Professional Documents
Culture Documents
net/publication/239392024
CITATIONS READS
67 4,522
4 authors, including:
Roberto Lot
University of Southampton
125 PUBLICATIONS 2,520 CITATIONS
SEE PROFILE
All content following this page was uploaded by Roberto Lot on 13 October 2015.
Corresponding author: Prof. Alberto Doria, Via Venezia 1, 35131 Padova (Italy)
Tel. +39 049 827 6803, Fax: +39 049 827 6785,
e-mail alberto.doria@unipd.it
A Frequency-domain Method for Evaluating the Ride Comfort of a Motorcycle 1
1. SUMMARY
In many European towns the demand for fast and efficient mobility is frequently
satisfied by means of two-wheeled vehicles. The improvement of comfort of two-
wheeled vehicles used by tired and busy workers can increase safety in ground
transport. Nowadays multi-body codes make it possible to predict the ride comfort of
two-wheeled vehicles by means of time-domain or frequency-domain simulations.
Comfort indices can be developed by post-processing the results of numerical
simulations. This task is difficult, because the indices should depend on vehicle’s
characteristics and should be independent of road quality and vehicle’s speed. Poor
quality roads may generate non-linear effects. Speed influences the trim of the vehicle
and the wheelbase filtering, which takes place because the same road unevenness
excites the front and rear wheel with a time delay which depends on the vehicle’s speed.
In this paper the comfort of two-wheeled vehicles is studied by means of a frequency-
domain approach. The wheelbase filtering is averaged considering typical missions of
the vehicle. The missions are journeys with a forward speed that assumes different
values according to a probability density function. Indices of comfort are calculated
taking into account human sensitivity. The examples show that the proposed comfort
indices depend on suspensions’ characteristics and, hence, are useful design tools.
Finally, some time-domain calculations are carried out to give emphasis to non-linear
effects and to show the limits of the frequency-domain analysis.
2. INTRODUCTION
Actual roads are characterized by a random fluctuation of surface elevation about the
nominal geometry, which is called road unevenness. When a two-wheeled vehicle runs,
road unevenness forces the vertical displacements of both wheels generating vibrations,
which are transmitted through the suspensions to the frame, rider and passenger.
Vibrations cause discomfort, noise and, in the worst cases, the failure of mechanical
components or electronic equipment. Moreover, road unevenness causes variations in
tire load and hence tire adhesion may be impaired.
The spectrum of the vehicle vibrations induced by road unevenness may be divided into
three ranges of frequencies: the quasi-static range (frequency ν < 0.5 Hz), the ride
range (0.5 < ν < 20 Hz) and the acoustic range (20 < ν < 20000 Hz). The quasi-static
range corresponds to the passage through hills and slopes and the frequency of
excitation is low enough in relation to the natural frequencies of the vehicle to assume
the motorcycle as a static system. The ride range roughly corresponds to the passage on
undulations of the road. In this band of frequency the modes of vibration of the vehicle
are characterised by the rigid motion of the front and rear frames, by the travel of the
suspensions and by the deformation of the tires. A full dynamic model is required to
study the vehicle behaviour in this frequency band. The ride range is the most important
2 Vittore Cossalter, Alberto Doria, Stefano Garbin, Roberto Lot
from the point of view of comfort for the following reasons. The human sensitivities to
whole-body vibrations and to arm-hand vibrations peaks in the ranges 1-8 Hz [1] and
12-16 Hz respectively [2]. In the ride range of frequencies the spectrum of vibrations
generated by road unevenness, which is a monotonically decreasing function, still
shows relevant amplitudes. Finally, the acoustic range corresponds to the effect of road
roughness. The response of the motorcycle in this frequency range has to be calculated
taking into account the structural deformation of the front and rear frames, because the
structural modes of vibration belong to this frequency band [3][4]. The human body is
less sensitive to high frequency vibrations, but these vibrations generate noise that is
perceived by the rider.
In two-wheeled vehicles there are other sources of excitation: engine shaking forces,
rotating imbalance and aerodynamic forces. The frequencies of engine shaking forces
are in the range 40-400 Hz, which belong to the acoustic frequency range. The forces
generated by rotating inbalance of the wheels belong to the ride range only if the speed
is small and the wheel radius is large. Finally, aerodynamic forces are important only if
the speed of the vehicle is high.
