ERDC/CRREL TR-01-16

US Army Corps
of Engineers®
Engineer Research and
Development Center

Finite Element Modeling of

Tire–Terrain Interaction
Sally A. Shoop November 2001
and Engineering Laboratory
Cold Regions Research

Approved for public release; distribution is unlimited.

 REPORT DOCUMENTATION PAGE Form Approved OMB No.
0704-0188
Public reporting burder for this collection of information is estibated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing
and reviewing this collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burder to Department of Defense, Washington
Headquarters Services, Directorate for Information Operations and Reports (0704-0188), 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302. Respondents should be aware that notwithstanding any other provision of
law, no person shall be subject to any penalty for failing to comply with a collection of information if it does not display a currently valid OMB control number. PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ADDRESS.
1. REPORT DATE (DD-MM-YYYY) 2. REPORT TYPE 3. DATES COVERED (FROM - TO)
01-11-2001 Technical xx-xx-2001 to xx-xx-2001
4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER
Finite Element Modeling of Tire-Terrain Interaction 5b. GRANT NUMBER
Unclassified 5c. PROGRAM ELEMENT NUMBER
6. AUTHOR(S) 5d. PROJECT NUMBER
Shoop, Sally A. ; 5e. TASK NUMBER
5f. WORK UNIT NUMBER
7. PERFORMING ORGANIZATION NAME AND ADDRESS 8. PERFORMING ORGANIZATION REPORT
U.S. Army Engineer Research and Development Center NUMBER
Cold Regions Research and Engineering Laboratory
72 Lyme Rd.
Hanover, NH03755-1290
9. SPONSORING/MONITORING AGENCY NAME AND ADDRESS 10. SPONSOR/MONITOR'S ACRONYM(S)
Office of the Chief of Engineers 11. SPONSOR/MONITOR'S REPORT
Washington, DC20314-1000 NUMBER(S)
12. DISTRIBUTION/AVAILABILITY STATEMENT
APUBLIC RELEASE
,
13. SUPPLEMENTARY NOTES
14. ABSTRACT
The desire to incorporate theoretical mechanics into off-road vehicle performance prediction has generated great interest in applying numerical
modeling techniques to simulate the interaction of the tire and terrain. Therefore, a full three-dimensional model simulating a tire rolling over
deformable terrain was developed. Tires were simulated using a rigid wheel, a deformable tire simplified with user-defined sidewall elements,
and modal analysis tire models. Model comparisons with measured, hard-surface tire deformation and contact stress showed very good
agreement. The simplified tire model was much more computationally efficient but the modal analysis model yielded better contact stress
distribution. Each of the tire models was then combined with rolling on deformable terrain. Fresh snow and compacted sand surfaces were
modeled using critical-state plasticity models. The rigid wheel model was validated on snow using field measurements of tire forces and snow
deformation and then compared to performance predictions using the NATO Reference Mobility Model. These comparisons indicate excellent
agreement between the model and the measurements. Preliminary results of the modal analysis tire model on snow show very little
deformation in the tire, indicating that the rigid wheel simplification may be a good approximation for soft terrain.
15. SUBJECT TERMS
Mobility; Tires; Tire Modeling
16. SECURITY CLASSIFICATION OF: 17. LIMITATION 18. 19. NAME OF RESPONSIBLE PERSON
OF ABSTRACT NUMBER Fenster, Lynn
Public Release OF PAGES [email protected]
69
a. REPORT b. ABSTRACT c. THIS PAGE 19b. TELEPHONE NUMBER
Unclassified Unclassified Unclassified International Area Code
Area Code Telephone Number
703767-9007
DSN
427-9007
Standard Form 298 (Rev. 8-98)
Prescribed by ANSI Std Z39.18
Abstract: The desire to incorporate theoretical ter contact stress distribution. Each of the tire
mechanics into off-road vehicle performance models was then combined with rolling on de-
prediction has generated great interest in ap- formable terrain. Fresh snow and compacted
plying numerical modeling techniques to simu- sand surfaces were modeled using critical-state
late the interaction of the tire and terrain. There- plasticity models. The rigid wheel model was
fore, a full three-dimensional model simulating validated on snow using field measurements of
a tire rolling over deformable terrain was devel- tire forces and snow deformation and then com-
oped. Tires were simulated using a rigid wheel, pared to performance predictions using the
a deformable tire simplified with user-defined NATO Reference Mobility Model. These com-
sidewall elements, and modal analysis tire mod- parisons indicate excellent agreement between
els. Model comparisons with measured, hard- the model and the measurements. Preliminary
surface tire deformation and contact stress results of the modal analysis tire model on snow
showed very good agreement. The simplified show very little deformation in the tire, indicat-
tire model was much more computationally effi- ing that the rigid wheel simplification may be a
cient but the modal analysis model yielded bet- good approximation for soft terrain.

How to get copies of ERDC technical publications:

Department of Defense personnel and contractors may order reports through the Defense Technical Informa-
tion Center:
DTIC-BR SUITE 0944
8725 JOHN J KINGMAN RD
FT BELVOIR VA 22060-6218
Telephone (800) 225-3842
E-mail [email protected]
[email protected]
WWW http://www.dtic.mil/
All others may order reports through the National Technical Information Service:
NTIS
5285 PORT ROYAL RD
SPRINGFIELD VA 22161
Telephone (703) 487-4650
(703) 487-4639 (TDD for the hearing-impaired)
E-mail [email protected]
WWW http://www.ntis.gov/index.html
For information on all aspects of the Engineer Research and Development Center, visit our World
Wide Web site:
http://www.erdc.usace.army.mil
Technical Report
ERDC/CRREL TR-01-16
US Army Corps
of Engineers®
Engineer Research and
Development Center

Finite Element Modeling of

Tire–Terrain Interaction
Sally A. Shoop November 2001

Prepared for
OFFICE OF THE CHIEF OF ENGINEERS

Approved for public release; distribution is unlimited.

