See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/285592400

Tire Simulations Using a Slip Velocity, Pressure and Temperature Dependent

Friction Law

Research · December 2015

DOI: 10.13140/RG.2.1.2678.9203

CITATIONS READS

5 3,756

3 authors, including:

Michael Kaliske
Technische Universität Dresden
572 PUBLICATIONS 6,578 CITATIONS

SEE PROFILE

All content following this page was uploaded by Abolhasan Giashi on 04 December 2015.

The user has requested enhancement of the downloaded file.

Tire Simulations Using a Slip Velocity, Pressure and Tem-
perature Dependent Friction Law

Korbinian Falk, Abolhasan Nazarinezhad, Michael Kaliske∗

Institut für Statik und Dynamik der Tragwerke (ISD), Technische Universität Dresden, Fakultät Bauingenieurwesen,
01062 Dresden, Germany

Abstract A detailed finite element (FE) model for the analysis of realistic force transmission between a rolling
tire and a rigid flat road is presented. In order to achieve a more precise investigation, enhanced friction models
for the contact analysis are employed. The model is based on an extensive program of material tests and
measurements to fit the models. The investigated high performance tire has been validated for its static and
dynamic stiffness as well as footprints, deformed shapes and temperature fields on the surface and inside the
tire. Based on the enhanced friction models, the tire’s frictional behaviour is computed according to its relative
sliding velocity, contact pressure and temperature. In this way, the contact formulation for the tire model is
significantly improved and a stable simulation with more accurate results is obtained. Temperature input is
generated by the thermo-mechanically coupled tire simulation which is developed at the ISD. The respective
experimental data out of the linear friction test on a dry granite surface are used for parameter identification
based on an optimization algorithm. Tire frictional behaviour is then characterized by a nonlinear, continuous
tanh-function to describe the stress-displacement interaction. The user subroutine FRIC written in FORTRAN
is applied to control the frictional parameters. Finally, the capability of the FE simulation for predicting the
fundamental frictional characteristics at braking and at cornering are widely verified for different speed, load
and temperature conditions based on the Savkoor, Huemer and Coulomb friction law.
Keywords: Finite Element Analysis (FEA), Friction Coefficient, Slip Velocity, Contact Pressure, Steady-State
Rolling (SSR), Cornering and Braking.

1 Introduction

Friction interaction between tire and road has an undeniable influence on the rolling tire performance.
In order to attain an accurate analysis of a moving tire, the prediction of tire friction characteristics has
been widely considered through academic and industrial research. Tires are designed and produced in
order to increase the durability, safety and controllability of driving performance. Therefore, an accu-
rate prediction of the tire-road friction coefficient is necessary for a better understanding of tire design
and simulation. The estimation of the frictional behaviour of a tire depends on contact mechanics,
tire component properties, road surface characteristics, environmental and driving conditions.
Many studies and approaches have recently been proposed in this area. The robust tire-road friction
force estimation was studied by Matusko et al. [6] who proposed a special method for tire friction
estimation using neural networks and Lyapunov stability analysis. Heinrich et al. [3] investigated tire
component deformation due to frictional stresses. They studied rubber hysteresis and adhesional fric-
tion, tread deformation and tire traction based on the tire geometry, applied load and velocity. Beside
the numerical analysis of tire performance phenomena, a comprehensive investigation for a steady
state rolling tire based on FE modeling was presented by Kaliske [5]. In this work, the fundamen-
tal characteristics of tire modeling such as rubber compound properties, in addition to geometrical
nonlinearities and rubber visco-elastic properties have been widely studied. Van der Steen [9] made
∗
Corresponding author
Tel.: +49-351-46334386
Fax: +49-351-46337086
Email: michael.kaliske@tu-dresden.de
a) b)

Fig. 1 Rubber material testing and fitting: a) tensile test ; b) fitting for tread rubber material model Yeoh

thorough studies about enhanced friction modeling for steady-state rolling tires. Using FE models,
the characteristics for tire simulation, friction and steady state rolling tire performance have been
studied. Another similar work presented by Pinnington [7] focused on the influence of road surfaces
on rubber friction characteristics.
In this work, the frictional behaviour of a special performance tire at steady state rolling situation
is studied. Tire mechanical and material characteristics based on an extensive program of material
testing are defined through a FE model using the commercial ABAQUS package. Enhanced friction
models are employed to describe the contact interaction between tire and a rigid flat road. A user
subroutine FRIC via FORTRAN is employed to control the solution. Finally, the capability of the FE
simulation predicting the fundamental frictional characteristics at braking and cornering is verified
for different speed, load and temperature conditions based on the Savkoor, Huemer and Coulomb
friction law.

