefcient of 0.3.
The pre-analysis positioning of CASIMIR, i.e. in the
pre-processor, is carried out in two steps: First, the posture
of CASIMIR is synchronised with the posture of the seat.
This is dened by the opening angle between the backrest
and seat cushions (correlation angle of cushion/angle of
backrest inclination). Second, the CASIMIR model is
positioned levitating above the seat with disappearing
interspaces to the cushion or backrest.
4.2. Static seating simulation
The computation of the seating process is carried out
under gravity loading in the negative z-direction. As a
result of the large displacements caused by the compliance
of the foam cushions, the simulation is geometrically
nonlinear. The seat structure is xed at the restraint points
of the seat rail. The boundary conditions of the occupant
model reproduce real possible movements. Accordingly,
feet and hand movements are dened by the operation
environment, that is, e.g. a translation of the feet in the
x-direction (movements footwell) as well as a rotation of the
hands around the y-axis (hold steering wheel) is possible.
The main results of the static simulation are:







Vertical displacement on the seat cushion:


Horizontal displacement on the backrest:

40.25 mm
21.95 mm

The computation correlates well with the measurements.


The analysis of the hip joint location (Fig. 15) gives
information about the position of the occupant model
after the static seating process. Thereby, it is possible to
enhance the package design of passenger cabins in the
digital development phase. The advantage, compared
to known procedures that are based only on the geometric surfaces of the components, is the consideration
of the materials mechanical properties, especially their
compliance.
4.4. Pressure distribution
The pressure distributions of the cushion and the
backrest are the most important evaluation values in static
seating comfort. Recently, they have increasingly become
part of more development specications for seats the Tier-1
suppliers have to meet. Early-phase prediction is possible
with static seating simulations. Fig. 16 shows exemplary
results with CASIMIR, percentile m50 (Pankoke and
Siefert, 2007) of a car passenger seat.
According to the posture variations of CASIMIR (due
to, e.g., a lordotic and kyphotic lumbar spine or a thigh
opening), an enhanced evaluation of the seat structure on
the criteria of the pressure distribution is possible. This
makes an optimisation of special structural features, e.g.
the lumbar spine support or the stiffness of cushion anks,
due to driver postures possible. Fig. 17 shows the
qualitative inuence of posture variation on seat pressure
distribution due to the opening of the thighs.

Displacement in the z-direction on the seat cushion


Displacement in the x-direction on the backrest
Location of the hip joint
Contact pressure on the seat cushion
Contact pressure on the backrest

4.3. Cushion displacement and location of the hip joint


The displacement at the surface of the cushion enables
one to validate the simulation compared to the real seating
behaviour. Obviously, the displacement depends on the

Fig. 15. Position of the hip joint after static seating (dark).

ARTICLE IN PRESS
A. Siefert et al. / International Journal of Industrial Ergonomics 38 (2008) 410424

419

Fig. 16. Pressure distributions of backrest (left) and cushion (right).

Fig. 17. Pressure distribution of cushion with opened thighs.

5. Seat vibrations and dynamic comfort


5.1. Evaluation of dynamic comfort
In the present procedure of seat development, besides
subjective assessments dynamic comfort is mostly evaluated by analysing of the seat-transfer function. The seattransfer function provides dynamic response on the surface
of the seat related to the vibration excitation at the seat
slide; thus it is a frequency-dependent complex function,
i.e. it shows magnitude and phase. Depending on the
direction of excitation and the direction and location of
response, it is differentiated between seat-transfer functions
(z) and backrest-transfer function (x and y).
The seat-transfer function is only one element in the
chain of dynamic comfort analysis. An evaluation of the
vibrational exposure on the driver can be carried out only
in combination with the excitation signals on the seat slide.
Then the superposing considers the application area and

the set-up of the whole vehicle, which strongly differs


between compact and luxury cars.
The simulation of the complete vehicle chain is in an
early-development state very complex. Therefore, an
alternative procedure is the evaluation of the seat in
addition to the experiences gained from previous seat type
series. Results of a previous benchmark analysis may then
serve as source for seat-transfer function.
Following this procedure a numerical computation of
seat vibrations must be capable of calculating the seattransfer function. Thereupon the user has, in the virtual
environment, the possibility of modifying structural parameters inuencing dynamic comfort.
Furthermore, an analysis of the dynamic loads on the
lumbar spine represents an important comfort aspect.
Especially in the case of long-distance drivers, such as taxi
drivers and travelling salesman, this is very important for
an evaluation of the performance. Accordingly, in addition
to the global dynamic behaviour, an occupant model must
reect the ow of forces caused by vibrational loads in this
area.
5.2. Dynamic mass
For a numerical computation of the seat-transfer
function, the applied human model must full the sodyn
called dynamic mass M
(Eq. (2)), an individual property
of the driver.
Dynamic mass:
M

dyn

FS
.
q S

(2)