PREFACE

This report was prepared by Sally A. Shoop, Research Civil Engineer, Applied and Mili-
tary Engineering Branch, U.S. Army Engineer Research and Development Center (ERDC),
Cold Regions Research and Engineering Laboratory (CRREL), Hanover, New Hampshire.
This work was funded by the Office of the Chief of Engineers under work unit Winter
Terrain Effects on Army Simulations.
The author thanks Ron Liston and George Ashton, CRREL, who provided financial and
moral support; Paul Richmond, George Blaisdell, and Byron Young, CRREL, who assisted
in gathering, providing, and analyzing experimental data; Robert Haehnel of CRREL and
Ken Kestler, independent consultant, who did much of the snow modeling; Ian Darnell,
University of Michigan, who helped in generating and debugging the Shoop-Darnell mod-
els; Jane Mason, Tom Vaughan, and Peter Keene, CRREL, for assistance with the graphics;
and David Cate, Donna Harp, and John Severance for helping with finishing touches on the
document. Mike Trinko at Goodyear and Paul Corcoran and Liqun Chi at Caterpillar pro-
vided many helpful discussions, and Dr. Roman Hryciw, University of Michigan, provided
much encouragement and advice. Technical reviews were provided by Dr. Hryciw, Dr.
Gregory Hulbert, and Dr. Kenneth Ludema, University of Michigan.
The contents of this report are not to be used for advertising or promotional purposes. Ci-
tation of brand names does not constitute an official endorsement or approval of the use of
such commercial products.

ii to contents
CONTENTS

Preface ............................................................................................................................... ii
Nomenclature .................................................................................................................... vi
Introduction ....................................................................................................................... 1
Background ........................................................................................................................ 1
Off-road vehicle performance prediction ............................................................... 1
Vehicle movement on snow ................................................................................... 2
Tire–terrain modeling ............................................................................................ 2
Tire terminology .................................................................................................... 4
Terrain material models ..................................................................................................... 5
General concepts for plasticity models .................................................................. 5
Modified Drucker–Prager cap model .................................................................... 10
Crushable foam model ........................................................................................... 12
Constitutive models for snow ................................................................................ 12
Soil model .............................................................................................................. 22
Tire models for a deformable substrate ............................................................................. 24
Tire model construction ......................................................................................... 24
Evaluation of tire models on a rigid surface .......................................................... 35
Combined tire–terrain model ............................................................................................. 41
Modeling the tire–terrain contact interface ............................................................ 41
Rigid wheel on snow ............................................................................................. 42
Deformable tire on soil .......................................................................................... 51
Deformable tire on snow ....................................................................................... 54
Conclusions ....................................................................................................................... 54
Summary ................................................................................................................ 54
Significant findings ................................................................................................ 55
Recommended applications and future research needs .......................................... 56
Literature cited ................................................................................................................... 56

ILLUSTRATIONS

Figure
1. Sand deformation under a towed wheel moving to the right ................................. 2
2. Shallow and deep snow under a wheeled vehicle .................................................. 2
3. Snow cross sections showing pressure bulb formation .......................................... 3
4. Two-dimensional modeling of tire-terrain interaction by Aubel (1993) ................ 4
5. Modeling of the effect of lug design on tread–terrain interaction
in two dimensions by Fervers (1994) .................................................................... 4
6. Tire direction convention ....................................................................................... 4
7. Construction components of a radial ply tire ......................................................... 5
8. Measurement of tire dimensions ............................................................................ 5
9. Modified Drucker–Prager yield surface in deviatoric space .................................. 7
10. Yield surfaces in deviatoric space ......................................................................... 7
11. Common yield surfaces along the hydrostatic axis in principal stress space ......... 7
12. Three-dimensional view of critical state yield surface .......................................... 8
13. Cap contraction or expansion reflecting the softening
or hardening of the material ................................................................................... 8
14. Exponential hardening law for material in hydrostatic compression ..................... 9
15. Piecewise linear modeling of the hardening law ................................................... 9
16. Isotropic and kinematic hardening ......................................................................... 10

iii
Figure
17. Modified Drucker–Prager cap yield surface in the p-t plane ................................. 11
18. Modified Drucker–Prager cap flow potential in the p-t plane ............................... 11
19. Yield surface for the crushable foam model .......................................................... 12
20. Plastic flow surface for the crushable foam model ................................................ 12
21. Compilation of Young’s modulus and Poisson’s ratio measurements on snow .... 13
22. Calculated c′ and φ′ using measurements from the CRREL Instrumented Vehicle
and the ring shear device and predicted using the Shallow Snow Model .............. 14
23. Compilation of uniaxial strength data .................................................................... 15
24. Compression of natural snow at 0° to –3°C ........................................................... 16
25. Compression of undisturbed snow and ice ............................................................ 17
26. Pressure–volume data from compression tests of Abele and Gow (1975) ............ 17
27. Comparison of model and experimental data for uniaxial compression tests
on low-density snow .............................................................................................. 18
28. Deformed meshes for field plate sinkage simulations ........................................... 19
29. Comparison of modeled and measured plate sinkage test
for laboratory and field tests .................................................................................. 20
30. Modeled snow density of plate sinkage test for Drucker–Prager cap material ...... 21
31. Measured displacement and snow density in laboratory plate sinkage test ........... 21
32. Deformation of fresh snow under a plate ............................................................... 22
33. Grain size curves for McCormick Ranch sand and Lebanon sand ........................ 22
34. Validation of McCormick Ranch sand Drucker–Prager cap model
with uniaxial strain test data .................................................................................. 23
35. National Tire Modeling Program model of a radial tire ........................................ 24
36. Tires used in experiments ...................................................................................... 25
37. Deflection measurements for the tires used in the experiments ............................. 26
38. Comparison of measured contact area for the three test tire at 241 kPa ................ 27
39. Construction elements of the Darnell tire model ................................................... 28
40. Cut tire sections used to determine composite tire properties ................................ 28
41. Tire geometric parameters for the sidewall model ................................................. 29
42. Shoop-Darnell tire model ...................................................................................... 30
43. Modal analysis tire model with smooth tread ........................................................ 32
44. Modal analysis tire model with ribbed tread .......................................................... 33
45. Measured deflection for three tire pressures .......................................................... 35
46. Comparison of measured and modeled deflection for the Shoop–Darnell,
smooth tread, and ribbed tread models for three inflation pressures ..................... 35
47. Comparison of all measured and modeled deflections
at 241 kPa inflation pressure .................................................................................. 36
48. Measured contact areas at three inflation pressures ............................................... 36
49. Comparison of measured and modeled contact areas at three inflation pressures . 37
50. Comparison of Shoop–Darnell model results with measured contact areas .......... 38
51. Comparison of the smooth and ribbed modal analysis tire models
with measured contact areas .................................................................................. 38
52. Measured and modeled contact stress distribution
for half carcass on a hard surface ........................................................................... 39
53. Irregular stress contours generated by the Shoop–Darnell model
on a rigid surface ................................................................................................... 39
54. Close-up view of contact for the ribbed tire model ............................................... 39
55. Measured and modeled sidewall profiles for the Shoop–Darnell tire model ......... 40
56. Model of wheel rolling on 20 cm of fresh snow .................................................... 43
57. Comparison of standard mesh and ALE mesh ....................................................... 43
58. The CRREL Instrumented Vehicle ........................................................................ 44
59. Configuration of speed sensors and axle-mounted load cells on the CIV .............. 44
60. Parameters used to predict motion resistance using the NRMM algorithm ........... 46

