From Tread Design To Tread Stiffness Matrices: F. Braghin, F. Cheli, and R. Sangalli

Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

F. Braghin,1 F. Cheli,2 and R.

Sangalli3

From Tread Design to Tread Stiffness


Matrices4

REFERENCE: Braghin, F., Cheli, F., and Sangalli, R., ‘‘From Tread Design to Tread
Stiffness Matrices,’’ Tire Science and Technology, TSTCA, Vol. 36, No. 3, July – September
2008, pp. 227-241.

ABSTRACT: The tread design, together with the tread compound, is responsible for the grip
characteristics of tires. The aim of this work is to develop an engineering tool that is able to
predict the stiffness distribution, i.e., the grip characteristics, of a given tread design without
the need of producing a prototype. This is done by combining a graphical preprocessor de-
veloped in Matlab that is able to convert a drawing into a mesh, and a commercial FEA code.
KEY WORDS: tread design, nonlinear stiffness, FEA analysis

In order to reliably simulate tire wear 共the same applies for tire-road fric-
tion兲, it is of fundamental importance to reproduce, as close to reality as pos-
sible, the real tire tread design.
If a 3D brush model is considered for the mapping of the tire rolling
surface, two different approaches can be followed to reproduce the real tire
tread design: either use brushes that have different heights or use brushes that
have different density. The second approach was followed since it was more
computationally efficient. The drawback of this approach is that the stiffness of
the brushes with respect to adjacent brushes and to the belts changes as a
function of the brush position. Thus, stiffness matrices have to be identified.
In the present research, a numerical identification procedure based on the
tire tread design and on the tire tread stiffness and damping characteristics 共that
change with frequency and preload 关1兴兲 is proposed. The tire rolling surface is
meshed with the so-called micromesh, having a minimum dimension of 0.1
⫻ 0.1⫻ 0.1 mm to be able to reproduce even the smallest details of the tread
design, and with the so-called macromesh, which corresponds to the adopted 3D
brush model. Then, through an automated procedure, a series of finite element
analyses 共FEAs兲 on each brush element and on the coupling of adjacent brush
elements is carried out and the equivalent stiffness with respect to the belts and
to adjacent brushes is determined.

1
Mechanical Engineering Department, Politecnico di Milano, Via La Masa 34, Milan 20158, Italy.
Electronic mail: [email protected]
2
Mechanical Engineering Department, Politecnico di Milano, Via La Masa 34, Milan 20158, Italy.
Electronic mail: [email protected]
3
Pirelli Tyres, PreDevelopment and Research, Viale Sarca 222, Milan 20124, Italy. Electronic mail:
[email protected]
4
Presented at VSDIA 2006, the Tenth Mini-Conference on Vehicle System Dynamics, Identification
and Anomalies, Budapest, Hungary, 6–8 November 2006.

227
228 TIRE SCIENCE AND TECHNOLOGY

FIG. 1 — Storage modulus as a function of frequency and temperature (at constant prestrain and
strain amplitude).

Note that the same procedure can be applied to determine the damping
characteristics of the tire tread design. However, transient FEA analyses are
required for the correct identification of damping characteristics. It was there-
fore decided to postpone these analyses due to the required computational
effort.

The Mechanical Properties of Tread Rubber Compound

As is well known, rubber mechanical characteristics are a function of sev-


eral quantities such as the preload, the load amplitude, the excitation frequency,
and the working temperature. Figures 1 and 2 show, as an example, the storage
and loss moduli of the considered tread compound as a function of the excita-
tion frequency and working temperature 共keeping prestrain and strain amplitude
constant兲. It can be clearly seen that, at increasing frequency and decreasing
temperature, both storage and loss moduli increase.
Using the WLF transformation 关2兴 it is possible to combine the various
curves shown in Figs. 1 and 2 into two curves, one for the storage modulus and
one for the loss modulus, at a given temperature 共usually ambient temperature
or 70 ° C for car tire tread compound兲. These two curves represent the mechani-
cal characteristics of the tire tread compound at that temperature.
BRAGHIN ET AL. ON TREAD DESIGN TO TREAD STIFFNESS 229

FIG. 2 — Loss modulus as a function of frequency and temperature (at constant prestrain and strain
amplitude).