Thereby, the force F S is acting between an occupant


and a seat (Fig. 18 ), while the system is accelerated at
the seat slides. At the same time q S is the acceleration at the
foam surface. The characteristics of the used compliant
seat structure can be described by the dynamic stiffness
kS;dyn (Eq. (3) ), which is dened by the force F S and

ARTICLE IN PRESS
A. Siefert et al. / International Journal of Industrial Ergonomics 38 (2008) 410424

420

the displacements at the foam surface q S and at the seat


slide q 0 .
Dynamic seat stiffness:
k S;dyn O

F S
.
qS  q 0

(3)

The aforementioned requirement for the occupant model


can then be easily shown by deriving the seat-transfer
S (Eq. (4)).
function H
Seat-transfer function:
k S O
S O q S
.
H

q0 k S O  O2 MO

(4)

Accordingly, the following can be derived:





the occupant model is essential for determining the seattransfer function,


applied occupant models must reect the dynamic mass
of operators and
occupant models not fullling these criteria should not
be used for dynamic seating simulations.

The consistency of CASIMIRs dynamic mass is shown


in Fig. 19, where it is compared to a number of m50individual dynamic mass measurements on real humans in
the driving posture. Eliminating all stiffness effects of the
seat, the measurement of dynamic mass is carried out on a
rigid plate, used as a seat cushion and backrest.
5.3. Dynamic simulation

Fig. 18. Interface force F S and acceleration q S acting between the seat and
the occupant during dynamic excitation.

The dynamic computation follows the static seating


process. This procedure is necessary since both the seat and
the driver are nonlinear and thus the operating point
calculated by the static seating process is of decisive
signicance. The vibration simulation is carried out by a
direct solution in the frequency domain. This analysis
procedure is linear, i.e. that the system properties are
linearised around the calculated operation point dened by
the static simulation. The assumption of linearisation is

Fig. 19. Dynamic mass: CASIMIR result and individual measurements for percentile m50 sitting on a rigid plate.

ARTICLE IN PRESS
A. Siefert et al. / International Journal of Industrial Ergonomics 38 (2008) 410424

admissible in the case of relatively small amplitudes under


vibration excitation.
Compared to modal analysis methods, the applied direct
solution has the advantage that frequency-dependent
behaviour and discrete damper elements can be considered,
which is important for modelling both foam and the driver.
Furthermore, it additionally allows the denition of
frequency-dependent viscoelastic material parameters, considering an appropriate solution also for higher vibrational
amplitudes.
The excitation is applied as base point excitation by
inhomogeneous boundary conditions at the four assembly
points of the seat slide at the vehicle body. In this example,
vertical excitation is investigated rst, with the considered
frequency domain being 120 Hz.
The simulation results for the seat-transfer function of
the cushion in the z-direction (zz) and the backrest in the
x-direction (xz) under vertical excitation for a luxury
passenger car are given in Figs. 20 and 21. The evaluation
was carried out corresponding to measurements at the
foam surface. Respective measuring results for an individual test driver are opposed to the results obtained with the
CASIMIR simulation.
Generally, the characteristics of the transfer functions
are reproduced very accurately, especially regarding the
frequency ranges of arising resonances. This is a necessary
prerequisite for an interpretation of the vibration levels
inside the human body.
The evaluation of the load values on the lumbar spine is
carried out by the forces on the intervertebral discs or

421

respective muscles. As the simulation is carried out with a


unit displacement, the results must be superposed with the
real excitation signals at the seat slide. In Fig. 22, the result
in the intervertebral disc L4L5 in the time domain
according to a road trial is shown exemplarily. As limit
values for the internal loads of the human body often are
not dened, evaluation must be carried out by a qualitative
comparison of different seat designs.
6. Application with construction vehicle seats
Considering the special requirements of the construction
vehicles, the present simulation method can be applied with
some minor changes. Thereby the following issues represent a challenge for computation:




Higher accelerations and amplitudes of excitation


Strong horizontal excitation

Accordingly, the properties of the CASIMIR and the


seat model must be adapted for an excitation with a higher
intensity. Therefore the viscoelastic parameters of the
lumbar spine, the abdominal musculature, the foam and
the compliant structural parts must be updated by
corresponding measurement data.
Obviously, there are differences between the set-up
of a car passenger seat and a construction vehicle seat
due to the suspension and the structural frame. Nevertheless, is the question remains as to weather modelling
of a vehicle seat structure by a nite-element code

Fig. 20. Seat-transfer function seat cushion zz.

ARTICLE IN PRESS
422

A. Siefert et al. / International Journal of Industrial Ergonomics 38 (2008) 410424

Fig. 21. Seat-transfer function seat backrest xz.

Fig. 22. Load forces (horizontal and vertical) of intervertrebal disc between L4 and L5.

considering special requirements, e.g. an active suspension


system, is possible.
Another problem is the integration of the operation
elements, which have a strong inuence on dynamic
behaviour. As the general set-up of the model CASIMIR
is symmetric to the sagittal plane, this presetting does not
consider the asymmetric drivers environment.

Considering the lateral vibration excitations caused


thereby, the model CASIMIR has the possibility of turning
out the extremities, see Fig. 23.
As the participating muscle of the shoulderhand
arm system is at the moment only through joint stiffness and damping simplied implemented, the human
model must be expanded by a more detailed approach

ARTICLE IN PRESS
A. Siefert et al. / International Journal of Industrial Ergonomics 38 (2008) 410424




423

Does the simulated hip-joint location coincide with the assumed position of the
geometrical package manikin?
What inuence do special seat features (e.g.
lumbar support) have on seat pressure
distribution, and do their adjustment ranges
consider the operator percentiles?

Dynamics:




Fig. 23. CASIMIR: unsymmetric posture of arms and legs.

Does the (occupied) seat in its resonances


match with the expected excitation spectra
coming from the entire vehicle (i.e. vehicle
body with power train and axles)?
What amount of inuence do different
tuning approaches (e.g. foam thickness)
have on the occupied seat?
In which load levels are the forces in the
lumbar spine?

Furthermore, it has to be mentioned that the digital


phase provides the possibility of investigating completely
new set-ups for the seat structure, where the realisation of a
hardware prototype would be difcult.
7.2. Hardware phase

to consider the higher requirements for the construction


vehicles.
7. Process integration
At present, the development process consists of two
parts, the digital phase and the hardware phase. As the
comfort evaluation of the seat structure in the digital phase
is currently limited to informations on the structure itself
without considering a predictive human model, many
modications are performed and investigated in the
hardware phase. This procedure is very time-consuming
and sometimes can hardly be realised, as the development
time is shortened by the market requirements. For that
reason the system occupant-seat-vehicle must be analysed
and optimised in the digital development phase. The
respective amount of work can be divided into an
investigation of a model of the occupied seat with and
without the entire vehicle, whereby it is possible to work
double-tracked.
7.1. Digital phase
In the digital phase, an analysis model for the simulation
of the seat occupied by the driver is set up according to the
procedure described in the previous sections. Accordingly,
it is now possible to determine both static and dynamic
comfort quantities. Among others the simulation results
serve to answer the following questions:
Statics:

Which foam material stiffness and thickness


optimise the seat pressure distribution?