The comfort of vehicles can be studied numerically both in the time-domain and in the
frequency-domain. The time-domain approach [5] makes it possible to take into account
the many non-linear phenomena that are present in vehicle dynamics. They are related
to the vehicle’s kinematics, stiffness and damping characteristics of suspensions,
presence of end-stroke pads, tire and aerodynamic forces. Actually, by means of a
detailed vehicle model and a time-domain analysis, the comfort can be evaluated in
extreme conditions too, for example off-road journeys. If a non-linear model is used the
superposition principle does not hold true and the comfort of the vehicle in different
roads and in different conditions can be evaluated only by means of different
simulations. This is the main limitation of the time-domain analysis, since many
cumbersome simulations have to be carried out to study vehicle comfort.
This inconvenience is not present in the frequency-domain approach, which is based on
linear models of the vehicle. The superposition principle makes it easy to evaluate the
vehicle’s behaviour in different conditions, such as different kinds of roads or different
speeds. The frequency-domain approach gives the possibility of formulating comfort
indices, as proposed in this paper, but non-linear phenomena are neglected.
This paper deals with the numerical study of motorcycle comfort in the ride range using
the frequency-domain method. The paper focuses on the excitation caused by road
unevenness and on the importance of suspensions, whereas other excitation sources (i.e.
engine shaking forces) are not considered. The analysis is carried out by means of a
multi-body model of the vehicle which includes rigid bodies, suspensions and
deformable tires. At a given speed, the steady state configuration of the vehicle is
evaluated using the full non-linear model, then the equations of motion are linearized
and the analysis is carried out in the frequency-domain. Some additional simulations in
the time-domain are presented to assess the validity and show the limitations of the
proposed frequency-domain analysis. The frame compliance is not included in the
A Frequency-domain Method for Evaluating the Ride Comfort of a Motorcycle 3
model because the “structural” modes of vibration do not belong to the ride range of
frequencies [3] and are less excited than the “rigid” modes.
The study of comfort becomes more complex owing to the wheelbase filtering [6][7].
This phenomenon takes place because, when a two-wheeled vehicle follows a path, both
the wheels pass over the same road unevenness but at different times. The excitation of
the rear wheel is the same as the front wheel, but has a time delay which is proportional
to the wheelbase and inversely proportional to the speed. The significance of the
wheelbase filtering is discussed and an original method for estimating the average
behaviour of the vehicle for different speeds is developed. Finally, indices for comfort
evaluation are proposed and are used for investigating the influence of the suspensions
stiffness and damping on the ride comfort.
The human body is sensitive to accelerations. Studies dealing with vehicle’s comfort are
based on the analysis of the accelerations of the vehicle sprung mass when the vehicle
runs at constant speed on an uneven road surface. In this paper the motorcycle model
described in reference [8] is used. The vehicle was modelled with four rigid bodies (the
frame with the rigidly attached rider, the swinging arm and the wheels, see figure 1) and
the non-linear equations of motion were derived. In the framework of this research the
out-of-plane degrees of freedom are not used.
For any given vehicle speed V, the steady state conditions are calculated according to
the non-linear equations of motion. From this configuration, the equations of motion are
numerically linearized. Due to the presence of the road excitation, the linearized
equations of motion become:
(1) M q + C q + K q = C*f + K *f
in which f = { f f , f r } is the vector of road unevenness that excites the front and rear
T
µ
S x
z
fr ff
At constant speed, all matrices of equation (1) are time-constant. If the drag force is
neglected, the trim, and therefore these matrices, do not depend on the vehicle speed.
The longitudinal and vertical displacement of the saddle ( x, z ) can be calculated as
linear combinations of coordinates q through matrix D :
x
(2) =
y = Dq .
z
(5) Y (ω ) = DQ (ω )
where F (ω ) , Q (ω ) and Y (ω ) correspond to f , q and y in the time-domain. The
Frequency Response Functions (FRFs) in terms of accelerations may be calculated as
follows:
(6) Y (ω ) =
−ω 2 Y (ω ) =
H (ω ) F (ω )
in which H (ω ) is the FRFs matrix of the system:
H (ω ) H x , r (ω )
−ω 2 D ( −ω 2 M + iω C + K ) ( iω C* + K * ) .