2 Tire model

Tire modeling is a complex task as large deformations, complex material interactions, different stiff-
ness and thin geometries need to be considered. The geometry of the undeformed outer contour, the
position and direction of fiber layers and rubber compounds are basic information for the FE model.
To predict global and local tire properties as well, it is essential to add all structural elements in the
model for a steady state tire analysis. Elements need to avoid locking effects and singularities. Thin
layers of rubber compounds define the small element width. To reach acceptable computational time,
it is necessary to reduce the number of segments around the circumference, especially outside the
contact area. This means, the element length in circumference reaches multible times the width in
cross-section.
Rubber shows nonlinear and time dependent material characteristics and it is almost incompressible.
To gain a suitable material model, experimental data of tensile and compression tests are evaluated.
A systematic fitting procedure is used to gain the right coefficients of the Yeoh material model to
describe the hyperelastic behaviour. Experiments are carried out for different strain amplitudes and
frequencies and show respectively a specific hysteresis. Hyperelastic material models can describe
only one stress-strain path. For this reason, it is essential to choose suitable experimental data for
the fitting procedure representing the material behavior for the simulation best. Fig. 1 shows the
fitting for a tread rubber compound which is in good agreement for the tensile part. The compression
part is represented too soft and could be improved including less accuracy for tensile fitting. The
chosen parameters have shown the best results for the global stiffness behaviour, footprint (especially
cornering) and deformation inside the tire. For visco-elastic properties, DMA results need to be
Fig. 2 Schematic relation of an enhanced friction model

available to find parameters for the Maxwell elements. The number of Maxwell elements is defined
by the range of the represented frequency within the simulation.

3 Friction models

To model the frictional behaviour of rubber materials in contact with rough surfaces is a challenging
topic. Since tire and road are in contact, the resulting forces can be transmitted in normal and tangen-
tial direction and the friction forces are directly related to the existing normal forces by the friction
ratio. Therefore, the computation of the coefficient of friction between tire and road is essential to
obtain a good basis for tire design and prediction of tire performance. Based on experiments, rubber
friction is comprehensively depending on its physical parameters such as relative sliding velocity,
contact pressure and working temperature. Due to this dependency, the global Coulomb friction ap-
proach as a constant value cannot exactly describe the frictional behaviour of tire-road interaction.
Based on existing enhanced friction models, the tire’s frictional behaviour is computed according
to its relative sliding velocity (v) and contact pressure (P ). Temperature (T ) is the third important
variable for friction, hence the given friction models need to be extended. This way, the friction
formulation for the tire model is significantly improved and a stable simulation with more accurate
results can be achieved.

3.1 Friction formulation

Throughout this study, two enhanced friction models presented by Huemer [4] and Savkoor [8] are
applied to express the friction coefficient. Equations (1) and (2) represent the mathematical definition
of Savkoor and Huemer friction models respectively,

 −k    
P 2 2 |v|
µ(v, P ) = µs + (µm − µs ) exp −h log (1)
p0 Vmax

where µs represents the static friction coefficient when no sliding velocity is present and µm is the
maximum resulting friction that occurs at Vmax in the Savkoor model. These assumptions include
a number of parameters to correlate the friction coefficient based on sliding velocity and contact
pressure.
The Huemer friction law is a phenomenological formulation

α |P |n−1 + β
µ(v, P ) = b c
(2)
a + v1/m + v2/m

in which the parameters α, β and n influence the pressure dependency whereas a, b, c and m control
the µ-function with respect to slip velocity. The introduced friction laws cannot consider a temper-
ature dependency, just a discrete temperature. For this reason, the parameters need to be a function
a) b)

Fig. 3 Parameter fitting for a temperature of 25◦ C for safety walk surface: a) Huemer friction law ; b) Savkoor friction law
m[-]

Experimental
Savkoor law
Huemer law

Pl
contact pressure [MPa]

