Professional Documents
Culture Documents
motorcycle simulations
DCT 2005.65
Master's thesis
Met de rekencapaciteit van huidige computers wordt het simuleren van het gedrag van voertu-
igen steeds belangrijker. Vooral bij motorfietssimulaties speelt het bandmodel een belangrijke rol
in het gedrag van het voertuig. Door TNO Automotive is daarom een bandmodel voor motor-
fietssimulaties ontwikkeld, MF-MCTyre. Ondanks het belang van een nauwkeurig bandmodel,
zijn bepaalde onderdelen van het MF-MCTyre model afgeleid van een autobandmodel. Daarom is
vooral de invloed van de camberhoek, die voornamelijk aanwezig is tijdens het rijden door bochten,
op deze onderdelen onderbelicht gebleven. Het doel van dit onderzoek is dan ook om het stationair
gedrag in bochten van het MF-MCTyre model te verbeteren.
Om vertrouwd te raken met het stationair gedrag van het bandmodel en een motorfiets in
een bocht, is een simulatiemodel van een motorfiets ontwikkeld aan de hand van het model van
Cornelis Koenen. Door middel van simulaties met dit model en het bestaand bandmodel is het
stationair gedrag van beide geanalyseerd. Een literatuurstudie toont aan dat het model van
Koenen een van de meest complete modellen is. Tevens laat deze literatuurstudie zien dat de
Magic Formula algemeen geaccepteerd wordt. Daarom wordt aangenomen dat het MF-MCTyre
model dat gebaseerd is op deze Magic Formula in staat moet zijn om het stationair gedrag van
een motorfietsband in een bocht correct weer te geven.
Tijdens de analyse van het bandmodel komen er drie problemen naar voren die de nauwkeurigheid
van het model reduceren. Ten eerste worden er vereenvoudigde aannames gemaakt voor de ver-
ticale stijfheid en de bandcontour. Dit leidt tot een onnauwkeurige hoogte van het wielcentrum
en de belaste bandstraal gedurende simulaties. Omdat de belaste bandstraal ook tijdens metin-
gen onnauwkeurig wordt bepaald zijn de transformaties van de momenten tussen het contactpunt
en het wielcentrum ook onnauwkeurig. Ten tweede wordt ook de effectieve rolstraal incorrect
bepaald door het bandmodel. De bandcontour wordt buiten beschouwing gelaten en het effect van
de verticale belasting wordt overgenomen van een autobandmodel. Daardoor zijn de rotatiesnel-
heden van de wielen onnauwkeurig tijdens voertuigsimulaties. Tenslotte wordt er ook een sterk
vereenvoudigde aanname gemaakt voor de rolweerstand. Als gevolg hiervan is weergave van de
langskracht en het terugstelmoment onnauwkeurig tijdens simulaties.
Een uitgebreid meetprogramma is opgezet, om bepaalde specifieke aspecten van een motorfiets-
band te bepalen. Door middel van deze metingen zijn de bandcontour, verticale stijfheid, effectieve
rolstraal en rolweerstand bepaald. Deze resultaten zijn gebruikt om de eerder genoemde problemen
van het bandmodel op te lossen en een verbeterd bandmodel te ontwikkelen. Uit tests blijkt dat
dit nieuwe model voornamelijk bij grote camber hoeken de krachten en momenten nauwkeuriger
representeert. Tevens is aangetoond dat het verbeterde bandmodel een duidelijke invloed heeft of
het stationair gedrag van het motorfietsmodel. Parameters als de camberhoek, het stuurkoppel,
de leunhoek van de bestuurder, de rotatiesnelheden van de wielen en de hoogte van de wielcentra
veranderen significant als het nieuwe bandmodel wordt gebruikt in plaats van het oude.
i
Abstract
With the computational capacity of computers available nowadays, simulating the behaviour of
vehicles becomes more and more important. Especially for motorcycle simulations, the tyre model
has a significant influence on the vehicle behaviour. TNO Automotive has therefore developed
a tyre model for motorcycle simulations, MF-MCTyre. Although the importance of an accurate
tyre model, several aspects of the MF-MCTyre model are derived from an automobile tyre model.
Therefore, especially the influence of the camber angle on these aspects has been underexposed.
The goal of this research is therefore to improve the steady state cornering behaviour of the
MF-MCTyre model.
To get familiar with the tyre model and motorcycle steady state cornering behaviour, a mo-
torcycle simulation model is developed for which the model of Cornelis Koenen is used as a basis.
From simulations with the motorcycle model and the existing tyre model the steady state behaviour
of the motorcycle model and its tyres is analyzed. A literature study shows that the model of
Koenen is one of the most complete models. Furthermore, this literature study shows that the
Magic Formula is broadly accepted. Therefore, it is believed that the MF-MCTyre model, which
is based on this Magic Formula, should be able to represent the steady state cornering behaviour
of a motorcycle tyre correctly.
If the tyre model is analyzed, three main problems are found which decrease the accuracy of the
tyre model. First of all, simplified assumptions are made for the motorcycle tyre contour and the
vertical tyre stiffness. Therefore the wheel center height and loaded radius during the simulations
are inaccurate. As the loaded radius is also inaccurately determined during measurements, the
transformation of the moments between the wheel center and the contact point are also inaccurate.
Secondly, the effective rolling radius is incorrectly determined by the model. The tyre contour is not
taken into consideration and the effect of the vertical load is copied from an automobile tyre model.
During vehicle simulations, this leads to inaccurate rotational velocities of the wheels. Finally, a
simplified assumption is made for the rolling resistance. During simulations, this assumption leads
to an inaccurate representation of the longitudinal force and the self aligning moment.
An extensive measurement program is conducted, in order to determine the behaviour of specific
motorcycle tyre aspects. Moreover, the tyre contour, vertical stiffness, effective rolling radius
and rolling resistance are determined within this measurement program. The results of these
measurements are used to overcome the problems mentioned above and to develop an improved
tyre model. From simulations it is learned that this new tyre model represents the forces and
moments more accurately, especially under large camber angles. Furthermore, it is shown that
the improved tyre model has a significant influence on the steady state cornering behaviour of
the motorcycle model. Moreover, parameters as the camber angle, steer torque, rider lean angle,
wheel rotational speeds and wheel center heights show significant differences if the improved tyre
model is used instead of the present one.
ii
Contents
Samenvatting i
Abstract ii
1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Problem statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Research layout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 Literature study 3
2.1 Motorcycle modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1.1 Model structure and features . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1.2 Mass and inertial parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.3 Stabilizing controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Motorcycle tyre modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
iii
CONTENTS iv
Sign conventions
Throughout this report the ISO sign convention for force, moment and wheel slip of a tyre is used.
This sign convention is depicted in figure 1.
x
V a F x z
g
F z
y F y
M z
y F y
M x
T o p v ie w R e a r v ie w
The MF-MCTyre model is in accordance with the standard TYDEX conventions. Two TYDEX
coordinate systems with ISO orientation are particularly important, the C- and W-axis systems as
depicted in figure 2. The C-axis system is fixed to the wheel carrier with the longitudinal xc -axis
Z
nr
nr
X
Zc Y
Zw Zc
Vx
Zw
Vc
Xc
Xc
O Xw
Xw Yc
-γ
Yc
Yw
Yw
C
C
parallel to the road and in the wheel plane (xc -zc -plane). The origin O of the C-axis system is
the wheel center. The origin of the W-axis system is the road contact point C defined by the
intersection of the wheel plane, the plane through the wheel spindle and the road tangent plane.
The xw -yw -plane is the tangent plane of the road in the contact point C, and it defines the camber
angle γ together with the normal nr to the road plane (xw -yw -plane). Furthermore, in chapter 6
the T -axis system is introduced. This axis system is defined in the contact and therefore it has
the same origin as the W -axis system. However, the T -axis system rotates with the camber angle
and is therefore parallel to the C-axis system.
SIGN CONVENTIONS AND SYMBOLS vi
Symbols
Symbol Description Unit
Capitals
A Actual contact point
B Magic Formula factor [−]
C Magic Formula factor [−]
C Fictitious tyre-road contact point
D Magic Formula factor [−]
N
Cz Vertical tyre stiffness [m ]
D Magic Formula factor [−]
E Magic Formula factor [−]
E Relative fitting error [−]
F Magic Formula factor [−]
Fa Centrifugal force [N ]
Fd Aerodynamic drag force [N ]
Fg Gravitational force [N ]
Fl Aerodynamic lift force [N ]
Fx Longitudinal force [N ]
Fy Lateral force [N ]
Fz Vertical force [N ]
Fzo Nominal vertical load [N ]
J Moment of inertia [kgm2 ]
cm J Inertia tensor with respect to the center of mass [kgm2 ]
oJ Inertia tensor with respect to an arbitrary point O [kgm2 ]
K Gain [−]
Kz Vertical tyre damping [ Nms ]
Mrr Rolling resistance moment magnitude [N m]
Ms Steering torque [N m]
Mx Overturning moment [N m]
My Rolling resistance moment [N m]
Mz Self aligning moment [N m]
O Wheel center position [m]
O Origin of motorcycle inertial frame
R Cornering radius [m]
Re Effective rolling radius [m]
Reo Effective rolling radius of an undeformed tyre [m]
Rl Loaded radius [m]
Ro Unloaded radius [m]
SH Horizontal shift in the Magic Formula
SV Vertical shift in the Magic Formula
V Velocity [m
s ]
Vo Reference velocity [m
s ]
Vr Linear rolling velocity [m
s ]
Vsx Longitudinal slip speed [m
s ]
Vsy Lateral slip speed [m
s ]
Vx Longitudinal velocity [m
s ]
Vy Lateral velocity [m
s ]
Ym Measurement point
Yf Description of the measurement point
Yn Parameter obtained by simulating the improved tyre model
Yo Parameter obtained by simulating the existing tyre model
SIGN CONVENTIONS AND SYMBOLS vii
Normal
a Ellipse parameter with respect to the x-direction [−]
b Ellipse parameter with respect to the y-direction [−]
d Relative difference [−]
fr Rolling resistance coefficient [−]
g Gravitational constant [ sm2 ]
g1x Distance from mass m1 to steer axis [m]
g1ux Distance from mass m1u to steer axis [m]
h Height of the wheel center [m]
m Mass [kg]
nr Normal to the road
pdz Vertical position of the aerodynamic application point [m]
plx Horizontal position of the aerodynamic application point [m]
qc Camber influence on vertical stiffness [−]
qf cx Longitudinal force influence on vertical stiffness [−]
qf cy Lateral force influence on vertical stiffness [−]
qf z1 Vertical deflection influence on vertical stiffness [−]
qf z2 Quadratic influence of vertical deflection on stiffness [−]
qsy1 Initial rolling resistance moment Mrr [−]
qsy5 Camber influence on Mrr [−]
qsy6 Fzw influence on Mrr [−]
qv2 Vertical stiffness increase with speed [−]
rcm
~ Vector from an arbitrary point O to the center of mass [m]
rt Tyre cross section radius [m]
t Time [s]
t Mechanical trail [m]
u Distance [m]
yM x Distance between actual and fictitious contact point [m]
Greek
Ω Tyre rotational velocity [ rad
s ]
α Side slip angle [rad]
β Twist angle [rad]
δ Steerangle [rad]
² Rake angle [rad]
γ Camber angle [rad]
κ Longitudinal slip [−]
λ Magic Formula scaling factor [−]
ω Yaw velocity [ rad
s ]
ρ Vertical tyre deformation [m]
ρd Dimensionless vertical tyre deflection [−]
ρF z0 Nominal tyre deformation [m]
σκ Longitudinal relaxation length [m]
σα Lateral relaxation length [m]
ξ Rotational tyre deformation [rad]
SIGN CONVENTIONS AND SYMBOLS viii
Numbers
1 Steerbody
1s Twistbody
1u Front unsprung part
1w Frontwheel
2 Mainbody
2s Swingarm
2w Rearwheel
3 Rider upper body
Subscript
f Front tyre parameter
r Rear tyre parameter
u With respect to the u-direction
v With respect to the v-direction
x With respect to the x-direction
y With respect to the y-direction
z With respect to the z-direction
c With respect to the c-axis system
t With respect to the t-axis system
w With respect to the w-axis system
Chapter 1
Introduction
1.1 Background
With the computational capacity of computers available nowadays, simulating the dynamic be-
haviour of vehicles has become more and more important. During its development process, sim-
ulations with a model of the vehicle make it possible to accurately predict its dynamic behaviour
without building a prototype. With this so-called ’virtual prototyping’ both design times and
costs are reduced. Already in the early years of vehicle modelling it has been concluded that
the behaviour of a vehicle is strongly depending on the tyre behaviour. This holds especially for
motorcycles, as single track vehicles are inherent to instabilities which are partly governed by the
tyre behaviour. Therefore, the quality of a motorcycle model strongly depends on the accuracy of
the tyre model that is implemented.
In 1987 the first version of the Magic Formula has been presented [1], which is able to represent
the stationary slip characteristics of an automobile tyre. However, from figure 1.1 it can be learned
that motorcycles have different camber and slip angle ranges than automobiles. Therefore the
Magic Formula has been adapted in 1997 [28], to be able to describe the stationary motorcycle
tyre characteristics for large camber angles. The improved description of the slip characteristics
4 5
M o to r c y c le s
3 0
C a m b e r a n g le [d e g ]
1 5 A u to m o b ile s
5 1 0
S lip a n g le [d e g ]
Figure 1.1: Comparison between the slip and camber angle of automobiles and motorcycles
by the adapted Magic Formula is implemented in a tyre model (MF-MCTyre) which is used for
motorcycle simulations. Although that the effect of large camber angles is introduced in the Magic
Formula, it is insufficiently incorporated in other aspects of the MF-MCTyre model and in the
processing of measurement data.
During tyre measurements forces and moments are determined in the wheel center. However,
1
CHAPTER 1. INTRODUCTION 2
the Magic Formula is developed to describe the forces and moments in the tyre-road contact
point. Therefore, the measured forces and moments are transformed from the wheel center to the
contact point. Furthermore, during simulations the Magic Formula is evaluated in the contact
point and the forces and moments are applied to the vehicle in the wheel center, so again a
transformation is required. Although the tyre loaded radius Rl plays an important role during
this transformation, it is not unambiguously defined during the processing of the measurements and
simulations. Furthermore, also the rolling resistance moment plays a role in these transformations.
