Professional Documents
Culture Documents
Iyengar3
REFERENCE: Song, G., Conard, B., and Iyengar, S. K. R., ‘‘Damping Characterization
Using Hysteresis on Static Nonrolling and Dynamic Rolling Behavior of Farm Tires,’’
Tire Science and Technology, TSTCA, Vol. 36, No. 2, April – June 2008, pp. 108-128.
KEY WORDS: tire, hysteresis, rubber, elastomer, Abaqus, finite element, hyperelastic,
damping, characterization, energy dissipation, rolling, implicit, explicit
1
Corresponding/presenting author. Dassault Systemes Simulia Corp., West Lafayette, Indiana
47906, USA. Electronic mail: guiyong.song@3ds.com.
2
Dassault Systemes Simulia Corp., West Lafayette, Indiana 47906, USA. Electronic mail:
brett.conard@3ds.com.
3
John Deere Product Engineering Center, Waterloo, Iowa 43840, USA. Electronic mail:
IyengarSeshadriKR@JohnDeere.com.
4
Presented at the twenty-sixth annual meeting of the Tire Society, Akron, Ohio, September 25–26,
2007.
108
SONG ET AL. ON DAMPING CHARACTERIZATION 109
Introduction
Farm tires differ from common automotive tires in several ways because
the application requirements are very different. At very low inflation pressures
and speeds tractor tires behave quite differently from automotive on-highway
tires. Automotive on-highway tire models are typically unsuitable for the sus-
tained high-torque conditions encountered by tractor tires for field operations.
The biggest difference between a tractor and an on-road vehicle is that most
tractors do not have a suspension system that is commonly found in an on-road
vehicle. Therefore, farm tires are required to provide more energy dissipation
than is common on-road vehicle tires. Farm tires meet this requirement by
introducing more material damping and thus provide ride comfort to farm
equipment to some degree. Tire slippage, wear, and energy dissipation are what
cushion the rest of the vehicle system. In deformable soil, the terrain also
provides damping to the vehicle system. In general, hard terrain operations
result in more severe loads than soft soil operations.
Finite element methods have been widely used in the tire industry for tire
simulations, including rolling simulation, temperature prediction, etc. Tire hys-
teresis resulting from viscoelastic properties of tire compounds and cords has
long been recognized as a principal cause of tire rolling resistance 关1兴. Quasi-
static tire-rolling simulations with hysteresis effect using a static finite element
method are reported in Refs. 关2,3兴.
All conventional tire models need nonrolling and rolling tests to establish
parameter values such as stiffness to put into models. For tires in the conceptual
stage, these are either guesses and/or extrapolations from similar tires. John
Deere has been measuring chassis accelerations over known terrains, e.g., In-
ternational Standards Organization 共ISO兲 tracks, Dewy ditch, etc., for 20 years
and tried to correlate models with measurements 关4兴. More recently wheel trans-
ducers have been used to measure hub forces and moments which are then used
to assess models. Significant efforts have been made to ensure the repeatability
of test runs. After reviewing 20 years of efforts, however, it is concluded that
inadequate tire characterizations are the major reason for the relatively poor
agreement between mathematical models and measurements for severe events.
Few publications on simulations of severe dynamic events such as tire
bouncing on/off ground are available in the literature. This paper presents the
characterization of damping behavior as characterized by hysteresis for farm
equipment tires and time domain numerical simulation of off-road tire rolling
using Abaqus finite element software. Both static nonrolling and dynamic
rolling/bouncing simulations are conducted using a hysteresis model available
in Abaqus/Standard and a newly developed Abaqus/Explicit hysteresis model
based on a VUMAT user subroutine.
110 TIRE SCIENCE AND TECHNOLOGY
F = FBeFBcr . 共1兲
SB
FBeḞBcrFBcr−1FBe−1 = ˙ Bcr , 共2兲
B
where ˙ Bcr is the effective creep strain rate in network B, SB is the Cauchy stress
deviator tensor in network B, and B is the effective stress in network B. The
effective creep strain rate in network B is given by the expression
S = 1.6,
5
A= 共s兲−1 共MPa兲−m ,
共冑3兲m
m = 4,
C = − 1.0,
E = 0.01. 共4兲
To verify the new hysteresis model, uniaxial tension and simple shear simu-
lations were conducted to compare the behavior of the explicit dynamics hys-
teresis model with implicit hysteresis model under quasistatic loading condi-
tions. The simple uniaxial tension and simple shear simulations were performed
at a cycling frequency of 0.1 Hz. A single-element unit-cube model with unity
dimensions and unity volume was used to test the hysteresis behavior. The
explicit dynamics unit cube model was the same as the implicit dynamics model
except that the hysteresis material models are different. A sinusoidal extension
displacement was applied to one face in the uniaxial tension model. A sinusoidal
shear displacement was applied to one edge in the shear model. The correspond-
ing reaction forces were recorded. The forces and displacements were plotted on
the same graph to form hysteresis loops. The displaced shape of the uniaxial
extension model and the comparison of force-displacement curves from implicit
and explicit solutions are shown in Figs. 2 and 3, respectively. A similar dis-
placed shape and a force-displacement curve comparison for the simple shear
model are shown in Figs. 4 and 5, respectively. As can be seen, the results
compare quite favorably for these simple modes of deformation at low cycling
frequencies where inertia effects are minimal.
Farm tires exhibit damping in cyclic loading. The damping is primarily due
to the hysteresis of the tire constituent elastomeric compounds. Tire tests can
provide the overall hysteretic behavior without having to test individual mate-
rials. Hysteretic load-displacement curves from nonrolling tests can be used to
calibrate the hysteresis material model for finite element analyses, whereas the
114 TIRE SCIENCE AND TECHNOLOGY
FIG. 3 — Comparison of force-displacement loops for unit-cube uniaxial tension models using
Abaqus/Standard and Abaqus/Explicit.
FIG. 5 — Comparison of force-displacement loops for unit-cube simple shear models using Abaqus/
Standard and Abaqus/Explicit.
displacement curves obtained from the nonrolling test constitute the basis for
characterization of hysteresis material models.
The tires were then rolled off a step to test the energy dissipations. The
emphasis was on large nonlinear displacements. Figure 7 shows a trailer vehicle
used for this test. The transient history data of vertical acceleration, hub load,
and vertical displacement were recorded at the wheel center using sensors.
These data are also required to characterize the hysteresis model parameters.
A simplified finite element model was used to represent the farm tire. The
model features solid elements strengthened by rebars for tire representation and
fluid elements for pressurization. Rigid surfaces represented the ground in non-
rolling simulations and the vertical step in drop-off dynamic simulations. A
polynomial hyperelastic material model was used to define the strain energy
potentials for the rubber components in the tire.
Nonrolling Simulations
Nonrolling finite element analyses were performed to simulate the nonroll-
ing tests of farm tires. The hysteretic load-displacement curves from the non-
rolling tests were used to calibrate the material constants in the hysteresis ma-
terial model for the rubber compounds in the tire model.
Figure 8 shows the setup of a nonrolling model. A rigid surface represents
the test stand. Contact constraint was defined at the interface between the tire
and the rigid surface. The nonrolling simulation was conducted using an im-
plicit solution and consisted of two steps. First, an inflation pressure was ap-
plied to the tire interior to pressurize it. Then a sinusoidal displacement bound-
ary condition was applied to the tire to move it up and down against the rigid
surface.
SONG ET AL. ON DAMPING CHARACTERIZATION 117
Three frequencies were analyzed with the finite element model—0.1, 5, and
10 Hz. The only thing different among the three cases was the cycling fre-
quency. The inflation pressure and the boundary conditions were the same for
all three cases. After adding the hysteresis definition, the hyperelastic materials
became strain-rate-dependent. Therefore, the step time specified for the static
general procedure became physically meaningful. It was directly related to the
loading frequency. To obtain the same number of cycles for different frequen-
cies, the step time must be changed relative to frequency. The step time for the
0.1, 5, and 10 Hz cases, were 30, 0.6, and 0.3 sec, respectively 共three cycles
each兲.