iv
Figure
61. Modeled resistance force in coefficient form (longitudinal/vertical) and vertical dis-
placement of the wheel hub in 20-cm snow ........................................................... 47
62. Finite element model, measured data, and NRMM predictions
for sinkage and motion resistance in fresh snow ................................................... 47
63. Finite element simulations at zero slip and at unlimited slip
and NRMM motion resistance predictions for 240-kg/m3 snow ........................... 48
64. Marking the snow to observe snow deformation after vehicle passage ................. 49
65. Comparison of measured displacement to modeled displacement
and deviatoric stress in the longitudinal direction ................................................. 50
66. Comparison of measured and modeled displacement
in a cross section transverse to the direction of travel ........................................... 50
67. Comparison of measured and modeled densities
in the cross section transverse to the direction of travel ........................................ 51
68. Rolling tire on an elastic material (sand). .............................................................. 52
69. Rolling Shoop–Darnell tire on a Drucker–Prager cap model
of the McCormick Ranch sand .............................................................................. 53
70. Deformed mesh of modal analysis tire model in snow after 8 cm of sinkage ....... 54
71. Close-up of a tire after sinking 8 cm into the snow at the beginning of roll .......... 54

TABLES

Table
1. Initial material model parameters .......................................................................... 15
2. Pressure–volume data calculated from the Abele and Gow (1975) data set
for 200- to 220-kg/m3 snow at 0° to –3°C ............................................................. 16
3. Hardening table for fresh snow with a density of 200 kg/m3 ................................. 18
4. Material constants for Drucker–Prager cap model of McCormick Ranch sand ..... 23
5. Hardening model for McCormick Ranch sand (HKS, 1996) ................................. 23
6. Tire characteristics for test tires at an average vertical load of 6227 N ................. 25
7. Bending stiffness of tire sections ........................................................................... 29
8. Properties of sidewall elements and tire model components ................................. 29
9. Features of the modal analysis models with smooth and straight ribbed tread ...... 34
10. Qualitative summary of runtime analysis .............................................................. 41
11. Measured sinkage and resistance in snow ............................................................. 45

v
NOMENCLATURE

a arclength in contact with snow

c cohesion
c′ apparent cohesion
d Drucker–Prager material cohesion
e void ratio
E Young’s modulus
E1 tread longitudinal shell modulus
E2 tread lateral shell modulus
Fc cap failure surface for compactive–dilatant failure
Ffoam foam model failure surface
Fs shear failure surface
Ft transitional failure surface
g flow potential surface for crushable foam model
Gs flow potential shear surface
Gc flow potential cap surface
G shear modulus defining recoverable shear strains
h snow depth
i slip
I moment of inertia
Ii invariants of normal stress
j relative slip distance between the wheel and the terrain
Ji invariants of deviatoric or shear stress
k constant in equation defining friction as a function of slip distance (eq 26)
K flow stress ratio (related to shape of yield surface in deviatoric plane)
K0 coefficient of earth pressure
l beam length
M slope of the critical state line in the effective stress plane
ncl normal compression line
csl critical state line
p mean (total) normal stress
p′ mean effective normal stress
pa defines the evolution of the hardening and softening of the cap (location of Drucker–
Prager cap along the mean pressure axis)
pb hydrostatic compression yield stress (Drucker–Prager model)
pc hydrostatic compression strength (foam model)
p c| 0 initial value of hydrostatic compressive stress (foam model)
pmax maximum contact pressure
pt hydrostatic tensile yield stress
pelt elastic tensile limit
q deviatoric stress
r tire radius
R cap eccentricity parameter (controlling the shape of the cap)
Rs motion resistance force
si component of stress deviation (σi – p)
t stress deviator parameter, t = q when R = 1
tbeam thickness of the beam
T temperature
u pore water pressure
v specific volume
vw wheel speed

vi to contents
vv vehicle speed
w maximum tire width
wbeam width of the beam
W applied load on the end of the beam
ymax maximum deflection at the free end of the beam
z sinkage

α parameter to smooth the transition between the shear and cap failure
β Drucker–Prager material angle of friction
ε invol volumetric plastic strain
εinvol|0 initial value of volumetric plastic strain
κ elastic slope on the pressure–volume curve
λ plastic slope on the pressure–volume curve
µ coefficient of friction
ρ density
ρ0 initial density
σ0 initial yield stress in uniaxial compression
σh horizontal stress
σ normal stress
σi normal stress in the i direction, where i = x,y,z (principal stress when i = 1,2,3)
σoct octahedral normal stress
σv vertical stress
τ shear stress
τij shear stress, where i,j = x,y,z
τoct octahedral shear stress
ν Poisson’s ratio
φ friction angle
φ′ apparent friction angle