Currently little is known about the rolling resistance of a motorcycle tyre, and it is therefore
assumed to be constant at any condition. Finally, during simulations the contact point, vertical
load and the effective rolling radius are determined without taking into account motorcycle tyre
features as the typical tyre contour and a camber dependent vertical tyre stiffness. All these
facts are expected to introduce considerable errors in the representation of the steady-state tyre
behaviour by the MF-MCTyre model.
• The transformations of the forces and moments during both measurements and simulations
• The determination of the contact point and vertical load during simulations
• The determination of the effective rolling radius, Re , during simulations
• The description of the rolling resistance of a motorcycle tyre
Literature study
In order to get a better idea of the quality of the motorcycle model developed by Koenen and of
the MF-MCTyre model, the history of both multi-body motorcycle modelling and motorcycle tyre
modelling is investigated. First of all the history of motorcycle models is discussed chronologically.
It becomes clear that, especially the last ten years, a lot of motorcycle research is being conducted
at the department of Mechanical Engineering of the University of Padua. Furthermore also a
section at the Imperial College of Science in Londen under leadership of Prof. Robin S. Sharp
has made important contributions. The basis of the motorcycle model described in chapter 3 is
formed by the PhD thesis of Cornelis Koenen [10]. Both the degrees of freedom (DoF’s) and the
mass and inertial parameters used in other models are compared with this model of Koenen. The
need for a stabilizing controller is explained and the controller types used in history are listed.
Secondly the modelling of motorcycle tyres over the years is investigated. The behaviour of a
vehicle strongly depends on the performance of the tyres. Therefore a lot of research has been
conducted to describe the behaviour of an automobile tyre, but surprisingly the motorcycle tyre
always has been underexposed, until the last decade.
3
CHAPTER 2. LITERATURE STUDY 4
g S te e r
B o d y le a n
T w is t
P itc h
P itc h
z
y
O x
running straight ahead and when cornering. Especially the analysis of the dynamic behaviour of
the cornering motorcycle has given new insights. Moreover, it has been concluded that during
cornering the in-plane and out-of-plane modes of motion start to influence each other. Therefore,
the in- and out of plane motion cannot be decoupled and the eigenmodes during cornering contain
both. Next to this, Koenen he has also investigated the influence of several design parameters
on this dynamic behaviour. In 1985 Sharp has written a paper [18] with a review of the state
P itc h P itc h
O th e rs K o e n e n
of knowledge and understanding of the steering behaviour of single-track vehicles. This with the
main accent on vehicle design, vehicle design analysis and behaviour prediction. In this paper, the
model of Koenen is referred to as the most comprehensive motorcycle model, which represents a
step change in the technology. As the model of Koenen is so comprehensive, little to no changes
in the modelling of motorcycles can be seen until nowadays.
The major breakthrough after this period has been the development of computers with more
and more memory and computational capacity. This leaded to the development of software pack-
ages with which it is possible to analyse complicated multi-body models with relatively little effort.
In recent history, a lot of research is conducted at the department of Mechanical Engineering of the
University of Padua. Next to modelling of motorcycles also a lot of measurements are performed
to confirm the results gained by modelling. In 2000 a paper [3] has been presented in which a
multibody simulation model of a motorcycle is present. The model contains the same DoF’s as
Koenen, only again a swing arm is modelled for the rear suspension and no frame compliance
CHAPTER 2. LITERATURE STUDY 5
was taken into account. With this model, the motorcycle handling is evaluated and the results
obtained by simulation show very good similarity with measurements. Two years later, an im-
proved model has been presented by Cossalter [6]. In this paper also tyre modelling has been
improved, as will be discussed later. The results of slalom simulations have again been compared
with measurement results, and show very good correspondence. Recently Cossalter [7] has shown
that this model (with again some changes) can also be used for modal analysis of a motorcycle.
The results for the frequency of the eigenmodes are similar to those obtained from measurements,
however the damping is of worse quality, for which no exact reason has been found.
In 2001 again the confidence of Sharp in Koenen’s work has been shown as the motorcycle
model has been used to develop a motorcycle model for computer simulations. In this research [21]
the Koenen model is build with the automated model-building platform AutoSim, with the only
difference that a swing arm is used. The striking part of this research is that the results obtained
by the AutoSim model only partly match the results of Koenen’s research. The explanation for
this is two-folded. By checking the AutoSim code of Sharp, it appears that the inertial parameters
of Koenen are generally misinterpret. Secondly, Koenen has analyzed the 28th order motorcycle
model completely by hand, which is almost impossible to do without minor mistakes. In the same
year Sharp has presented a complete sensitivity and stability analysis of the same model [19]. In
this paper, also the power of present multibody software has been discussed.
Concluding, it can be said that the model of Koenen is still one of the most comprehensive
motorcycle models. Frame compliance is taken into account by a twist degree of freedom, the rider
behaviour is partly modelled as well as the aerodynamic effects. The only doubtful point is the
way in which the rear suspension is modelled. The results of Koenen show good comparison to the
results of others, especially for the dynamic behaviour when running straight ahead. Some doubt
remains about the dynamic behaviour while cornering, but also some errors might be present in
his work because the complexity of the analysis.
Masses [kg] Koenen [10] [8] [9] [5] [22] [20] [20] [20] [20]
mmain 209.6 170.3 154.5
mrider 44.5 50.0 44.0
mrearwheel 25.6 25.0 16.5
mmaintotal 279.7 245.3 215.0 153.4 355.0 245.3 218.0 221.5 273.9
in the model of Koenen are in good agreement with the masses of other models. For a modern
motorcycle, the entire motorcycle mass would be somewhat large, but the model is already 20 years
old. Especially the distribution of masses over the separate bodies seems good in comparison with
other models.
Next to the masses, also the moments of inertia of the bodies have an important role in the
dynamic behaviour of the motorcycles. The inertias of the models mentioned above are found
in table 2.2. All inertias are given in the local center of mass of each body, with respect to the
local axis system. Furthermore, the axis system is defined as in figure 2.1, with the x-axis in
longitudinal direction, the y-axis in lateral direction and the z-axis in vertical direction. In the
nominal situation all bodies are assumed to be symmetrical with respect to the vehicle center
plane. Because the local y-axis of each body is directed perpendicular to this plane, the product
of inertia Jxy , Jyx , Jyz and Jzy are all zero. The models of [10], [8] and [9] all assume the rider
Inertias [kgm2 ] Koenen [10] [8] [9] [5] [22] [20] [20] [20] [20]
JmainX 37.75 38.40 23.00 22.46 30.61 34.20 31.20 32.35 34.34
JmainY 42.65 32.90 18.90 36.24 0.00 0.00 0.00 0.00 0.00
JmainZ 22.69 8.26 26.00 17.36 19.78 22.53 21.00 19.84 22.81
JmainXZ -7.89 -3.00 -3.70 -4.40 0.00 0.00 0.00 0.00 0.00
JsteerX 4.80 3.97 4.00 1.80 1.58 3.97 3.70 3.45 3.58
JsteerY 4.10 0.00 2.31 1.77 0.00 0.00 0.00 0.00 0.00
JsteerZ 0.49 0.00 0.65 0.35 1.00 0.36 0.36 0.30 0.41
JsteerXZ -0.29 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Jf rontwheelY 0.58 0.58 0.46 0.49 1.04 0.58 0.72 0.57 0.68
JrearwheelY 0.74 0.41 1.67 0.71 1.04 1.06 1.06 1.13 1.25
to be non-rigidly attached to the mainframe. Therefore, the moments of inertia of the rider are
separately defined. As the other models assume a rigidly attached rider, the moments of inertia
of the main-bodies of those three models need to be compensated in order to make a correct
comparison. For this comparison the rider upper bodies are assumed to be rigidly attached for
those models. From [25] it is learned that the formula of Huygens-Steiner can be used to write
the inertia tensor of a certain mass m with respect to an arbitrary point O if the vector between
both points rcm
~ is known.
oJ =cm J + m(rcm
~ .rcm
~ I − rcm
~ rcm
~ ) (2.1)
This equation tells us that the inertia matrix with respect to an arbitrary reference point o J equals
the inertia matrix with respect to the center of mass cm J plus the inertia matrix with respect to
the arbitrary point of the the mass m concentrated at the center of mass. With this known it
is possible to translate the inertia tensor of the rider upper body to the center of mass of the
main-body. At this point it is added to the inertia tensor of the main-body. The same has been
done for the moments of inertia of the steer body of the models of [10] and [9]. Those models
define separate bodies that rotate around the steering axis, which have been defined as if they
were one body to make a fair comparison. With some exceptions, also the moments of inertia are
in good correspondence with other models. In general the moments of inertia used are somewhat
higher than those used by others, which is in agreement with the higher masses in the motorcycle
CHAPTER 2. LITERATURE STUDY 7
of Koenen. Some doubt is present towards the parameters of the rider upper body. The moments
of inertia of [9] are in the same order as those of Koenen, those of [8] are much higher. The
problem with the rider upper body is of course how to define the division between the lower and
upper body of a rider and how to determine the moments of inertia of the upper body.
Tδ = K1 γ + K2 γ̇ + K3 u + K4 u̇ (2.2)
Both Sharp [21] and Kamata [9] only use the camber angle of the mainframe to stabilize the
motorcycle. Where Sharp uses a hand-tuned PID-controller, Kamata uses a more sophisticated
method. System identification is applied in order to design a stabilizing H∞ -controller. Finally,
also Ruijs [16] conducted research to the stability of motorcycles. As he is the successor of Koenen,
his research is based on the motorcycle model of Koenen. To stabilize this model, not only the
feedback of again camber and its time-derivative are needed but also the time derivative of the
steer angle. With Nyquist plots it is explained that feedback of the camber angle is needed to
stabilize the capsize mode, its time-derivative is needed to stabilize the weave mode and the time
derivative of the steering angle is needed to stabilize the wobble mode.
which seems of large influence on the dynamic behaviour of a motorcycle. Only two years later,
also non-linear tyre forces have been introduced in motorcycle models [15].
Next to the comprehensive motorcycle model Koenen has also developed a sophisticated tyre
model. The tyres are modelled as radially flexible, and their cross-sectional shape is accounted for
to some extent. Inputs for the tyre model are the slip-angle, camber angle, turn-slip and vertical
load. As the vertical load is depending on the lateral force and the lateral force on its turn is
depending on the vertical load, the equations are solved with a Newton-Raphson iteration. Also
the non-stationary tyre behaviour is taken into account with a first order relaxation model.
In 1987 the first version of the Magic Formula tyre model has been presented [1]. This model
with an empirical background contains a set of mathematical formulae, which are partly based
on a physical background. Those formulae are able to accurately approximate the typical force
characteristics that are generated by a pneumatic tyre. At first, the Magic Formula only considers
passenger car and truck tyres, i.e. tyres with relatively large slip angles and camber angles which
only exceptionally exceed 10 degrees. First developments of a Magic Formula tyre model applicable
for motorcycle tyres by De Vries are seen in 1997 [28]. The results show that the adapted model
is able to accurately describe the stationary lateral force and aligning torque. Some doubts are
present about the correctness of the first-order relaxation model at high (≥ 100 [ km h ]) velocities.
At high velocities the model is not capable of taking into account the gyroscopic effects on the tyre
belt. A year later De Vries has investigated the influence of tyre modelling on the stability analysis
of a motorcycle [29]. The root loci of a motorcycle model which is equipped with different tyre
models are analyzed. First of all, the motorcycle is equipped with a front and a rear tyre which
have a constant relaxation length. Secondly, the rear tyre is replaced with a rigid-ring tyre model,
while the front tyre is left unchanged. The first impression is that the implementation of the rigid
ring model does not give large differences in the weave mode of a motorcycle in comparison to
the constant relaxation length model. At high speeds stabilizing contributions of the gyroscopic
couple can be found in this mode if the rigid ring tyre model is used. At moderate speeds (80
[km/h]) this tyre model gives little less damping than the constant relaxation length model. It is
mentioned that the effect of the rigid ring tyre model of a front tyre still has to be studied, and
that its influence on the wobble mode is expected to be larger. The general acceptance of the
Magic Formula tyre model is shown by the use of the model by several researchers [23] [3].
S p in d le a x is
T y re c o n to u r
C A R o a d s u rfa c e
y M x
In the meanwhile, a tyre model is developed at the University of Padua to overcome several
problems of the first-order relaxation model. The first signs of this model are seen in the work of
Cossalter in 1999 [5]. The Magic Formula is adapted in such a way that it is suitable to act in
the ’actual’ contact point (A). The original Magic Formula is used in the ’fictitious’ contact point
(C) defined by the wheel-plane and the road surface, as can be seen in figure 2.3. One of the
advantages of moving the contact point is that the overturning moment Mx in this point can be
neglected. This model is further extended by Cossalter [4] with the elastic modelling of the tyre
carcass. In order to describe the contact point accurately two linear models are developed. The
CHAPTER 2. LITERATURE STUDY 9
simpler model has two DoF’s, the lateral and radial displacements of the tyre with respect to the
hub. The tyre stiffness is modelled by linear springs, which are aligned with the radial and lateral
direction. The second model has one extra DoF with respect to the first model; the rotation of the
tyre. By measurements all parameters needed for the description of such models were gained. The
graphical representation of those models can be seen in figure 2.4. In 2002 Cossalter has extended
the two DoF model with rotational deformation in the wheel spin direction (See figure 2.5) and
implemented it in a multi-body motorcycle model. It is proven that such a model represents the
dynamic tyre behaviour in a way equivalent to the relaxation tire models. However, this approach
has several advantages. First of all, it explains the physical behaviour of the tyre in a more
realistic manner. Furthermore, only static and steady state experimental tests are required to
characterize the tyre behaviour in both static and dynamic conditions. The latest developments
tir e
r im
to this tyre model are recently presented by Lot [12]. Again the three elastic deformations of the
tyre are taken into account. However one should interpret the determined tyre stiffnesses critically,
as the measurement range of the applied force is limited (0-300 [N]). In this model a very good
approximation of the tyre geometry is introduced. The tyre cross section is photographed and
the contour of the tyre is fitted. Finally, the presented model is compared with the relaxation
model finding that both models are equivalent in absence of camber angle. When a camber angle
is introduced the new model is able to properly fit experimental tests, while the relaxation model
is not. The phase lag of the relaxation model is too large in comparison with the new model.
