ARTICLE IN PRESS

International Journal of Industrial Ergonomics 38 (2008) 396–409

www.elsevier.com/locate/ergon

Modelling and simulation of a fore-and-aft driver’s seat suspension

system with road excitation
G.J. Steina,, R. Zahoranskýa, T.P. Gunstonb,1, L. Burströmc,2, L. Meyerd,3
a
Institute of Materials and Machine Mechanics, Slovak Academy of Sciences, Račianska 75, SK-831 02 Bratislava 3, Slovak Republic
b
Human Factors Research Unit, Institute of Sound and Vibration Research, University of Southampton, Highfield, Southampton SO17 1BJ, UK
c
National Institute of Working Life, Box 7654, SE-907 13 Umeå, Sweden
d
Test Laboratory, Isringhausen GmbH and Co. KG, An der Bega 58, D-32657 Lemgo, Germany

Received 13 July 2007; received in revised form 12 October 2007; accepted 22 October 2007
Available online 3 December 2007

Abstract

This paper describes a simplified simulation of two configurations of the fore-and-aft seat suspension system. A fore-and-aft
suspension system model was proposed based on the laboratory measurements of the seat vibration isolation performance. Friction was
identified as an important parameter, so different approaches to simulating the suspension friction were investigated. Predicted seat
vibration mitigation properties were compared with those measured in the laboratory in response to the recordings of the fore-and-aft
vibration measured at the base of the driver’s seat in an on-road tractor–trailer combination (articulated truck). Optimisation of the
suspension elements parameters was then performed to identify the maximum attainable attenuation. A solution incorporating
supercritical suspension damping predicted to give an improvement of the order of 10% in the x-direction mitigation properties as
compared to a fixed (locked) horizontal suspension system.
Relevance to the industry: Simulations conducted in this study are of use to seat manufacturers in developing the fore-and-aft seat
suspension systems with improved vibration mitigation properties and for predicting its dynamic performance. The optimisation study
shows the attainable vibration mitigation limits for a horizontal suspension system.
r 2007 Elsevier B.V. All rights reserved.

Keywords: Driver’s seat; Horizontal suspension; Friction modelling; Whole-body vibration

1. Introduction fatigue and in some cases to injury. Workplace epidemiolo-

gical studies (Dupuis et al., 1988; Schwarze et al., 1999;
In various ground vehicles, for example on-road, off-road Griffin, 1990; Mansfield, 2005) have stressed the need for
vehicles, industrial trucks, agricultural tractors, railway proper suspension design. Most seat development effort has
vehicles, etc., the isolation of the seated operator from been concerned with isolating vibration in the vertical
vibration and shock is of considerable importance. Exposure direction (z-axis). However in real operating conditions the
to whole-body vibration in working environment may lead to driver is exposed to multi-axis vibration, as analysed, for
example, recently in Mansfield and Maeda (2006, 2007). Little
Corresponding author. Tel.: +4212 5930 9422; fax: +4212 5477 2909. research has been conducted on vibration isolation in the non-
E-mail addresses: [email protected] (G.J. Stein),
vertical axis. The aim of the European project VIBSEAT was
[email protected] (T.P. Gunston), to determine how the operators of various vehicle types could
[email protected] (L. Burström), be protected from horizontal and rotational vehicle vibration
[email protected] (L. Meyer). by using suspension seating through the development of
1
Now at: SIG SCP Noise and Vibration Laboratory, Units 5-7, Lower performance assessment, subjective discomfort valuation and
William Street, Southampton, SO14 5QE, UK.
2
Now at: Department of Public Health and Clinical Medicine, Umeå vibration isolation strategies (using theoretical models and
University, SE-901 87 Umeå, Sweden. new prototype seats) for multi-axis random excitation, as
3
Tel.: +49 5261 210 293; fax: +46 5261 210 55 293. practically measured in real operating conditions.

0169-8141/$ - see front matter r 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.ergon.2007.10.016
 ARTICLE IN PRESS
G.J. Stein et al. / International Journal of Industrial Ergonomics 38 (2008) 396–409 397