Fig. 4 Contact pressure effect on the friction coefficient

of temperature to achieve a pressure, slip velocity and temperature dependent friction formulation.
Huemer’s friction law can be written as
α(T ) |P |n(T )−1 + β(T )
µ(v, P, T ) = b(T ) c(T )
(3)
a(T ) + v 1/m(T )
+ v 2/m(T )

and Savkoor’s friction law as

 −k(T )    
P 2 |v|
µ(v, P, T ) = µs (T ) + (µm (T ) − µs (T )) exp −h(T )2 log .
p0 (T ) Vmax (T )
(4)

A systematic parameter identification is then applied. The experimental data out of the linear friction
test on a dry safety walk and granite surface are used. At the end, a full set of parameter functions
in dependency of temperature is obtained. At this point, two major characteristics of the different
friction laws can be stated: Savkoor’s law uses parameters with clear and separable influence on the
shape of the friction field. This is advantageous if no or just some experimental data are available. The
parameters of Huemer’s friction law show a complex interaction requiring a suitable fitting algorithm.
The main advantage however, is the flexibility of the 3D-function enabling a good fitting over large
areas of the friction field. Fig. 3 clarifies this characteristics as the pronounced maxima with strong
gradients are not sufficiently representad by Savkoor’s friction law.
 stress-velocity correlation

τcrit

shear stress

tanh-function
ideal stick condition
penalty methode

γcrit
slip velocity

Fig. 5 Stress-motion functions

The influence of the contact pressure on the resulting friction coefficient is presented by Fig. 4. For
both models, the resulting friction follows a decreasing tendency at higher pressure. If contact pres-
sure reachs values close to zero, the friction coefficient gets considerably high due to the mathematical
definition of the models. In order to control this effect, the minimum contact pressure is limited by a
specific value Pl . Furthermore, the friction values near to zero pressure can be achieved by a linear
approximation as indicated by the dashed lines in Fig. 4.
The contact stiffness and stress in the footprint are directly related to the stress-slip definition. This
function controls the stick-slip state or the elastic and plastic slip for the penalty method. The de-
scribed frictional characteristic can be controlled using the FRIC subroutine [1]. Fig. 5 shows three
approaches to define stick and slip condition. For the ideal sticking condition, no relative motion oc-
cours, if the frictional shear stress τ < τcrit for τcrit = µ ∗ P . For this case, a very high initial contact
stiffness is assumed and a sudden change of stiffness occurs between sticking and sliding. These two
features lead to strong difficulties in numerical simulations. The penalty approach approximates the
condition of no sliding with a linear relation of elastic slip. If the value of τcrit is exceeded, a constant
plateau of sliding is assumend. Several functions exist to define this elastic region and it depends
on the problem description, numerical solution process, computational time and accuracy improve-
ment. One of the main studies of this research work is based on a new regularization to determine the
stress-slip velocity relationship, which is expressed by a continuous function

γ 
i
τi = tanh τcrit . (5)
ε

The tanh-function is characterized by an asymptotic expansion, adjustable gradient at the origin and
continuous shape. These characteristics make it suitable to describe the stress-motion relationship
which is identified as well by [10]. Fig. 5 shows, that the continuous regularization expresses a stiffer
frictional behaviour at small slip velocities. Furthermore, it converges to τcrit at higher values of slip
velocities. Therefore, a specific predetermined value for sticking and slipping is not present. The
obtained numerical results concerning this modification are presented in the next section.
a) b)

Fig. 6 Contact patch at maximum cornering: a) shear stress ; b) slip velocity for test tire

Longitudinal Forces
4000

3000

2000
longitudinal forces [N]

1000

-1000

-2000

-3000 Coulomb fric. law

Huemer fric. law
Savkoor fric. law
-4000
-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2
slip ratio [-]

Fig. 7 Resulting longitudinal force according to various applied friction laws. (Vx = 100 [km/h]; Fz = 3000 [N ])