Chapter 3
In the year 1983 Cornelis Koenen has published his PhD work in the report ’The dynamic be-
haviour of a motorcycle’ [10]. In this report a physical model of a motorcycle is described with
a multibody approach. In order to get a better understanding of the dynamics of both a motor-
cycle and its tyres, Koenen’s work is made operational by developing a model in the multibody
toolbox of Matlab/Simulink, SimMechanics. The model of Koenen is explained in this chapter,
the parameters used in this model can be found in Appenix B.
g 3 S te e r
B o d y le a n
T w is t
1 s
P itc h
2
2 s 1 u
1 w
z
2 w
y
O x
10
CHAPTER 3. A MOTORCYCLE SIMULATION MODEL 11
All parts that are shown in figure 3.1 are assumed to be infinitely stiff. The most relevant
elasticity property of the frame is accounted for in the ’twist’ degree of freedom. The main frame
(2) of the motorcycle forms the basis part of the model. In the SimMechanics model, the connection
to the ground plane is made with this body by means of a 6 DoF joint i.e. the motorcycle can
freely with respect to the inertial frame. In some studies concerning motorcycle dynamics the
rider body is assumed to be rigid and rigidly connected to the main frame, which gives a poor
representation of the reality. To avoid too large differences between the model and reality, in this
case the rider body is split up in two parts. The lower segment of the rider body is assumed
to be rigidly attached to the main frame (2), the upper part (3) to rotate about an axis which
is horizontal in the initial condition, see figure 3.1. This rotation is both sprung and damped.
Furthermore, the rear wheel (2) is connected to the main mass with a sprung and damped swing
arm. This massless swing arm makes it possible for the rear wheel to rotate around a point on
the main body and in the plane of symmetry, the ’pitch’ movement. This is the only point at
which the SimMechanics model differs from the Koenen model because Koenen uses a different
joint for the pitch motion, as explained in section 2.1.1. As this joint is difficult to model in the
SimMechanics environment the choice is made to use a swing arm, which is normally used in
motorcycle models. The rear wheel (2w) is of course also given a DoF in such a way that it is
able to rotate around its own axle. At the front end of the main mass the steer pivot is located.
The steer body (1), twist body (1s), front unsprung mass (1u) and front wheel (1w) together
rotate as a whole relative to the main mass, about an inclined steering axis. As said, the main
elastic property of the frame has been accounted for in the twist degree of freedom. The twist
axis, which is perpendicular to the steering axis, allows the twist body (1s), front unsprung mass
(1a) and front wheel (1w) to rotate out of the plane of symmetry of the motorcycle. Also this
rotation is sprung and damped. The front suspension is modelled as a translatory movement of
the front unsprung mass (1u) and front wheel (1w) perpendicular to the steering axis if no twist
angle is present. Again this movement is both sprung and damped. Finally, the front wheel (1w)
is given one DoF, to be able to rotate around its spindle. All parameters considering this model
are documented in Appendix B, together with a geometrical figure of the model.
F d
P d z
P lx
F l
Additional to the parts of the model that are depicted in figure 3.1, the environment of the
motorcycle needs to be modelled. This comprises the road surface and the air through which
the vehicle moves. The road surface is assumed to be a flat and even plane perpendicular to the
direction of the local gravitational field. The air surrounding the vehicle is assumed to be initiallly
still relative to this ground plane. The motion of the vehicle will give rise to both stationary
and non-stationary forces acting on it. From these forces only two components are regarded, the
CHAPTER 3. A MOTORCYCLE SIMULATION MODEL 12
stationary drag and lift forces. The direction and lines of application of these forces can be seen
in figure 3.2. The motorcycle is modelled in such a way that the aerodynamic forces act at a
specified point of the main mass (2).
d
g
S te e r to rq u e
M o to r c y c le m o d e l
g
K d
0
K g
0
K g
re f
In order to steer the motorcycle into a corner, the reference of the camber angle γ is changed
from 0 [rad] to a certain value. The references of the camber velocity and the steer angle velocity
remain 0 [rad/s]. The feedback gains of the controller are depending on the forward speed of the
motorcycle. For several forward velocities, these gains are determined empirically.
Next to the stabilizing controller, also a controller needs to be developed to keep the forward
velocity constant during a simulation. The controller therefore needs to compensate for the rolling
resistance of the tyres and the aerodynamic drag. To overcome this problem a PD-controller is
implemented, with the feedback of the actual forward velocity of the motorcycle. The output of
this controller is a driving torque that is applied at the rear wheel.
Figure 3.4: Implementation of the MF-MCTyre model in the motorcycle simulation model
to the rest of the vehicle by means of the ’Vehicle model connection’. The first input of the
tyre model (road) defines the road profile during the simulation. The height of the road and
the slope of the road in both the x and y direction are specified. As only flat road surfaces are
considered, all three input variables will remain 0. The second input dis is defined as the x, y
and z position of the wheel in the global axis system. This position is measured by means of a
SimMechanics element; the wheel sensor. This sensor also defines the input tramat, which is a
9 component transformation matrix defined in the global plane. Both vel and omega are also
defined by a sensor connected to the wheel center. This sensor defines the global velocity (vel)
and the global angular velocity (omega) of the wheel center with respect to the wheel’s local
axis system. The signals angtwc and omegar give the rotation angle and the relative rotational
velocity between the tyre belt and the wheel body. As this motion is not taken into consideration
within the simulations within this research, both signals remain 0. Finally, seven coefficients to
scale several Magic Formula coefficients are introduced. For example the longitudinal and lateral
peak friction coefficients and the pneumatic trail coefficients can be adapted. The output of the
interface consists of the forces (force) and torques (torques) defined in the local axis system of
the wheel. By means of the actuators those forces and moments are fed back to the center of the
wheel. The third output varinf contains 40 signals representing the tyre behaviour. In this signal
for example the forces and moments in the contact center, the longitudinal slip κ, the side slip
angle α and the camber angle γ are given.
−20
−40
−60
Y−distance [m]
−80
−100
−120
−140
−160
−180
50 100 150 200 250
X−distance [m]
torque. In figure 3.6 the camber angle and the steering torque during the simulation are depicted.
If one examines the steering torque while the motorcycle is driving into the corner, one sees the
50
Camber angle, γ [deg]
40
30
20
10
−10
0 10 20 30 40 50 60 70 80 90 100
10
0
Steer torque [Nm]
−10
−20
−30
−40
−50
0 10 20 30 40 50 60 70 80 90 100
Time [s]
Figure 3.6: Camber angle and steering torque response during a simulation
typical paradox in the motorcycle steering behaviour. When the motorcycle needs to drive into
a right hand corner (γ ≥ 0), first a steering impulse to the left is needed. The origin of this
behaviour, called gyroscopic precession, lies in the inertia of the rotating front wheel. A spinning
CHAPTER 3. A MOTORCYCLE SIMULATION MODEL 15
wheel has a very stable axis of rotation, i.e. a strong tendency to maintain its plane of rotation.
If a twisting moment is applied which tries to change this plane of rotation (as happens when the
wheel is steered) this leads to a tilting moment around an axis at 90 degrees to that of the twisting
moment. If, for example, a spinning wheel is steered to the left this results in a moment which
tilts the wheel to the right. Therefore, a large steering torque to the left is needed for an instant
to steer the motorcycle into a right hand corner.
Once the motorcycle is cornering in a steady state, several equilibriums of forces and moments
are present which can be analysed. The signals used for the analysis of these equilibriums are
averaged over the last 10 seconds of the simulation. The tyre in- and output quantities can be
found in table 3.1 and the variables measured from the motorcycle are presented in table 3.2.
Note that the tyre parameters are given an extra index f of r to define the difference between
the front and rear tyre respectively. Furthermore, note that the front and rear suspension have a
pretension, so the forces and moments cannot be directly derived from the deformation and the
stiffness. The pretension is of such magnitude that when the motorcycle is in an upright position
with no forward velocity, the geometrical parameters of the motorcycle that are given in Appendix
B are present. Moreover, the suspension deflection and the pitch angle are given with respect to
this initial state.
Table 3.1: The tyre in- and output for the front and rear tyre during steady state cornering
In figure 3.7 the deformation of the motorcycle during steady state cornering is graphically
represented. As can be seen from the rear view, the positive rider lean angle leads to the rider
upper body leaning into the right hand corner. Furthermore, also the twist angle is positive due
to which the front fork and the front wheel are rotated with respect to the main body. Therefore,
the camber angle of the front wheel is larger than the camber angle of the rear wheel. From the
CHAPTER 3. A MOTORCYCLE SIMULATION MODEL 16
gr
B o d y le a n a x is
3 S te e r
B o d y le a n
T w is t a x is 1
T w is t
g
1 s
b 2
1 u
fro n t ty re 2 s
P itc h
2 w 1 w
re a r ty re
Figure 3.7: Rear view(left) and side view(right) of motorcycle deformations during steady state
cornering
side view the deformation of the rear suspension (pitch) and the front suspension can be seen. The
front suspension is somewhat compressed and therefore the front suspension deflection is negative.
Finally, the pitch angle is positive, so the wheel is rotated towards the main body and the rear
suspension is compressed.
If this equilibrium is computed a residual force of -1.3 [N] is present most probably due to averaging
and rounding. With respect to the 332 [N] of the rear tyre, this residual force is only 0.4% which
is negligible.
The overall vertical force of both tyres balances the overall gravitational force and the compo-
nent of the aerodynamic lift in the z-direction.
When the equilibrium is determined with the averaged forces, a residual force of 4.6 [N] is present.
Again this is a negligible 0.15% with respect to the overall vertical tyre force of 3068.5 [N].
The lateral tyre forces are negative (pointing inwards the corner) and for an equilibrium in the
lateral direction they need to compensate for the centrifugal forces (pointing outwards the corner)
that are acting on each body. Moreover, these centrifugal forces Fa can be calculated for each
CHAPTER 3. A MOTORCYCLE SIMULATION MODEL 17
individual body.
Fa = mω 2 R (3.3)
with m the mass of the body, ω the rotational velocity and R the cornering radius of the center
of gravity of the body. Therefore the equilibrium is formed by:
n
X
Fyf + Fyr + Fai = 0 (3.4)
i=1
n
X
Fyf + Fyr + ω 2 mi Ri = 0 (3.5)
i=1
If this equilibrium is determined, a residual force of -26.85 [N] is present, which means that the
total of the tyre forces is 0.97% larger than the total centrifugal force.
S m
S F a
H S F g
S M x w
D R
Figure 3.8: Equilibrium of moments of the complete motorcycle around its local x-axis
The camber angle that is present during steady state cornering can be obtained from the
equilibrium of the moment around the local x-axis. This equilibrium is schematically depicted in
the left part of figure
P 3.8. The center of gravity of the
P motorcycle is depicted in which the sum
of the centrifugal ( Fa ) and gravitational forces ( Fg ) act. Together with the height of the
centers of gravity of the bodies, the centrifugal forces lead to a moment which tries to decrease
the camber angle γ. Due to the camber angle, the bodies are cornering on a different radius than
the contact points of the tyres. Therefore, the gravitational forces lead to a moment which tries to
increase the camber angle. Finally, there is an overturning moment Mxw in the contact point of
each tyre which has already been explained with the aid of figure 2.3. The equilibrium is formed
by:
n
X n
X
Mxwf + Mxwr + Fgi ∆Ri = Fai Hi (3.6)
i=1 i=1
CHAPTER 3. A MOTORCYCLE SIMULATION MODEL 18
n
X n
X
Mxwf + Mxwr + gmi (Rf − Ri ) − ω 2 mi Ri Hi = 0 (3.7)
i=1 i=1
For the determination of this equilibrium, the cornering radius of the front tyre Rf is used. With
this assumption, the residual torque is 2.6 [Nm]. Again, this is relatively small as the total moment
around the local x-axis is around 1135 [Nm]. It is shown that the camber angle is determined by
he overturning moments of the front and the rear wheel. Therefore, if these moments would be
altered also the camber angle will change.
g 1 x
m
M s
s te e r a x is 1
tw is t a x is
F a 1 m 1
F g 1
g
e g
m
1 u x
F a 1 u
1 u
m 1 u
M z c
F y c F g 1 u
R o
w h e e l c e n te r
Figure 3.9: A side view(left) and a rear view(right) of the equilibrium of moments around the
steer axle
The equilibrium of the steer torque is somewhat more complicated. As can be seen from
figure 3.9, several components are present which have an influence on the equilibrium. For the
determination of the equilibrium the twist angle and the steer angle are not taken into consideration
as they are negligibly small (See table 3.2). Furthermore, the tyre forces and moments are firstly
transformed to the C-axis system which has its origin in the wheel center. With the forces and
moments in this axis system, the equilibrium is somewhat easier to explain and understand. In
figure 3.10 the forces and moments in the tyre road contact point O and the wheel center C are
depicted. As the forces in the tyre road contact point are known from table 3.1, the forces in the
wheel center can be derived with this figure.
F y c = 2 3 4 .1 [N ] F z c = 1 7 1 2 .0 [N ] M y c = 0 [N m ] M z c = -2 1 .0 [N m ]
O O
F z w = 1 3 1 0 .3 [N ] M z w = -3 8 .4 [N m ]
g g
R l R l
C C
F y w = -1 1 2 6 .4 [N ] M y w = -5 .9 [N m ]
Figure 3.10: The forces (left) and moments (right) in the W and C-axis system of the front wheel
Next to the tyre forces and moments, also the gravitational (Fg1 and Fg1u ) and centrifugal
(Fa1 and Fa1u ) forces of the masses m1 and m1u influence the equilibrium. The force of each mass
that effectively influences the steer torque equilibrium are calculated by:
Ms − Fyc t − Fm1 g1x − Fm1u g1ux − Mzc cos ² + Mxc sin ² = 0 (3.13)
If the equilibrium is determined a residual torque of 0.3 [Nm] is present which is again negligible.
Most probably, this residual torque is caused by the fact that the twist angle and the steer angle
are neglected. Moreover, it is remarked that the largest contribution in this equilibrium are the
moment caused by the force Fyc in combination with the mechanical trail t and the moment caused
by Mzc in combination with the rake angle ². With -15.4 [Nm] and 18.0 [Nm] these torques are
relatively large and opposite to each other. Therefore if one of both would be altered, a significant
change in steer torque would be seen. Moreover, with this equilibrium it is also proven that the
mechanical trail t and the rake angle ² have a significant influence on the steering behaviour of a
motorcycle.