The response of the human body to vibration in the fore- than to a complex apparent mass loading to concentrate
and-aft direction (x-axis) and in the lateral direction attention on the x-direction seat suspension properties
(y-axis) have been widely reported (e.g. Fairley and Griffin, rather than on the whole system. Given this restriction, an
1990; Holmlund and Lundström, 1998; Mansfield and indicative performance prediction is possible for the single
Lundström, 1999a; Fleury, 2004; Nawayseh and Griffin, axis excitation in laboratory conditions as is shown below.
2005b; Mandapuram et al., 2005; Mansfield and Maeda, This can aid the seat designer in the design phase prior to
2007). The papers by Fairley and Griffin (1990), Holmlund development of a prototype seat and reduce laboratory
and Lundström (1998) and Mansfield and Lundström tests with human test subjects, with the associated costs
(1999a) involved the human body sitting upright on a rigid and ethical concerns. This approach helps to decrease the
seat without a seat cushion in a well-defined biodynamical time to market and helps identify problems early in the
position without interaction with controls or the seat back. development process.
Only a few papers are concerned also with the influence of This paper describes the analysis and optimisation of the
the seatback support (Fairley and Griffin, 1990; Fleury, fore-and-aft (x-direction) suspension system of a driver’s
2004; Nawayseh and Griffin, 2005a; Mandapuram et al., seat for use in road transport. The vibration isolation
2005; Mansfield and Maeda, 2007), mostly assuming a rigid performance of a currently produced seat with a fore-and-
vertical seatback support. aft suspension system was first measured in laboratory and
To be able to model the complex interactions of the field conditions. A simplified mathematical model of the
seated human body and a cushioned seat it is necessary first seat suspension system including friction effects was then
to analyse the various interactions in a modern driver’s seat developed and used to simulate the performance of the real
with both vertical and fore-and-aft suspension systems. A seat. The simulation results were compared with the
good starting point is the model of Nawayseh (2002), which measured seat performance for two suspension configura-
accounts for the fore-and-aft rotational movement of the tions and for a configuration with the suspension disabled
seated human body, as observed in various experiments, (locked). Finally, the modelled stiffness and damping
e.g. Fairley and Griffin (1990), Mandapuram et al. (2005) parameters were optimised to suggest an improved config-
and Fleury (2004). Fleury and Mistrot (2006) describe uration for reducing the transmission of vibration through
x-direction human body model using rotary and translatory the seat.
mechano-mathematical elements. They used this model for
prediction of x-direction vibration suppression by a driver’s 2. Modelling assumptions
seat equipped with a fore-and-aft suspension system with
low inherent friction. They account for the damping of the The intention of the model of the system, the x-direction
rotational movement of the upper torso by parameters seat suspension system (inclusive of friction)—cushioned
describing conditions with no back support, with two seat upper part—sitting human body, used in this study
methods of back support, and with a steering wheel. The was to simplify the complex dynamic behaviour of the seat-
investigators observed that if a full back contact is person system into a mechano-mathematical model suffi-
maintained the rotational degree of freedom may be cient to describe approximately the main problem features.
discounted, whereas if back support in the lumbar region The model should allow indicative prediction of driver’s
only is maintained then the parameters of the rotational seat performance in the x-direction in controlled labora-
degree of freedom are completely changed. For comparison tory environment. Here, contrary to real field conditions,
they used measured and simulated x-direction apparent single axis excitation was assumed and used throughout.
mass. It is noted both in Fleury (2004) and Fleury and Following simplifying assumptions were made:
Mistrot (2006) that the agreement between the simulated
and the measured human response is poor, which was (A) Vertical vibration was not analysed and the motion of
attributed to the non-linear behaviour of the human body. the vertical seat suspension is assumed to be largely
This was especially so for the case of backrest contact in the independent of the fore-and-aft suspension system.
lumbar region only, as it is difficult to obtain reproducible Only the x-direction motion was simulated. Influence
backrest contact conditions. of both vertical and horizontal kinematic excitation
Measurements with human subjects supported by the was analysed at some depth by Zahoranský and Stein
seat-back in the lumbar region (Hinz et al., 2005; Stein et (2007b).
al., 2007a) have shown that due to the back support the (B) The seated human body was modelled as an inert
rotary motion is strongly reduced, so a simplified model reduced mass, as explained above.
employing linear mechano-mathematical elements was (C) The seat components were assumed to be sufficiently
found to be sufficient, as developed in (Stein et al., 2006). rigid not to take into consideration any torsion
A thorough study of various models of human body, movement of the rigid armchair frame. High frequency
sitting on a rigid seat without seat cushion and seat-back oscillations of seat upper part structural elements were
support is reported by Mansfield and Lundström (1999b). also not taken into account.
In view of the above, the present study further simplified (D) The seat back position was assumed to be close to
the human body model to a simple mass loading rather vertical, inclined at a small angle of less than 101, as
 ARTICLE IN PRESS
398 G.J. Stein et al. / International Journal of Industrial Ergonomics 38 (2008) 396–409

common in industrial practice (Nawayseh and Griffin, Interaction of the seated human body with the seat
2005a, b). suspension system and seat upper part is not well under-
stood, as discussed in (Stein et al., 2006). For the purpose
If all these assumptions are taken into consideration, of the present study, it was assumed as the first
the simple fore-and-aft seat suspension model according approximation to this complex problem, based on (Stein
to Fig. 1a emerges. However, the laboratory measure- et al., 2007a), that the seated human body and the
ments described below indicated an influence of the cushioned seat upper part can be reasonably modelled in
end-stops, if the linear range xL of the seat suspension a simplified way using an inert mass mr, being the sum of
travel is exceeded. Furthermore, if structural limits xop are the mass of the upper part of the seat and the reduced mass
reached strong impacts with the seat structure may be of the seated driver. The reduction in mass of the seated
observed in some exceptional cases. The expanded model driver by 18–20% accounts for the legs and feet not being
shown in Fig. 1b accounted for these effects and was used supported by the seat but rather by the cabin floor.
throughout, as described in more detail in Zahoranský The input to the seat-occupant model was the fore-and-
et al. (2005). € at the seat base and the output was the
aft acceleration uðtÞ
It should be stressed that this is a simplified system € at the seat/driver interface.
fore-and-aft acceleration xðtÞ
model, catering for the principal features only and not
going into detailed description of respective physical 3. Measurement and identification of the horizontal
components of the system and the complex influence of suspension system parameters
the seated human body interaction with steering-wheel and
pedals reaction. Detailed description of a real hydraulic 3.1. Measurement
damper is far more complicated than an idealised simple
damping constant cx can encompass. The description of a The fore-and-aft seat suspension system together with
steel spring by a spring constant kx is usually sufficient. The the seat upper part was tested by the manufacturer to
system friction, which is always present in such a system, is determine the force–deflection characteristics. The appara-
discussed in more detail below. tus used is shown in Fig. 2 and consisted of a hydraulic
actuator with built-in displacement transducer acting onto
the seat upper part via a load cell, while the seat base is
fixed to a supporting rail.
The force–displacement characteristics obtained for two
different seat suspension configurations without a hydrau-
lic damper are depicted in Fig. 3. The area enclosed by the
hysteretic curve represents the energy dissipated in one
complete cycle due to friction and structural damping. The
influence of the end-stops and their non-linear force–de-
flection characteristics is also clearly seen. Fig. 3b (the ‘low
friction’ system) shows results for a suspension system
where friction is significantly reduced compared to the
system shown in Fig. 3a (the ‘high friction’ system). No
further measurements were available so no further detailed

Fig. 1. Two models of the fore-and-aft suspension system (a) a simple

model and (b) a complex model with inclusion of the end-stops. Fig. 2. Apparatus for measurement of the hysteretic curves.
 ARTICLE IN PRESS
G.J. Stein et al. / International Journal of Industrial Ergonomics 38 (2008) 396–409 399

component kNL may be found for the (xop, xL) and

(+xL, +xop) regions. Here the influence of the rubber end-
stops with progressive force–deflection characteristics
combines with the linear spring. As shown in Zahoranský
and Stein (2007a), a polynomial approximation of the form
(1) gives an appropriate fit to the rubber end-stop buffer
force–deflection characteristic:

F s;NL ¼ kNL lðjxr j  xL Þ2 , (1)

while the unit of kNL is [N/m2] and l is a binary variable,
which has unity value for xLp|xr|oxop and zero for
|xr|oxL. More details on chaotic suspension system
response due to reaching the seat suspension system
operational limits can be found in Zahoranský and Stein
(2007a). The estimated parameters are given in Table 1.
A hydraulic shock absorber was installed in the low
friction system before vibration isolation tests were con-
ducted. The damping force Fd versus relative velocity of the
damper ends vr was measured by the damper manufacturer
and supplied as a set of values as shown in Fig. 4. As the
relative velocity vr in the analysed case was low, a linear
approximation within the range (0.066,+0.066 m/s)
was used, leading to a linearised damping coefficient of
cx ¼ 606 Ns/m. This damper model was known to be a
simplistic description of the real dynamic system.
The summarised models parameters are given in Table 1.
Due to the large difference in the friction force value (being
of an order of magnitude) and presence/absence of the
hydraulic damper, each system was expected to have
different vibration mitigation properties when subjected to
Fig. 3. Measured hysteretic curves for two different horizontal suspension
systems of a driver’s seat (a) the high friction seat and (b) the low friction
seat; ( ) measured, ( ) mean curve. Table 1
Identified parameters of the two suspension systems

Suspension kx (kN/m) kNL (N/m2) cx (Ns/m) mk (1)

research could have be made, e.g. on possible friction force
load dependence. High friction 9.14 6.68  106 – 0.102
Low friction 8.51 11.85  106 606.0 0.009

3.2. Identification of suspension system parameters

For the high friction system the friction force magnitude

|Ff| was estimated to be 80–85 N with a load of 82.5 kg
consisting of 60 kg of inert load and of the seat upper part
of mass 22.5 kg, while the low friction system showed
friction of only 5–10 N for the same loading. The friction
coefficients were calculated as mkE0.102 and 0.009,
respectively. A middle curve, from which the stiffness
linear range (xL, +xL) was determined, was constructed
for both systems (solid line). This range was710 mm and
the estimated linear stiffness coefficient was kx ¼ 9.14 kN/m
for the suspension system shown in Fig. 3a. For the
suspension system shown in Fig. 3b the linear range was
also710 mm and the estimated linear stiffness coefficient
was kx ¼ 8.51 kN/m.
Not taking into account the strong impacts onto the seat
structure when the relative displacement reaches the system Fig. 4. Damper force–velocity characteristics (supplied values denoted by
operational limits xop (see Fig. 1b) the non-linear stiffness dots).
 ARTICLE IN PRESS
400 G.J. Stein et al. / International Journal of Industrial Ergonomics 38 (2008) 396–409

realistic random excitation. A detailed analysis of the a larger force is required than when the two surfaces are
friction influence in the high friction seat suspension continuously slipping on each other with non-zero slipping
configuration was therefore undertaken first to arrive at velocity vr (Timoshenko et al., 1974). The friction force at
usable simulation results. The mathematical description of vr ¼ 0 has to be described as a function of a limit force FL,
the friction force present was approached as discussed in external to the friction interface. The limit force FL is
the following section. obtained by analysing the force balance across the sliding
surfaces interface and has to be compared to the static
4. Assessment of friction friction force value Ffs
if jF L jpF fs ) vr ¼ 0. (3)
In analysing an oscillatory system with both viscous and
frictional damping it is of considerable interest to under- If this condition is met the system is in standstill in the
stand the relative contribution of both dissipative terms to so-called stick state, indicated in Fig. 5 by the vertical line
the total vibratory energy dissipation, the transfer function segment. If at a certain time instant the external forces
and time domain response. Despite considerable research exceed the static friction force Ffs, the oscillatory system
the mathematical description of this phenomenon is not starts abruptly to move and the relative velocity vr attains
fully established and the underlying physical processes are some non-zero value. Smooth movement is maintained
not practically reproducible as they depend on surfaces until vr eventually decreases to zero and the system stops.
state, surface lubrication, asperities, temperature, normal In employing friction models for oscillatory systems
force magnitude, surfaces’ relative velocity, and other analysis, two approaches are generally feasible:
factors as described e.g. by Berger (2002) and Petrov and
Ewins (2002). Many friction models are based on the (i) An approximate analytical approach, based on the
relation of friction force Ff to relative velocity vr of sliding harmonic balance method. This method assumes
surfaces in a phenomenological way. The basic approach, harmonic excitation by vibrations with variable angular
as described by Berger (2002) and Petrov and Ewins (2002), frequency ox in vicinity of the system natural frequency
is shown in Fig. 5. o0. The method is fully explained in standard text-
The Coulomb type friction is mathematically described by books e.g. Timoshenko et al. (1974), Klotter (1980),
the so-called relay characteristics and depends on the Inman (2001). However this approximate method
sliding surfaces’ relative velocity vr according to cannot account for the stick-slip phenomena and is
F f ¼ F fk signðvr Þ; F fk ¼ mk mgn , (2) not applicable when random excitation is assumed, as
occurs in driver’s seat research.
where Ff is the friction force and gn is the standard gravity (ii) A non-linear time domain numerical simulation ap-
acceleration and mk is the kinetic friction coefficient. The proach, employing contemporary simulation software
friction force Ff cannot be determined for vr ¼ 0 and can (e.g. MATLABs), which makes use of conditioned
have any value in the interval (Ffk,+Ffk). switching between solutions on a short time-scale in
The normal force FN ¼ mgn is usually assumed to be comparison to the excitation signal dominant period.
constant; however, it is not necessary so for horizontal The simulation approach enables to use either the
suspension system of a driver’s seat. As described in signum function approach given in Eq. (6) below
Zahoranský and Stein (2007b) no marked influence of or a more physically correct approach using Eq. (4) for
variable FN on the horizontal suspension system perfor- vr 6¼ 0 and the limit force analysis by Eq. (7) for vr ¼ 0.
mance was found in this particular case. In Stein et al. (2007b) it is shown that for low intensity
In reality a larger force is needed to start the sliding random acceleration excitation the physically correct
motion. For overcoming adhesion at zero relative velocity model, employing rigorously the limit force analysis,
gives more reliable results. This approach was used for
the present study.