To be able to introduce these material properties into the selected commer-


cial FEA code 共ABAQUS version 6.3 关3兴兲, it was decided to fit the experimental
curves using a generalized linear Maxwell viscoelastic model 共Fig. 3兲 since
such a model is fully equivalent 关4兴 to a viscoelastic model already imple-
mented into the code 共Prony series兲:

FIG. 3 — Generalized linear Maxwell viscoelastic model.


230 TIRE SCIENCE AND TECHNOLOGY

FIG. 4 — Experimental (thin line) and identified (thick line) storage (upper graph) and loss (lower
graph) moduli at 50 ° C (5% of strain amplitude).

N
r2i ␻2ki
E ⬘共 ␻ 兲 = k ⬁ + 兺
i=1 k2i + r2i ␻2
共1兲
N
ri␻k2i
E⬙共␻兲 = 兺
i=1 k2i + r2i ␻2

where ki and ri are the stiffness and damping values of the ith Maxwell element
and k⬁ is the nonlinear hyperelastic stiffness value in parallel with all other
Maxwell elements.
Note that the present version of ABAQUS 共version 6.7兲 allows the charac-
terization of the mechanical properties of rubber compounds in the frequency
domain also in tabular format, thus avoiding the need to identify the coefficients
of a Prony series. Using eight linear Maxwell viscoelastic elements in parallel,
it is possible to fit the experimental storage and loss curves in the frequency
range of interest, i.e., from 0 to 600 Hz 共Fig. 4兲. Better identification results
BRAGHIN ET AL. ON TREAD DESIGN TO TREAD STIFFNESS 231

could have been achieved using a higher number of elements, but the compu-
tational effort would have increased significantly.
Instead of E⬘ and E⬙ vs frequency, the graphs shown in Fig. 4 represent G⬘
and G⬙ vs frequency. To pass from E⬘ 共E⬙兲 to G⬘ 共G⬙兲, the following relations
hold:

E⬘ E⬘
G⬘ = =
2共1 + ␯兲 3
共2兲
E⬙ E⬙
G⬙ = =
2共1 + ␯兲 3
where ␯ is Poisson’s ratio. It should be observed that, going from 0 Hz 共static
deformation of the rubber element兲 to 200 Hz, the stiffness properties of rubber
increase more than 60% while the damping properties increase almost 200% at
50 ° C. This demonstrates the importance of correctly determining the rubber
characteristics as a function of frequency and temperature.

Tread Design Preprocessor

After having characterized the tread rubber mechanical properties, it is


possible to start with the identification of the stiffness characteristics of the tread
design. The first step is to import the tread design from a 2D drawing. Then, the
tread depth is introduced, thus obtaining a 3D model of the tread design. Fi-
nally, a FEA preprocessor enables the splitting of the 3D model into single
brushes and triplets of brushes, meshes them, and prepares input files for
ABAQUS analysis.

Importing the Tread Drawing


Tire manufacturers usually use 2D CAD software for the design of the tread
共Fig. 5兲. To be able to develop a solid model of the tread for FEA simulations,
two main steps are necessary: before importing the drawing into the preproces-
sor, the tread design has to be isolated; once imported, the tread depth has to be
introduced. The tread depth is introduced through a color scale as shown in Fig.
6. Note that some preliminary operations on the 2D tread drawing are usually
necessary such as the closing of areas.