Chapter 4
The basis of the MF-MCTyre model is the Magic Formula. Research [28] proves that the mo-
torcycle version of this formula is able to accurately describe the characteristics of the forces and
moments generated by a motorcycle tyre. The model surrounding the Magic Formula has been
derived from an automobile tyre model, which is therefore suited especially for the slip and cam-
ber angle range of these tyres. As the camber angle range is significantly larger for motorcycle
tyres than for automobile tyres, the model is known to be less accurate when it is evaluated under
large camber angles. Furthermore, for the processing of the measurements several assumptions are
made which negatively influence the accuracy. In this chapter the weak points and error sources
of the MF-MCTyre model are revealed and possible solutions are mentioned. In this chapter the
results of an elaborate measurement program are presented. These measurement results are used
to elaborate the solutions presented in this chapter and to evaluate their correctness. In this
chapter only several aspects of the MF-MCTyre model are discussed. For more information about
the tyre model the reader is referred to Appendix A.
nr
X
Zc Y
Zw Zc
Vx
Zw
Vc
Xc
Xc
O Xw
Xw Yc
-γ
Yc
Yw
Yw
C
C
to the road and in the wheel plane (xc -zc -plane). The origin O of the C-axis system is the wheel
20
CHAPTER 4. PROBLEMS OF THE MF-MCTYRE MODEL 21
center. The origin of the W-axis system is the road contact point C defined by the intersection of
the wheel plane, the plane through the wheel spindle and the road tangent plane. The xw -yw -plane
is the tangent plane of the road in the contact point C, and it defines the camber angle γ together
with the normal nr to the road plane (xw -yw -plane). Furthermore, the forces and moments are
described in both axis systems. Therefore all forces and moments have a c or w index, which
points out with respect to which reference axis system they are defined.
In section 3.3 the implementation of the MF-MCTyre model by means of the Standard Tyre
Interface (STI) is described. Each time step, several kinematic parameters are retrieved from
the wheel center (the C-axis system) and used as an input for this interface. Furthermore the
interface returns the forces and moments in the C-axis system as a feedback to the wheel center.
The processing of the input parameters within the STI is schematically depicted in figure 4.2.
First of all the contact routine uses the position and orientation of the wheel and the road profile
K in e m a tic s F o rc e s & M o m e n ts
C - a x is C - a x is
V e h ic le m o d e l
W - a x is W - a x is
C o n ta c t r o u tin e M a g ic F o r m u la C o n v e r s io n
S ta n d a rd T y re In te rfa c e
T y r e p r o p e r ty file
M F -to o l
W - a x is
M -ty re C o n v e r s io n
C - a x is
M e a s u re m e n ts (D T T )
to determine the exact position of the contact point. This point is used as the origin of the W-
axis system. In this origin the input variables of the Magic Formula, the vertical load Fzw , the
longitudinal slip κ, the side slip angle α and the camber angle γ are determined by this routine.
With these input parameters known, the Magic Formula is evaluated and the forces and moments
in the contact point are determined. As these forces and moments are applied to the vehicle model
at the wheel center, they are converted from the W - to the C-axis system. As said, the loop with
the vehicle model and the STI is evaluated at each time step of the vehicle simulation.
Next to the simulation loop, also the measurement line is depicted in figure 4.2. In contra-
diction to the simulation loop this line is only evaluated once. First of all, the tyre behaviour is
experimentally determined by the Delft-Tyre Test Trailer (DTT). During tests the test trailer has
a certain forward velocity and the tyre is pushed against the road at various loads, orientations
and motion conditions. During these tests, all forces and moments are measured at the wheel
center (the C-axis system). As the Magic Formula is evaluated in the W -axis system, the M-Tyre
software is used to convert the measurements to the contact center. Finally, the MF-Tool software
is used to fit the Magic Formula parameters to the measured forces and moments. This leads to
a set of parameters that represents one single tyre. This set of parameters and some general tyre
parameters are captured in a tyre property file which is used by the STI during each time step of
a vehicle simulation, in order to describe the momentary tyre behaviour.
CHAPTER 4. PROBLEMS OF THE MF-MCTYRE MODEL 22
O
R
l R o
g
h
ro a d ta n g e n t
p la n e
r t
C
r
r o a d p r o file
Figure 4.3: Determination of the contact point C and the vertical compression ρ
known, also the vertical deformation ρ of the tyre can be determined. In order to do this the
tyre contour is approximated by a circle with radius rt . With the aid of figure 4.3, the normal
compression of the tyre can be determined.
ρ = max(0, ρ0 ) (4.2)
As the tyre is assumed to have a constant vertical stiffness Cz and damping Kz , the normal load
Fzw can then be calculated as:
Fzw = Cz ρ + Kz ρ̇ (4.3)
with ρ̇ the deflection velocity. If during a time step the vertical load is for example higher than the
calculated Fzw , the axle height will slightly decrease. In the next time step the vertical compression
and therefore also the vertical load will increase and the vehicle load is again supported. With
this ’loop’ the equilibrium between tyre load and calculated vertical force is constantly balanced
during a simulation.
parts of the vehicle will be incorrect. The axle height is especially critical for motorcycles as these
vehicles are very sensitive to mass distribution. Moreover, motorcycles are inherent to instabilities
which are partly governed by the positions of the centers of gravity of several parts. Despite the
importance of an accurate axle height, the contact routine contains assumptions which decrease
the accuracy of the axle height during simulations. First of all, the approximation of the road by
a tangent plane at the point on the road below the wheel center introduces an error. Especially
when the tyre model is evaluated under a large camber angle the distance between the point below
the wheel center and the contact point becomes rather large. Therefore, the road profile at the
contact point may be different from the estimated tangent plane. Furthermore, the tyre contour
is assumed circular which is not the case in practice. Each tyre has its individual shape which
should be taken into consideration if a correct axle height is requested. Finally, the vertical tyre
stiffness Cz is modelled as a linear spring that is independent of the camber angle and the vertical
compression ρ, which is not validated.
To overcome these problems a new contact routine should be developed. First off all, this
routine should contain a more accurate determination of the contact point. Therefore, the actual
contact point should be determined without the approximation of the tangent road plane. Fur-
thermore, also an accurate description of the tyre contour is needed. Finally the assumption of a
constant vertical tyre stiffness should be replaced with an accurate description of the vertical load
as a function of the vertical compression. To be able to make these improvements, first of all the
vertical tyre contour and the vertical tyre stiffness should be measured.
10
α [deg]
0
−10
0 10 20 30 40 50 60 70
2000
1000
Fyc [N]
−1000
0 10 20 30 40 50 60 70
500
Mxc [Nm]
−500
0 10 20 30 40 50 60 70
Time [s]
the axis systems are present and the the tyre model will produce incorrect moments on the wheel
axle during a simulation. The obvious solution to this problem is to use similar parameters which
determine the loaded radius during both the processing of the measurements and simulations. If
this is done, the loaded radius can be unambiguously defined and is always equal during equal
circumstances.
Vr = ΩRe (4.5)
The longitudinal slip of a tyre is defined as the relative difference between this linear rolling velocity
and the actual longitudinal velocity Vx .
Vr − Vx
κ= (4.6)
Vx
As becomes clear from (4.5) and (4.6), the effective rolling radius Re of a tyre is of considerable
importance for the correct determination of this slip. This radius determines the ratio between
the forward velocity Vx and the rotational speed Ω when no longitudinal slip is present. Therefore,
if the effective rolling radius is incorrectly implemented the rotational speed of the wheel will be
incorrect during a simulation.
For automobile tyres the effective rolling radius has a specific dependency on the vertical load
Fzw which is depicted in figure 4.6. At first, an increase in vertical load will decrease the effective
CHAPTER 4. PROBLEMS OF THE MF-MCTYRE MODEL 25
R o
W
V x
R l
R e
rolling radius in an almost similar way as the loaded radius Rl . The vertical load will deform the
tyre rubber, the circumference of the tyre will decrease and with that also Re decreases. As the
vertical load further increases the tyre will be compressed but the effect on the rolling radius is
getting smaller. This is explained by the fact that the steel carcass of a tyre is radially flexible,
but in the circumferential direction very stiff. Therefore at increasing vertical load the tyre still
deforms, but its circumference remains almost equal. As the circumference does not alter, also the
effective rolling radius will not decrease significantly. Although this effect is thoroughly checked
for automobile tyres, little information is present on this effect for motorcycle tyres. In the present
tyre model the behaviour of an automobile tyre is therefore implemented.
304
Effective rolling radius Re
302 Loaded radius R
l
300
298
Radius [mm]
296
294
292
290
288
286
0 500 1000 1500 2000 2500
Vertical force, Fzw [N]
Figure 4.6: The effective rolling radius and loaded radius as a function of the vertical load
For the estimation of the effective radius Re as a function of the vertical load Fzw a Magic
Formula approach is used.
In this formula, Ro is the unloaded free radius when no camber is present. The nominal tyre
deflection ρFz0 is defined by the vertical tyre stiffness Cz and the nominal wheel load Fz0 .
Fz0
ρFz0 = (4.8)
Cz
CHAPTER 4. PROBLEMS OF THE MF-MCTYRE MODEL 26
Finally the dimensionless radial tyre deflection ρd can be calculated with the momentary tyre
deflection ρ, which is calculated with (4.1)-(4.2) in the previous section.
ρ
ρd = (4.9)
ρFz0
The factor B in (4.7) determines the slope at Fzw = 0, the factor D defines the height of the asymp-
tote at high wheel loads and the factor F defines the ratio between the tyre radial deformation
and the effective tyre deformation.
It has been proven that this method describes the effective rolling radius of an automobile
tyre with sufficient accuracy. Therefore it has also been implemented in the present motorcycle
tyre model, without any adaptations. However, several assumptions that are made specifically for
automobile tyres negatively affect the accuracy. First of all, it has already been said that little
information is known on the effect of the vertical load on the effective rolling radius for motorcycle
tyres. Copying the parameters B, D and F from automobile tyre measurements is doubtful as the
construction of a motorcycle tyre is completely different. Therefore, motorcycle tyre measurements
are needed to determine these parameters accurately. Furthermore, no attention is paid to the tyre
contour and its effect on the effective rolling radius when the wheel is cambered. The unloaded free
radius Ro is used in (4.7) instead of the effective rolling radius of a cambered and undeformed tyre,
Reo . In figure 4.7 an idea is given of the impact of this assumption. Especially under large camber
R o
R e o = R o
g
R e o
angles the effect is considerable. The effective rolling radius is estimated too large and therefore
the longitudinal slip is determined too small during vehicle simulations under high camber angles,
which affect the rotational velocity of the wheel. Furthermore, the vertical stiffness used in (4.8)
is again assumed as being independent from the camber angle.
Summarizing, several types of measurements are needed to correctly implement the effective
rolling radius in the tyre model. The Magic Formula coefficients that describe the influence of the
vertical load need to be determined from measurements at different vertical loads. Furthermore
a correct description of both the tyre contour and the vertical tyre stiffness are needed to include
the camber dependency of the effective rolling radius. When these changes are implemented, an
extensive measurement program is needed to check the correctness of the new description of the
effective rolling radius.
10
−5
−10
0 10 20 30 40 50 60 70
100
MF−MCTyre
0 Measurements
Fxw [N]
−100
−200
−300
0 10 20 30 40 50 60 70
Time [s]
negligible. The difference between the output of the MF-MCTyre model and measurements results
is depicted in figure 4.8.
The cause of this problem can be found in the processing of the measurement data. During
testing the moment around the wheel axle (Myc ) cannot be measured and therefore an assumption
has to be made. To overcome this problem an estimation is made for the rolling resistance moment
Myw . In the past, tests have been performed with a non-cambered freely rolling automobile tyre.
As the wheel is freely rolling the rotational velocity Ω is almost constant, which leads to the
conclusion that there is no effective moment around the wheel axle (2nd law of Newton). With
this assumption, the equilibrium of a non-cambered wheel can be formulated.
A schematic representation of this equilibrium can be seen in the left part of figure 4.9. As the
F z c
F y c
M z c
M y c
M y c O
O F x c
g
F z w
R l
M z w R l
M y w
C F x w F y w M y w C
longitudinal force Fxw and the loaded radius Rl can be determined from test results an estimation
CHAPTER 4. PROBLEMS OF THE MF-MCTYRE MODEL 28
of the rolling resistance Myw has been made. This estimation has been defined as:
Myw = −Fzw fr Rl (4.11)
with fr the rolling resistance coefficient of the tyre. Despite the good results with this assumption
for automobile tyres, problems arise with motorcycle tyres. More specifically, the problems shown
in figure 4.8 arise due to the large camber angle range of a motorcycle tyre. The problem can
be explained with the equilibrium of moments around the spindle axle for a freely rolling and
cambered wheel. In the right part of figure 4.9 the forces and moments of a cambered wheel in
the (y-z)-plane are depicted, from which the equilibrium is derived.
Myc = Myw cos γ + Mzw sin γ − Fxw Rl (4.12)
From an α-sweep, the specific characteristic of the self aligning moment Mzw as a function of the
slip angle is determined and fitted with the Magic Formula. It is therefore known that this self
aligning moment fluctuates during an α-sweep. In contradiction, the measured longitudinal force
and the estimated rolling resistance do remain almost constant during the same measurement. If
the forces and moments in the contact point are determined with the current rolling resistance
assumption and the equilibrium of 4.12 is determined, the calculated Myc cannot remain constant.
In figure 4.10 the fluctuations of the calculated Myc are shown for a measurement that is processed
with the current rolling resistance assumption and a rolling resistance coefficient fr of 1.5 %.
However, in practice Myc ≈ 0 as the wheel has an almost constant rotational speed during an
α-sweep.