The equation of motion of a damped system with a

linear spring with spring constant kx, progressive end-
stops, described by the non-linear spring force Fs,NL for
xLo|xr|oxop is then, for vr 6¼ 0:
mr x€ þ cx x_ r þ kx xr þ F s;NL þ F f ¼ 0, (4)
where Fs,NL is given by Eq. (1) and
xr ¼ x  u; vr  x_ r ¼ x_  u.
_ (5)
Friction damping force Ff for vr 6¼ 0 has the form:

Fig. 5. Friction force Ff as a function of the relative velocity vr. F f ¼ F fk signðx_ r Þ ¼ mk mgn signðx_ r Þ. (6)
 ARTICLE IN PRESS
G.J. Stein et al. / International Journal of Industrial Ergonomics 38 (2008) 396–409 401

For vr ¼ 0 the limit force FL analysis has to be Fore-and-aft accelerations previously recorded in the
preformed. The limit force FL is field on a road tractor–trailer combination (articulated
truck) were reproduced on the vibrator platform. The
jF L j ¼ jmr x€ þ kx ðx  uÞj. (7)
tractor–trailer combination used for the field measure-
ments was loaded with 22 t of ballast and driven on a
(i) When: |vr|oe AND |FL|oFfs the oscillatory system is highway with a constant speed of 70 km/h. The accelera-
in standstill in the stick state; tions had been measured using a B&K 4368 accelerometer
(ii) When: |vr|4e OR |FL|4Ffs the oscillatory system is in connected to a B&K 2635 conditioning amplifier and
the slip state and Eq. (4) holds. recorded on-board by a Sony 204A DAT recorder with
sampling frequency of 12 kHz. The test signal time history
The variable e determines the goodness of fit of the and the corresponding PSD are shown in Fig. 6.
relative velocity actual numerical value with zero and has Fore-and-aft accelerations on the vibrator platform at
to be selected with some care, as discussed in depth in Stein the base of the seat and on the seat surface were measured
et al. (2007b). with Entran EGCS-D0-10 piezo-resistive accelerometers,
the seat surface accelerometer mounted in a semi-rigid disk
5. Validation of the seat suspension with simple human body as described in ISO 10326-1 (1992). The accelerations were
model digitally acquired via an HVLab v3.81 data acquisition and
analysis system with an anti-aliased bandwidth of greater
5.1. Description of laboratory measurements than 50 Hz. The digitised platform and seat surface
accelerations were then band-limited using a 2-pole high-
To be able to give confidence in the use of the models for pass Butterworth filter with a 3 dB cut-off point of 0.20 Hz
indicative prediction of vibration mitigation properties and and a 6-pole low-pass Butterworth filter with a 3 dB point
for guidance for real system optimisation, comparisons of 17.78 Hz. Both filters were implemented in software.
were made against laboratory measurements under con-
trollable conditions.
The measurements were performed on the Institute of
Sound and Vibration Research Human Factors Research
Unit electro-hydraulic horizontal vibrator. This apparatus
is capable of a peak acceleration of 10 m/s2 and has a total
operating stroke of 1 m.
The seat was securely attached to the vibrator platform.
The backrest was reclined by less than 101, in line with the
observations of Nawayseh and Griffin (2005a). The seat
height and fore-and-aft position adjustments were set as
close to the mid-point of the available range as the discrete
position settings of the adjustment mechanisms permitted.
An appropriate compressed air supply was connected to
the seat and the vertical suspension air spring was allowed
to self-level to the operating position within the suspension
stroke each time the seat was loaded.
Before the tests with human subjects, the seat was
subjected to a run-in of 2000 cycles at 1.0 Hz in the fore-
and-aft direction while loaded with an inert mass of 75 kg.
The load mass during run-in consisted of lead shot bags
secured to the seat surface. The run-in exercised the
suspension over approximately 75% of the available stroke
and was interrupted as required to prevent the damper
overheating.
The laboratory measurements were made with two test
subjects only, one female and one male, of body mass 55
and 106 kg and body heights 174 and 191 cm, respectively.
An adjustable footrest was provided and the height of the
footrest was set to allow the subjects to assume the posture
described in International Organization for Standardiza-
tion (ISO) 10326-1 (1992). Subjects rested their hands on
their thighs. The measurements were conducted with the
approval of the ISVR safety and ethics committee. Fig. 6. Base fore-and-aft acceleration aB (a) time domain and (b) PSD.
 ARTICLE IN PRESS
402 G.J. Stein et al. / International Journal of Industrial Ergonomics 38 (2008) 396–409

The suspended mass of the test seat, excluding the error in the SEAT factor the formula has the form:
occupant, was 22.5 kg. Two seat configurations were
SEATxsim  SEATxmeas
tested—one with a high friction and no hydraulic damper d¼  100 ½%. (10)
and the second with a low friction and a hydraulic damper SEATxmeas
installed. The high friction and the low friction fore-and-aft
In the first seat configuration the horizontal suspension suspension configuration parameters are given in the
system could be locked in the middle of available Table 1, those appropriate to loading by different test
horizontal travel, i.e. blocked. Except of minute relative subjects in Table 2.
displacements of the order of 0.1 mm, due to the actual The results obtained for the high friction suspension
stiffness of the seat structure, no real movement was system are depicted in Fig. 7. Those for the low friction
observed. So no simulation of the locked seat was deemed system are presented in Fig. 8. Values of the characteristic
as necessary. variables are given in Tables 3 and 4, respectively.