Meshing the Tread Design


After the import phase of the preprocessor is completed, a solid model of
the tread design is available. This model is meshed using 8-node linear brick
elements 关3兴. Two types of mesh are created: a refined mesh, from now on
called micromesh, whose dimensions can be as small as 0.1⫻ 0.1⫻ 0.1 mm in
232 TIRE SCIENCE AND TECHNOLOGY

FIG. 5 — Original 2D drawing provided by tire manufacturers.

order to be able to describe even the smallest details of the tread design, and a
coarse mesh, from now on called macromesh, whose dimensions correspond to
the dimensions of the 3D brush model that is used to schematize the tire rolling
surface.
Note that the dimensions of the micromesh significantly affect the com-
plexity of the FEA model: for a tread design 220 mm wide, 210 mm long, and
10.5 mm high, a micromesh having dimensions equal to 0.1⫻ 0.1⫻ 0.1 mm
leads to a FEA model of the whole tread design with almost 500 million ele-
ments. Thus, a 64-bit supercomputer is necessary. It is therefore of fundamental
importance to carry out a preliminary sensitivity analysis on the micromesh
dimensions considering, for example, a single macroelement.
To speed up the meshing procedure, a graphical user interface in the Matlab
environment has been developed 共Fig. 7兲. This interface allows the easy modi-
fication of both the micro- and macromesh dimensions, as well as immediate
visualization of the effects of the changing of the mesh dimensions. The output
of the interface are the input files for FEA simulations 共at the moment, the input
file format is for ABAQUS but the environment is very general and any input
file format could be easily implemented兲 to assess the stiffness distribution of
BRAGHIN ET AL. ON TREAD DESIGN TO TREAD STIFFNESS 233

FIG. 6 — Tread drawing with tread depth in gray scale.

the considered tread design. Thus, material hyper- and viscoelastic characteris-
tics of the tread rubber compound have to be linked 共see paragraph 2兲 and loads
and boundary conditions have to be applied.
For boundary conditions, it was decided to fix the upper surface of the
tread共s兲, i.e., the tire rolling surface, to reduce the likeliness of buckling of the
fine parts that are due to the cutting of the tread design into rectangular shaped
macroelements.
Instead, loads 共concentrated loads on mesh nodes兲 are applied to lower
surface, i.e., in correspondence of the connection to the liner, of the tread under
investigation. In fact, four different tread configurations are considered to de-
termine the 3D stiffness matrices of the tread design:

1. a single macroelement 共Fig. 8兲;


2. three subsequent macroelements in the longitudinal direction where the
tread element under investigation is the middle one 共Fig. 9兲;
3. three subsequent macroelements in the transversal direction where the
tread element under investigation is the middle one 共Fig. 10兲;
4. five macroelements with the tread element under investigation at the
center and the two adjacent macroelements in the longitudinal and in
234 TIRE SCIENCE AND TECHNOLOGY

FIG. 7 — MATLAB GUI to speed-up the meshing procedure.

FIG. 8 — Single macroelement (1: longitudinal direction; 2: transversal direction; 3: vertical


direction).
BRAGHIN ET AL. ON TREAD DESIGN TO TREAD STIFFNESS 235

FIG. 9 — Three longitudinally coupled macroelements (1: longitudinal direction; 2: transversal


direction; 3: vertical direction).

FIG. 10 — Three transversally coupled macroelements (1: longitudinal direction; 2: transversal


direction; 3: vertical direction).
236 TIRE SCIENCE AND TECHNOLOGY

FIG. 11 — Five cross-coupled macroelements (1: longitudinal direction; 2: transversal direction; 3:


vertical direction).

the transversal direction 共Fig. 11兲; it was found that this fourth con-
figuration did not add any additional information to the 3D stiffness
matrices with respect to the previous three, and it was therefore de-
cided not to carry out the FEA simulations involving these configura-
tions to reduce the computational effort.