10
Slip angle, α [deg]
−5
−10
0 10 20 30 40 50 60 70
0
−20
−40
Myc [Nm]
−60
−80
−100
−120
0 10 20 30 40 50 60 70
Time [s]
Figure 4.10: Myc during an α-sweep with the present rolling resistance estimation
As all measurements are processed with this incorrect rolling resistance assumption, also the
fluctuations in the longitudinal force Fxw of figure 4.8 can be explained. The MF-MCTyre model
contains information about the forces and moments in the W -axis system which lead to a fluctu-
ation Myc during simulations. Due to this moment, the wheel is accelerated while the rest of the
vehicle has a constant forward velocity. Therefore, longitudinal slip is generated which affects the
longitudinal force Fxw . To solve these problems, measurements of a cambered and freely rolling
motorcycle wheel need to be analysed. With this analysis a new axis system can be introduced in
the contact point in which the rolling resistance can be defined correctly. Furthermore, the camber
angle is expected to have an effect on the rolling resistance factor fr which is nowadays considered
to be constant. Therefore, the measurements are also used to describe the camber dependance of
the rolling resistance coefficient fr .
Chapter 5
In the previous chapter some deficiencies of the MF-MCTyre model are revealed. It is shown that
several important motorcycle tyre parameters should be measured to be able to include them in the
tyre model. Therefore, an elaborate measurement program is conducted in order to get a better
understanding of the tyre contour, vertical tyre stiffness, effective rolling radius and the rolling
resistance of motorcycle tyres. Especially the effect of the camber angle on these parameters will
be investigated. The results of the measurements are presented in this chapter and in the following
chapter they are used to improve the tyre model.
In order to be as consistent as possible the research within this master thesis has been performed
with one single type of motorcycle tyres. Of this type a front and a rear tyre are taken and their
properties are determined. TNO Automotive has recently executed an extensive measurement
program with the Delft Tyre Test trailer in order to determine the Magic Formula parameters
of these tyres. The results of these measurements are also used within this research, but in a
later stadium. Additional to the measurements conducted by TNO a measurement program was
set-up for this master thesis. Vertical stiffness and effective rolling radius measurements have been
conducted on the Flatplank Tyre Tester at the Eindhoven University of Technology. The Flatplank
Tyre Tester has been chosen for these measurements as the road surface disturbances are negligible
in comparison to the disturbances measured with the Delft Tyre Test trailer. Furthermore the
forward velocity and rotational velocity are measured with a higher accuracy compared to the test
trailer, which is especially favourable for the effective rolling radius measurements.
29
CHAPTER 5. MOTORCYCLE TYRE MEASUREMENTS 30
60 100
Round contour Round contour
Measured contour 90 Measured contour
50
80
70
40
60
Height [mm]
Height [mm]
30 50
40
20
30
20
10
10
0 0
0 10 20 30 40 50 60 0 20 40 60 80 100
Width [mm] Width [mm]
Figure 5.1: The round and measured tyre contour of a front(left) and a rear(right) tyre
During a measurement significant fluctuations in the measured vertical force Fzw are present
although the vertical deformation ρ is almost constant. This can be declared by the un-roundness
of the rolling tyre which is confirmed by the fact that the vertical force reaches its minimum and
maximum at the same points of the tyre circumference. This is shown in figure 5.2 where the
measured vertical displacement and force are depicted of a front tyre which is cambered over 25
degrees. To exclude the influence of this fluctuation both Fzw and ρ the signals measured over
the last complete tyre rotation are used.
6.01
Deformation, ρ [mm]
5.99
5.98
5.97
5.96
0 100 200 300 400 500 600 700
1000
Vertical force, Fzw [N]
950
900
850
800
750
0 100 200 300 400 500 600 700
Rotation angle [deg]
Figure 5.2: Fluctuating Fzw over the tyre circumference of a front tyre
For the graphical representation of the vertical tyre stiffness, the vertical load Fzw and the
vertical deformation ρ are averaged over the last complete tyre rotation. From these averages
an average vertical tyre stiffness can be determined for each measurement, which is depicted in
figure 5.3. The results of the stiffness measurements of both tyres generally show a consistent
behaviour for the influences of the camber angle. As the camber angle rises, the overall vertical
stiffness is decreasing which is in consistence with the known behaviour of automobile tyres. For
automobile tyres it is known that without camber the lateral stiffness is significantly smaller
than the vertical stiffness. Therefore, the overall vertical tyre stiffness perpendicular to the road
decreases with an increasing camber angle. For both the front and rear motorcycle tyres this effect
is also present. Moreover, also the effect of the vertical deformation ρ on the vertical tyre stiffness
Cz is in consistence with the behaviour of automobile tyres. Especially for the front tyre, the
vertical stiffness increases with an increase of the vertical deformation. However, for the rear tyre
this effect is not consistent over the camber angle range and overall it can be concluded that is
has less influence than on the stiffness of the front tyre.
The vertical load Fzw during simulations is defined from the calculated vertical deformation ρ.
From (5.1) it is learned that the relationship between both is given by the vertical tyre stiffness.
Therefore, the vertical tyre stiffness measurements presented in this section are used in chapter 6
to describe the relationship between the vertical deformation and the vertical tyre force. As the
results show, this description should be a function of at least the camber angle γ and the vertical
deformation ρ.
CHAPTER 5. MOTORCYCLE TYRE MEASUREMENTS 32
170
ρ = 3 [mm]
ρ = 6 [mm]
ρ = 10 [mm]
160
Vertical stiffness, Cz [kN/m]
150
140
130
120
110
0 5 10 15 20 25 30 35 40 45 50
Camber angle, γ [deg]
225
ρ = 3 [mm]
220 ρ = 6 [mm]
ρ = 10 [mm]
215
210
Vertical stiffness, Cz [kN/m]
205
200
195
190
185
180
175
0 5 10 15 20 25 30 35 40 45
Camber angle, γ [deg]
Figure 5.3: Measured vertical stiffnesses for a front(top) and a rear(bottom) tyre
CHAPTER 5. MOTORCYCLE TYRE MEASUREMENTS 33
302
Measurements
MF−MCTyre
300
298
Effective rolling radius [mm]
296
294
292
290
288
0 500 1000 1500 2000 2500
Vertical force, Fzw [N]
Figure 5.4: Effect of the vertical load on the effective rolling radius of a front tyre
In the previous chapter it has been explained that the influence of the vertical load on the
effective rolling radius is nowadays copied from automobile tyres. To correctly introduce the effect
of the vertical load in the new motorcycle tyre model, this effect needs to be measured. Therefore
measurements have been performed on the Flatplank Tyre Tester with an un-cambered freely
rolling wheel. The procedure for these measurements is similar to the procedure used for the
vertical stiffness measurements. During these measurements both the forward velocity Vx and the
rotational velocity Ω are determined, and as no driving or braking torque is applied the effective
rolling radius Re can be determined as:
Vx
Re = (5.2)
Ω
Furthermore, the measurements are again performed over two complete tyre rotations and the
measured signals during the last complete tyre rotation are used. For the graphical representation,
the results are obtained by averaging the signals over this last complete tyre rotation. Figure 5.4
shows the measured relationship between the rolling radius and vertical load and the relationship
which is presently implemented in the tyre model. As one can see large differences are present
between the model and the measurement results. The results of these measurements will therefore
be used in the next chapter to correctly implement the effect of the vertical load on the effective
rolling radius.
camber angle and vertical load the measurements are again repeated three times, in order to reduce
the influence of random errors. Furthermore, the main steps of the measurement procedure are:
• Inflate the tyre to the prescribed pressure
• Apply a prescribed camber angle
• Place the tyre exactly against the road so that there remains zero vertical load
• Remove all offsets by zeroing the measurement channels
780
Measurement
Vertical load, Fzw [N]
Average
760
740
720
0 100 200 300 400 500 600 700
0.32
Effective roll.rad., Re [m]
0.31
0.3
0.29
0.28
0.27
0 100 200 300 400 500 600 700
Rotation angle [deg]
Figure 5.5: Effective rolling radius measurement of a front tyre under 30 degrees camber
Similar to the measurements described earlier, the measurements are performed over two com-
plete tyre rotations. In figure 5.5 the measured vertical load Fzw and the effective rolling radius
calculated with (5.2) are depicted for one measurement. From this figure it can be concluded that
especially the vertical load fluctuates with the tyre circumference. Therefore, the results are again
averaged over the last complete tyre rotation, from which the results are also shown.
Equal measurements have been performed for the front and rear tyre from 0 to 50 degrees
camber with intervals of 5 degrees. The results of these measurements and therefore the effect of
camber on the effective rolling radius are depicted in figure 5.6. Furthermore, also the effective
rolling radius that is calculated by the present tyre model is depicted and as one can see, large
differences are present between both. These differences can be clarified by the fact that the tyre
contour is not taken into account by the tyre model. The effect that cannot be explained are the
large differences between the effective rolling radius at low and high vertical load. At high camber
angles (45 and 50 degrees) the difference between both is suddenly much larger than it is a lower
camber angles. Most probably, the unreliable results at high camber angles are caused by the lack
of grip on the Flatplank surface. As the camber angle of the tyre increases, also the lateral force
of the tyre increases. At a certain point, the grip in the contact patch of the tyre is not sufficient
CHAPTER 5. MOTORCYCLE TYRE MEASUREMENTS 35
anymore and the tyre partly begins to slide over the surface. With the sliding, a lot of stick-slip
appears which can be noticed during the measurements by its specific sound. Most probably, this
stick slip effect negatively influences the measurements and therefore the results at large camber
angles are not completely trusted on their validity.
300
295
Effective rolling radius, R [mm]
290
e
285
280
275
270
Measured
MF−MCTyre
265
0 10 20 30 40 50
Camber angle, γ [deg]
330
320
Effective rolling radius, R [mm]
310
e
300
290
280
Measured
MF−MCTyre
270
0 10 20 30 40 50
Camber angle, γ [deg]
Figure 5.6: Measured and calculated effective rolling radius for a front(top) and a rear(bottom)
tyre
will be defined, with which the measurements described in this section are already processed. It
will be explained that with this new definition for a freely rolling wheel it follows that:
Fxw = Fzw fr (5.3)
With this known, the measured longitudinal force Fxw and vertical force Fzw can be used to
obtain the rolling resistance coefficient fr at different circumstances. For each measurement, the
average rolling resistance coefficient is determined to give an impression of its dependence on
the vertical force and camber angle. If the measurement results that are depicted in figure 5.7
are studied, the strong camber dependence of the rolling resistance becomes clear. For both the
front and rear tyre, the rolling resistance is approximately quadrupled over the camber range.
Furthermore, also the vertical load plays an important role. The influence of the vertical load also
increases with the camber angle, especially for the rear tyre. With these results it is shown that
the present assumption of a constant rolling resistance factor fr of 1.5% should be replace with
a more sophisticated description. Therefore, in section 6.3, these measurement results are used
to generate an accurate description of the rolling resistance which takes the effects of camber and
vertical load into account.
0.09
Fzw = 750 [N]
Fzw = 1500 [N]
0.08
Roll. resistance coefficient [−]
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0 10 20 30 40 50
Camber angle, γ [deg]
0.07
Fzw = 750 [N]
Fzw = 1500 [N]
0.06
Roll. resistance coefficient [−]
0.05
0.04
0.03
0.02
0.01
0 10 20 30 40 50
Camber angle, γ [deg]
Figure 5.7: Rolling resistance coefficients for both a front and a rear tyre
Chapter 6
In chapter 4 the deficiencies of the MF-MCTyre model have been discussed, and possible solutions
are proposed. These solutions are further elaborated in this chapter and their correctness is
checked. To do so, the results of the measurements presented in the previous chapter are used.
Moreover, they are used to improve the determination of the contact point and the vertical load,
the description of the effective rolling radius and the description of the rolling resistance moment.
Throughout the present and the following chapter it is attempted to describe the results of
several measurements. The accuracy of these descriptions is indicated with the relative error E in
percentages. For each measurement and its description, this relative error is determined as:
sP
n
(Y − Ymi )2
E = 100 Pn f i
i=1
2
(6.1)
i=1 (Ymi )
with n the number of measurement points, Yf the description and Ym the measurement.
u2 v2
+ =1 (6.2)
a2 b2
In this case the ellipse is described with respect to its local axis system, with axis u representing
the direction of the tyre width and the v axis describing the tyre height direction. Within this axis
system the tyre contour is described independent from its orientation, due to which it is easier to
implement the tyre contour in the contact routine. The ellipse equation is used to describe the tyre
contour and the error definition of (6.1) is used to calculate the error made by this description.
This error can be minimized for the ellipse parameters a and b, which is done with the Matlab
built-in function f minunc.m. In figure 6.1 a fit of the rear tyre contour is depicted from which it
can be seen that an elliptic fit represents the tyre contour very well. The minimal fitting errors
are 2.14% and 3.46% for respectively the front and rear tyre contour. Moreover, these errors are
21.24% and 22.58% if the tyre contours are describe with a circle, which used to be done. Large
improvements are therefore made and the major advantage is that this fit is described only by the
two ellipse parameters a and b. As the tyre contour is accurately described it is possible to correctly
37
CHAPTER 6. IMPROVEMENTS TO THE MF-MCTYRE MODEL 38
80
Measured curve
500 Fitted ellipse
70
60
400
50
Height [mm]
Height [mm]
300 40
30
200
20
100 10
0
0 0 20 40 60 80 100
−100 0 100
Width [mm] Width [mm]
determine the contact point between an undeformed tyre and the road. For simplicity, the contact
routine developed in this research is limited to simulations with a flat road surface. With a flat
road surface, the problem with the inaccurate assumption of the tangent wheelplane beneath the
wheel center is avoided. In a further stage, the new contact routine should be improved in such a
way that it can handle non-flat road surfaces.
During a simulation the [3 × 3] rotation matrix which defines the orientation of the wheel with
respect to the global axis system is known. At each simulation time-step, the camber angle γ is
defined from this matrix. Next to the camber angle, the tyre contour is described in the already
introduced (u, v)-axis system which rotates with the camber angle. With these parameters known
it is possible to calculate the actual contact point A, which is the point at which the tyre touches
the road if no deformation is present, see figure 6.2. This point is computed with the derivative
v
u v
u a
z z
g
y
R o
y
b R o
ro a d
r
A A
u2 v2
+ =1 (6.3)
a2 b2
r
u2
v 2 = ±b 1− (6.4)
a2
du b2 u u b
=− 2 q = − ( )2 (6.5)
dv a b 1− u2 v a
a2
The point at which this derivative equals tan γ is the contact point A(uA ,vA ) of the undeformed
tyre.
uA b 2
− ( ) = tan γ (6.6)
vA a
a( ab ) tan γ
uA = − q (6.7)
1 + ( ab )2 tan2 γ
b
vA = − q (6.8)
1+ ( ab )2 tan2 γ
As the height of the wheel center h is also retrieved from the vehicle model, also the vertical
deformation of the tyre at point A can be determined. With the aid of figure 6.2 it follows that:
Now that the vertical deformation ρ of the tyre is known, also the vertical force Fzw can be
determined. The relationship between these parameters is known as the vertical tyre stiffness,
which has been measured. Therefore, an accurate description of these measurements would be
sufficient to describe the vertical tyre force during a simulation. At TNO automotive, also the
MF-Swift tyre model is developed with which the tyre dynamic behaviour can be simulated.