5.2. Methods used for comparison of the simulations with 5.3. Results and their discussion
the laboratory measurements
It can be seen from Tables 3 and 4 that the difference in
The following characteristic values were used to compare measured and simulated values of the SEATx factor, which
the performance of the simulated seat responses with those was the seat performance primary indicator, was marginal. It
measured in the laboratory: was therefore assumed that the simplified simulation model
adequately predicted the real system performance under low
 The acceleration power spectral density (PSD) at the intensity random excitation. This conclusion is also sup-
seat surface in response to the same seat base excitation ported if the simulated and measured PSDs and TFEs are
acceleration. compared (see Figs. 7 and 8). For the heavy subject sitting in
 The transfer function estimate (TFE) from the seat base the seat with a low friction horizontal suspension system the
to the seat surface. model results agree with the measured response around
 The coherence function between the seat surface damped natural frequency. The other excitation peaks,
acceleration and seat base excitation acceleration. especially those in the frequency band 10–15 Hz are not so
 The SEATx factor defined conventionally for the well approximated. The approximation in this frequency
x-direction in analogy with the definition of the SEAT band is much better for the light driver sitting in the same
factor in the vertical direction (z-direction) as seat; however, there is a marked difference in the damped
axwS natural frequency of the simulated results when compared to
SEATx ¼ , (8)
axwB the measurements. This difference cannot be explained from
available data and calls for careful interpretation. Such
where the seat (index S) and base (index B) RMS
deviations are also present when the Tax values are
acceleration values in the x-direction are weighted by the
compared, which indicates that this simple model does not
same weighting filter Wd according to ISO 2631-1 (1997)
represent well the real seat behaviour at higher frequencies,
for the x-direction.
well beyond main human body resonance, as described by
 The transmissibility Tax, defined as the ratio of
the Wd filter. The model does not account well for occasional
unweighted RMS acceleration value at the seat surface
high amplitude acceleration maxims, as there is a large
axS to that at the seat base axB
difference between the measured and calculated on-seat crest
factor CS values, which are larger for the high friction seat
axS
T ax ¼ . (9) configuration than for the low friction seat configuration
axB For comparison with a seat without any horizontal
suspension system the high friction seat with the horizontal
 The crest factors CB for base x-direction acceleration suspension system locked was measured in the same way. The
and CS for seat x-direction acceleration. course of PSD and TFE were much the same as those for the
seat with high friction suspension and hence are not depicted.
The relative percentage errors d between the simulated The high friction oscillatory system is in the stuck state
and measured values were calculated. For example for the for most of the time and behaves as a rigid body, so the

Table 2
Test subjects and friction influence

Subject Mass m (kg) Height (cm) Reduced mass mr (kg) High friction Ffk (N) Low friction Ffk (N) FN (N)

Light subject 55 174 63.6 82.5 5.0 393.9

Heavy subject 106 191 105.4 137.0 8.0 759.1
 ARTICLE IN PRESS
G.J. Stein et al. / International Journal of Industrial Ergonomics 38 (2008) 396–409 403

Fig. 7. PSD, TFE and coherence for the high friction suspension system configuration (a) light subject and (b) heavy subject; measured ( ), simulated
( ).

coherence function hovers around the unity value as seen system under low intensity vibrations except for some
on Fig. 7. In the frequency bands where insufficient occasional instants.
vibratory power is supplied to the oscillatory system, the (ii) There is minimal improvement in the fore-and-aft
coherence decreases, indicating a poor signal-to-noise ratio vibration mitigation properties of the high friction
(e.g. in frequency band around 5 and 10 Hz). horizontal suspension system compared to the locked
For the low friction configuration, the departure of the state. This is so for both measures—the SEATx factor
coherence from unity value is much larger (Fig. 8). It and for the transmissibility Tax. The SEATx is around
appears that enough vibratory power was supplied only in unity in both cases, while there is a marginal
the frequency bands 0.5–3.0 and 11.5–14.5 Hz (as also seen improvement in the Tax.
from the PSDs) to sufficiently excite the suspension system. (iii) The additional peaks around frequency 12 Hz and a
Further non-linear effects may come into action (e.g. in the smaller one around 6 Hz do not contribute markedly
hydraulic damper, not accounted for by the simplified to the overall seat suspension performance.
model), which could account for the reduced coherence.
The analysis of laboratory and field measurements of the The low friction seat configuration exhibits an increase in
seat with the high friction suspension configuration the SEATx factor, whereas the unweighted transmissibility
revealed that: Tax decreased. This reduction is more evident for the light
subject. These results are understandable, if the TFEs of the
(i) The horizontal seat suspension system exhibits high fore-and-aft suspension system are compared with the Wd
friction, which prevents movement of the suspension weighting filter course after ISO 2631-1 (1997)—both
 ARTICLE IN PRESS
404 G.J. Stein et al. / International Journal of Industrial Ergonomics 38 (2008) 396–409

Fig. 8. PSD, TFE and coherence for the low friction suspension system configuration (a) light subject and (b) heavy subject; measured ( ), simulated
( ).

Table 3
Comparison of measured and calculated response characteristics for the high friction suspension system

Characteristic variable Light subject Heavy subject

Measured Calculated d (%) Measured Calculated d (%)

axB (m/s2) 0.600 0.600 — 0.597 0.597 —

axwB (m/s2) 0.203 0.203 — 0.202 0.202 —
CB (1) 5.65 5.65 — 5.81 5.81 —
axS (m/s2) 0.532 0.501 5.8 0.501 0.580 15.8
axwS (m/s2) 0.205 0.205 0.0 0.204 0.203 0.5
CS (1) 5.83 3.42 45.0 5.53 3.24 41.0
SEATx (1) 1.01 1.01 0.5 1.01 1.01 0.2
Tax (1) 0.889 0.836 6.0 0.841 0.852 1.3

maximums are nearly coincident. There is vibration The performances of the two seat configurations with
amplification rather than reduction at the suspension fore-and-aft suspensions were compared with the perfor-
damped natural frequency. The reduction in the unweighted mance of the seat with locked suspension system. The
transmissibility Tax at higher frequencies, above 5 Hz characteristic values are given in the Table 5 in which the
(especially the marked peak around 12 Hz) did not weighted and unweighted RMS acceleration values, as well
substantially affect the SEATx factor. as the SEATx factor values are given. The differences in the
 ARTICLE IN PRESS
G.J. Stein et al. / International Journal of Industrial Ergonomics 38 (2008) 396–409 405

Table 4
Comparison of measured and calculated response characteristics for the low friction suspension system

Characteristic variable Light subject Heavy subject

Measured Calculated d (%) Measured Calculated d (%)