3.2 Results

The results obtained for the frictional behaviour of a steady state rolling tire using FE simulation are
shown subsequently. The analyse is focused on local contact characteristics as well as the study of
global reaction forces. With the help of the user subroutine FRIC, the described advanced friction
approach can be implemented in ABAQUS. The code allows extended monitoring and output to
control, analyse and improve the formulation. One advantage of this contact model is a more robust
tire simulation enabling less computational time, larger force transmission from tire to road and larger
car velocity. Fig. 6 shows the contact patch for Coulomb’s and Huemer’s law for the same load and
speed configuration (maximum cornering). For Coulomb friction, the maximum shear stress can be
found in the boundary region of the contact patch. The nodes in this region alternate between contact
to no contact, hence they reach very low contact pressures and high slip velocities. These pronounced
slips can be well interpreted using the enhanced friction law and a more realistic contact patch can be
simulated. For a visco-elastic material description, the pronounced slips can be even better simulated.
Focusing on the resulting horizontal forces, Coulomb’s friction law shows a smaller longitudinal
stiffness in comparison to the enhanced approaches. The stiffness which is predicted by Savkoor’s
 Ve
3000
3000

rti
ca
2500

l lo
Cornering force [ N]

a
d[
2000

N]
2000

1500
6° 7° 8° 9° 10 ° 11 ° 12 ° 13 ° 14 ° 15 ° 16 ° 17 ° 18 °
5°
4°
n gle
pa

.74
3°
Sl i

9
09

65.8
1000

y 7
C=

.17
y 6
C=

92
2°

.65
y 5
C=

65
=4
y
C
Cornering st iffness, Cy [ N/ deg]
1°

0 0°

Fig. 8 Resulting cornering force characteristics according to Tanh method and Huemer’s friction law (Vx = 100 [km/h])

3000

Ve
3000

rti
ca
2500

l lo
Cornering force [ N]

a
d[
N]
2000
2000

1500
8° 9° 10 ° 11 ° 12 ° 13 ° 14 ° 15 ° 16 ° 17 ° 18 °
7°
6°
5°
n gle
1000 4° pa
3°
Sl i
Penalt y Huem er law
2° Penalt y Savkoor law

1° Penalt y Coulom b law

0 0°

Fig. 9 Resulting cornering force characteristics according to various applied friction laws (Vx = 100 [km/h])

and Huemer’s model is almost similar to each other. For small values of slip velocity around free
rolling, a dominant sticking situation in the contact patch can be assumed. For these slip velocities and
smooth pressure distribution in the contact patch, higher friction coefficients can be reached leading
to a stiffer behaviour. At full braking and acceleration, a significant difference between the values of
different friction laws is seen. Coulomb’s friction law follows a constant ratio instead the enhanced
friction approaches tendency is being decreased. Furthermore, the temperature in the contact area
is increasing between 10 to 20 ◦ C (depending on car velocity), because the whole contact patch can
be assumed as sliding. It can be stated, that the whole longitudinal force is clearly different for the
enhanced friction laws and a global friction coefficient, even if they reach the same maximum value.
The resulting lateral force based on the tanh formulation for the enhanced friction model is shown
in form of a carpet plot according to the applied vertical forces and slip angles in Fig. 8. According
to Fig. 8, for small slip angles the achieved cornering force approximately follows a linear tendency.
Because of the inertia effect and the capability of lateral deformation of the tire structure in balance
with deflecting lateral forces generated by increasing slip angle, an approximately linear trend of the
cornering function can be stated for small slip angles. The deflection is gradually increased up to a
transition point where the nodes in contact with the road increasingly start sliding. At this situation,
a nonlinear behaviour can be predicted until the plateau is reached and, hence, the friction in the
footprint plays the most important role in generation of cornering forces. The resulting cornering
stiffness based on various friction approaches is seen in Fig. 9. The enhanced friction models illustrate
a higher and stiffer behaviour for cornering due to higher friction coefficients for small slip angles.
The reason for a higher initial friction coefficients correlates with the motivation of the characteristics
of longitudinal forces: lower temperature, lower slip velocity and lower pressure. For higher slip
angle, a maximum cornering force is reached, which decreases with further increasing slip angle
Fig. 10 Footprint shape (left: experiment, right: simulation) and pressure distribution in lateral section (top right: experi-
ment, bottom right: simulation)

Fig. 11 Vertical stiffness for different global tire temperature

for the enhanced friction laws. In comparison, the approach including Coulomb’s global friction
coefficient remains constant for a lateral force:

Fy = µCoulomb ∗ Fz (6)