−1
(7) H (ω ) =
x, f
=
H z , f (ω ) H z , r (ω )
H (ω ) has dimension 2x2 and the rows correspond to the outputs x, z (first index),
whereas the columns correspond to the inputs f f , f r (second index).
Throughout this paper the numerical calculations are performed by means of the
FastBike code [8] considering a commercial scooter with 150 cc displacement. The trim
of the scooter, which is the geometric configuration that the motorcycle assumes in
steady state condition [9], was calculated assuming a constant forward speed of 20 m/s,
then the equations of motion were linearized and the FRFs matrix was calculated.
A Frequency-domain Method for Evaluating the Ride Comfort of a Motorcycle 5
Plots a) and b) of figure 2 show the magnitudes and phases of the FRFs between the
horizontal acceleration and the front wheel displacement H x , f (ω ) and between the
horizontal acceleration and rear wheel displacement H x , r (ω ) . With the front input the
magnitude of the horizontal acceleration response regularly increases with the frequency
and reaches a maximum near the front hop mode natural frequency (the modal
characteristics of the vehicle are summarized in table 1), whereas with the rear input the
magnitude of the response suddenly increases near the pitch mode natural frequency. In
the whole range of frequencies the phase difference between the two FRFs is about π ,
this result agrees with the physical behaviour that horizontal vibrations due to front and
rear excitation have opposite directions.
Plot c) of figure 2 shows the magnitude and phase of the FRF between the vertical
acceleration of the saddle and the front wheel displacement H z , f (ω ) . It highlights two
peaks at 3.2 Hz and 15.7 Hz, the first one is close to the natural frequencies of the
bounce and pitch modes, the second one is close to the frequency of the front hop mode.
Plot d) shows the FRF between the vertical acceleration of the saddle and the rear wheel
displacement H z , r (ω ) . This FRF has larger magnitude at low frequency than the
previous one, in particular the magnitude is large in the range of frequencies that
includes the bounce, pitch and rear hop modes. The phase diagrams of H z , f (ω ) and
Frequency [Hz] Frequency [Hz]
2
800
a) b)
fron t hop
600 600
rear hop
bounce
pitch
bounce
400 400
pitch
200 200
0 0
+π +π
0 0
−π −π
800 800
Rear contact point disp. [m]
phase [rad] front contact point disp. [m]
vertical acceleration [m/s2]
c) d)
600
rear hop
bounce
pitch
front hop
bounce
400 400
pitch
200
0 0
+π +π
phase [rad]
0 0
−π −π
0 5 10 15 20 0 5 10 15 20
Frequency [Hz] Frequency [Hz]
H z , r (ω ) are similar and show that at low frequency the vehicle acceleration is in
opposition with the contact point displacement, whereas at higher frequency the vehicle
acceleration is almost in phase with the contact point displacement.
It is worth highlighting that, at frequency higher than wheel hop resonances, the
magnitude of acceleration response decreases as the frequency increases in the FRFs
shown in the figure 2.
In order to model the vibration of the vehicle when it runs at different speeds, the non-
linear equations of motion should be linearized in every condition and the FRFs matrix
should be recalculated because the variations in speed may cause variations in the trim.
In this paper the FRFs matrix calculated at the mean speed is assumed to hold true for a
range of speeds, since the variations in the trim are negligible. Some calculations were
carried out to determine the range of speeds in which the frequency response matrix is
approximately constant; these results will be shown in section 3.5.
− iω
p
H x* H x, f + H x, r e V
(11) * = p
H z − iω
H z , f + H z , r e
V
This correlation between front and rear excitation is also called wheelbase filtering. In
order to understand the physical meaning of wheelbase filtering, it is useful to consider
the genesis of the correlated FRFs H* (ω ,V ) by means of a Nyquist plot, i.e the locus
described in the complex plane by a vector having the magnitude and phase angle of the
p
− iω
FRF. The term e V in equation (11) has the meaning of a rotational operator in the
complex plane, therefore, a correlated FRF is the resultant of the vector due to the front
p
excitation and the vector due to the rear excitation rotated by angle −ω , which is
V
caused by the wheelbase effect. Figure 3 refers to the saddle vertical acceleration. Plot
a) is the Nyquist plots of H z , f (ω ) , H z , r (ω ) , the complex vectors at a frequency ν = 13.2
Hz are also represented, the phase difference is about π 2 . Plot b) shows the correlated
FRF H z* (ω ,V ) which is given by the composition of the previous two FRFs when the
forward speed is 20 m/s. The phase difference between the vectors corresponding to
front and rear excitation is strongly modified by the wheelbase effect: the two
components that give H z* (ω ,V ) are almost in phase.