vii to contents
 Finite Element Modeling of Tire–Terrain Interaction

SALLY A. SHOOP

INTRODUCTION terrain are highly deformable. A rigid approximation

of either is an oversimplification of the problem and
Vehicle mobility on unpaved surfaces is important would yield erroneous contact conditions, which are
to the military as well as to agriculture, forestry, min- critical to obtaining correct solutions. The incorpora-
ing, and construction industries. The problems can be tion of the third (lateral) dimension is similarly nec-
grouped into two major categories: predicting vehicle essary, as the lateral movement of the terrain (in ad-
performance on various terrains (will it get stuck; dition to deformation in front and beneath the tire) is
how much traction or pull is available to climb or readily apparent for most soft terrain materials (soils
pull) and estimating the consequences of the vehicle and snow). The effect of the lateral dimension is also
passage (rut formation, shearing/tearing of roots, soil evident in equations for rolling resistance, which is a
compaction, and the effects of these on vegetation function of the width of the rolling object (Bowden
and erosion). and Tabor 1964, Ludema 1996a). Similarly, empiri-
The objective of this project is to produce a three- cal sinkage and motion resistance equations by Bek-
dimensional finite element model of tire-terrain inter- ker (1969) for soil and Richmond (1995) for snow
action that can be used to explore the effects of tire include the tire or track width, indicating the impact
and terrain variables on vehicle mobility (traction and of the third dimension on the performance of the tire
motion resistance) and terrain deformation (rut for- or track.
mation and shearing). Such a model would enable The details of the tire–soil modeling problem are
detailed analysis of the complex interactions resulting divided into three topic areas: 1) material models for
from contact friction and would further the under- the terrain material, 2) tire models for use on a de-
standing of off-road vehicle mobility by defining formable substrate, and 3) the combined tire–terrain
critical mechanisms involved in vehicle traction and model and the treatment of the interface. These are
motion resistance. Ultimately the model generated presented after a brief historical perspective on tire–
would be used for tire design and specification for terrain interaction.
off-road vehicles (for construction, mining, and rec-
reation), for vehicle performance prediction, and for
terrain damage prediction and reduction of the BACKGROUND
environmental impact of off-road travel. Previously a
three-dimensional simulation of contact between a Off-road vehicle performance prediction
deformable tire and deformable terrain had been too Off-road vehicle performance analysis is a three-
difficult and computationally time consuming. Re- dimensional, nonlinear, dynamic problem. Major
cent advancements in the contact formulations of early efforts were led by Bekker, at the University of
general-purpose finite element codes (e.g. ABAQUS, Michigan and the U.S. Army Land Locomotion
HKS 1998) and increases in computer processing Laboratory (Bekker 1956, 1960, 1969), and his con-
speeds have brought such a model into the realm of temporaries at the National Tillage Laboratory at
possibility. Auburn, Alabama (Gill and Vanden Berg 1967).
The originality of this research lies in accounting Karafiath and Nowatzki (1978) extended the state of
for the deformable nature of both the tire and the ter- the art to include modeling of soil deformation under
rain in a fully three-dimensional model. Both tire and the wheel or track based on theoretical soil mechan-

to contents
ics using plasticity theory. They developed a two- In this study, only fresh snow within a limited
dimensional finite difference model of tire–terrain range of initial density (less than 250 kg/m3) was
interaction that simulated the plastic soil deformation modeled. The snow depth, however, ranged from
under a tire, accurately reproducing the experimental “shallow” to “deep.” Shallow snow is defined as
measurements of sand deformation under a rigid snow where the “pressure bulb” (the snow compacted
wheel by Wong and Reece (1967), as shown in Fig- due to the weight of the vehicle) intersects a rigid
ure 1. More-recent seminal books on the subject are interface. In deep snow the pressure bulb does not
those by Yong et al. (1984), which includes some of intersect a rigid interface below the snow (Fig. 2).
the early finite element modeling of tire–terrain inter-
action, and Wong (1989), who approaches numerical
modeling of tracked vehicles based on Bekker’s
semi-empirical representation of the terrain.

Figure 2. Shallow and deep snow under a wheeled

vehicle.

Since the pressure bulb forms in response to a load

on the snow, the classification of deep or shallow
snow is vehicle dependent and therefore is a function
of the vehicle’s weight, load distribution, tire or track
Figure 1. Sand deformation under a towed wheel geometry, tire pressure, etc. Nonetheless, a rule of
moving to the right. Dashed lines show the soil thumb is that snow deeper than two thirds of the
failure zones. (From Wong and Reece 1967.) wheel radius is usually “deep” snow. Except for very
specialized vehicles, greater snow depths will impede
Because vehicle–terrain interaction is compli- wheeled vehicle travel because of excessive sinkage.
cated, comprehensive modeling of off-road mobility Generally, tracked vehicles with low ground contact
is usually done using empirical algorithms based on pressure are used to traverse deep snow. Figure 3
large data sets, such as the NATO Reference Mobil- shows cross sections of the pressure bulb under a
ity Model (Ahlvin and Haley 1992). These are sup- vehicle for shallow and deep snow.
plemented with high-fidelity simulations using so-
phisticated vehicle dynamics models, but in the Tire–terrain modeling
future these simulations will use vehicle–terrain nu- Because of the complexities involved in vehicle–
merical models. terrain interaction and the desire to incorporate more
of the physics into performance models, there is great
Vehicle movement on snow interest in applying numerical modeling techniques.
Vehicle mobility on snow has some unique char- Several researchers have produced numerical models
acteristics and terminology. The performance of a simulating tire–terrain interaction over the last 20
vehicle in snow is usually limited by the bearing years. Many of the models are two-dimensional
strength of the snow, which controls the amount of simulations of a rigid wheel on a deformable surface.
sinkage. Sinkage is a function of the compressibility Yong and Fattah (1976) and Yong et al. (1978) intro-
and shear strength of the snow in response to the ve- duced the first finite element models of tire–terrain
hicle load. Since snow strength is difficult to measure interaction using a two-dimensional rigid tire on de-
in the field on a routine basis, vehicle performance is formable terrain and then progressing to an elastic
often related to index parameters that are easier to tire and calculating the contact using Hertz contact
measure, such as the snow depth and density. theory. This model was difficult to use, as it needed the

2 to contents
a. Snow deformation under a Jeep Cherokee (2.7 b. Snow deformation under a BV202 (a 3.5-ton, 31-
ton, 24 kN) in 19-cm “shallow” snow. kN, articulated tracked vehicle) in a 122-cm
“deep” snow cover. (From Harrison 1975.)
Figure 3. Snow cross sections showing pressure bulb formation.