In the hardware phases, i.e. with rst prototypes of the


seat and/or the vehicle being available, verication
measures will be performed to further enhance the quality
of the seat model. Occupied seat simulation will then be
performed to support test personnel in identifying, understanding and reducing any unwanted phenomena, e.g. local
vibrations of structural components under certain ride
conditions.
8. Conclusion
Simulation of the occupied seat has strategic relevance
for all vehicle manufacturers and Tier-1 suppliers. In
particular, a consequent integration of the simulation of
the vibration system humanseatvehicle in the digital
development process shows the following advantages:
1. Avoidance of expensive late modications at the seat or
construction vehicle
2. Allowance for a possible optimisation of weight of the
seat structure and the seat coupling
3. Noticeable reduction of the test expenditure in case
of unwanted vibration phenomena arising, since the
analyses can be effected parallel to the testing phase
References
ABAQUS Inc. /www.abaqus.comS.
Buck, B., 1997. Ein Modell fur das Schwingungsverhalten des sitzenden
Menschen mit detaillierter Abbildung der Wirbelsaule und Muskulatur
im Lendenbereich. Shaker Verlag, Aachen, Germany.
Cakmak, M., Delavoye, C., Fuhrlinger, L., 2006. Sitzkomfortsimulation:
eine neue methode in der pilotanwendung (A seating comfort

ARTICLE IN PRESS
424

A. Siefert et al. / International Journal of Industrial Ergonomics 38 (2008) 410424

simulation: a new method in pilot application). Numerical Analysis


and Simulation in Vehicle Engineering. VDI-Berichte 1967, VDIVerlag, Dusseldorf, Germany.
Choi, H.Y., et al., 2006. Human Body Modeling for Virtual Seat Comfort
Testing. SAE DHMC, Lyon.
Digital Human Modeling Conference (Digitale Mensch-Modellierung),
2002. VDI-Berichte 1675, VDI Verlag, Dusseldorf, Germany.
DIN 45676, 2003. Mechanical impedances at the driving point and
transfer functions of the human body, DIN.
Fairley, T.E., Grifn, M.J., 2003. The apparent mass of the seated human
body: vertical vibration. Journal of Biomechanics 22, 8194.
Human Solutions GmbH /www.human-solutions.comS.
ISO 3386-1, 1986. Polymeric materials, cellular exibleDetermination of
stressstrain characteristics in compression. Part 1: low density
materials, ISO 1986.
Knoblauch, J., 1992. Entwicklung und Bau eines physikalischen Schwingungsmodells des sitzenden Menschen. Dissertation, Shaker Verlag,
Aachen, Germany
Mergl, C., 2006. Entwicklung eines Verfahrens zur Optimierung des
Sitzkomforts auf Automobilsitzen. Dissertation, TU Munchen,
Munich, Germany
Moes, N.C.C.M., Horvath, I., 2002. Using nite elements model of the
human body for Shape optimization of seats: optimization material properties. In: International Design ConferenceDesign 2002, Dubrovnik.
Pankoke, S., 2003. Numerische simulation des raumlichen Ganzkorperschwingungsverhaltens des sitzenden Menschen unter Berucksich-

tigung der individuellen Anthropometrie und Haltung. FortschrittBerichte VDI No. 522, VDI-Verlag, Dusseldorf, Germany.
Pankoke, S., Siefert, A., 2007. Virtual simulation of static and dynamic
seating comfort in the development process of automobiles and
automotive seats: application of nite-element-occupant-model
CASIMIR. In: SAE Digital Human Modelling Conference, Seattle,
USA.
Pankoke, S., Hofmann, J., Wolfel, H.P., 1998. Dynamic FE model of
sitting man adjustable to body height, body mass and posture, used for
calculating internal forces in the lumbar vertebral discs. Journal of
Sound and Vibration 215 (4).
Pankoke, S., Siefert, A., Breitfeld, Th., 2005. Numerische simulation von
sitz-schwingungen in oberklasse-pkw: einsatz des nite-elementemensch-modells CASIMIR (Numerical simulation of seat vibrations
in luxury-class automobiles: application of the nite-element-manmodel CASIMIR). Reifen-Fahrwerk-Fahrbahn, VDI-Berichte 1912,
VDI-Verlag, Dusseldorf, Germany.
Siefert, A., Delavoye, C., Cakmak, M., 2006. CASIMIR: Human niteelement-model for static and dynamic assessment of seating comfort.
In: IEA-Conference 2006, Maastricht, Netherlands
Verver, M.M., 2004. Numerical tools for comfort analyses of automotive seating. Dissertation, Technische Universiteit Eindhoven,
Eindhoven
Wolfel, H.P., Rutzel, S., Mischke, Ch., 2004. Biodynamische Modelle des
Menschen. In: VDI-Conference on Human Vibrations; Darmstadt,
Germany.