The influence of the vehicle speed and excitation frequency on the wheelbase filtering
may be highlighted by calculating the magnitudes of the correlated FRFs as follows:
pω
H x* (ω , V ) H x, f (ω ) + H x, r (ω ) + 2 H x, f (ω ) H x, r (ω ) cos + ∆ϕ x (ω )
2 2
(12) =
V
pω
H z* (ω , V ) H z , f (ω ) + H z , r (ω ) + 2 H z , f (ω ) H z , r (ω ) cos + ∆ϕ z (ω )
2 2
(13) =
V
in which ∆ϕ x (ω ) is the phase difference between H x , f (ω ) and H x , r (ω ) and ∆ϕ z (ω ) is the
phase difference between H z , f (ω ) and H z , r (ω ) . Equations (12) and (13) show that the
response reaches the maximum value when the argument of cosine is 2kπ ( k is an
integer number) and reaches the minimum value when the argument is ( 2k + 1) π .
90° 90°
1200 1200
Amplitude [1/s 2]
a) b)
900 45° 900 45°
-iω p/V
H e
600 600 z,r
H
z,f H
300 300 z,f
0 0 H'
z 0°
H
z,r
270° 270°
Figure 3: Saddle acceleration. Nyquist plot of the front and rear FRFs and of the correlated FRF.
8 Vittore Cossalter, Alberto Doria, Stefano Garbin, Roberto Lot
By substituting the equations above in the expressions (12) and (13) one obtains
pω
H x* (ω , V ) H x (ω ) 2 1 − cos
(16) =
V
pω
H z* (ω , V ) H z (ω ) 2 1 + cos
(17) =
V
These equations highlight that the response of the vehicle is modulated with a constant
frequency ∆ν = V p and the maximums of horizontal acceleration correspond to the
minimums of vertical acceleration and vice versa.
These considerations are still valid for real vehicles. Indeed, figure 4 shows the
magnitude of the correlated FRFs of the longitudinal and vertical saddle acceleration for
three different values of the forward speed (5, 20 and 35 m/s). The wheelbase
modulation is clear and in agreement with the above considerations. In addition, it may
be noticed that, since the number of maximums and minimums decreases when the
forward speed increases, the FRFs are more regular at high speed then at a low speed.
5 m/s a) 5 m/s b)
20 m/s 20 m/s
road uneveness [m]
road uneveness [m]
700 700
350 350
0 0
0 5 10 15 20 0 5 10 15 20
Frequency [Hz] Frequency [Hz]
k =
n n2 k > k0
in which k0 = 1 is the cut-off wavenumber,
= n1 2,=n2 1.5 are exponents and S0 is a
constant that depends on the quality of the road, see table 2.
If the speed of the vehicle is constant, the PSD in the wavenumber domain is
transformed into a new PSD in time domain by means of the following equation [14]:
S0 Vk0 = n n1 ω V ≤ k0
n
If forward speed V increases, road unevenness having a given range of wavelengths are
transformed into wider ranges of frequencies and for this reason the factor 1 V is present
in the equation (19). At the same time, the magnitude of the excitation at a given
frequency increases with the speed because it corresponds to a shorter wavenumber and
larger amplitude unevenness. The second effect is dominant since n > 1 and Srr (ω ,V )
increases as V increases, see figure 5.
Ride quality may be evaluated looking at the PSD of vehicle acceleration. Since the
correlated FRFs have been calculated in the previous section, the PSDs of the horizontal
and vertical saddle acceleration can be calculated as follows [14]:
2
(20) S xx (ω , V ) = H x* (ω ) S rr (ω , V )
2
(21) S zz (ω , V ) = H z* (ω ) S rr (ω , V )
Figure 6 shows the PSDs of the horizontal and vertical saddle acceleration of the scooter
that runs at different speeds on an average quality road. Some phenomena are evident.