contact displacements or stresses as input values. IKK, including the developments by Aubel and
Nevertheless, it was a major advance in applying Fervers, is given in Schmid (1995). Fervers further
numerical methods to tire–terrain interaction. Model- developed his model to a pseudo-three-dimensional
ing the interface between a deformable soil and a representation and used it in several applications,
deformable wheel has been both problematic and including travel over rough surfaces, soil compaction
computationally time consuming. Pi (1988) modeled beneath a rolling tire, and the influence of tread de-
a two-dimensional elastic wheel (connected springs) sign and slip on tire performance (Fervers 1997,
on a viscoelastic soil for high-speed landing of air- 1999a, 1999b).
craft on soil. Others have used estimates of the con- Current efforts toward the development of a three-
tact stress distribution predefined on the soil surface dimensional model of a deformable tire on deform-
(Saliba 1990, Chi and Tessier 1995). This approach able terrain are being undertaken by 1) the U.S. Army
simplifies the problem and is suitable for analyzing Cold Regions Research and Engineering Laboratory
different terrain parameters but has the major disad- (CRREL), Goodyear, and Caterpillar through a Co-
vantage of requiring an estimate of the contact dis- operative Research and Development Agreement; 2)
placements or stresses a priori when this is nearly Tordesillas (1996), using a mathematical contact me-
impossible to predict and very difficult to measure. chanics approach at the University of Melbourne; and
Continued advances in computing capabilities and 3) Shoop, using the Darnell tire model at the Univer-
in general-purpose finite element codes have enabled sity of Michigan, Automotive Research Center, and
researchers to concentrate on the physics of the CRREL (Mousseau and Hulbert 1996, Darnell et al.
model rather than on code development. Sophisti- 1997, Alverez Sanz 1999, Darnell, in progress). The
cated two-dimensional models of a rigid wheel on Automotive Research Center (IKK) in Hamburg,
deformable soil have been described by Liu and Germany, and the Transport Technology Research
Wong (1996) and Foster et al. (1995). Models have Laboratory, Carleton University, Ottawa, Canada,
recently been extended to three dimensions by Chi,* may also be continuing their research. Researchers at
Chiroux et al. (1997), and Shoop et al. (1999). the Virtual Proving Ground at the National Automo-
To date, the most realistic representation of a tive Dynamic Simulator in Iowa have recently begun
pneumatic tire on deformable soil was done in two to pursue this area as well. The research described in
dimensions at IKK (Automotive Research Institute) this report uses the Darnell tire model, which is effi-
in Hamburg, Germany (Aubel 1993, 1994) (Fig. 4). cient enough to make simulations of a rolling tire
Fervers (1994) extended Aubel’s model to study the achievable in near-real time, so the addition of a de-
effect of lug design on tire performance (Fig. 5). A formable terrain is computationally feasible. This
good review of the tire–terrain research program at study also includes the development of a terrain ma-
terial model for snow.

* Personal communication, L. Chi, Caterpillar, Inc.,

Peoria, IL, 1996.

3 to contents
 Figure 4. Two-dimensional modeling of tire–terrain interaction
by Aubel (1993).

Figure 5. Modeling of the effect of lug design on tread–terrain

interaction in two dimensions by Fervers (1994).

Tire terminology
Because of the interdisciplinary nature of this
work, some of the tire terminology and conventions
are presented here, as they are used throughout the
Longitudinal
text. The tire direction convention is based on the
SAE (Society of Automotive Engineers) standard
definitions for vehicle dynamics. Since the tire–
terrain models are currently formulated for straight- Lateral or
ahead rolling (zero slip angle), the tire longitudinal Transverse
axis is in the direction of travel and the lateral or
transverse direction is perpendicular to travel (in the
horizontal plane), as shown in Figure 6.
The tires modeled are modern radial tires, which are
constructed with numerous components, as shown in Fig- Vertical
ure 7. Each of these elements contributes to the structural
behavior of the tire and is considered either individually
or en masse when creating deformable tire models. Figure 6. Tire direction convention.

4
to contents
 Tread Rubber
Belt

Belt Cushion

Inner Liner

Sidewall Rubber Ply

Filler
Rubber Chafer Bead
Apex
Bead
Filler
Wrapping
Bead Wires

Figure 7. Construction components of a radial ply tire.

(After Goodsell 1995.)

Tire deflection is a primary measure of the tire TERRAIN MATERIAL MODELS

structural response to load and is often reported based
on inflation pressure and vertical load. Deflection is Terrain substrate subjected to vehicle loads has
the difference between the unloaded and the loaded been represented using a wide variety of material
section height (and is sometimes reported in percent- models, including elastic, non-linear elastic, visco-
age, normalized by the unloaded section height). The elastic (Pi 1988), and elastic-viscoplastic (Saliba
tire cross section and the deformation of the cross sec- 1990). Recent studies have concentrated on using
tion under load are illustrated in Figure 8. A thorough either cap Drucker–Prager plasticity (Aubel 1993,
discussion of tire mechanics is given in Clark (1981). 1994, Fervers 1994) or critical state models such as
the Bailey and Johnson (1989) soil compaction model
implemented by Foster et al. (1995) or a new critical
state model (similar to Lade and Kim 1995) imple-
mented by Liu and Wong (1996). Comparisons be-
tween the Drucker–Prager and Cam–clay models for
tire–terrain interaction have been published by
Meschke et al. (1996) for snow and by Chi and
Tessier (1995) for soil. Although the Cam–clay
model is perhaps the most widely used soil model,
both of the above studies comment on convergence
problems when using Cam–clay. The choice of mate-
rial model is based on balancing the type of behavior
desired in the model with the information available to
determine model parameters. The terrain materials
simulated in this study were fresh snow and com-
pacted sandy soil. The two material models used
were a modified Drucker–Prager with a cap, and a
critical state, crushable foam model. Both models
were considered suitable for capturing the highly
compressible behavior of natural fresh snow. The
compacted sand was modeled using the modified
Drucker–Prager cap only. A description of these
models, model parameter determination, and valida-
tion of the material models with test data follows.

