Beruflich Dokumente
Kultur Dokumente
0022--4898(94)00012-3
MEASURING
SOIL PROPERTIES
TO PREDICT
TRACTIVE
PERFORMANCE
OF AN AGRICULTURAL
DRIVE TIRE
SURI T H A N G A V A D I V E L U , * R. T A Y L O R , t S. CLARK~ and
J. S L O C O M B E ~
Summary--The coefficient of traction for a 9.5-16 R-1 bias ply tire was measured and
compared with predictions using equations developed by Janosi and Hanamoto [Proc. 1st Int.
Conf. on Mechanics, p. 707-736 (1961)]; Wismer and Luth [J. Terramechanics 10, 49-61
(1973)] Gee-Clough et al. [J. Terramechanics 15, 81-84 (1978)] and Brixius [ASAE Paper No.
87-1622 (1987)]. For the soft soil condition, with a cone index of 120 kPa, Gee-Clough's
equation predicted the coefficient of traction better, but predictions using the Brixius equation
were better for soil with a cone index of 225 kPa. An experimental device was developed to
simultaneously measure the horizontal and vertical stress-strain relationships of soil. The use
of resultant stress from the experimental device data failed to show any improvement in
coefficient of traction prediction over using the cone index. The resultant of the normal and
shear stress from the experimental device data did not adequately represent the soil properties
involved in terrain-vehicle mechanics.
INTRODUCTION
Theoretical analysis of soil-vehicle interaction is complicated because of the nonhomogeneous, anisotropic and inelastic nature of soil. Therefore, empirical methods
are used widely to define the stress-strain relationship of soil for predicting off-road
vehicle performance.
Tractive performance predictions generally are based on stress-strain relationships
of soil using soil cohesion (c), soil-soil friction angle (q~), and sinkage parameter (k)
or the strength of soil usually measured as cone index (CI). Bekker [1] adopted the
M o h r - C o u l o m b theory of soil failure to develop equations to predict the rolling
resistance of and traction developed by pneumatic tires. Janosi and H a n a m o t o [2]
developed tractive force prediction equations based on stress-strain relationship of
soil for the case of uniform pressure distribution under the tractive device.
Wismer and Luth [3] developed equations to predict tractive performance of
pneumatic tires using CI and tire parameters. Brixius [4] developed tractive performance prediction equations for bias ply tires operating in cohesive-frictional soils also
based on soil strength (CI) and tire parameters. Witting and Alcock [5] developed a
single-wheel tester to establish traction conditions of topsoil by measuring the
maximum transferable torque under known axle loads. They found that tractive
performance prediction using soil parameters like bulk density and/or soil water
content was more accurate than the standard Wismer and Luth [3] equations.
For prediction equations to be effective, the terrain and traction device parameters
216
S. Thangavadivelu et al.
M A T E R I A L S AND M E T H O D S
An experimental device (ED) was fabricated to simultaneously measure the vertical
and horizontal stress-strain behavior of soil. It was designed to be simple, easy to
use, and provide data that could be used in existing traction prediction equations. The
ED is a modified cone penetrometer with two rectangular vanes welded 180 apart
along the slant edge of the standard cone (Fig. 1). The vanes provide a twisting action
that shears the soil in contact while the cone is pushed into the ground, thereby
providing a measure of both vertical and horizontal stress-strain behaviors of soil.
The ED was attached to a shear graph in place of the shear head for recording the
normal and shear stress. A study was conducted to compare different traction
prediction equations and determine if the predictions were improved by using the ED
data.
Experiments were conducted in a soil bin 9 m long and equipped with a soilprocessing carriage and an instrumented test carriage. Carriage movement was
controlled by a winch mechanism driven by a hydraulic motor. The processing
carriage had a roto-tiller, levelling blade, and compaction drum. The test carriage
contained a hydraulic power unit to drive a pneumatic tire and was instrumented to
record pull, input torque, and tire and test carriage speeds. Speed was measured by
pulse counters mounted on the axles of the wheel and test carriage.