The PSD of horizontal acceleration at low frequency (<5 Hz) is much smaller than the
PSD of vertical acceleration, because at low frequency the FRFs of the horizontal
acceleration are smaller than the ones of vertical acceleration. The wheelbase filtering
found in the FRFs is still present in the PSDs of accelerations. As the frequency
-2
10
PSD of road profile [m2/Hz]
-3 35 m/s
10
-4
10
-5
20 m/s
10
-6 5 m/s
10
-7
10
0.1 0.5 1 5 10 20
Frequency [Hz]
Figure 5: PSD of road profile. Average quality road.
10 Vittore Cossalter, Alberto Doria, Stefano Garbin, Roberto Lot
0.4 4
[(m/s ) /Hz]
0.2 2
2 2
2 2
0.1 1
0 0
0 5 10 15 20 0 5 10 15 20
Frequency [Hz] Frequency [Hz]
Figure 6: PSD of saddle acceleration, average quality road.
a) saddle horizontal acceleration, b) saddle vertical acceleration.
increases, the PSDs of accelerations decreases because the road excitation strongly
decreases. When the forward speed increases the PSDs of accelerations increases
because road excitation is larger.
4. EVALUATION OF COMFORT
If equation (23) is introduced into equation (22), the magnitudes and phases of the FRFs
can be moved outside the integral because they do not depend on the speed:
A Frequency-domain Method for Evaluating the Ride Comfort of a Motorcycle 11
Szz (ω ) H z , f (ω ) + H z , r (ω ) wa (ω )
2 2
=
(24)
+2 H z , f (ω ) H z , r (ω ) cos ∆ϕ z (ω ) wr (ω ) − sin ∆ϕ z (ω ) wi (ω )
with:
Vmax pω
(25) wr (ω ) = ∫ Vmin
Srr (ω , V ) cos
V
p (V ) dV
Vmax pω
(25) wi (ω ) = ∫Vmin
Srr (ω , V ) sin
V
p (V ) dV
Vmax
(27) wa (ω ) = ∫ Vmin
Srr (ω , V ) p (V ) dV .
and for the acceleration of any other point which is relevant to comfort (e.g. a point on
the handle-bars). Therefore the weight functions wr (ω ) , wi (ω ) , wa (ω ) make it possible to
evaluate the statistic behaviour of the vehicle moving at different speeds.
It is worth to mention that the integration does not eliminate the wheelbase filtering, but
calculates a “mean” wheelbase filtering. If the vehicle operates in the range of high
speeds, pω V is about zero in the whole frequency range and the integration of equation
(22) leads to:
Szz (ω ) ≈ H z , f (ω ) + H z , r (ω ) + 2 H z , f (ω ) H z , r (ω ) cos ∆ϕ z (ω ) wa (ω )
2 2
(29)
because wr (ω ) ≈ wa (ω ) and wi (ω ) ≈ 0 . Equation (29) has the meaning of PSD of vertical
saddle acceleration neglecting the wheelbase filtering.
The simplest, but not trivial, p (V ) is the uniform distribution: the probability of forward
speed taking every value of the range of speeds is equal. Figure 7 shows the mean PSD
of the horizontal and vertical acceleration considering an uniform distribution in the
range 5-35 m/s.
12 Vittore Cossalter, Alberto Doria, Stefano Garbin, Roberto Lot
0.3 3
a) uniform b)
2 2
2 2
0.2 2
wheelbase filtering
neglected uniform
0.1 1 distribution
0 0
0 5 10 15 20 0 5 10 15 20
Frequency [Hz] Frequency [Hz]
Figure 7: Mean PSD of saddle acceleration, average quality road and uniform distribution.
a) horizontal acceleration, b) vertical acceleration.
The mean PSD of horizontal acceleration (solid line) shows a large maximum at about 5
Hz. The vertical acceleration shows a large maximum at low frequency (close to the
frequencies of the pitch and bounce modes) and a secondary maximum at about 12 Hz.
The shape of the curve and the values are similar to the ones of vertical acceleration
PSD at the average speed of 20 m/s (figure 6), but the sequence of maximums and
minimums typical of the wheelbase filtering is less evident.