Within this model, the relationship between the vertical tyre force and the vertical compression
is described as:
ΩRo qF cx Fx 2 qF cy Fy 2 ρ ρ
Fz = {1 + qv2 | |−( ) −( ) }.{qF z1 + qF z2 ( )2 }.Fzo (6.10)
Vo Fzo Fzo Ro Ro
From this formula it can be observed that:
• a parabolic approximation is used to describe the load-deflection curve
• the vertical stiffness increases almost linearly with the angular velocity
• a reduction of the vertical stiffness appears when longitudinal and/or lateral forces are ap-
plied in the contact patch
In order to make the different tyre models as consistent as possible, this formulation is also used
within this research to describe the vertical stiffness measurements presented in the previous
chapter. When the parameters are fitted to the measurements, it appears that the effect of the
longitudinal and lateral force on the vertical stiffness are negligible. Furthermore, the flatplank
measurements are performed at a very low and constant velocity, so also the effect of the rotational
velocity is negligible. As appears from the measurements the camber angle does have a significant
influence on the vertical stiffness which should be included in the description. Therefore (6.10) is
rewritten to:
ρ ρ
Fz = (1 − qc |γ|).(qf z1 + qf z2 ( )2 ).Fzo (6.11)
Ro Ro
CHAPTER 6. IMPROVEMENTS TO THE MF-MCTYRE MODEL 40
with which the flatplank measurements are described. Following the error description of (6.1)
the fitting errors for the front and rear tyre are 7.3% and 5.0%. If the camber dependence is
neglected during the fit procedure, the fit errors increase to 7.9% and 5.2% for the front and rear
tyre respectively. Although this effect is smaller then expected, it is still significant and therefore
the choice is made to remain the inclusion of the camber dependence by means of factor qc . In
the tyre property file the vertical stiffness Cz at the nominal load is specified instead of qF z1 . It
can be calculated as:
Fzo q 2
Cz = qf z1 + 4qf z2 (6.12)
Ro
With this description of the vertical force the contact routine fulfills the demands. The contact
point is determined as a function of the tyre contour, and the vertical load is accurately determined.
Therefore, the effects of this method on the wheel axle height h and the tyre loaded radius Rl can
be discussed.
With the actual measurements as an input, (6.13) is used to calculated the effective rolling radius
at each measurement point. The relative error defined by (6.1) is determined and minimized for the
parameters B, D and F with the Matlab built-in function f minunc.m. In this way, the effective
rolling radius can be represented accurately with as can be seen for a front tyre in figure 6.3.
0.304
Fit
Measurements
0.302
0.3
Effective rolling radius, Re [mm]
0.298
0.296
0.294
0.292
0.29
0.288
0 500 1000 1500 2000 2500 3000
Vertical force, Fzw [N]
Figure 6.3: Measurements and fit of the effect of the vertical load on the effective rolling radius
Next to the correct Magic Formula parameters, also the effective rolling radius of an undeformed
tyre Reo is needed to determine the actual effective rolling radius. In section 6.1 the coordinates
of the actual contact point A have been determined with the tyre contour described by an ellipse.
With these coordinates, the free rolling radius Ro and the aid of figure 6.4, Reo can be defined.
Reo = Ro − b − vA (6.14)
The effect of the tyre contour can now be taken into account by using Reo instead of the free
rolling radius Ro in 6.13. Furthermore, the vertical tyre deformation only partly affects the
effective rolling radius because of the camber angle of the wheel. This effect is also taken into
account by multiplying the deformation term with cos(γ).
u v
z
b g
R o
y
R
R e o
A
Figure 6.4: Effect of camber on the effective rolling radius of an undeformed tyre
For the determination of the nominal tyre deflection ρFz0 the constant stiffness described in (6.12)
is used.
Fz0
ρFz0 = (6.16)
Cz
Finally the dimensionless radial tyre deflection ρd is calculated with the momentary tyre deflection
ρ. With the improvements in the determination of the contact point and the vertical load Fzw that
have been shown in the previous section, also the momentary tyre deflection is determined more
accurately. Of course, this also positively influences the accuracy of the effective rolling radius.
ρ
ρd = (6.17)
ρFz0
As the determination of the effective rolling radius is now expected to be correct, its accuracy
can be checked. The effective rolling radius of both the front and rear tyre have been determined
with measurements, from which the results have been presented in section 5.4. These measure-
ments are now used as a reference, and will be compared with the new description of the effective
rolling radius. The camber angle γ and the vertical load Fzw are averaged over the last complete
tyre rotation and used as input parameters for the calculation method. By subtracting the cal-
culated average radius from the measured one, the error of the calculation method is determined.
In figure 6.5 the absolute errors made by the new effective rolling radius calculation method are
depicted for both the front and rear tyre.
The results obtained by the new calculation method are very promising from 0 up to 30
degrees camber. Within this range the absolute error remains around 1 [mm], which corresponds
to relative errors within 0.5 % for both the front and rear tyre. However, at higher camber angles
the error strongly increases to an unacceptable magnitude. The reason for this rising error can
most probably be found in the measurement results. In section 5.4 it has already been observed
that the measurements at larger camber angles were inconsistent and not completely trusted on
their validity. Unexpected sudden jumps of the effective rolling radius are seen, although the
deformation and the vertical force remain almost constant. Therefore, the calculation method is
trusted on its validity and the increasing error at larger camber angle is expected to be caused by
the measurements.
CHAPTER 6. IMPROVEMENTS TO THE MF-MCTYRE MODEL 43
2
Fzw = 1500 [N]
1 Fzw = 750 [N]
−1
Error [mm]
−2
−3
−4
−5
−6
−7
0 10 20 30 40 50
Camber angle, γ [deg]
5
Fzw = 1500 [N]
Fzw = 750 [N]
−5
Error [mm]
−10
−15
−20
0 10 20 30 40 50
Camber angle, γ [deg]
loaded radius Rl for the conversion of the moments is changed to the free tyre radius Ro , this is
also done for the magnitude of the rolling resistance. Moreover, the rolling resistance coefficient
is assumed as a contant 1.5% which is also done for automobile tyres.
To define the direction of this moment a new axis system is generated in the tyre-road contact
point, the T -axis system. In contradiction to the W -axis system the T -axis system is rotated with
the camber angle with respect to the road plane. Therefore, this new axis system is perpendicular
to the C-axis system. All three axis systems are depicted in figure 6.6. In chapter 4 it has been
F z c
F y c
M z c
M y c
g
F z w
M M R
y t z w M z t l
F y w M y w C
shown that for automobile tyres the assumption has been made to define the rolling resistance
Mrr in the W axis system, moreover to define that:
For motorcycle tyres this assumption has been copied, and it has been shown that this leads to
the fluctuating moment around the spindle axis of a freely rolling wheel. As the camber angles
remain small for automobile tyres, the difference between the W - and T -axis system remains
small. For motorcycles though, camber angles become significant and large differences are present
between both axis systems. To overcome the problems it is therefore suggested to define the rolling
resistance moment Mrr in the T -axis system. With this definition, the moments in the T -axis
system are defined as:
In the previous chapter it has been shown that assuming the rolling resistance in the W -axis system
leads to fluctuations in the moment Myc . This is because of the fact that with this assumption a
fluctuating Mzw is present during measurements, while all other variables are almost constant. If
the rolling resistance definition in the T -axis system is now examined, it appears from equations
(6.22)-(6.23) that the fluctuating moment Mzc is present in both Myw and Mzw . Therefore, this
might lead to a more constant Myc , which is expected from the 2nd Law of Newton. With both
rolling resistance definitions a large number of test trailer measurements with a freely rolling wheel
under several camber angles have been analyzed. Significant improvements were found with the
new definition, with the difference between both definitions increasing with the camber angle. The
result of a representative α-sweep under a camber angle of 50 degrees is shown in figure 6.7. This
10
Side slip angle, α[deg]
−5
−10
0 10 20 30 40 50 60 70
100
Mrr defined in T
50 Mrr defined in W
Myc [Nm]
−50
−100
0 10 20 30 40 50 60 70
Time [s]
Figure 6.7: Difference in Myc for different definitions of the rolling resistance
figure clearly shows the large differences between both methods. A large and fluctuating residual
moment on the wheel axle is present when the rolling resistance is defined in the W -axis system.
When the rolling resistance is defined in the T -axis system the moment Myc is very close to zero,
as expected with a freely rolling wheel. Therefore, the rolling resistance is from now on defined in
this axis system.
However, the definition of the rolling resistance in the T -axis system also affects the description
of the moment Mz in the contact point. This moment is nowadays fitted in the W -axis system,
but with the definition of the rolling resistance in the T -axis system it would be more convenient
to fit the moment Mzt . If this is possible no transformation of the moment Mz is needed anymore,
following (6.21). Therefore the measurements of a front and a rear tyre are processed with the new
rolling resistance definition and the moments Mzw and Mzt are fitted with the Magic Formula. For
the rear tyre, the fit error of Mzw is 7.6% which decreases to 6.7% for the moment Mzt . For the
front tyre though, the fit of the moment Mzw is more accurate than the fit of the moment Mzt as
the fitting errors are 7.1% and 8.0% respectively. Although the fit of the front tyre is less accurate
in the T -axis system, the choice is made to fit all moments in this axis system. The increase in the
error percentage is relatively small, and for the rear tyre even a smaller fit percentage is achieved
with fits in the T -axis system. Moreover, the advantage of a more convenient definition of the
moments in the T -axis system is decisive.
CHAPTER 6. IMPROVEMENTS TO THE MF-MCTYRE MODEL 46
Fzw 2
Mrr = −Fzw Ro λmy {qsy1 + qsy5 γ 2 + qsy6 ( )γ } (6.28)
Fz0
The factor λmy is introduced to give the tyre model users the freedom to scale the rolling resistance
with little effort. The Flatplank measurements have been used to determine the correct parameters
qsy1 , qsy5 and qsy6 for the front and rear tyre. With (6.28) it is possible to fit the rolling resistance
moment of the front tyre measurements with an error percentage of 13.28%, following the definition
of (6.1). The error when fitting the rolling resistance of the rear tyre is 11.08%. With an optimal
but constant rolling resistance coefficient these percentages are 44.84% and 42.28%. Therefore,
it is now possible to describe the rolling resistance of a tyre sufficiently accurate with only three
parameters.
Chapter 7
In the previous chapter a number of improvements to the tyre model are developed. These
improvements are implemented in an updated version of the MF-MCTyre model which can be
used during simulations. In this chapter, several simulations are used to show the accuracy of the
improved tyre model and the effect of the improvements. To give an idea of the differences and
their cause, first of all a short summary of the improvements to the tyre model are given. Then
the improved and the old tyre model will be used in a SimMechanics simulation which mimics
the tyre measurements performed with the Delft Tyre Test trailer. With this model it is possible
to compare accuracy of the models with respect to the actual measurements. Finally, also the
motorcycle simulation model of chapter 3 is used. The results of simulations with this model are
used to determine the effect of the improvements to the tyre model on the steady state cornering
behaviour of a motorcycle.
47
CHAPTER 7. ANALYSIS OF THE IMPROVED TYRE MODEL 48
Figure 7.1: The SimMechanics model which mimics the Delft Tyre Test trailer measurements
First of all, a model made in SimMechanics is being used to evaluate the different tyre models.
This simulation model (which is depicted in figure 7.1 has the same DoF’s as the testing tower
of the Delft Tyre Test trailer. As can be seen, the forward velocity Vx , the side slip angle α, the
longitudinal slip κ, the camber angle γ and the vertical load Fzw are input parameters for the
simulation. These parameters are also determined during measurements and these signals can be
used as an input for the simulation. In this way, a tyre measurement can be accurately represented
by the simulation model.
The forces and moments during measurements and simulations used to be compared in the
W -axis system. However, both the output of the forces and moments by the tyre model as
measured forces and moments are in the C-axis system. The representation of the forces and
moments in this axis system by the tyre model therefore determines its accuracy. Therefore, in
this section the results of the simulations are compared with the measurements in the C-axis
system, which is much more logical. It should be remarked that for these specific simulations,
the rolling resistance moment measured with the Delft Tyre Test trailer is used. It is known
that during measurements an incorrect amplification factor is used for the longitudinal force Fxw .
However, the signal cannot be corrected as the the correct amplification factor is not known.
Although the longitudinal force is incorrect in magnitude, the effects of the camber angle are still
seen in the rolling resistance. In order to compare the simulation results with the measurements,
therefore the incorrect measurements are used in the improved tyre model.
First of all, an α-sweep at 5 degrees of camber and a nominal load Fzw of 2000 [N] is used
to check the correctness of the tyre models. In figure 7.2 the forces and moments in the C-axis
system during this simulation are depicted. Moreover, in table 7.1 the relative errors made by
both models during this simulation can be found. For the determination of these errors, again the
definition of (6.1) is used.