2
axB (m/s ) 0.600 0.600 — 0.597 0.597 —
axwB (m/s2) 0.203 0.203 — 0.202 0.202 —
CB (1) 5.65 5.65 — 5.81 5.81 —
axS (m/s2) 0.308 0.258 16.2 0.375 0.241 35.7
axwS (m/s2) 0.225 0.221 1.8 0.248 0.222 10.5
CS (1) 5.87 4.64 21.0 4.65 3.63 22.0
SEATx (1) 1.12 1.09 1.7 1.23 1.10 10.0
Tax (1) 0.513 0.430 16.2 0.628 0.404 35.7

Table 5
Comparison of measured response characteristics for the locked seat, high friction seat configuration and the low friction seat configuration

Characteristic variable Light subject Heavy subject

Locked High friction Z (%) Low friction Z (%) Locked High friction Z (%) Low friction Z (%)
2
axB (m/s ) 0.600 0.600 — 0.600 — 0.597 0.597 — 0.597 —
axwB (m/s2) 0.203 0.203 — 0.203 — 0.202 0.202 — 0.202 —
axS (m/s2) 0.560 0.532 5.0 0.308 45.0 0.573 0.501 5.0 0.375 34.6
axwS (m/s2) 0.208 0.205 1.4 0.225 8.2 0.211 0.204 1.0 0.248 17.5
SEATx (1) 1.02 1.01 1.0 1.12 9.8 1.05 1.01 1.2 1.23 17.1
Tax (1) 0.935 0.890 4.8 0.513 45.1 0.961 0.841 4.3 0.415 56.8

Minus sign — improvement; plus sign — deterioration.

weighted RMS values, and hence also in the SEATx values, the x-direction vibration mitigation. However, having
are marginal and well within measurement and analysis developed horizontal seat suspension model and noting
error. The relative difference Z of variables entered in Table that the predicted values of the SEATx values are within
5 in comparison to the locked seat is calculated, for acceptable limits the model, despite of its simplicity, can be
example, as used for further suspension system analysis and theoretical
SEATunblock  SEATblock optimisation. Two aims were followed:
Z¼  100 ½%, (11)
SEATunblock
(i) To investigate the minimum attainable SEATx value,
and indicates the improvement (minus sign) or deteriora- irrespective of other constraints.
tion (plus sign) in respect to the situation without the (ii) To investigate attainable SEATx value, if only the
horizontal suspension system. The results, shown in Table damping would be modified.
5, indicate that:
6.1. Damping and stiffness optimisation
(i) The high friction horizontal seat suspension configura-
tion brings no real improvement to a seat without The lower limit of attainable SEATx factor and of
horizontal suspension system. transmissibility Tax value for the horizontal seat suspension
(ii) The low friction horizontal seat suspension configura- system was investigated first assuming an unlimited
tion amplifies the response around the damped natural variability in damping coefficient cx and spring stiffness
frequency, leading to deterioration of the SEATx factor kx and constant low friction force Ff. The minimisation of
by 10–17% and at the same time to marked improve- the chosen goal function (SEATx or Tax) was performed by
ment in the Tax value by 40–60%. the function fminsearch, implemented in Matlabs, which
uses an n-dimensional simplex method of the search of
local minimum in the neighbourhood of initially chosen
6. The fore-and-aft seat suspension system optimisation point.
The first problem can be approached using the standard
The result of above laboratory measurements and vibration isolation theory. It is possible to identify a single
simulations indicated that none of the fore-and-aft seat degree of freedom oscillatory system with a damped
suspension configurations brought about improvement in natural frequency f0 sufficiently low to give acceleration
 ARTICLE IN PRESS
406 G.J. Stein et al. / International Journal of Industrial Ergonomics 38 (2008) 396–409

transmissibility below unity in the frequency band where

the excitation exhibits most of the relevant vibratory
energy. The Wd weighting filter together with the asso-
ciated band-limiting filters of ISO 2631-1 (1997) has a
maximum around 1.1 Hz, hence the oscillatory system
damped natural frequency should be below 1.1 Hz/O2E
0.78 Hz. This approach was consequently followed and
following results were obtained by the two-parameter
optimisation:

(A) For the heavy subject:

kxopt ¼ 10 N/m; cxopt ¼ 433 Ns/m (f0 ¼ 0.05 Hz);
SEATx ¼ 0.499; Tax ¼ 0.210.
(B) For the light subject:
kxopt ¼ 363 N/m; cxopt ¼ 221 Ns/m (f0 ¼ 0.06 Hz);
SEATx ¼ 0.565; Tax ¼ 0.243.

It can be seen that the SEATx factor can be roughly

halved to that one of approximately unity value for the
unsuspended seat and the x-direction transmissibility can
be reduced to a quarter of the original values; however, at
the expense of having a seat with an extremely low natural
frequency and extremely low stiffness. This means that the
seat upper part would almost freely move between the end-
stops, hitting these from time-to-time, with effects de-
scribed in Zahoranský and Stein (2007a). Obviously, for
vehicle control and safety this is the most unwanted
consequence and practically unacceptable solution. Hence,
this approach is not viable.
Fig. 9. Dependence of (a) SEATx value and (b) transmissibility Tax on the
value of cx for the low friction suspension system: light subject ( ),
6.2. Optimisation of damping for given seat suspension heavy subject ( ).
stiffness