Tests were conducted with a 9.5-16 R-1 bias ply tire carrying a 4 kN axle load on
two different soil conditions of five slip levels. The specifications of the test tire are
listed in Table 1. Soil was prepared for each test by tilling the top 0.19 m, levelling
and passing the compaction drum twice to create the first soil condition
217
Fig. 1. The modified ASAE standard cone mounted on a shear graph to simultaneously measure the
normal and shear stress of soil.
Table 1. Specifications of the 9.5 16 Firestone traction field and
road tire used for testing
Tire parameter
Value
0.24
0.85
0.02
0.21
23
(CI = 120 kPa) and six times for the second soil condition (CI = 225 kPa). Wheel
speed was held constant at approximately 1.34 m/s, while the test carriage speed was
varied to create different slip levels. A n IBM PC-XT equipped with Omega
Engineering, Inc. DAS-16 and ComputerBoards, Inc. CIO-CTR05 data acquisition
boards was used to record data. Voltage signals for the torque and pull load cells
were amplified with Calex MK-160 bridge conditioning cards and recorded at a
frequency of 100 Hz by the DAS-16. The pulse signals from optical encoders for the
218
S. Thangavadivelu et al.
tire and test carriage speeds were counted by the CIO-CTR05 for 2.5 s for each test.
This arrangement yielded 250 readings for torque and pull for each test, which were
averaged to obtain a single value. Tire and test carriage speeds were determined from
the number of pulses counted during the 2.5 s interval.
Soil properties were measured for each condition tested. Soil-cohesion (c) and
soil-soil friction angle (~0) were determined from shear graph data. The sinkage
parameters (kc and k~) were determined from plate sinkage data obtained by
pressing 0.1 and 0.12 m square plates into the soil while measuring force with an
Alphatron 9 kN capacity load cell and sinkage with a Celesco 0.75 m range position
sensor. The CI for each soil condition was determined using an ASAE standard cone
with the above mentioned load cell and position sensor for a depth of 0.15 m. The
experimental device was used with a shear graph calibrated for simultaneous
application of torque and normal force. The vertical and horizontal stress-strain
relationships were recorded for each condition over depths of 0.125 and 0.075 m for
the first and second soil conditions, respectively. Soil property measurements were
replicated three times for each soil condition and slip level tested. The average values
used for tractive performance predictions are shown in Table 2.
Experimental data
Wheel slip was determined from the carriage and test wheel forward speeds
recorded using pulse counters. Wheel slip and coefficient of traction (CT--ratio of
pull to axle load) are plotted for the two soil conditions in Figs 2 and 3. A non-linear
model of the form given in equation (1) was fitted to the slip and coefficient of
traction data using the Marquardt's method in the SAS NLIN procedure [8]. The
estimate (Pexp = {aexp, bexp, Cexp}) and confidence interval of each parameter for the
model fitting the experimental data are shown in Table 3. The fitted model showed
good agreement with the experimental data for both soil conditions (Table 3).
P - a ( 1 - e bs)- c,
W
(1)
Soil parameter
Cone Index, CI (kPa)
Soil cohesion, C (kPa)
Soil-soil friction angle, ~ (deg)
Sinkage parameter, kc
Sinkage parameter, k
Sinkage exponent, n
Resultant stress, R (kPa)
Wheel numeric, Cn
Mobility number, Bn
Mobility number, M
measured/
Condition
No. 1
Condition
No. 2
120.0
6.3
18.3
2.1
4.7
1.4
170.5
8.5
8.7
1.7
225.0
14.5
23.8
1.9
9.2
1.8
325.0
17.2
17.6
3.1
0.4
Actual
Soil condition 1
......
T i r e 9.5-1 6
.......
0.3
C
0
219
Fitted
Brixius
Wis. & Luth
G e e Clough
Janosi
.m
I0.2
Q
,"
0
0
.7
0.1
. . . . . . )..