By adjusting the material constants in the hysteresis material model, very
good correlations between the nonrolling analyses results and test results were
obtained. It was found that the following set of constants gives a good fit with
the test results in the 0.1, 5, and 10 Hz cases: S = 2, A = 5.55555556⫻ 10−21, m
= 3.5, C = −1, and E = 0.01. A comparison of the analysis load-displacement re-
sults with the test data at 0.1 Hz is shown in Fig. 9. The comparisons at 5 and
10 Hz are shown in Figs. 10 and 11, respectively.
The correlations of analysis results with test data are very good at 0.1 and
5 Hz. The comparison is reasonably good at 10 Hz in the sense that the analysis
curve envelopes the test curve. It should be noted that test data at 10 Hz is
difficult to achieve, and thus the test curve provided at 10 Hz does not have
high fidelity.
A 5 in. vertical step was used to test the tire and simulate the drop-off test.
The curb was modeled as a rigid surface in the finite element model. The
drop-off simulation was conducted using implicit dynamics because the tire
model size and the deformation were moderate. It consisted of two steps. In step
1, the tire was pressurized using fluid elements and a footprint was established
through the contact between the tire and ground. In step 2, a rotational velocity
was applied to the tire center to make it roll on the ground.
The same hysteresis model constants from nonrolling simulations were
used in the implicit dynamic procedure to simulate the rolling of the tire. Figure
12 shows the initial and the intermediate configurations of the tire rolling on a
surface with a 5 in. step. Due to the hysteresis of the hyperelastic materials, the
vertical reaction force at the road reference node resulting from the drop-off
gradually decreases as the tire is rolling and bouncing on the road surface with
constant line speed.
Two translational speeds were analyzed—2.4 m / s 共5.2 mph兲 and 1.1 m / s
共2.3 mph兲. Figure 13 compares the vertical acceleration at the wheel hub from
the analysis and that measured in the test for the 2.4 m / s case. The comparison
of the vertical reaction force measured at rigid ground reference node for the
2.4 m / s case is shown in Fig. 14. The reaction force includes the inertia effect
in the dynamic analyses. At the first few cycles, the correlation for both accel-
eration and vertical reaction force are good. After a longer period of time the
analysis results appear to be slightly overdamped compared to the test results.
Similar observations were made for the 1.1 m / s case.
Since the parameters result in excessive damping, they need to be modified
to improve the correlation between the analysis and test results. After a few
iterations, it was found that the following set of hysteresis constants gives rise
to better correlation for vertical acceleration and vertical reaction force for the
rolling tire simulations: S = 10, A = 9.55555556⫻ 10−15, m = 2.9, C = 0, and E
= 0.01.
FIG. 12 — Initial and intermediate configurations of tire rolling off a 5 in. step.
120 TIRE SCIENCE AND TECHNOLOGY
FIG. 13 — Comparison of vertical accelerations between drop-off analysis and drop-off test at
2.4 m / s line speed.
The comparison of the acceleration and vertical reaction force for the
2.4 m / s case are shown in Figs. 15 and 16, respectively. As seen in the figures,
the correlation between analysis and test results is very good in both cases.
Good correlation is also observed for the 1.1 m / s line speed case.
FIG. 14 — Comparison of reaction forces between drop-off analysis and drop-off test at 2.4 m / s
line speed.
SONG ET AL. ON DAMPING CHARACTERIZATION 121
FIG. 15 — Comparison of vertical accelerations between drop-off analysis using recalibrated pa-
rameters and drop-off test at 2.4 m / s line speed.
from high vertical accelerations. A target was set to attenuate the acceleration
after the first bounce following the drop-off by 50%. This is expected to sig-
nificantly improve ride performance of the farm equipment.
The tire used for these simulations is larger than that used in the nonrolling
simulations and the drop-off simulations in previous sections. First, three dif-
ferent inflation pressures were analyzed using implicit dynamics at the same
cycling frequency of 0.1 Hz to study the effect of inflation on hysteresis behav-
ior under quasistatic cyclic loading conditions. The hysteresis parameters de-
rived in the previous section were chosen as the starting parameters for the
FIG. 16 — Comparison of reaction forces between drop-off analysis using recalibrated parameters
and drop-off test at 2.4 m / s line speed.