Load steps are applied as follows: at first, a normal load is applied to


reproduce a uniform contact pressure; then, a 共small兲 longitudinal distributed
force is applied, followed by a 共small兲 transversal distributed force and finally
by a 共small兲 normal distributed force 共Fig. 12兲. These load steps are repeated for
any considered contact pressure.
Since the aim of these load steps is to determine the linear elastic behavior
of the macroelement around the configuration determined by the applied contact
load, the applied force in any of the three directions should be lower than 0.1 N
for the considered tread compound. Higher values would result in a nonlinear
characteristic curve 共applied force vs resulting displacement兲 of the tread ele-
ment considered. Thus, the hardening characteristic of rubber is neglected in the
3D brush model.
BRAGHIN ET AL. ON TREAD DESIGN TO TREAD STIFFNESS 237

FIG. 12 — Applied loads on the three longitudinally coupled elements (1: longitudinal direction; 2:
transversal direction; 3: vertical direction).

Identification of the Stiffness Characteristics

Once FEA simulations have been carried out, through simple postelabora-
tion of the FEA results carried out through a Python script, it is possible to
extract the stiffness of each macroelement with respect to the liner and with
respect to adjacent elements. The stiffness kij of each ijth macroelement with
respect to the liner 共in both longitudinal, transversal and normal directions兲 is
immediately determined from FEA simulation of the single macroelement,

Fij
kij = 共3兲
dij

Fij being the applied force along the considered direction and dij being the
resulting mean displacement of the lower tread element surface along that di-
rection. The stiffness of each macroelement with respect to the adjacent ele-
ments requires the solution of a linear system. In fact, the force applied to a
macroelement when coupled to adjacent elements determines a deformation of
all elements 共Fig. 13兲. Thus, the following equations hold for longitudinally FijL
and transversally FijT coupled elements, respectively:
238 TIRE SCIENCE AND TECHNOLOGY

FIG. 13 — Scheme of the load application of longitudinally and transversally coupled


macroelements.

FijL = kijdij + ki−1jdi−1j + ki+1jdi+1j + ki÷i−1j共dij − di−1j兲 + ki÷i+1j共dij − di+1j兲


共4兲
FijT = kijdij + kij−1dij−1 + kij+1dij+1 + kij÷j−1共dij − dij−1兲 + kij÷j+1共dij − dij+1兲
where ki÷i−1j, ki÷i+1j, kij÷j−1, and kij÷j+1 are the unknown mutual stiffness values
in the longitudinal and transversal directions, respectively 关the forces are im-
posed, the stiffness values kij have been determined through Eq 共3兲, and the
displacements dij, di−1j, di+1j, dij−1, and dij+1 of the macroelements are deter-
mined by taking the mean displacement of the nodes at the lower tread element
surface along the force application direction兴. Considering one single macroele-
ment, we have four unknown stiffness values and just two equations.
However, in the case of transversal coupling, macroelements at the tire
boundaries will have just one adjacent macroelement. Therefore, the second Eq
共4兲 will have only one unknown that can be easily determined, thus allowing the
determination of all kij÷j+1 values one after the other.
A similar reasoning allows us to determine ki÷i+1j values: considering the
macroelement i + 1, we can add an equilibrium equation similar to the first Eq
共4兲 with the additional unknown ki+1÷i+2j 共the stiffness value ki+1÷ij is equal to
the stiffness value ki÷i+1j兲. After one entire wheel turn we will be able to write
an equilibrium equation similar to the first Eq 共4兲 with no additional unknown.
In fact, we have written n equilibrium equations 共n being the number of mac-
BRAGHIN ET AL. ON TREAD DESIGN TO TREAD STIFFNESS 239

FIG. 14 — Longitudinal force distribution at three different displacement values and with a uniform
contact pressure equal to 0.92 bar.