0.0
10
20
30
40
Wheel s l i p
Fig. 2. Regression curves of the experimental and predicted data for soil condition 1 (CI = 120 kPa).
0.6
Soil condition 2
......
....
Tire 9,5-1 6
.......
0.5
.- . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
C
0
Actual
Fitted
Brixius
Wis. & Luth
G e e Clough
Janosi
0.4
I.-
"6
03
fJ
0.2
0,
/11//
0.0
0
10
15
20
25
30
Wheel slip
Fig. 3. Regression curves of the experimental and predicted data for soil condition 2 (CI = 225 kPa).
S. Thangavadivelu et al.
220
Table 3. Estimate of model parameters for the experimental and predicted values of
coefficient of traction for the test tire using C! and RES for the two soil conditions
Data set
Re
0.44
[0.04, 0.84]
0.27
0.37
0.26
0.41
0.74
0.75
0.43
-27.73
[-44.96, -10.50]
-11.61
-9.84
-13.96
-13.96
-1.88
-2.55
-3.38
0.16
[-0.25, 0.56]
0.23
0.16
0.00
0.00
0.24
0.18
0.00
0.87
- 11.86
[-14.23, -9.49]
-8.77
-8.34
-28.65
-28.65
-3.03
-4.88
-2.84
0.03
[-0.03, 0.09]
0.12
0.09
0.00
0.00
0.14
0.11
-0.01
0.98
0.36
[0.31, 0.41]
0.47
0.59
0.50
0.59
0.79
0.75
0.46
Bold faced numbers are estimates of model parameters for predicted data falling
within the confidence interval of the model parameters for the experimental data.
[*]95% confidence interval.
c o n f i d e n c e i n t e r v a l o f t h e c o r r e s p o n d i n g m o d e l p a r a m e t e r for the e x p e r i m e n t a l d a t a
({aexp, bexp, Cexp}) and
(2) the closeness o f t h e r e g r e s s i o n curves for p r e d i c t e d a n d e x p e r i m e n t a l data.
H=(AC+
P
Wtan(q0)(1-
-0.75(1-
e3C")-
1(1-e-J)),
l
(2)
( 12 + 0.04)),
co
(3)
Cn --
Clbd
W
e-m),
0.92
M
(4)
K(CT)ma x - 4.838 + 0 . 0 6 1 M ,
M
CI bd ~ [
1
.~ l - - I
W ~/---h\l + ( b / 2 d ) )
221
Bn
where
1 + 0.5_._____Ls
(5)
Tl\l + 3(b/d)!
j iol/k;
i0 degree of slip;
k deformation modulus;
6 tire deflection;
soil-soil friction angle;
A contact area;
Bn dimensionless mobility number;
C soil cohesion;
CI cone index;
Cn wheel numeric;
CT coefficient of traction;
H traction force;
K sinkage parameter; and
M mobility number.
The Gee-Clough equation (4) predicted coefficient of traction better than the
others for the softer soil condition (CI = 120 kPa), with all three parameter estimates
(apre, bpre, pre} falling within the confidence interval of the parameter estimates of
the experimental data {aexp, bexp, Cexp}. Estimates of the model parameters for the
predicted data (Ppre) using the Brixius equation were within the confidence interval of
the model parameters for the experimental data (Pep) but the regression curve was
not close to the experimental data regression curve. Although the regression curves
(a)
0.6
0.5
Soil condition 1
Tire 9.5-16
FI'I-I'ED MODEL
Brlxlus (Cl)
Brlxlus (RES)
Cl = 120 kPa
RES = 170 kPa
__o
E
"6
0.4
I-
0.3
O
0.2
O
3
0.1
//
0.O
0.0
....
0.1
......
. . . . . . . . . . . . . . .
0.2
Wheel slip
Fig. 4.
Continued overleaf.