The error percentage of the forces Fyc and Fzc made by the existing and improved tyre model
are small and only differ tenths of percents. These forces are determined by the lateral and
vertical forces in the W -axis system (Fyw and Fzw ), which are rotated over the camber angle. As
the vertical force Fzw is prescribed for the simulation and the description of the lateral force is
similar for both tyre models, these forces are also expected to show large correspondence. As the
simulation results are almost equal and the differences between these results and the measurements
are very small, the differences can hardly be seen in the graphical representation of figure 7.2. It
is clear that the major difference lies in the longitudinal force Fxc . In paragraph 4.4 it has been
explained that the old rolling resistance assumption leads to the large and fluctuating longitudinal
force which can also be seen in figure 7.2. Moreover, these fluctuations are known to increase
with the camber angle which will also be shown in this paragraph. With the definition of the
moments in the T -axis system and the improved description of the rolling resistance factor fr
these problems are overcome. The results presented in figure 7.2 show that the longitudinal force
CHAPTER 7. ANALYSIS OF THE IMPROVED TYRE MODEL 49
10
Measurements
Alpha [deg]
5 Improved model
Existing model
0
−5
−10
0 5 10 15 20 25 30 35 40 45 50
0
Fxc [N]
−20
−40
−60
0 5 10 15 20 25 30 35 40 45 50
2000
1000
Fyc [N]
−1000
−2000
0 5 10 15 20 25 30 35 40 45 50
2400
Fzc [N]
2200
2000
1800
0 5 10 15 20 25 30 35 40 45 50
500
Mxc [Nm]
−500
0 5 10 15 20 25 30 35 40 45 50
50
Mzc [Nm]
−50
0 5 10 15 20 25 30 35 40 45 50
Time [s]
Figure 7.2: Measured and simulated forces and moments in the C-axis system during an α-sweep
under 5 degrees camber
CHAPTER 7. ANALYSIS OF THE IMPROVED TYRE MODEL 50
is much more constant, which is also the case during measurements. The noise on the longitudinal
force generated by the improved model is declared by the fact that the new rolling resistance
coefficient fr is dependent on the vertical force Fzw . The vertical force is used as an input for
the simulation and the noise level on this signal is significant. As the longitudinal force of a
freely rolling wheel is determined by the rolling resistance, the noise can also be seen on this
force. Therefore, the relative error is still 57.0% for the improved model. In comparison with the
relative error of 148.9% with the existing model, this is a significant improvement but if the noise
level is lower during the simulation (which is normally the case), the relative error will decrease
significantly.
Next to the forces, also the moments Mxc and Mzc are depicted in figure 7.2. Due to the small
camber angle, the differences between both models and the differences with the measurements
remain relatively small. The moment Mxc is determined by the moment Mxw , the lateral force
Fyc and the distance between the wheel center and the contact point. It is already explained
that the difference in Fyc is negligible. Furthermore, the moment Mxw generated by the tyre
models under a small camber angle is comparable. As the loaded radius of the existing model
is reasonably accurate under a small camber angle and Ro is used by the improved tyre model,
the difference in the relative error in moment Mxc is only 0.1%. The moment Mzc in the contact
point is fitted in the T -axis system by the improved tyre model. With this definition no conversion
between the wheel center and the contact point is needed anymore, which increases the accuracy
of the improved tyre model. The error made in Mzc therefore decreases from 11.1% to 8.7% for
the existing and improved tyre model respectively, as depicted in table 7.1.
Table 7.1: Errors made by the existing and improved tyre model during an α-sweep simulation
under 5 degrees camber, following the definition of (6.1)
The forces and moments during a measurement and simulations with the two tyre models
during an α-sweep under a camber angle of 50 degrees can be found in figure 7.3. As expected,
the difference in the longitudinal force between both models strongly increases. As the camber
angle is much larger, also the difference between the W - and T -axis system increases. Therefore,
the error made by the old rolling resistance assumption is increasing. As the tyre model uses the
longitudinal force Fxw to remain the equilibrium around the spindle axle of a freely rolling wheel,
the error of this force is increasing with the camber angle. The error made in Fxc by the existing
tyre model is now 582.4%, while the error of the improved model is only 36.2%. The error made
by the improved model decreases with respect to the error made with the simulation at 5 degrees
camber, which can be declared by two facts. First of all, the noise level on the vertical load Fzw is
relatively smaller than during the measurement at 5 degrees camber. Also less noise is therefore
present on the rolling resistance coefficient fr , which also leads to less noise on the longitudinal
force Fxw . Furthermore, the rolling resistance coefficient increases quadratically with the camber
angle. As the influence of the vertical load on the rolling resistance coefficient increases only
linearly with the camber angle, the relative influence of the vertical load is decreasing with the
camber angle. Therefore, the relative effect of the noise on the vertical load is also decreasing with
the camber angle and the relative error decreases.
During the simulation with the existing tyre model, large differences arise in the moment Mxc as
can also be seen in figure 7.3. The cause of this can be found in the inaccurately and unambiguously
defined loaded radius Rl . By substituting the loaded radius with the undeformed tyre radius Ro ,
these problems are overcome. The conversions are ambiguously defined and accurate, which also
CHAPTER 7. ANALYSIS OF THE IMPROVED TYRE MODEL 51
10
Measurements
Alpha [deg]
5 Improved model
Existing model
0
−5
−10
0 10 20 30 40 50 60 70
0
−100
Fxc [N]
−200
−300
−400
0 10 20 30 40 50 60 70
1500
Fyc [N]
1000
500
0
0 10 20 30 40 50 60 70
3000
2500
Fzc [N]
2000
1500
1000
0 10 20 30 40 50 60 70
600
Mxc [Nm]
400
200
0
−200
0 10 20 30 40 50 60 70
0
Mzc [Nm]
−50
−100
0 10 20 30 40 50 60 70
Time [s]
Figure 7.3: Measured and simulated forces and moments in the C-axis system during an α-sweep
under 50 degrees camber
CHAPTER 7. ANALYSIS OF THE IMPROVED TYRE MODEL 52
leads to an accurate representation of the overturning moment by the Magic Formula fit. As
a result, the relative error of 41.7% made by the existing model decreases to 8.8% with the
improved model. Next to the improvements in Mxc , also significant improvements are made with
the representation of the moment Mzc . The new rolling resistance assumption has a significant
influence, especially when the camber angle increases. Furthermore, it seemed that the Magic
Formula is better suited to fit the moment Mzt than the moment Mzw . Therefore, the error
percentage made by the improved tyre model is 8.7% instead of 11.1% with the existing model.
Table 7.2: Errors made by the existing and improved tyre model during an α-sweep simulation
under 50 degrees camber, following the definition of (6.1)
0
Existing model
Improved model
−20
−40
−60
Y−distance [m]
−80
−100
−120
−140
−160
Figure 7.4: Motorcycle trajectory during simulations with the existing and improved tyre model
Rearwheel
κ [-] 0.004 0.004 -2.3%
α [deg] 1.37 1.35 -1.5%
γ [deg] 48.26 50.16 3.9%
Fxw [N] 331.98 328.59 1.0%
Fyw [N] -1635.90 -1639.00 -0.2%
Fzw [N] 1758.20 1749.60 0.5%
Mxc [Nm] -115.96 -144.02 -24.2%
Myc [Nm] 146.38 152.54 -4.2%
Mzc [Nm] -17.51 -15.78 9.9%
Table 7.3: Comparison between tyre in- and output for the front and rear wheel
CHAPTER 7. ANALYSIS OF THE IMPROVED TYRE MODEL 54
front and rear tyre is larger with the improved tyre model, the camber angle is also larger to
obtain a steady state cornering situation. The slip angle of the front wheel is mainly determined
by the moment Mzc . The negative Mzc which is present is turning the wheel into the corner. A
positive side slip angle is turning the wheel out of the corner. As the moment Mzc is 32.5% smaller
with the existing tyre model, the tyre sideslip angle is larger during the simulation with this tyre
model. Finally, the rolling resistance of the existing and the new front tyre are comparable as
the longitudinal forces are almost similar. As the front wheel is freely rolling the moment Myc of
the front wheel is zero during the simulations. The driving torque Myc of the rear tyre is 4.2%
larger during the simulation with the improved tyre model. This moment is present in order to
overcome the rolling resistance of both tyres and the aerodynamic drag. The forward velocity is
equal during both simulations, and therefore also the aerodynamic drag is equal. As the rolling
resistance of the front tyre is also almost similar, the difference is caused by the rolling resistance
of the rear tyre.
Next to the tyre in- and output, also the response of the motorcycle can be compared. The
results of this comparison can be found in table 7.3. It has already been explained that the camber
angle of both wheels is 3.9% larger due to the larger moment Mxc . The overall camber angle is
therefore of course also 3.9% larger. The steer torque on the other hand is much smaller with
the improved tyre model than with the existing one. The systems which cause a torque around
the steering axle have been explained in section 3.4. As the lateral force and the mechanical trail
are similar during both simulations, the smaller steering torque is mainly caused by the moment
Mzc . Although the steering torque is significantly larger, the steer angle is almost unaltered.
Although the fact that a relatively large difference is present (225.8%), the absolute steer angle
is too small to be significant. As the camber angle is larger with the improved tyre model, the
equilibrium around the twist angle also changes. With a larger camber angle the vertical force
Fzw gets more influence on the twist equilibrium, while the lateral force Fyw gets less influence.
As the vertical force is larger than the lateral force, the twist angle increases with the improved
tyre model. Within table 7.3 it has been shown that the overal moment Mxc and therefore also
the camber angle are larger with the improved tyre model. The equilibrium of the motorcycle
around its local x-axis therefore changes which leads to a larger rider lean angle.
The introduction of the tyre contour and the correct effect of the vertical load, should lead to
significant changes in the effective rolling radius during simulations. If the front wheel is taken into
consideration, it is learned from table 7.3 that the longitudinal slip κ does not change significantly.
This is also expected as there is no effective driving or braking torque around the wheel axle.
In figure 7.5, the effective rolling radius of the front wheel during both simulations is depicted.
Already in the first simulation seconds, there is a significant difference between the effective rolling
radius of the existing and improved tyre model. Research shows that the difference, which is in
the order of 10 [mm], is caused by the different Magic Formula parameters that introduce the
CHAPTER 7. ANALYSIS OF THE IMPROVED TYRE MODEL 55
Existing model
Improved model
0.295
Effective rolling radius, Re [m]
0.29
0.285
0.28
0.275
0.27
0 20 40 60 80 100
Time [s]
Figure 7.5: The effective rolling radius Re of the front wheel during the simulations
effect of the vertical load. Furthermore, after 10 seconds the motorcycle drives into the corner
and the difference increases significantly. This can be explained by the fact that the tyre contour
is implemented in the improved model, while this contour was absent in the determination of the
effective rolling radius in the existing model. In section 4.3 the definition of linear rolling velocity
Vr and the longitudinal slip κ has already been presented:
Vr = ΩRe (7.2)
Vr − Vx
κ= (7.3)
Vx
As the longitudinal slip κ and the forward velocity Vx are almost constant during the simulation,
also the linear rolling velocity Vr has to be almost constant. As the effective rolling radius strongly
decreases with the camber angle, the rotational velocity needs to increase to obtain an almost
constant linear rolling velocity. From table 7.3 it is learned that the rotational velocity of the
front wheel is 9.4% larger during steady state cornering with the improved tyre model. For the
rear tyre this increase is even larger, namely 16.5%. Moreover the rotational velocity of the wheels
is known to have a significant influence on the handling behaviour of motorcycles. Together with
the inertia of the wheel around the spindle axis, the rotational velocity of the wheel determines the
gyroscopic precession which has been explained in chapter 3. This effect not only has a significant
influence on the steering behaviour of a motorcycle, it also partly defines the instabilities of the
motorcycle.
Finally also the wheel center height is investigated. In table 7.3 it is shown that the difference in
wheel center height during steady state cornering is 5.5% for the front and 2.9% for the rear wheel.
However, those differences are also caused by the different camber angles during the simulations
with the existing and improved tyre model. To get the effect of the changes in the tyre model on
the wheel center height in the right perspective, also a simulation with a constant camber angle of
50 degrees is conducted. The axle height of the front wheel during these simulations can be seen
in figure 7.6. With the implementation of the improved description of the tyre contour and the
vertical tyre stiffness, the axis height is known to be more accurate in the improved tyre model.
Due to the tyre contour implementation and the decreasing vertical stiffness with an increasing
CHAPTER 7. ANALYSIS OF THE IMPROVED TYRE MODEL 56
camber angle, a smaller axle height is present with the improved tyre model than with the existing
tyre model. The difference during steady state cornering is around 6 [mm], which is in the order of
3.1% of the total axis height. In section 4.2.1, it has been explained that the motorcycle behaviour
is partly governed by the height of the wheel axles. Therefore, an accurate axle height is of large
importance and a difference of 3.1% is significant.
Existing model
0.29 Improved model
0.28
0.27
axle height, h [m]
0.26
0.25
0.24
0.23
0.22
0.21
0.2
0 5 10 15 20 25 30 35
Time [s]
Figure 7.6: Front wheel center height during simulations with the existing and improved tyre
model
Chapter 8
8.1 Conclusions
The goal of this master thesis is to improve and validate the motorcycle tyre model, MF-MCTyre.
An improved model should be able to correctly represent the behaviour of a motorcycle tyre for its
complete working area at steady-state conditions. Therefore, deficiencies of the model have been
localized and solutions are developed. It appears that the MF-MCTyre model has to be improved
in three areas:
57
CHAPTER 8. CONCLUSIONS AND RECOMMENDATIONS 58
wheels. Not only does this have an impact on the dynamic behaviour of the motorcycle, the
rotational speed of the wheels also play an important role if for example the driveline of the
motorcycle is modelled.
As an overall conclusion it can be said that the motorcycle tyre model MF-MCTyre has un-
dergone significant improvements. Due to these improvements, this tyre model is now able to
accurately represent the steady state cornering behaviour of a motorcycle tyre.
Automotive has developed a model which is able to represent the dynamic behaviour of an
automobile tyre up to 100 [Hz]. It is therefore recommended that the present knowledge
on tyre dynamics is extended and combined with the improved MF-MCTyre model. This
combination can lead to a model which is able to correctly represent both the dynamic and
steady state behaviour of a motorcycle tyre.
• Motorcycle model parameters
The motorcycle simulation model which is developed in this research is derived from the
model of Koenen [10]. His research has been completed in 1983 and since then motorcycles
have undergone significant developments. Within the literature study presented in chapter 2
the correctness of the range of parameters used in the Koenen model is shown. However, these
parameters should be updated in such a way that they represent a nowadays motorcycle.
To do so, too little information is present in the available literature. Therefore, an elaborate
measurement program is needed to find the geometrical, mass and inertial parameters of a
recent motorcycle.
[1] E. Bakker, L. Nyborg, and H.B. Pacejka. Tyre modelling for use in vehicle dynamic studies.
SAE paper, (870421), 1987.