In figure 7 two limit curves are represented as well. They are the mean PSD of
accelerations that are calculated neglecting the wheelbase filtering (equation (29) for the
vertical acceleration). The differences between the mean curves in the presence of the
wheelbase filtering and the limit curves highlight that equation (24) does not eliminate
the wheelbase filtering. In particular figure 7 shows that the wheelbase filtering
decreases the vertical acceleration and contributes to generate the horizontal
acceleration.
The β probability density function [15] is well suited to define a mission profile of a
vehicle, since it can be defined on the interval [Vmin Vmax ] and the mean value can be
moved within the interval. The mean value and the variance of the β probability density
function are defined by setting the values of two parameters ( a, b ). Figure 8 shows the β
probability density function for different choices of a, b . It is worth highlighting that, if
a = b , the probability density function is symmetric, the mean coincides with the center
of the interval and the variance decreases if a increases. If a= b= 1 the β probability
density function coincides with the uniform probability density function.
0.1
urban mission mean speed main road mission
0.08 (a=2, b=4.6) (a=4, b=4) (a=4.6, b=2)
Probability density
0.06
0.04
0.02
0
5 10 15 20 25 30 35
Forward speed [m/s]
Figure 8: β probability density function of forward speed with different exponents
A Frequency-domain Method for Evaluating the Ride Comfort of a Motorcycle 13
From a practical point of view the missions with symmetric β probability density
function are well suited to study the comfort of a vehicle which runs principally at
speeds that are close to the mean of the range of possible speeds of the vehicle (mean
speed mission). However some users of the same vehicle may run, on average, at lower
speeds (urban mission) or at higher speeds (main-road mission). In these cases a non-
symmetric β probability density function is more suited to study the comfort.
Figure 9 deals with three missions: a mission with mean speed 20 m/s and symmetric
probability density function ( a= b= 4 ) ; a main-road mission with mean speed 26 m/s
=( a 4.59
= b 1.97 ) ; ( a 1.97
a urban mission with mean speed 14 m/s= = b 4.59 ) . The
variances of the distributions for each mission are equal, the matrix of FRFs is
calculated at the mean speed. The symmetric β probability density function gives results
similar to the ones obtained with the uniform probability density function, but the
undulations due to the wheelbase filtering are more evident. The mean curves are fairly
similar to the curves of acceleration PSD calculated at the mean speed (20 m/s, figure
6).
The main-road mission, with large probability densities of high speeds, leads to an
enlargement of the peak of the horizontal acceleration PSD and to an increase of the
peak of the vertical acceleration PSD. These effects take place because the road PSD at
high speed is large (see figure 5). The urban mission, with large probability densities at
the low speeds, leads to a reduction of main peak of both the vertical and horizontal
acceleration PSDs.
0.3 3
a) main road mission b)
mean PSD of horizonatal
2 2
0.2 2
0 0
0 5 10 15 20 0 5 10 15 20
Frequency [Hz] Frequency [Hz]
Figure 9: Mean PSD of saddle vertical acceleration, average quality road and some β distributions.
a) horizontal acceleration, b) vertical acceleration.
14 Vittore Cossalter, Alberto Doria, Stefano Garbin, Roberto Lot
0.3 3.0
mean PSD of horizontal a) non-weighted PSD b)
acceleration [(m/s ) /Hz]
0.2
2 2
2.0
PSD weighted PSD weighted
1.5
according to according to
0.1 human sensitivity 1.0 human sensitivity
0.5
0.0 0.0
0 5 10 15 20 0 5 10 15 20
Frequency [Hz] Frequency [Hz]
Figure 10: Effect of human sensitivity. a) horizontal acceleration, b) vertical acceleration.
In which ωR is the upper limit of the ride range. k x = 1.4 and k z = 1 are the coefficient
proposed in ISO 2631 standard [1] for combining weighted accelerations in the two
directions.
The proposed comfort index takes into account the vehicle dynamic behavior, road
quality and mission profile and also has the physical meaning of the root-mean-square
value of the acceleration of the time history which corresponds to the mission.