0.3
0.4
222
S. Thangavadivelu et al.
(b)
0.6
0.5
t,-
_o
3
Soil condition 1
Tire 9.5-16
CI = 120 kPa
RES = 170 kPs
FITTED MODEL
Wis. & Luth(CI)
Wls. & Luth(RES)
0.4
I-
"6
0.3
O
O
0.2
0.1
jjt
0.0
0.0
jjj
jJ
.//
,,/
0.1
0.2
0.3
0.4
Wheel slip
0.6
(c)
0.5
P
o
o
Soil condition 1
Tire 9.5-16
CI = 120 kPe
RES = 170 kPa
FITTED MODEL
Gee Clough(CI)
Gee Clough(RES)
0.4
I-
"6
0.3
fJ
_o
@
o
O
J I
....................................
0.2
0.1
0.0
0.0
///.
0.1
0.2
0.3
0.4
Wheel slip
Fig. 4, (a) Regression curves of the experimental data and data predicted with the Brixius equation using
cone index (CI) and resultant stress (RES) for soil condition 1 (CI = 120 kPa and RES = 170 kPa). (b)
Regression curves of the experimental data and data predicted with the Wismer & Luth's equation using
cone index (CI) and resultant stress (RES) for soil condition 1 (CI = 120 kPa and RES = 170 kPa). (c)
Regression curves of the experimental data and data predicted with the Gee-Clough equation using cone
index (CI) and resultant stress (RES) for soil condition 1 (CI = 120 kPa and RES = 170 kPa).
for the Wismer and Luth and Janosi equations did not resemble the experimental
data curve, some of the model parameter estimates (Ppre) were within the confidence
interval of the model parameters of the experimental data (Pexp), because of the wide
confidence interval.
The confidence interval of Pexp for the second soil condition (CI = 225 kPa) was
narrower and very few model parameter estimates of Ppre fell within it. The model
parameter estimate that fell in the confidence interval was the constant Cpre, which is
of lesser importance than the other two (apre and bpre). The Brixius equation (5) was
223
closer to the experimental data than others for this soil condition (CI--225 kPa),
even though none of Ppre were within the confidence interval of Pexp.
0.6
0.5
Soil condition 2
Tire 9.5-16
Cl : 225 kPs
RES = 325 kPe
......
e.
mo
/ / J
j/-
0.4
./
./
I-
"6
FITTED MODEL
Brlxlus(CI)
Brlxlus(RES)
/
J
0.3
....
...........
O
mtJ
"6
0.2
o
O
//
0.1
//
0.0
O.0
"
/
I
0.1
0.2
0.4
0.3
Wheel slip
(b)
0.6
0.5
Soil condition 2
Tire 9.5-16
CI = 225 kPs
RES = 325 kPs
....
j f
0.4
.....'""
/ J
I-
"6
0.3
0.2
0.1
0.0
0.0
FITTED MODEL
Wls. & Luth(Cl)
WIs. & Luth(RES)
.....
.~
//
///.... "
0.1
0.2
Wheel slip
0.3
014
224
S. Thangavadivelu et al.
(c)
0.6
j i
0.5
--ira
0.4
i--
"6
/ "
_Q
_u
q)
o
. . . . . . . . .
.m
FITTED MODEL
Gee Clough(Cl)
Gee CIough(RES)
0.3
/
/
Soil condition 2
Tire 9.5-16
CI = 225 kPa
RES = 325 kPa
0.2
I
0.1
0.0
0.0
/
0.1
0.2
0.3
0.4
Wheel slip
Fig. 5. (a) Regression curves for experimental data and data predicted with the Brixius equation using cone
index (CI) and resultant stress (RES for soil condition 2 (CI = 225 kPa and RES = 325 kPa). (b) Regression
curves for experimental data and data predicted with the Wismer & Luth equation using cone index (CI)
and resultant stress (RES) for soil condition 2 (CI = 225 kPa and RES = 325 kPa). (c) Regression curves for
experimental data and data predicted with the Gee-Clough equation using cone index (CI) and resultant
stress (RES) for soil condition 2 (CI = 225 kPa and RES = 325 kPa).