roelements along the tire longitudinal direction兲 in n unknown stiffness values


that can thus be determined solving a simple matrix equation.
Figure 14 shows, as an example, the force distribution due to the applica-
tion of a uniform longitudinal displacement equal to 0.5, 1.0, and 2.0 mm and
with a contact pressure equal to 0.92 bar. Due to confidentiality reasons no
force values are reported. However, the global reaction forces are equal to
1485.97, 2971.95, and 5943.89 N, respectively. To validate these results, FEA
simulations of the whole tread design were carried out. While the stiffness
matrices were calculated using a micromesh of 0.4 mm in the longitudinal
direction, 0.3 mm in the lateral direction, and 0.61 mm in the vertical direction,
it was impossible to carry out a FEA simulation of the whole tread design with
a mesh having such dimensions 共it would have had almost 4 million elements兲.
Thus, the FEA simulation of the whole tread design was carried out with a mesh
of 1.9 mm in the longitudinal direction 共rolling direction兲, 1.5 mm in the lateral
direction, and 0.61 mm in the vertical direction 共radial direction兲. The obtained
global reaction forces are equal to 1652.70, 3312.73, and 6654.35 N, respec-
tively. The reaction forces determined through the FEA model of the whole
tread design are therefore higher than the corresponding forces determined
through the stiffness matrices by approximately 10%. This may be due to the
different meshes adopted. Note that the tire tread design is clearly visible.
Similarly, Fig. 15 shows the force distribution due to the application of a
uniform transversal displacement equal to 0.5, 1.0, and 2.0 mm and with a
contact pressure equal to 0.92 bar. Again, due to confidentiality reasons no force
240 TIRE SCIENCE AND TECHNOLOGY

FIG. 15 — Transversal force distribution at three different displacement values and with a uniform
contact pressure equal to 0.92 bar.

values are reported but the global reaction forces are equal to 1314.58, 2629.16,
and 5258.33 N, respectively. These values are smaller than the corresponding
values in the longitudinal direction as expected. In this case, the global reaction
forces obtained through the FEA model of the whole tread design are equal to
1532.82, 3087.25, and 6220.53 N, respectively, thus almost 15% higher than
those determined with the stiffness matrices. Again, this difference may be due
to the different meshes adopted and may be considered as acceptable in most
cases, especially when comparing the simulation times.

Conclusion

The stiffness matrices of the 3D brush model with respect to the belts and
to the adjacent brushes have been identified through a FEA model. These ma-
trices take into account the influence of the preload due to pressure and of the
load amplitude, thus allowing evaluation of the tire grip and wear characteristics
共using the PaRRT model 关4–6兴兲 due to its tread design.
A powerful pre- and post-processing tool has been developed to speed up
the meshing and stiffness extraction operations. However, too much computa-
tional time is still required. Further improvements in the direction of reducing
the number of FEA simulations are necessary.
Further research is being carried out in order to assess the corresponding
damping matrices.
BRAGHIN ET AL. ON TREAD DESIGN TO TREAD STIFFNESS 241

Acknowledgments
The authors wish to gratefully acknowledge Pirelli Tires for having funded
this research project.

References

关1兴 Yoshida, J., Abe, M., and Fujino, Y., “Constitutive Model of High-Damping Rubber Materials,”
Journal of Engineering Mechanics, Vol. 130, No. 2, 2004, pp. 129–141.
关2兴 Ferry, J. D., Landel, R. F., and Williams, M. L., “Extensions of the Rouse Theory of Viscoelas-
tic Properties to Undiluted Linear Polymers,” Journal of Applied Physics, Vol. 26, No. 4, 1955,
pp. 359–362.
关3兴 ABAQUS Theory Manual.
关4兴 Lupker, H., Cheli, F., Braghin, F., Gelosa, E., and Keckman, A., “Numerical Prediction of Car
Tire Wear,” Tire Science and Technology, TSTCA, Vol. 32, No. 3, 2004, pp. 164–186.
关5兴 Braghin, F., Cheli, F., Melzi, S., and Resta, F., “Tyre Wear Model: Validation and Sensitivity
Analysis,” Meccanica, Vol. 41, No. 2, 2006, pp. 143–156.
关6兴 Braghin, F., Cheli, F., Scaltritti, D., and Tomasini, G., “Rigid-Ring Tyre Model: Out-of-Plane
Parameter Identification from Experimental Tests on Rolling Tyre,” Proceedings of the 8th
VSDIA Conference on Vehicle System Dynamics Identification and Anomalies, Budapest, Hun-
gary, 2002.

You might also like