General concepts for plasticity models

The purpose of a plasticity model is to define the
Figure 8. Measurement of tire dimensions. permanent deformation of a plastic material. This is

5
to contents
done with the following basic components (Wood and Goodier 1970; Jaeger and Cook 1969), the stress
1990): invariants are:
• Elastic properties to define the recoverable de-
Ii = invariants of normal stress:
formation;
• A mathematical surface to define the yield
I1 = σ x + σ y + σz = σ1 + σ2 + σ3 (1)
boundary between elastic behavior and plastic
material behavior;
• A plastic flow potential to mathematically de- ( )
I 2 = − σ y σ z + σ z σ x + σ x σ y + τ yz 2
(2)
fine the plastic deformation (also called a plas-
tic flow law); and +τ zx 2 + τ xy 2 = − (σ 2 σ3 + σ3 σ1 + σ1σ 2 )
• A hardening/softening rule defining the move-
ment (expansion or contraction) of the yield I 3 = σ x σ y σ z + 2τ yz τ zx τ xy − σ x τ yz 2
surface during plastic deformation. (3)
−σ y τzx 2 − σz τ xy 2 = σ1 σ 2 σ3
A good overview of the development of plasticity
theory and constitutive modeling of soil is given in
Scott (1985). Schofield (Schofield and Wroth 1968) Ji = invariants of deviatoric or shear stress:
extended plasticity theory to the critical state concept,
defining either contractile or dilatant deformation of J1 = sx + sy + sz (4)
porous material as a function of its specific volume or
void ratio. In critical state theory this rule is devel-
(
J 2 = 16  σ y − σz ) + (σz − σx )2 + (σx − σy ) 
2 2
oped around the concept of a “critical state,” where 
the plastic shearing deformation occurs at a constant
volume. Perhaps the most famous critical state +τ yz 2 + τzx 2 + τ xy 2
model, the Cam–clay model, was developed based on
the behavior of clays. The concepts are equally appli- = 16 (σ 2 − σ3 ) + (σ3 − σ1 ) + (σ1 − σ2 ) 
2 2 2
  (5)
cable to defining the shearing and volumetric behav-
ior of granular materials such as granular soils or
snow (Wood 1990). Although the concepts are appli-
cable for both cohesive and granular materials, the (
= I12 + 3I 2 / 3 )
behavior of the granular materials has not been ex-
plored as thoroughly, particularly regarding the influ- J 3 = sx sy sz + 2τ yz τ zx τxy − s x τ yz 2 (6)
ence of deviatoric stress on the yield surface, which
2
is less clearly defined in soils but may take on a much − sy τzx − sz τ xy2
different shape than the yield surface of metals.
Wood (1990) considers this a difference in detail
rather than a difference in concept. (
= 2 I13 + 9 I1 I 2 + 27 I 3 / 27)
The critical state class of model used in this study is
the crushable foam model, specifically designed for where the components of stress deviation si are
highly compressible materials, is which a characteristic
of fresh snow. The modified Drucker–Prager cap si = σi − p for i = x, y, z, and i = 1, 2, 3. (7)
model also has the features of a critical state model
(i.e. regions of constant volume shear deformation, and
Model parameters are defined in either the pres-
compactive–dilatant flow). Both models use non-
sure–deviatoric plane (also called the meridinal, p–q,
associative flow (i.e. the flow potential is not associ-
or p–t plane) or the pressure–volume plane (v–p or v–
ated with the yield surface), except on the cap surface
ln p plane) according to the following definitions.
of the Drucker–Prager model.
1. Mean (total) normal stress, also called equiva-
lent pressure stress, or the octahedral normal stress,
Yield surface
which determines uniform compression or dilation:
The yield surface for both material models is de-
scribed in terms of stress invariant functions of the
normal σi and shear stress τij, where i and j are direc- p = σoct = 13 I1 . (8)
tions x, y, and z and represent principal stresses when
i = 1, 2, or 3. Based on elasticity theory (Timoshenko

6 to contents
For soils, p is often taken as the mean effective stress
p', taking into account the pore water pressure, u:

p' = p − u . (9)

2. Deviatoric stress, also called the Mises equiva-

lent stress, or the octahedral shear stress, which de-
termines distortion:

q = τoct = 2
9 (I1
2
)
+ 3I 2 = 2
3
J2
.
(10)

3. An additional deviatoric stress measure, t (HKS

1998):

q LM
1 1 FG IJ FG J3 IJ 3 OP
t = 1+ − 1−
2 KMN K H K H q K PQ (11)
Figure 10. Yield surfaces in deviatoric space.
(After HKS 1998.)

where K is a material parameter (between 0.778 and

Hydrostatic Hydrostatic
1.0) that controls the yield dependence on the third σ1 Axis σ1 Axis
stress invariant and defines the shape of the yield
surface in the deviatoric plane. The K value allows
for different stress failures in tension and compres-
sion. For K = 1, the yield surface is circular (von
Mises yield), as shown in Figure 9, and the failure σ3 σ3
stress is the same in tension and compression. K is
limited to values of 0.778 or greater for the surface to
remain convex. Some of the more common failure σ2 Tresca σ2 von Mises

surfaces are shown in Figure 10 in deviatoric space Hydrostatic Hydrostatic

and in Figure 11 in principal stress space. σ1 Axis σ1 Axis

σ3 σ3

σ2 Mohr-Coulomb σ2 Drucker-Prager

Hydrostatic
σ1 Axis

σ3

σ2 Lade
Figure 9. Modified Drucker–Prager yield sur-
face in deviatoric space. (After HKS 1998.) Figure 11. Common yield surfaces along the hy-
drostatic axis in principal stress space. Caps
defining compactive failure are not shown. (After
Shen and Kushwaha 1998.)

7 to contents
 Cap plasticity models incorporate both the com- law (Fig. 14). In cases where exponential hardening
pression and shearing of the material. Critical state is not a good fit, as is common in soils, the hardening
models are a type of cap model where a critical state law can also be represented in a piecewise linear ap-
line (csl) is located at the peak of the yield surface proach using the experimental data (Fig. 15).
and divides material yield by compaction on one side
and yield by dilation on the opposite side. The mate-
rial deforms at a constant volume at the critical state.
The critical state line (curve) in three-dimensional
space defined by normal compressive stress p', devia-
toric (shear) stress q, and specific volume v is shown
in Figure 12. For simplicity in gathering the model
parameters, the yield surface is generally viewed in
planar form in the compressive plane (v–p or v–ln p)
or the deviatoric plane (p'–q or p–t).

Hardening law
The cap on the yield surface defines the compac-
tive–dilatant behavior of the material failure as the
cap expands and contracts (Fig. 13). The cap is gen-
erally spherical or ellipsoidal, and the material either
hardens or softens by expanding or contracting the
cap. This behavior is defined in the pressure-volume
relationship called a hardening law. The pressure– Figure 12. Three-dimensional view of critical state
volume relationship is often exponential and there- yield surface. The critical state line defines failure
fore can be modeled using an exponential hardening at constant volume. (After Wood 1990.)

Figure 13. Cap contraction or expansion reflecting the softening (left)

or hardening (right) of the material. (After HKS 1998.)

8 to contents
 Figure 14. Exponential hardening law for material in hydro-
static compression.

Figure 15. Piecewise linear modeling of the

hardening law. (After HKS 1998.)

Hardening can be modeled as isotropic hardening, some of these features without the cumbersome
where the yield surface grows equally in all direc- tracking of yield surfaces by engaging a series of
tions, or as kinematic hardening, where the yield sur- yield surfaces (failure mechanisms). The MMM con-
face moves as a rigid body in principal stress space cept has been successfully applied to modeling the
(Fig. 16). Kinematic hardening is more accurate for cyclic loading of pavements by Smith (2000). This
modeling cyclic loading of soils but is more compu- constitutive model is still under development and is
tationally intensive. A combination of the two types currently being calibrated for loose, thawing soil be-
of hardening is probably the best representation for havior for future applications to tire–terrain model-
cyclic loading of soil. The Multi-Mechanical Model ing. The material models implemented in this study
(MMM) proposed by Peters (1991)* implements use isotropic hardening, which is considered a rea-
sonable approximation for monotonic loading (one
pass of the wheel).
* Also personal communication, Geotechnical and
Structures Laboratory, ERDC, Vicksburg, MS, 1999.

9 to contents
 M = slope of the critical state line in the p–q
plane, which controls the shape of the
yield ellipse
E = Young’s modulus of elasticity defining
recoverable strain
ν = Poisson’s ratio.
The following sections explain the Drucker–
Prager cap model and the crushable foam model in
more detail.

Modified Drucker–Prager cap model

For the modified Drucker–Prager cap model used
in this study, the yield surface is a modified von
Mises yield (i.e., the material constant K = 1.0),
which is circular in the deviatoric plane (Fig. 9). In
the p–t plane the yield surface has two major seg-
ments: the Drucker–Prager shear portion of the curve
(analogous to the Mohr–Coulomb line) defines shear
deformation, and the cap portion of the surface de-
fines the intersection with the pressure axis (Fig. 17).
The following equations define the yield criteria in
each section of the yield surface.
For the Drucker–Prager shear or distortional fail-
ure:

Fs = t − p tan β − d = 0 . (12)

Figure 16. Isotropic and kine- For the cap region of compactive–dilatant failure:
matic hardening. (After Shen
and Kushwaha 1998.) 2
 Rt 
Fc = ( p − pa ) 2
+ 
Parameters  (1 + α − α / cos β )  (13)
A relatively small number of parameters are
needed for these material models. Although more − R ( d + pa tan β ) = 0
detailed models may provide a more accurate solu-
tion, the parameters needed to define the model where α is a transition parameter, ranging typically
become prohibitively numerous and are often un- from 0.0 to 0.05, that smooths the transition between
available. The objective of any model is to capture the shear failure and the cap failure, R is a material
the most important material behaviors without being parameter controlling the cap eccentricity, and pa
unduly cumbersome. The following descriptions of defines the evolution of the cap along the pressure
the material parameters are based on the models as axes (via pb–εvol hardening law) according to
implemented in ABAQUS (HKS 1998). Additional
information on soil plasticity and critical state soil pb − Rd
pa = . (14)
mechanics is given in Wood (1990). (1 + R tan β )
The model parameters used in this study are:
The transition between the shear and the cap fail-
d = Drucker–Prager material cohesion ure is
β = Drucker–Prager material angle of friction
λ and κ = slopes of the loading (compression) and 2
  α  
unloading (elastic) lines in the compression Ft = ( p − pa ) 2
+ t −  1 −  ( d + pa tan β )
plane as indicated in Figure 14. This hard-   cos β  
ening or softening of the material can also
be defined in a piecewise manner (Fig. 15). −α ( d + pa tan β ) = 0 . (15)

10 to contents
 Figure 17. Modified Drucker–Prager cap yield surface in the p–t
plane. (After HKS 1998.)

The pressure–volume relationships define both mal to the surface) in the cap region; therefore, the
hardening and softening through volume changes equation for the flow surface is identical to the equa-
based on how the cap portion of the yield surfaces tion for the cap yield surface. In the transition and
expands and contracts. The hardening of the cap can shear regions, the flow is non-associative (the flow
be defined as an exponential relationship in the pres- potential is not perpendicular to the failure surface)
sure–volume space as shown in Figure 14, or it is and is defined by the following equation for the flow
entered in a piecewise fashion (Fig. 15) as a table of potential:
pb and ε plvol pairs. The piecewise approach is recom-
mended by HKS (1998) for a better fit to the data and
better model performance. Both methods were used 2
2  t 
in this study. Gs = ( pa − p ) tan β  +   . (16)
The plastic flow is defined by an elliptical 1 + α − α / cos β 
potential surface (Fig. 18). Flow is associative (nor-

Figure 18. Modified Drucker–Prager cap flow potential in the p–t

plane. (After HKS 1998.)

11 to contents
 Figure 19. Yield surface for the crushable foam model.
(After HKS 1998.)

pl
dε

σ
g

Figure 20. Plastic flow surface for the crushable foam model.