Since a linear behaviour of the vehicle is assumed, the quality of the road influences the
accelerations only through the constant S0 (see equation (19)), which can be moved
outside the integrals. Therefore a slightly different definition of the comfort index that
does not depend on the road quality is the following:
ωR
∫k w
1
=
(31) I 0 x
2
x (ω ) Sxx (ω ) + k z wz2 (ω ) Szz (ω ) dω
S0
0
This index is useful to compare vehicle comfort characteristics regardless of the quality
of the road, but it loses its meaning as the overall level of accelerations.
The peaks of the PSD have a limited influence on the index based on the overall level, if
their bandwidth is narrow. Hence, a vehicle with a small overall level and relevant
resonance peaks may be perceived as very comfortable. In order to overcome this
problem, quadratic indices that are more sensitive to the PSD peaks may be defined
using a second power of the PSD:
4 ωR
∫
2 2
=(32) Q k x wx2 (ω ) Sxx (ω ) + k z wz2 (ω ) Szz (ω ) d ω
0
4 ωR
∫
1 2 2
k x wx2 (ω ) Sxx (ω ) + k z wz2 (ω ) Szz (ω ) d ω
=(33) Q0
S02 0
Table 3 shows the influence of the kind of mission on the comfort indices. The speed
range is always 5-35 m/s. The missions with uniform distribution and β symmetric
distribution give very similar indices. The non symmetric β distribution
16 Vittore Cossalter, Alberto Doria, Stefano Garbin, Roberto Lot
=( a 4.59
= b 1.97 ) that corresponds to a main-road mission shows the largest indices,
( a 1.97
whereas the non symmetric β distribution= = b 4.59 ) that corresponds to a urban
mission shows the smallest indices. The quadratic index Q0 is more sensitive to the kind
of mission, because figure 9 shows that the most important difference between the
uniform, urban and main-road mission is the height of the main peak.
2 2
0.1 0.1
very good road very good road
0.01 0.01
linear linear
non-linear non-linear
0.001 0.001
0 5 10 15 20 0 5 10 15 20
Frequency [Hz] Frequency [Hz]
Figure 11: Comparison between linear and non-linear analysis, mean speed mission.
a) horizontal acceleration, b) vertical acceleration.
There is a good agreement between the two sets of results both in terms of numerical
values of PSD at a certain frequency and in terms of the shape of the curves. The linear
analysis slightly underestimates the main peak of the horizontal acceleration PSD and
overestimates the main peak of the vertical acceleration PSD. The non-linear analysis,
like the linear analysis, predicts PSD values that are roughly proportional to the road
quality; the curves that derive from the non-linear analysis shows irregularities which
may be related with the excitation of non-linear phenomena a certain frequencies.
C.I.=335 350
345 355 345
18 340
1600 C.I.=335
340 reference
350
360 16 case
1400 reference
case 14
1200
12
350
1000 10
500 600 700 800 900 1000 10 12 14 16 18 20
Front suspension damping [Ns/m] Front suspension stiffness [kN/m]
Figure 12: Effect of suspensions’ characteristics on the comfort index. a) effect of damping, b) effect of stiffness.
18 Vittore Cossalter, Alberto Doria, Stefano Garbin, Roberto Lot
Figure 12 deals with the numerical results. Plot a) is the contour plot of I o against the
front and rear values of damping. Since a linear model is used the damping values are
the averages of the values in compression and extension. The front and rear suspension
stiffnesses are held at the reference values of 17650 and 16000 N/m respectively. The
comfort index depends more on the front suspension damping than on the rear
suspension damping and reaches the minimum value when the front suspension
damping is small. The black circle shows the reference setting of damping values and
the corresponding value of the comfort index, which is close to the minimum value. Plot
b) of figure 12 is the contour plot of I o against the front and rear suspension stiffnesses.
The front and rear values of damping are held at the reference values of 750 and 1350
Ns/m respectively. The black circle shows the reference setting of stiffnesses. The rear
suspension stiffness is close to the optimum value, whereas the front suspension
stiffness has to be reduced to improve the comfort. Actually, the choice of suspensions’
characteristics has to take into account other aspects of motorcycle behaviour (trim
variations, handling properties) and the front suspension stiffness cannot be decreased
too much.