[2] P. Bayle, J.F. Forissier, and S. Lafon. A new tire model for vehicle dynamics simulations.
Automotive Technology International, pages 193–198, 1993.
[3] R. Berritta, F. Biral, and S. Garbin. Evaluation of motorcycle handling with multibody
modelling and simulation. High tech engines and cars, 6th international conference, May
25th-26th 2000.
[4] V. Cossalter and A. Doria. Model simulation: The latest dynamic simulation developments
for motorcycle tires. Tire Technology International, pages 38–41, September 2001.
[5] V. Cossalter, A. Doria, and R. Lot. Steady turning of two-wheeled vehicles. Vehicle System
Dynamics, 31:157–181, 1999.
[6] V. Cossalter and R. Lot. A motorcycle multi-body model for real time simulations based on
the natural coordinates approach. Vehicle System Dynamics, 37(6):423–447, 2002.
[7] V. Cossalter, R. Lot, and F. Maggio. The modal analysis of a motorcycle in straight running
and on a curve. Meccanica, 39(1):1–16, 2004.
[8] S. Evangelou and D.J.N. Limebeer. Lisp programming of the ’sharp 1994’ motorcycle model,
March 28th 2004. Department of Electrical and Electronic Engineering, Imperial College of
Science, Technology and Medicine.
[9] Y. Kamata and H. Nishimura. System identification and attitude control of motorcycle by
computer aided analysis. JSAE Review, 24(4):411–416, October 2003.
[10] C. Koenen. The dynamic behaviour of a motorcycle when running straight ahead and when
cornering. PhD thesis, Delft University of Technology, 1983.
[11] C. Koenen and H.B. Pacejka. The influence of frame elasticity, simple rider body dynamics
and tyre moments on free vibrations of motorcycles in curves. Vehicle System Dynamics,
pages 53–65, 1982. Proceedings of IAVSD symposium.
[12] R. Lot. A motorcycle tire model for dynamic simulations: Theoretical and experimental
aspects. Meccanica, 39:207–220, 2004.
[13] H.B. Pacejka. Tyre and vehicle dynamics. Butterworth Heinemann, 2002.
[14] J.S. Rankine. Dynamical principles of the motion of velocipedes. The Engineer, 28, 1869.
[15] R.D. Roland. Computer simulation of bicycle dynamics. ASME symposium on mechanics
and sport, pages 1115–1121, 1973.
[16] P.A.J. Ruijs and H.B. Pacejka. Recent research in lateral dynamics of motorcycles. Vehicle
System Dynamics, 15:467–480, 1985. Proceedings of IAVSD symposium.
60
BIBLIOGRAPHY 61
[17] R.S. Sharp. The stability and control of motorcycles. Journal of Mechanical Engineering
Science, 13(5):316–329, 1971.
[18] R.S. Sharp. The lateral dynamics of motorcycles and bicycles. Vehicle System Dynamics,
14:265–283, 1985.
[19] R.S. Sharp. Stability, control and steering response of motorcycles. Vehicle System Dynamics,
35(4-5):291–318, 2001.
[20] R.S. Sharp and C.J. Alstead. The influence of structural flexibilities on the straight-running
stability of motorcycles. Vehicle System Dynamics, 9:327–357, 1980.
[21] R.S. Sharp and D.J.N. Limebeer. A motorcycle model for stability and control analysis.
Multibody System Dynamics, 6(2):123–142, September 2001.
[22] P.T.J. Spierings. The effects of lateral front fork flexibility on the vibrational modes of straight
running single-track vehicles. Vehicle System Dynamics, 10:21–35, 1981.
[23] Y. Tezuka, H. Ishii, and S. Kiyota. Application of the magic formula tire model to motorcycle
manoevrability analysis. JSAE Review, 22(3):305–310, July 2001.
[24] TNO Automotive. Tyre models users manual; Using the MF-MCTyre model, May 2002.
[25] N. van de Wouw. Multibody dynamics, lecture notes, 2003. page 25.
[26] M.K. Verma, R.A. Scott, and L. Segel. Effect of frame compliance on the lateral dynamics
of motorcycles. Vehicle System Dynamics, 9(4):181–205, 1980.
[27] B. von Schlippe and R. Dietrich. Zur mechanik des luftreifens. Zentrale fur wissenschaftliches
Berichtwesen Berlin Adlershof, 1942.
[28] E.J.H. De Vries and H.B. Pacejka. Motorcycle tyre measurements and models. In Proceedings
of the 15th IAVSD Symposium, pages 280–298, 1997.
[29] E.J.H. De Vries and H.B. Pacejka. The effect of tire modelling on the stability analysis of a
motorcycle. In Proceedings of the 4th international symposium on advanced vehicle control,
pages 063/1–063/6, 1998.
Appendix A
This master thesis has been devoted to improve the MF-MCTyre model. In this appendix, its main
features will be explained on the basis of the user manual of the model [24]. For the description
of the Magic Formula also the book ”Tyre and vehicle dynamics” of Pacejka [13] is used. For a
more comprehensive description of the tyre model and the Magic Formula the reader is therefore
referred to those works. The implementation of the tyre model into a vehicle model has been
discussed in paragraph 3.3.
The normal load Fz is then calculated with the tyre vertical stiffness Cz and damping Kz ;
Fz = Cz ρ + Kz ρ̇ (A.2)
62
APPENDIX A. THE MF-MCTYRE MODEL 63
Z
nr
O
X
Zc Y
Zw
R0 Rcs
Vx
Xc γ
ha
tyre cross
Xw section
Yc rt
Mx
Yw
C Fy C
ρ road tangent
plane
Fz
Figure A.1: The contact point C and tyre compression ρ in this point
304
Effective rolling radius Re
302 Loaded radius Rl
300
298
Radius [mm]
296
294
292
290
288
286
0 500 1000 1500 2000 2500
Vertical force, Fzw [N]
This radius decreases with increasing vertical load at low loads, but around its nominal load the
influence of the vertical load is small as can be seen in figure A.2 for a tyre with a nominal load
of 1475 [N]. When assuming a constant vertical tyre stiffness Cz , the radial tyre deflection can be
calculated with;
Fz
ρ= (A.4)
Cz
For the estimation of the effective radius Re a Magic Formula approach is chosen;
in which R0 is the unloaded free radius and the nominal tyre deflection ρFz0 is defined by the
vertical tyre stiffness Cz and the nominal wheel load Fz0 ;
Fz0
ρFz0 = (A.6)
Cz
Factor B in equation A.5 determines the slope at Fz = 0, factor D defines the height of the
asymptote at high wheel loads and factor F defines the ratio between the tyre radial deformation
and the effective tyre deformation.
Vx Vsx
Vy α
V Vr Vs Vsy
With figure A.3 the tyre slip quantities can be derived. The longitudinal slip speed is defined
as;
Vsy = Vy (A.9)
With these slip speeds the practical slip quantities (κ and α) that are used as an input for the
Magic Formula are defined as;
Vsx
κ=− (A.10)
Vx
Vsy
α = arctan( ) (A.11)
|Vx |
With Vsx and Vsy the components of the slip speed that may be defined as the velocity of point
S in the W -axis system (see figure A.3). With Ω denoting the rotational speed of the tyre, the
linear rolling speed becomes;
Vr = R e Ω (A.12)
The original form of the formula that holds for given values of vertical load and camber angle
reads;
y(x) = D sin[C arctan{Bx − E(Bx − arctan Bx)}] (A.13)
with
Y (X) = y(x) + SV (A.14)
x = X + SH (A.15)
This original form is still being used for the representation of both the longitudinal (Fx ) and
lateral forces (Fy ) of automobile tyres. For motorcycle tyres the original formula is only used for
the longitudinal force. The Magic Formula y(x) typically produces a curve that passes through the
origin, reaches a maximum and subsequently tends to a horizontal asymptote. For given values
of the coefficients B, C, D and E the curve shows an anti-symmetric shape with respect to the
origin. To allow the curve to have an offset with respect to the origin, two shifts SH and SV are
introduced. A new set of coordinates Y (X) arises as shown in figure A.4. The formula is capable
of producing characteristics that closely match measured curves for the longitudinal force as a
function of the longitudinal slip κ with the effect of load Fz and camber angle γ included in the
parameters.
Figure A.4 illustrates the meaning of some of the factors used in the formula. Obviously,
coefficient D represents the peak value (for C ≥ 1) and the product BCD corresponds to the
slope at the origin. The shape factor C controls the limits of the range of the sine function
appearing in formula A.13. Thereby it determines the shape of resulting curve as also the height
of the horizontal asymptote ya . The factor B is left to determine the slope at the origin and is
called the stiffness factor. The factor E is introduced to control the curvature at the peak and at
the same time the horizontal position of the peak, xm .
y Y
X m
S H
D
y a
S V a rc ta n (B C D )
Next to κ, also the momentary vertical load Fz and camber angle γ are used as an input as the
coefficients are determined dependently on those parameters.
The pneumatic trail on its turn is represented by the cosine version of the magic formula;
With;
αt = α + SHt (A.20)
This version of the Magic Formula is able to produce the characteristic hill shaped curve, which
can be seen in figure A.5. In this figure the basic properties of the cosine based curve have been
indicated. Again, D is the peak value, C is a shape factor determining the level ya of the horizontal
asymptote and now B influences the curvature at the peak (illustrated with the inserted parabola).
Factor E modifies the shape at larger values of slip and governs the location x0 of the point where
the curve intersects the x-axis.
Y
y
-S h
-y a
X x
2 -x 0
B C
Figure A.5: Curve produced by the cosine version of the Magic Formula
S p in d le a x is
T y re c o n to u r
C A R o a d s u rfa c e
y M x
with qsx1 the factor with which a vertical offset can be introduced at 0 sideslip and camber.
Furthermore qsx2 and qsx3 are the factors introducing the overturning couple induced by camber
and lateral force respectively. Both λ values are scale factors. The main contribution of the
overturning moment while cornering will be induced by the camber factor. This is because of the
fact that the forces and moments are not exactly applied in the contact center A, but in the point
C, as can be seen in figure A.6. The moment introduced by this lateral shift, Fz .yM x , therefore
needs to be taken into account. As can be seen in figure A.6, this moment is considerable at large
camber angles.
My = −Fz Rl fr (A.22)
The vertical load and the loaded radius are determined in the contact routine, the rolling resistance
coefficient is determined as a function of the longitudinal force Fx and the longitudinal velocity
Vx . This in contradiction to the processing of the measurements, where fr is taken as 1.5%.
APPENDIX A. THE MF-MCTYRE MODEL 68
φ V
elastic foundation
a α
a
σ
v2 v1
path of
contact points stretched string
Mz
Fy
Fx = Fx (α0 , γ, κ0 , Fz ) (A.27)
Fy = Fy (α0 , γ, κ0 , Fz ) (A.28)
Mz = Mz (α0 , γ, κ0 , Fz ) (A.29)
APPENDIX A. THE MF-MCTYRE MODEL 69
F o r c e s & M o m e n ts in C F o r c e s & M o m e n ts in W M a g ic F o r m u la p a r a m e te r s
* .m a t file s * .td x file s * .tir file s
M -ty re M F -to o l
the program M-tyre generates TYDEX (*.tdx) files. These files contain the forces and moments
that are converted to the contact point C. Next to these forces and moments also other variables
derived from the measurements are present, for example the loaded radius, longitudinal slip etc.
On their turn all TYDEX files are loaded in MF-Tool, which is a fitting program. In this program
all Magic Formula parameters are fitted to the measurement data and saved in a *.tir file. This
file is used by the MF-MCTyre model to read the Magic Formula parameters during simulations.
Appendix B
To be able to implement the Koenen motorcycle model in SimMechanics all parameter values
need to be specified. In ’The dynamic behaviour of a motorcycle’ [10] all parameters are specified
in Appendix G (p. 162-163). These parameters are listed in this Appendix, together with a
geometrical figure of the motorcycle.
s te e r a x is
g 2 x g 1 x m
g
1
tw is t a x is
1 z
d x
m 3 g 1 s z
a 1 z
g 1 s x
g 1 u z
r id e r le a n a x is g 3 z m 1 s
m 2
fz e g 1 u x
r
d m
z g 2 z
1 u
m 2 w m 2 u m 1 w
p itc h a x is
R o r R o f
s x m
s x j
a 1 x
f x
For these eight masses the moments of inertia that are being used are;
Jx1 = 0.46 [kgm2 ] Jx1s = 0.0 [kgm2 ]
2
Jy1 = 1.2 [kgm ] Jy1s = 0.0 [kgm2 ]
70
APPENDIX B. MOTORCYCLE MODEL PARAMETERS 71
Finally, an aerodynamic drag force FD and lift force FL are acting on the motorcycle. These forces
are determined as:
1
FD = ρCDA v 2 (B.1)
2
1
FL = ρCLA v 2 (B.2)
2
with ρ the density of air, v the forward velocity and CDA and CLA the effective drag and lift areas
respectively. These areas are;
With the Delft Tyre Test Trailer, the forces and moments that are generated by a tyre can be
measured. This can be done at two different measurement towers on the right and left side of the
test trailer. The tyre forces and moments are measured in the hub of the wheel (in the C-axis
system) and transformed to the forces and moments in the tyre road contact point (the W -axis
system). Furthermore, the tyre model converts the forces and moments from the contact point to
the wheel center. Therefore, in this Appendix the conversions will be given.
G y
G z 1 G z 2
G x 2 G x 1
b a
With the output of the strain gauges, the forces and moments in the C-axis system can be
determined. With the distances a and b known, the forces (Fc ) and moments (Mc ) are defined as:
Fyc = Gy (C.2)
72
APPENDIX C. PROCESSING MEASUREMENT DATA 73
3 2
4 1
All four of these cells measure a force in the x-, y- and z-direction. These forces are sometimes
combined and all measurement data is given in 7 measurement channels. The output of these
channels are defined as:
b b
F z 1 2 3 4 F x 1 2 3 4
F y 2 3 F y 1 2
F z C F x C
F y F y
v C
h C
M x C M z C
F y 1 4 F y 3 4
F z c
F y c
M z c
M y c
g
F z w
M z w R lo a d
F y w M y w C
The equilibrium of forces is formulated with figure C.4 and from this equilibrium the forces in
the W -axis system are defined as:
The friction factor fr is estimated to be 1.5 %. With this assumption, the moments in the W -axis
system become: