Professional Documents
Culture Documents
Popio2
REFERENCE: Luchini, John R. and Popio, James A., ‘‘Modeling Transient Rolling Resis-
tance of Tires,’’ Tire Science and Technology, TSTCA, Vol. 35, No. 2, April – June 2007, pp.
118-140.
ABSTRACT: The transient rolling resistance of several tires is predicted from the behavior of
each tire on an equilibrium test. The objective of the study was to determine if model predic-
tions would fall within the lab-to-lab experimental error described in the J-2452 test standard.
The model used for this study, to predict the results of J-2452 from J-1269 data, was presented
in Ref. 关1兴. The study used data from one tire to establish the model parameters. Then those
parameters were used with tire and rolling resistance measurements from 12 other tires of
another size. These tires were from four samples of each of three different tire constructions
共tread patterns兲 and manufacturer. The reasons for the differences between the predictions and
the test results were investigated and two significant sources of error were identified. First, the
simplistic assumption that one set of heat-transfer parameters would apply to all tires was
found to be inadequate. Second, the matrix of test conditions and regression model in J-1269,
which must be used to extrapolate to the test conditions of J-2452, is inadequate. The model-
ing work in this article also found some of the details of the test protocols that may have
significant effects on reported rolling resistance.
KEY WORDS: tire rolling resistance, tire rolling loss, transient coastdown testing, SAE
J-1269, SAE J-2452
The theoretical approach, and principal reference, used in this work was
previously published 关1兴. A good summary of research on tire rolling resistance
through 1980 may be found in Schuring’s monograph 关2兴 which was substan-
tially revised for 1990 by Schuring and Futamura 关3兴. A later review of mea-
surement and modeling technology was presented for the Department of Energy
and at the International Tire Exhibition and Conference in 2000 关4兴.
Background
In 1998 a theoretical method for using equilibrium test data from SAE
J-1269 to model transient tire rolling resistance was presented by Mars. At the
time of the presentation, the SAE J-2452 test had not yet been established. The
technique was developed from an understanding of the known behavior of tire
1
Cooper Tire and Rubber Company, Findlay, Ohio 45840. Electronic mail:
JRLuchini@CooperTire.com
2
Smithers Scientific Services, Ravenna, Ohio 44266. Electronic mail: JPopio@Smithersmail.com
3
Presented at the twenty fifth annual meeting of The Tire Society, Akron, Ohio, September 11–12,
2006.
118
LUCHINI AND POPIO ON MODELING TRANSIENT ROLLING 119
rolling resistance and basic thermodynamics 关5–9兴. The Mars model uses tire
measurement data, such as diameter and mass, as well as data from an equilib-
rium rolling resistance test of a tire. This information is then used in a thermal
model to predict the transient rolling resistance of the tire for an arbitrary
history.
This article will use the Mars model, with the established SAE procedures,
to evaluate how well it can predict transient rolling resistance from measured
equilibrium rolling resistance.
Motivation
The model is constructed with the assumption that the transient rolling
resistance is a function of only the steady-state rolling resistance 共at given
instantaneous operating conditions兲 and the instantaneous tire temperature. The
input rolling resistance data can be obtained from SAE J-1269.
For a given set of conditions, the regression equation specified in J-1269
provides the steady-state rolling resistance. Using this value for rolling resis-
tance, a steady-state operating temperature 共representative of the average tire
temperature兲 is computed. At steady state, the tire’s internal heat generation
exactly balances the heat energy dissipated to the surroundings. This internal
heat generation is also equal to the work required to maintain constant rolling
velocity. The first-order differential equation for the time evolution of tempera-
ture in the tire is Mars’ Eq. 16
120 TIRE SCIENCE AND TECHNOLOGY
dT ␣ 共2−p兲 *2
mc = 共AV p + ␣VR*兲共T⬁ − T兲 + VR* + V R
dt A
The model then predicts a rolling resistance that depends on the instantaneous
temperature, using the empirically known decrease in rolling resistance for an
increase in temperature. For the case of a rapidly changing transient speed test,
the tire temperature is presumed constant and the equation becomes Mars’ Eq.
18
冋
R = R* 1 +
␣ 1−p *
A
共V R − Veq Req兲
1−p *
册
The following assumptions were used in developing the simulation model:
The tire’s transient rolling resistance is more sensitive to speed than is the
tire’s equilibrium rolling resistance. Some testing done within the SAE commit-
tee showed that the rolling resistance may vary by as much as 8% at a single
speed, given different speed histories. The Mars’ simulation model can use the
estimated temperature for a particular speed history of the tire to predict this
result.
The article by Mars used an experimental program with a single OE tire, a
tire that was supplied as “Original Equipment” for a given vehicle. The input
information came from various sources including SAE J-1269 data for the tire.
The output transient results were compared with a coastdown test similar to
SAE J-2452 before it was an established standard.
Approach
To evaluate the model, data was acquired from the set of tires tested by
Popio and Luchini 关15兴. The experiment examined the fidelity of the rolling
LUCHINI AND POPIO ON MODELING TRANSIENT ROLLING 121
resistance results from the two SAE tests, on each tire in a set of 12, where all
tires were the same size. The selection of tires was not intended to be represen-
tative of any particular tire market segment.
Consider the factors ␣ and  from Mars’ Eq. 16, shown previously. The
temperature factor, ␣ = 0.006/ ° C, was taken directly from J-1269. The input
parameter  is 共h0兲 / 共V p兲. Also note that the steady-state rolling resistance Req
*
in
Mars’ Eq. 18 was assumed to be a quadratic function of speed, based on ex-
tremely limited empirical data. For the current study, this function was
“switched” using a speed factor 共SF兲 which could be 0 or 1
*
Req = R80kph 关for SF = 0兴
or
*
Req = R80kph共V2/V80
2
兲 关for SF = 1兴
The heat-transfer parameters used in the original Mars model were well
founded, but there was no reason to assume that those parameters were appro-
priate for other tires. The parameters would be reestablished for the current
study, using a tire that would not be part of the set of tires in the primary
evaluation.
This tire is identified here as the “First Tire.” The First Tire in this study
was a P255/45ZR17 of known construction and manufacturer that had been
used in an interlaboratory study. The appropriate rolling resistance data was
available to set the values of parameters in the model.
The following model parameter values were used for all tires.
• c = 1300 J / 共kg° C兲 共average specific heat of a tire that is 20% steel 80%
rubber兲.
• h0 = 40 W / 共m2 ° C兲 共heat-transfer coefficient for the whole tire surface兲.
• p = 0.5 共heat-transfer exponent, where textbook values are: 0.5⫽laminar,
0.8⫽turbulent兲.
• Quadratic velocity effect 共SF= 1, in the parameter study兲.
The test results for each tire and other information were obtained from the
article by Popio 关15兴. The data acquired for the tires in the study are summa-
rized in Table 1, and the J-1269 and J-2452 test conditions are compared in
Table 2.
122 TIRE SCIENCE AND TECHNOLOGY
Note that the 12 sets of tire data were supplied to this study “blind.” The
tire identification was coded, with no reference to construction or manufacturer.
Only the size, P235/70R16, weight, geometry 共outside diameter, rim diameter,
and section width兲, and J-1269 test data were supplied to the model. After the
model predicted the J-2452 results, they were compared to the measured data
for the same tire. The results presented here have identification to match the
source article for clarity and consistency.
TABLE 2 — P-tire test conditions at regulated pressures for J-1269 and J-2452.
J-1269 J-2452
Base Base
Test % Max Pressure % Max Pressure
Point Load ±kPa Speed Load ±kPa Speed
1 90 −30 80 kph, 30 +10 115 kph to
2 90 +70 steady speed 60 −40 15 kph,
3 50 −30 共equilibrium兲 90 +60 in 180
4 50 +70 90 −40 seconds
LUCHINI AND POPIO ON MODELING TRANSIENT ROLLING 123
TABLE 3 — Parameter variations evaluated for the model of the First Tire.
3, and the model results and data for the first tire were plotted. The best values
for tire heat-transfer coefficient and velocity dependence for the first tire were
the same as those in Mars’ original article.
Hypotheses Testing
The statistics of the experimental tests and the assumptions in the modeling
must be considered before establishing evaluation criteria.
Statistical Considerations
Assumptions
Hypotheses
2. The output of the model would match the test data from SAE J-2452
for the same tire.
3. The value from the model should fall within the limits of the SAE
J-2452 test specification for lab-to-lab reproducibility.
4. The sensitivity of the model to load, pressure, and test sequence 共his-
tory兲 should match the variations seen for specific tires.
Acceptance Criteria
1. ±0.5 N 共±0.1# 兲 for the range of measured values due to alignment and
control inaccuracy 共precision兲.
2. 0.5 N 共0.1#兲 for the standard deviation for individual measured values
共repeatability兲.
3. 0.8 N 共0.2#兲 for the standard deviation of the regression residuals 共fit
accuracy兲.
4. ±1.0 N 共±0.25# 兲 for a 99% confidence band of the expected range of
measured values from different laboratories 共reproducibility兲.
4
The precision and bias for J-1269 is based on documents circa 1980 and the precision and bias
statement in J-2452 is statistically ambiguous. The references suggest a 共properly run兲 J-1269 test
would have a standard deviation of 0.25 N and a 99% confidence band of 2 N. The variability of
J-1269 data for the model should have a statistically insignificant effect on the predictions of the
simulation. The precision and bias statement from SAE J-2452 is ambiguous since there is an
indefinite number of data available from a continuous test, and the outlier removal process affects
the statistical confidence. There is no published data, or references, for J-2452 showing how the
statistical criteria were established.
126 TIRE SCIENCE AND TECHNOLOGY
Assuming the SAE documents use a 95% confidence band and applying the
statistics to a paired T-test for data from 12 tires, we anticipated a 95% confi-
dence band of ±3.2 N 共0.7#兲. Therefore, the criteria used in the hypothesis
testing for this study were:
1. Hypothesis 1, parameters from first tire can be used for other tires:
A single tire would be modeled with the theoretical inputs from
Ref. 关1兴.
The parameters for the model would be adjusted to provide results
within ±1 N of matching J-2452 results. The parameters selected
would be used for 12 other tires.
2. Hypothesis 2, simulation results are within the lab-to-lab
reproducibility:
Using the parameters from hypothesis 1, the model predictions will be
evaluated using a target of ±1 N average difference from the J-2452
test data and 3.2 N standard deviation or RMS difference.
Test Sequence Study. The results of the two model runs, along with the test
data, are presented in Fig. 1共a兲. The graph compares the four steps of test data,
with a tare preceding the last step, and two model predictions for sequences
with and without the tare preceding the last step. The model results are identical
until the end of step 3.
The experimental tare results were used as a zero condition for the other
test data. As a result of the experiment setting the rolling resistance at zero for
all speeds of the tare coastdown, the model shows a “negative rolling resis-
tance” appearing during the tare step. This model prediction is the result of a
real, potentially measurable, rolling resistance effect for the tire at the low loads
of the tare step. The raw data for the tare step should show a similar effect, but
the tare values are automatically subtracted from each matching data point in
the reported test results. Therefore, the experimental data show “zero” for each
LUCHINI AND POPIO ON MODELING TRANSIENT ROLLING 127
FIG. 1 — “First Tire” test data and model results. Model parameters at default values; two test
sequences. (a) Full coastdown with five steps of load and inflation pressure, including tare. (b)
Expanded and overlapping view of last non-tare step.
tare datum. This predicted tare step effect in the data had never been noted
before.
Because this sequence effect is small on the scale of Fig. 1共a兲, an expanded
graph of the results for the last step, step 4, is presented in Fig. 1共b兲 so they
overlap on the time scale. Even with the expanded vertical scale in Fig. 1共b兲 it
128 TIRE SCIENCE AND TECHNOLOGY
FIG. 2 — “First Tire” test data and model results (heat-transfer coefficient varied from default
value, H0 = 40). (a) Full coastdown with five steps of load and inflation pressure, including tare. (b)
Expanded and overlapping view of last loading step.
Figure 4共b兲 expands the results and data for the last step. Comparing the
simulation curves for the various values of the h0 exponent and the test data, it
appears that the best match to the shape of the data curve was for a quadratic
speed effect.
130 TIRE SCIENCE AND TECHNOLOGY
FIG. 3 — “First Tire” test data and model results (heat-transfer coefficient exponent varied from
default value, Px = 0.5). (a) Full coastdown with five steps of load and inflation pressure, including
tare. (b) Expanded and overlapping view of last loading step.
We will note here that the constant value for h0 共SF= 0兲 causes the coast-
down curve in Fig. 4共b兲 to become a straight line. This feature of the model will
be of interest later when one group of the tires in the “blind” part of the study
is found to exhibit this behavior.
LUCHINI AND POPIO ON MODELING TRANSIENT ROLLING 131
FIG. 4 — “First Tire” test data and model results [velocity exponent 共VSx兲 varied from default
value, SF = 1]. (a) Full coastdown with five steps of load and inflation pressure, including tare. (b)
Expanded and overlapping view of last loading step.
The equilibrium test data for each of the 12 tires were fit with the regres-
sion equation recommended in J-1269. These regression coefficients were pro-
132 TIRE SCIENCE AND TECHNOLOGY
vided to the model for the simulation of the J-2452 test. The results of the
simulations are shown with the test data in the graphs of Figs. 5共a兲–5共c兲. Tabu-
lation of the test data for the 12 tires appears in Ref. 关15兴. The results have been
sorted into tire groups “A,” “B,” and “C” as used in that article.
These graphs are presented to illustrate: 共1兲 the general ability of the model
to match the test data; 共2兲 each tire had similar but different results; and 共3兲 the
variation in test data among tires is similar to the variation in model results for
the same tires.
Hypothesis Testing
The SAE J-2452 data for the 12 tires consisted of six measurements at
decreasing speeds for each of four load-pressure conditions. The data were
paired with simulation results for each tire and the paired difference 共simulation
minus test兲 and standard deviation were computed. Figure 6 shows the average
differences for each of the 12 tires.
Using the criterion that the average difference for a valid model should be
no greater than ±1.0 N, the model was acceptable for 8 of the 12 tires. The
standard deviations should have been no larger than 3.2 N. This criterion was
met by 11 of the 12 tires.
The average differences for each of the 12 tires are plotted versus the
standard deviation in Fig. 7共a兲 and against the RMS value of the differences in
Fig. 7共b兲. The box on the plots represents the target range for the values and
variations. The figures show that seven of the tires met both criteria. Only one
tire, B3, fell outside of the target criteria for both average value and variation.
Figure 8 shows the differences between the simulation and test data were
not evenly distributed among the steps or tires. Note that group B tires have
more variation from each step than either of the other groups of tires. To explain
this variation, we reexamined the shape of the rolling resistance versus speed
plot for group B tires.
The group B coastdown plots in Fig. 5 were similar to the SF= 0 curve of
Fig. 4, while group A and C plots were similar to the SF= 1 curve of Fig. 4. This
suggests that the original idea of using a single set of heat-transfer properties for
all similar tires is not an adequate assumption. Since the tire data was provided
“blind” no cause could be identified. Future research may look for differences in
material or tread pattern affecting the velocity dependence.
LUCHINI AND POPIO ON MODELING TRANSIENT ROLLING 133
FIG. 5 — Simulation and test data for 12 tires (h0 = 40, p = 0.5, SF = 1, for all simulations). Squares
are experimental test values and diamonds are matching simulation model values. (a) Group A tires.
(b) Group B tires. (c) Group C tires.
134 TIRE SCIENCE AND TECHNOLOGY
Using the parameter values from the first tire, the discrepancy between test
and simulation was 1 N to 4 N, with the simulation generally lower than the test
data. This result could have been improved by modifying inputs to the simula-
tion; however, the consistency of this condition was of concern. Therefore,
further investigation of both the simulation model and the test was carried out.
The transient model internal parameters affect the shape curves. The equi-
FIG. 7 — Model error measures for 12 tires: (a) standard deviation versus difference and (b) RMS
difference versus difference.
LUCHINI AND POPIO ON MODELING TRANSIENT ROLLING 135
librium, asymptotic, value of rolling resistance at 50 mph for each load and
pressure is an input to the model. Since the test points for J-1269 and J-2452 are
different, this input was not a test datum but an extrapolation from the recom-
mended curve fit to the J-2452 load and pressure conditions. The behavior of the
model suggested that this input was not adequate, and was observed to have the
following characteristics:
1. The shape of both the model and data for the transient coastdown
curves appeared to be similar.
2. The value of the rolling resistance at low-load/high-pressure was very
close. The value of the rolling resistance at the tare load may be too
high.
3. The value of the rolling resistance at low inflation pressure points for
the model was always too low.
4. Any of the model parameters that would increase the predicted rolling
resistance for the high-load/low-pressure condition would also change
the values at the other conditions, or the curve shape, or both.
We note, also, that the initial modeling and validation work of Mars used
transient coastdown results for load and pressure conditions that were the same
as, or interpolated from, the test conditions of J-1269. Due to the nonlinear
curve fit recommended in J-1269, it was hypothesized that there was insufficient
136 TIRE SCIENCE AND TECHNOLOGY
For some regressions, the use of a zero-zero point, as per the tare datum,
provided the additional input data that prevented mathematical overflow.
The mathematically successful regressions provided “R-squared” values
above 0.995 and residuals of about 0.8 N 共0.2#兲 or less. The coefficients from
these regressions were then used to provide the input rolling resistance for the
transient tire model. Figure 9 shows the effect of using different regression
equations for the coastdown predictions. The equilibrium data for each of the
regression equations was the same but different equations were input to the
simulation. The selection of the regression equation used with the test data
affects the simulated rolling resistance.
While each of the regression equations gave a good fit to one or more of the
coastdown steps, overall the best match appears to be regression 4. This has the
aesthetic appeal that it is the regression equation recommended in J-1269. From
detailed examination of the results from one tire, it was apparent that neither the
statistically best fit to the data nor the most complex regression equation would
provide better predictions from the simulation.
138 TIRE SCIENCE AND TECHNOLOGY
Conclusions
1. The parametric inputs for the model can be set based on simplified
theoretical considerations and limited amount of validation data from a
sample tire. The sample tire should be selected to have a similar be-
havior with respect to the heat-transfer parameters as the tire to be
modeled.
2. The output of the simulation model for the transient test, using data
from the equilibrium test SAE J-1269, matches the qualitative behavior
of SAE J-2452 for the same tire. However, the model output did not
meet the lab-to-lab reproducibility specifications in SAE J-2452 when
compared to all of the test data.
Future Work
The SAE is aware that there is need for review of both SAE J-1269 and
SAE J-2452. A task group has been formed to review and evaluate the test
protocols and determine how to update the standards in order to serve the needs
of government, industry, and consumers. There is a need to develop simulation
models with a larger scope that can apply to tire operation on vehicles and work
with existing vehicle models. The simulations should consider a broad range of
factors and have parametric inputs for tires and vehicles based on known phys-
ics, as illustrated in the Mars’ model.