5. 4.CONCLUSIONS
The frequency-domain method proposed in this paper makes use of four FRFs, which
describe the longitudinal and vertical response of the saddle to road excitation at the
front and rear wheel. The wheelbase filtering is taken into account by combining the
complex FRFs due to front and rear excitation. The PSD of saddle acceleration is then
calculated assuming random properties of the road profile.
The comfort is evaluated by introducing the concept of vehicle mission, which is a
typical journey with a forward speed that assumes different values according to a
probability density function.
The proposed method is very fast and makes it possible to define a numeric comfort
index that depends on the characteristics of the suspensions and can be used for
optimization purposes. With simple modifications the method may take into account
hand-arm vibrations, foot vibrations and the vibrations transmitted to the passenger. The
proposed method can be used for elaborating the FRFs measured in laboratory by means
of shakers and accelerometers, so it is able to predict the comfort on the road starting
from stationary tests. Since the numerical code also calculates the FRF between tire
loads and wheels’ displacements, the proposed method makes possible the definition of
adhesion indices based on the variations in tire loads during typical missions.
The main limit of the method coincides with the limit of linear analysis: the method is
well suited to study the comfort on asphalt roads, but cannot be used for studying the
comfort in off-road missions, which are characterized by large suspension travels.
A Frequency-domain Method for Evaluating the Ride Comfort of a Motorcycle 19
6. ACKNOWLEDGEMENTS
This research was partially supported by funds from the University of Padova (2003
Research Projects) and from the Ministry of Education, University and Research (PRIN
2004).
REFERENCES
1.
2. ISO 2631: Evaluation of Human Exposure to Wholebody Vibration, 2003.
3. ISO 5349: Guidelines for the Measurement and Assessment of Human Exposure to Hand Transmitted Vibration.
4. Cossalter, V., Doria, A. and Mitolo, L.: Inertial and Modal Properties of Racing Motorcycles. In: Proc. of Motorsports
Engineering Conference & Exhibition, Indianapolis, Indiana, USA, 2002, SAE Paper Number 02MSEC-5, pp 605-612.
5. Cossalter, V., Doria, A. Basso, R. and Fabris, D.: Experimental Analysis of out-of-plane Structural Vibrations of Two-
wheeled Vehicles. Shock and Vibration 11, IOS press, Amsterdam (2004), pp 433-443.
6. Goncalves, J. P. C., Ambrosio, J. A.C.: Optimization of Vehicle Suspension Systems for Improved Comfort of Road
Vehicles using Flexible Multibody Dynamics. Nonlinear Dynamics 34, Kluwer Academic Publishers, Dordrecht (2003),
pp 113-131.
7. Hunt, H.E.M.: Modelling of road vehicles for calculation of traffic-induced ground vibration as a random process,
Journal of Sound and Vibration, 144, Academic Press, London (1991), pp. 41-51.
8. Hunt, H.E.M.: Stochastic modelling of traffic-induced ground vibration, Journal of Sound and Vibration, 144, Academic
Press, London (1991), pp. 53-70 .
9. Cossalter, V., Lot, R.: A Motorcycle Multi-Body Model for Real Time Simulations Based on the Natural Coordinates
Approach. Vehicle System Dynamics, 37, Swets & Zeitlinger, Lisse (2002), pp. 423-447.
10. Cossalter, V.: Motorcycle Dynamics. Race Dynamics, Greendale WI, 2002.
11. Nigam, N., C., Narayanan, S.: Applications of Random Vibrations. Springer-Verlag, Berlin, 1994.
12. Dixon, J. C.: Tires, Suspensions and Handling, second edition. SAE, Warrendale PA, 1996.
13. Sun, L.: On human perception and evaluation to road surfaces, Journal of Sound and Vibration, 247, Academic Press,
London (2001), pp. 547-560.
14. ISO. Draft Standard ISO/TC 108/WG9: Proposals for Generalized Road Inputs to Vehicles, 1972.
15. Davis, B.R.Thompson, A.G: Power spectral density of road profiles, Vehicle System Dynamics, 35, Swets & Zeitlinger,
Lisse (2001), pp. 409-415.
16. Mood, A. M., Graybill, F. A. and Boes, D. C.: Introduction to the Theory of Statistics. McGraw-Hill, New York, 1974.
17. Griffin, M. J.: Handbook of Human Vibration. Elsevier Ltd, Amsterdam, 2004.
18.