122 TIRE SCIENCE AND TECHNOLOGY
Stress
Set of hysteresis scaling Effective stress Creep strain
parameters factor S Creep parameter A exponent m exponent C
Baseline from 2.0 5.5555556⫻ 10−21 3.5 −1.0
calibration of the small
tire
Calibrated from 2.0 2.07621823269⫻ 10−15 2.5 −1.0
nonrolling of large tire
at 160 kPa
Recalibrated from 30.0 3.46680637175⫻ 10−36 6.0 −1.0
large tire drop-off
simulation
calibration because they gave very good correlation to the test data of the
smaller tire. The starting parameters are listed in the first row of Table 1. These
parameters were first used in the large tire nonrolling, sinusoidal loading im-
plicit simulations at 160 kPa to determine the hysteresis response. They did not
correlate well to the test data because the tire was different in both size and
compounds.
Based on the experience gained from the nonrolling and drop-off simula-
tions, the stress exponent m and the parameter A were slightly modified to
obtain better correlation to the nonrolling test hysteresis behavior at 160 kPa.
After several test runs, a set of new hysteresis parameters resulted in perfect
correlation to the test data at 160 kPa, as shown in Fig. 17. The new parameters
are listed in the second row of Table 1.
FIG. 17 — Comparison of 0.1 Hz analysis load-displacement loops with those of nonrolling tests at
83, 160, and 241 kPa inflation pressures.
SONG ET AL. ON DAMPING CHARACTERIZATION 123
FIG. 18 — Initial and intermediate configurations of the large tire rolling off a 5 in. step.
FIG. 19 — Comparison of vertical accelerations between drop-off analysis using recalibrated pa-
rameters and baseline model at 160 kPa inflation pressure.
ous section and the recalibrated parameters—at 160 kPa inflation pressure.
Similar comparisons for 83 and 241 kPa are shown in Figs. 20 and 21, respec-
tively. The vertical acceleration was reduced 51.0% after the first bounce fol-
lowing the drop-off at 160 kPa inflation pressure. At 83 kPa inflation pressure,
the vertical acceleration was attenuated 57.5%. The attenuation was 46.3% at
241 kPa inflation pressure.
FIG. 20 — Comparison of vertical accelerations between drop-off analysis using recalibrated pa-
rameters and baseline model at 83 kPa inflation pressure.
FIG. 21 — Comparison of vertical accelerations between drop-off analysis using recalibrated pa-
rameters and baseline model at 241 kPa inflation pressure.
126 TIRE SCIENCE AND TECHNOLOGY
accelerations. The dissipated energy increased 38.2% at 160 kPa inflation pres-
sure. The increases at 83 and 241 kPa inflation pressures are 23.5 and 19.2%,
respectively.
Acknowledgments
The authors thank engineers at the John Deere PEC and Moline Technology
Innovation Center who have shared their test data and experiences. Thanks also
go to John Deere management who granted permission to disseminate this
information and SIMULIA management who supported this effort.
References
关1兴 Walter, J. D., and Conant, F. S., “Energy Losses in Tires,” Tire Science and Technology,
TSTCA, Vol. 2, No. 4, 1974, pp. 235–260.
关2兴 Luchini, J. R., Peters, J. M., and Arthur, R. H., “Tire Rolling Loss Computation with the Finite
Element Method,” Tire Science and Technology, TSTCA, Vol. 22, No. 4, 1994, pp. 206–222.
关3兴 Shida, Z., Koishi, M., Kogure, T., and Kabe1, K., “A Rolling Resistance Simulation of Tires
Using Static Finite Element Analysis,” Tire Science and Technology, TSTCA, Vol. 27, No. 2,
1999, pp. 84–105.
关4兴 Kazempour, A. S., Volfson, B. P., and Iyengar, S. K. R., “On the Development and Verification
of High-Fidelity Agricultural Tire Models,” SAE Paper No. 2002-01-1467.
关5兴 Bloch, R., Chang, W. V., and Tschoegl, N. W., “The Behavior of Rubberlike Materials in
Moderately Large Deformations,” Journal of Rheology, Vol. 22, 1978, pp. 1–32.
关6兴 Ferry, J. D., Viscoelastic Properties of Polymers, John Wiley & Sons, Inc., New York, 1980.
关7兴 Bergstrom, J. S., and Boyce, M. C., “Constitutive Modeling of the Large Strain Time-
Dependent Behavior of Elastomers,” Journal of the Mechanics and Physics of Solids, Vol. 46,
No. 5, 1998, pp. 931–954.
关8兴 SIMULIA, “Abaqus Theory Manual,” Abaqus Version 6.7 Documentation, 2007.