References
关1兴 Mars, W. V. and Luchini, J. R., “An Analytical Model for the Transient Rolling Resistance
Behavior of Tires,” Tire Science and Technology, TSTCA, Vol. 27, No. 3, 1999, pp. 161–175.
关2兴 Schuring, D. J., “The Rolling Loss of Pneumatic Tires,” Rubber Chemistry and Technology,
Vol. 153, 1980, pp. 602–606.
关3兴 Schuring, D. J. and Futamura, S., “Rolling Loss of Pneumatic High-Way Tire in the Eighties,”
Rubber Chemistry and Technology, Vol. 63, No. 3, 1990, pp. 315–367.
关4兴 Luchini, J. R., “Measuring and Modeling Tire Rolling Resistance,” Paper 24A/25A, Presented
at Oak Ridge National Laboratory, Aug. 1999; and at ITEC 2000, Sept. 2000, Akron, OH, USA.
关5兴 Schuring, D. J., “Transient versus Steady-State Tire Rolling Loss Testing,” SAE 790116.
关6兴 Schuring, D. J. and Speyer, A. G., “Tire Rolling Loss And Pressure Increase,” SAE 831027.
关7兴 Kenny, T. M., “Prediction of Contained Air Temperature from SAE Standard Rolling Resis-
tance Test Data,” SAE 831796.
关8兴 Schuring, D. J., Siegfried, J. F., and Hall, G. L., “Transient Speed and Temperature Effects on
Rolling Loss of Passenger Car Tires,” SAE 850463.
关9兴 Clark, J. D. and Schuring, D. J., “Load, Speed, and Inflation Pressure Effects on Rolling Loss
Distribution in Automobile Tires,” Tire Science and Technology, TSTCA, Vol. 16, No. 2, 1988,
pp. 78–95.
关10兴 “Tires and Passenger Vehicle Fuel Economy,” Transportation Research Board, Special Report
286, Washington, DC, 2006.
关11兴 Whicker, D., Browne, A. L., and Segalman, D. J., “The Structure and Use of the GMR
Combined Thermo-Mechanical Tire Power Loss Model,” SAE 810164; Whicker, D. and
Rohde, S. M., “Modeling Tire Deformation for Power Loss Calculations,” SAE 810161; Segal-
man, D. J., “Modeling Tire Energy Dissipation for Power Loss Calculations,” SAE 810162;
Browne, A. L. and Arambages, A., “Modeling the Thermal State of Tires for Power Loss
Calculations,” SAE 810163; “A Thermo-Mechanical Approach to Tire Power Loss Modeling,”
General Motors Research Laboratories Research publication GMR-3310, May 30, 1980, and
The General Problem of Rolling Contact, AMD Vol. 40, 1980, published by the American
Society of Mechanical Engineers.
关12兴 Warholic, T. C., “Tire Rolling Loss Prediction from the Finite Element Analysis of a Statically
140 TIRE SCIENCE AND TECHNOLOGY
Loaded Tire,” M.S.E. Thesis, University of Akron, Akron, OH, 1987, available from University
Microfilms.
关13兴 Becker, A., Dorsch, V., Kaliske, M., and Rothert, H., “A Material Model for Simulating the
Hysteretic Behavior of Filled Rubber for Rolling Tires,” Tire Science and Technology, TSTCA,
Vol. 26, No. 3, 1998, pp. 132–148.
关14兴 Kennedy, R. H., Englehardt, M., and Day, G. L., “RPA Measurement of Hysteresis for CAE
Rolling Resistance Prediction,” ITEC ’98, Paper No. 30A.
关15兴 Popio, J. A. and Luchini, J. R., “The Fidelity of SAE J-1269 and SAE J-2452 Rolling Resis-
tance Testing,” Presented at the 2006 Tire Society Conference on Tire Science and Technology.