Further simulations investigated the optimum damping

coefficient cx value for the low-friction seat suspension marked improvement around damped natural frequency
system and given spring stiffness kx. This sub-optimal case (Fig. 10).
brings some improvement in vibration mitigation. To The relative difference Z was also calculated according to
illustrate the optimisation limits, the weighted SEATx formula (11). From Table 6, it can be seen that the
factor (a) and the unweighted transmissibility Tax (b) as a introduced supercritical damping coefficient leads to
function of variation of cx for the heavy and light subjects, improvement of both measures in respect to the locked
using the same spring stiffness kx of 9.15 kN/m are shown suspension system and in most cases also to original low
in Fig. 9. From the SEATx factor course given in Fig. 9a friction suspension system.
it can be seen that there is not a distinctive minimum The Tax values also exhibits a minimum, however, the
in the SEATx factor as a function of cx. The minimum values of optimal cxopt are quite different: cxopt ¼ 0.66 kNs/m
SEATx factor for the light subject corresponding to for the light subject, respectively (corresponding value of
cxopt ¼ 2.09 kNs/m was 0.94. For the heavy subject for Tax ¼ 0.430) and cxopt ¼ 1.11 kNs/m for the heavy subject
cxopt ¼ 2.59 kNs/m, SEATx was 0.91. The respective (corresponding value Tax ¼ 0.388). Again, the minimum is
critical damping constants cc are 1.47 and 1.89 kNs/m; not distinct.
hence, supercritical damping is required. An average
damping coefficient of 2.30 kN/s may be a practical 7. Discussion
solution. The resulting PSD and TFE curves for the
excitation signal used above for this damping coefficient It was concluded from the comparison of the results
and for the stipulated light and heavy subjects are shown in obtained by the simulation and those measured in
Fig. 10. The corresponding numerical values for this laboratory with low intensity random excitation in the
damper setting are given in Table 6. Note the difference x-direction that the simulation results, in spite of using a
of the simulated TFE curves to those measured, with simple seat suspension model, corresponded to the
 ARTICLE IN PRESS
G.J. Stein et al. / International Journal of Industrial Ergonomics 38 (2008) 396–409 407

laboratory measurements under the single axis excitation to low damping of the system, as well as to the fact that the
well. This gave confidence into the developed seat maximum of the Wd weighting filter and the damped
suspension model, including physically correct model of natural frequency of the suspension system were nearly
friction as the main feature. coincident.
The high friction seat suspension system configuration Neither the high friction nor the low friction seat
was observed to be mostly friction-locked and exhibited a suspension configurations brought about improvement in
SEATx factor close to the unity value. respect to the system without horizontal suspension (e.g. a
The low friction seat suspension system configuration locked horizontal suspension system). The result of the
gave a SEATx factor larger than unity. This was attributed laborious laboratory measurements and development of

Fig. 10. PSD and TFE for the low friction suspension system with optimal linear damper (a) light subject and (b) heavy subject; excitation (——),
measured ( ), simulated ( ).

Table 6
Comparison of simulated response characteristics for the real low friction seat, low friction seat with optimal damper and the locked seat

Characteristic variable Light subject Heavy subject

Locked Low friction Z (%) Optim. damp. Z (%) Locked Low friction Z (%) Optim. damp. Z (%)

cx (Ns/m) – 606 2089 – 606 2587

axB (m/s2) 0.599 0.600 – 0.599 – 0.596 0.597 – 0.597 –
axwB (m/s2) 0.203 0.203 – 0.203 – 0.202 0.202 – 0.202 –
axS (m/s2) 0.560 0.258 53.9 0.339 39.5 0.501 0.241 51.9 0.294 41.3
axwS (m/s2) 0.208 0.221 6.3 0.191 8.2 0.204 0.222 8.8 0.184 9.8
SEATx (1) 1.02 1.09 6.9 0.94 7.8 1.05 1.10 4.8 0.91 13.3
Tax (1) 0.935 0.430 54.0 0.566 39.5 0.961 0.404 58.0 0.492 48.8

Minus sign—improvement; plus sign—deterioration.

 ARTICLE IN PRESS
408 G.J. Stein et al. / International Journal of Industrial Ergonomics 38 (2008) 396–409

suitable simulation program code led to further exploita- performance was also compared with the performance of a
tion of the gained expertise with the aim to aid further the seat with no suspension fitted.
seat design. Comparisons between the friction-inclusive suspension
An optimal set of seat suspension stiffness and damping model with a simple human body load and the laboratory
was found, assuming low friction influence. This led to measurements gave confidence in the model performance.
halving the SEATx value at the expense of having a seat As an application example, the model was then used to
with extremely low natural frequency, calling for using a identify optimal suspension damping coefficients and
very soft spring. This approach is not acceptable on indicate the maximum isolation performance that might
ergonomic and psychological reasons. It just indicates the be achieved in practice by varying the damper.
theoretically attainable vibration mitigation limits for the It may be reasonable to expect the model structure as
fore-and-aft suspension system under the test motion used presented to be useful in simulating the performance of
in this study. similar configurations of seat suspension units in response
An alternative solution was found, based on using a to similar motions, but it should be recognised that
linear spring with the same stiffness as in the low friction substantially different motion characteristics would require
seat. The SEATx factor was improved by some 10% and further validation trials to give confidence that reasonable
the unweighted x-direction acceleration transmissibility predictions will be obtained.
decreased by 40% compared to the seat with no suspension Further development of this model may also be required
system at all. This was obtained by using a damper with a to account for the seated human body dynamic properties,
supercritical linear damping coefficient, e.g. a much harder its inherent non-linearity and the various specific interac-
damper than used in the tested seat. tions occurring in practice, especially those by the legs and
Results presented in the last section indicate possible feet, in a more exact way.
limits of vibration mitigation in the fore-and-aft direction
by passive means for the x-direction random excitation Acknowledgements
used in this study. They suggest that no more than 10%
improvement in the SEATx factor is feasible for the given This work has been conducted within the 5th Frame-
spring stiffness. work programme RTD project with acronym VIBSEAT,
The presented approach has some limitations. One funded by the European Commission (Contract no.
is the use of a single axis low intensity excitation instead G3RD-CT-2002-00827) and coordinated by Professor
of a multi-axis one. Although the suspension system Michael J. Griffin of the Human Factors Research Unit
response is, by design, constrained to a single axis, a (HFRU) of ISVR, Southampton University, UK, as well
multi-axis excitation would cause the cross-axial forces, as within the Grant no. 2/6161/26 of the Slovak VEGA
which would affect the suspension motion to some extent. Grant agency. The support of both Agencies is gratefully
Secondly, the use of a rigid mass as a human body acknowledged, as well as efficient and fruitful cooperation
surrogate may be an oversimplification of the real with VIBSEAT project partners concerned.
situation, especially for more shock-like motions. Also,
due to the available project time and resources, it was not References
possible to undertake a more thorough study with a cohort
of test subjects. Berger, E.J., 2002. Friction modeling for dynamic system simulation.
The purpose of the present study was to demonstrate the Applied Mechanics Reviews 55, 535–577.
Dupuis, H., Hartung, E., Christ, E., Konietzko, H., 1988. Mechanische
effectiveness of the friction-inclusive model in response to Schwingungen. Bundesanstalt für Arbeitsschutz, Dortmund, Ger-
the low magnitude motions. Authors’ experience has many, 174pp.
shown that standard simulation tools of MATLAB/ Fairley, T.E., Griffin, M.J., 1990. The apparent mass of the seated human
Simulinks programming environment do not provide body in the fore-and-aft and lateral directions. Journal of Sound and
Vibration 139 (2), 299–306.
acceptable results. Further research is required to develop
Fleury, G., 2004. Experimentelle Untersuchung der dynamischer Masse
a more complex system model sufficient to reproduce the einer sitzender Versuchsperson bei Schwingungen in der X-Richtung
most relevant responses of the seat-load system while zur Bildung eines Modells. Tagung ‘‘Humanschwingungen’’ Darm-
remaining sufficiently simple to be practically useful for stadt, VDI Berichte No. 1821, 301–316.
product development. Fleury, G., Mistrot, P., 2006. Numerical assessment of fore-and-aft
suspension performance to reduce whole-body vibration of wheel
loader drivers. Journal of Sound and Vibration 298 (3), 672–687.
8. Conclusions Griffin, M.J., 1990. Handbook of Human Vibration. Academic Press,
London.
Two horizontal driver’s seat suspension configurations, a Hinz, B., Blüthner, R., Seidel, H., Menzel, G., 2005. Apparent mass
high friction one and a low friction one, were measured in functions of the combination subject/seat—An experimental study.
laboratory. The vibration mitigation performance was Federal Institute for Occupational Safety and Health, Berlin, Germany
(Private communication).
compared to that one simulated using a simple seat Holmlund, P., Lundström, R., 1998. Mechanical impedance of the human
suspension model loaded with a rigid mass, including body in the horizontal direction. Journal of Sound and Vibration 215,
the influence of the friction. The vibration mitigation 801–812.
 ARTICLE IN PRESS
G.J. Stein et al. / International Journal of Industrial Ergonomics 38 (2008) 396–409 409