RES instead of CI. Because RES for each soil condition was higher than the
corresponding CI, the predictions using RES gave higher coefficient of traction at all
slip levels.
The coefficient of traction predictions were also compared with the model fitted to
the experimental data in terms of the mean absolute difference expressed as a
percentage of the mean coefficient of traction for slip levels ranging from 5-25% at
1% increments. The criterion for the predictions to be acceptable was arbitrarily set
to +5%. The results are summarized by soil condition for predictions made using CI
and RES in Table 4. Using RES improved the predictions by the Wismer and Luth
equation (3) for both the soil conditions. The predictions by the Brixius equation (5)
showed improvement in the softer soil condition (CI = 120 kPa, RES = 170 kPa).
Predictions by Gee-Clough's equation wer.e not improved by using RES over CI for
Table 4. Average absolute differences between observed and predicted values of coefficient of traction for
the two soil conditions
Prediction equation
Brixius
Gee-Clough
Wismer & Luth
Janosi
Parameter
CI
R
CI
R
CI
R
C, ~
(105.7)
(59.1)
(18.3)
(31.3)
(121.9)
(79.8)
(35.8)
(19.7)
(25.3)
(89.8)
(123.1)
(47.1)
(20.0)
(32.7)
Values in parentheses are mean absolute differences expressed as percentages of the mean coefficients of
traction (0.26 and 0.25 for soil conditions 1 and 2, respectively).
225
either soil condition. Although improvements were observed in some of the predictions, none of them fell in the acceptable range because the average coefficients of
traction in both soil conditions were low (0.27 and 0.26).
CONCLUSIONS
For the softer soil condition (CI = 120 kPa), the coefficient of traction prediction
with Gee-Clough's equation using CI gave better results and Brixius's equation with
CI predicted well for the other soil condition (CI -- 225 kPa). Use of RES instead of
CI to predict coefficient of traction failed to show any consistent improvement in the
prediction equations. However, predictions by the Brixius equation in the softer soil
condition (CI = 120 kPa, RES = 170 kPa) and predictions by the Wismer and Luth
equation in both soil conditions were improved by using RES. Although the
experimental device measures the horizontal and vertical stress-strain relationships,
the resultant of the normal and shear stress does not adequately represent soil
properties involved in terrain-vehicle mechanics to predict tractive performance.
REFERENCES
[1] M. G. Bekker, Theory o f Land Locomotion. The University of Michigan Press, Ann Arbor, MI (1956).
[2] Z. Janosi and B. Hanamoto, The analytical determination of drawbar-pull as a function of slip for
tracked vehicles in deformable soils. Proc. 1st Int. Conf. on Mechanics of Soil-Vehicle Systems, Torino,
Italy, pp. 707-736 (1961).
[3] R. D. Wismer and H. J. Luth, Off-road traction prediction for wheeled vehicles. J. Terramechanics 10,
49-61 (1973).
[4] W. W. Brixius, Traction prediction equations for bias ply tires. ASAE Paper No. 87-1622. ASAE,
St. Joseph, MI 49085 (1987).
[5] V. Wittig and R. Alcock, An empirical method of predicting traction. ASAE Paper No. 90-1570.
ASAE, St. Joseph, MI 49085 (1990).,
[6] M. G. Bekker, Introduction to Terrain-Vehicle Systems. The University of Michigan Press, Ann Arbor,
MI (1969).
[7] R. N. Yong, A. F. Youssef and E. A. Fattah, Vane-cone measurements for assessment of tractive
performance in wheel-soil interaction. Proc. 5th International Society of Terrain Vehicle Systems,
Detroit 3:769-788 (1975).
[8] SAS Institute Inc., SAS User's Guide: Statistics, Version 5 Edition. Cary, NC: SAS Institute Inc.
(1985).
[9] D. Gee-Clough, M. McAllister, G. Pearson and D. W. Evernden, The empirical prediction of
tractor-implement field performance. J. Terramechanics 15, 81-84 (1978).