4 Validation

The quality of the simulation and input parameter can be confirmed by extensive validations. The
introduced tire has been tested on a flat track and a drum test rig for static and dynamic behaviour.
The following validations could be realised:
– unloaded, deformed tire contour including inner pressure,
– deformed sidewall during rolling,
– vertical stiffness,
– longitudinal forces,
– lateral forces and
– surface and inner temperature.
Validation provides informations about the tire behaviour which is necessary to interpret results and
local phenomena of the tire simulation. The footprint is one validation example as it represents the
major influence factors for tire performance focusing on the frictional behavior. The shape of the
a) b)

Fig. 12 Application of FE tire model: a) tire performance optimisation using neuronal networks ; b) analyse of rim forces

Fig. 13 Damage investigation

footprint is strongly influenced by fiber and sidewall stiffness as well as by tread material character-
istics. The pressure distribution and footprint size influence each other and represent important input
variables for the friction formulation. The rectangular footprint shape from experiments is in good
agreement with simulation results, as well as length and width. The contact pressure in the contact
patch is smoothly distributed and shows a characteristic maximum in the shoulder region influenced
by the sidewall stiffness. The distribution and values of the contact pressure from experiments and
simulations are in good agreement as shown in Fig. 10. The vertical stiffness for cold (25◦ C) and hot
(120◦ C) tire is shown in Fig. 11. compared to experiments, the cold tire shows larger deflection of
1.1 mm than the hot tire (0.3 mm). This acceptable difference is motivated by better material fitting
for hot rubber compounds as they show smoother tensile-compression change.

5 Outlook

The model described in Section 2 includes thermo-mechanical couping [2] and frictional contact as
explained in Section 3. Extensive validations have been done, so it can be expected that a confi-
dential prediction of tire performance, reaction forces, temperature developement and deformation is
possible. This can save time and large experimental efforts which are one reason for structural anal-
ysis in tire construction. One example for the use of this model is a procedure for tire performance
optimization using neuronal networks. Advanced aerodynamic investigations can be done with the
information of deformed tire shapes.
Beside this examples, tire simulation enables investigations which are difficult or impossible to test
with nowadays available measurements. Temperature prediction inside the tire structure, energy loss
during driving, damage investigation and crack prediction are examples that have been realised. An-
other example is the analysis of rim forces for different speed, load and camber configurations to
improve rim construction. To include high vertical loads, achieve stable simulations for large lon-
gitudinal and lateral forces (µmax > 2.0), it is essential to start simulations with high quality FE
models. High speed rolling analysis can be carried out with a base described above to exclude contact
singularities and take temperature and time depending effects into account.
 Acknowledgement: For the experimental results with respect to friction and footprint analyses, we
kindly acknowledge the contributions of Leibniz-Institut für Polymerforschung Dresden, Deutsches
Institut für Kautschuktechnologie and Forschungsgesellschaft Kraftfahrwesen Aachen.

References

[1] ABAQUS. Abaqus Analysis-User Manual 2012

[2] Behnke, R., Kaliske, M.: Simulation of inelatic steady state rolling tires including self-heating.
in: Deutsche simulia anwenderkonferenz , dresden (2014)

[3] Heinrich, G., Klüppel, M.: Rubber friction, tread deformation and tire traction. Wear 265,
1052–1060 (2008)

[4] Huemer, T., Liu, W., Eberhardsteiner, J., Mang, H.: A 3d finite element formulation describing
the frictional behavior of rubber on ice and concrete surfaces. Engineering Computations 18,
417–437 (2001)

[5] Kaliske, M.: Numerical modeling in tire mechanics. In: LS-DYNA Forum, Bamberg (2010)

[6] Matuško, J., Petrović, I., Perić, N.: Neural network based tire/road friction force estimation.
Engineering Applications of Artificial Intelligence 21, 442–456 (2008)

[7] Pinnington, R.: Rubber friction on rough and smooth surfaces. Wear 267, 1653–1664 (2009)

[8] Savkoor, A.: Some aspects of friction and wear of tyres arising from deformations, slip and
stresses at the ground contact. Wear 9, 66–78 (1966)

[9] Van der Steen, R.: Enhanced friction modeling for steady-state rolling tires. Ph.D. thesis, Eind-
hoven University of Technology, Department of Mechanical Engineering (2007)

[10] Wriggers, P.: Computational Contact Mechanics. Wiley & Sons, Chichester (2002)

View publication stats