Inman, D.J., 2001. Engineering Vibration, second ed. Prentice-Hall Inc., Nawayseh, N., Griffin, M.J., 2005b. Non-linear dual-axis biodynamic
Upper Saddle River, NJ. response to fore-and-aft whole-body vibration. Journal of Sound and
International Organization for Standardization ISO 10326-1, 1992. Vibration 282 (3–5), 831–862.
Mechanical vibration and shock—Laboratory method for evaluating Petrov, E.P., Ewins, D.J., 2002. Models of friction damping with variable
vehicle seat vibration—Part 1: Basic Requirements. normal load for time-domain analysis of vibrations. In: Proceedings of
International Organization for Standardization ISO 2631-1, 1997. the ISMA 2002 Conference. Leuven, Belgium, pp. 421–430.
Mechanical vibration and shock-Evaluation of human exposure to Schwarze, S., Northbohm, G., Hartung, E., Dupuis, H., 1999. Epidemio-
whole-body vibration-Part 1: General requirements. logische Studie ‘‘Ganzkörpervibrationen’’ AbschluXbericht. HVBG,
Klotter, K., 1980. Technische Schwingungslehre—Teil B. Springer, Berlin, St. Augustin, Germany, 288pp.
Germany. Stein, G.J., Múčka, P., Chmúrny, R., 2006. Preliminary results on an
Mandapuram, S.C., Rakheja, S., Shiping, M.A., Demont, R.G., Boileau, x-direction apparent mass model of human body sitting in a
P.-É., 2005. Influence of back support conditions on the apparent mass cushioned, suspended seat. Journal of Sound and Vibration 298 (3),
of seated occupants under horizontal vibration. Industrial Health 43 688–703.
(3), 421–435. Stein, G.J., Múčka, P., Chmúrny, R., Hinz, B., Blüthner, R., 2007a.
Mansfield, N.J., Lundström, R., 1999a. The apparent mass of the human Measurement and modelling of x-direction apparent mass of the
body exposed to non-orthogonal horizontal vibration. Journal of seated human body-cushioned seat system. Journal of Biomechanics
Biomechanics 32, 1269–1278. 40, 1493–1503.
Mansfield, N.J., Lundström, R., 1999b. Models of the apparent mass of Stein, G.J., Zahoranský, R., Múčka, P., 2007b. On dry friction modelling
the seated human body exposed to horizontal whole-body vibration. and simulation in kinematically excited oscillatory systems. Journal of
Aviation and Space Environmental Medicine 70, 1166–1172. Sound and Vibration, doi:10.1016/j/jsv.2007.08.017, in press.
Mansfield, N.J., 2005. Human Response to Vibration. CRC Press, Boca Timoshenko, S., Young, D.H., Weaver Jr., W., 1974. Vibration Problems
Raton. in Engineering. Wiley, New York.
Mansfield, N.J., Maeda, S., 2007. The apparent mass of the seated human Zahoranský, R., Stein, G.J., 2007a. Modelling and evaluation of the
exposed to single-axis and multi-axis whole-body vibration. Journal of influence of end-stops impacts on response signal in a vibration
Biomechanics 40, 2543–2551. isolating system. Engineering Mechanics 14, 55–68.
Nawayseh, N., 2002. Modelling the vertical and fore-and-aft forces caused Zahoranský, R., Stein, G.J., 2007b. Modelling a simple oscillatory system
by whole-body vertical vibration. In: Proceedings of the 37th United with friction under bi-axial base random excitation. Strojnı́cky časopis.
Kingdom Conference on Human Responses to Vibration. Department Journal of Mechanical Engineering 58, 1–16.
of Human Sciences, Loughborough University, Leicestershire, UK, Zahoranský, R., Stein, G.J., Meyer, H., 2005. Modelling and simulation
pp. 302–320. of kinematically excited vibration isolation systems with friction, under
Nawayseh, N., Griffin, M.J., 2005a. Effect of seat surface angle on forces random and harmonic excitation. In: Proceedings of the European
at the seat surface during whole-body vertical vibration. Journal of Non-linear Conference. Eindhoven University, Eindhoven, the Neth-
Sound and Vibration 284 (3–5), 613–634. erlands, pp. 2663–2671.