Crushable foam model This assumes a flow direction identical to the

The crushable foam model is a class of critical stress direction for stress paths along the p axis,
state model and has many similarities to the form of which suggests that loading in any principal direction
the Drucker–Prager cap model. The crushable foam results in insignificant deformation in the other direc-
model, however, has a single ellipse forming the tions (HKS 1998). The validity of this assumption for
yield surface (Fig. 19), defined by (HKS 1998) snow was examined as part of this study.
The hardening law takes the same form as the
Ffoam = f − f0 (17) hardening law for the Drucker–Prager cap model.
1
 p − p   t  
2 2
p + pt
2
Constitutive models for snow
=  t c
+ p +   − c
 2   M   2 Two material models were judged to have per-
formance characteristics most similar to natural fresh
snow: a Drucker–Prager cap plasticity model and a
where pt is the material strength in hydrostatic ten- critical state plasticity, crushable foam model. Liu
sion, pc is the material yield stress in hydrostatic (1994) used the Drucker–Prager plasticity model for
compression, and M is the slope of the critical state modeling the sliding of rubber blocks on snow.
line in p–t plane. As indicated in Figure 19, the yield Mundl et al. (1997) extended Liu’s work using a
surface hardens and softens with yield but the pt is multi-surface plasticity model to optimize the snow
assumed constant (isotropic hardening). behavior under both shear and compression. Their
In the crushable foam model the plastic flow is intent, however, was to simulate shearing forces on
non-associative and perpendicular to the flow surface compacted snow on roads (with a density of 500
g (Fig. 20): kg/m3), which is a significantly different material
9 2 than encountered during cross-country mobility on
g= p + q2 . (18) fresh snow (with a density of 200–300 kg/m3). For
2

12 to contents
this study the plasticity model used was the well- Poisson’s ratio ν, were estimated based on data com-
documented Drucker–Prager cap. piled by Mellor (1975) and expanded on by Shapiro
The second material model, a crushable foam et al. (1997) (Fig. 21). Since the elastic modulus is a
model, was chosen based on work by Johnson et al. strong function of density, varying over four orders
(1992), where a crushable foam model was success- of magnitude, the elastic modulus should ideally be
fully applied to natural snow. Johnson used modeled as a function of the density as a state vari-
PRONTO, an explicit finite element code from San- able. For this study, however, Young’s modulus was
dia National Lab, to simulate shock wave propaga- held constant at a value equivalent to snow with a
tion in one dimension. For tire–snow interaction, density of 300–400 kg/m3. At snow densities less
however, the model must simulate snow deformation than this, the elastic contribution will be minimal.
in a three-dimensional stress field. Initial simulations
using the ABAQUS crushable foam model showed
that no lateral deformation resulted from an axial
load. This issue prompted questions regarding the
deformation behavior of snow under a purely vertical
load and whether the predicted response from a
crushable foam model was suitable for snow in a
three-dimensional stress state. Experiments measur-
ing the lateral deformation of snow under a purely
vertical load were conducted in cooperation with the
Keweenaw Research Center in Houghton, MI (Shoop
and Alger 1998). Results indicate that vertical load-
ing of non-sintered snow causes primarily vertical
deformation. Fukue (1979) performed experiments
that showed that lateral deformation was minimal for
high load rates. Both of these studies support our
choice of crushable foam models for vehicle loading.

Material parameter determination

Extreme changes in snow properties with time, tem-
perature gradients, applied load, and deformation pro-
hibit model parameter acquisition from the same snow.
Therefore, approximations were made by estimating
parameters using test data from snow of similar charac-
teristics (density, age, snow type, and location). The
type of snow modeled was a fresh snow with a density
of 200–250 kg/m3 at moderate temperatures (between
–10° and –1°C). Test data were gathered from the field
and from the literature to match this snow type as
closely as possible. Because data were gathered from
A = Pulse propagation or flexural vibration at high
several studies, the snow was not exactly the same in
frequencies, –10° to –25°C
crystal structure, but this effect is minimized in fresh
B = Uniaxial compression, strain rate approximately 3
snow where sintering (bonding between snow grains)
× 10–3 to 2 × 10–2 s–1, temperature –25°C
has not occurred. Also, most of the tests were performed
C1 = Uniaxial compression and tension, strain rate ap-
on “lake effect” snow from the Keewenaw Peninsula in
proximately 8 × 10–6 to 4 × 10–4 s–1, –12° to –15°C
the Upper Peninsula of Michigan, which is a fairly con-
C2 = Static creep test, –6.5° to –19°C
sistent snow type under normal winter conditions. A
D = Complex modulus, 103 Hz, –14°C
discussion of selecting the initial values of the material
K = Static Young’s modulus and quasi-static Pois-
parameters follows.
son’s ratio
Elastic properties. For snow deformation under a
S = Quasi-static measurements of Poisson’s ratio
tire, the plastic deformation is much greater than the
elastic deformation; however, the elastic deformation Figure 21. Compilation of Young’s modulus and
contributes to energy losses during rolling resistance. Poisson’s ratio measurements on snow. [After
The basic elastic parameters, Young’s modulus E and Mellor (1975) and Shapiro et al. (1997)]

13 to contents
 Porous elasticity accounts for the elastic properties compaction beneath the tire, a plane strain response
of materials that undergo large volume changes, such may be more appropriate for deep snow than for shal-
as fresh snow. To implement a porous elasticity low snow, as illustrated in the snow deformation
model, the additional parameters needed are the slope measurements in Figure 3. The Drucker–Prager mate-
of the elastic portion of the pressure–volume curve rial cohesion d and the Drucker–Prager material an-
(during an unload–reload event, as illustrated in Fig- gle of friction β can be calculated from
ure 14); κ, representing the elastic portion of compac-
tion; and the elastic tensile limit, p elt . The elastic tan β = 1.73 sin φ (19)
compaction value was not available since compaction
tests were performed under monotonic loading (no d = c 1.73 cos φ. (20)
unloading), so κ was chosen to provide realistic elas-
tic behavior within the stress ranges of the simula- Acquiring apparent values for cohesion c' and fric-
tions. The elastic tensile limit p elt was chosen to be tion angle φ' from shear loading of snow under a ve-
the same as the same as tensile limit pt used to define hicle tire and from a ring shear device is discussed by
the yield surface. Blaisdell et al. (1990) and Alger and Osborne (1989).
Yield surface. The parameters defining the shear Both methods were applied to field measurements on
portion of the yield surface for the Drucker–Prager undisturbed snow with a density ranging from 60 to
model can be calculated from the Mohr–Coulomb 250 kg/m3 and temperatures from –2° to –16°C. The
cohesion c and the internal angle of friction φ, assum- results are shown in Figure 22, along with the values
ing plane strain response and non-dilatant flow. Al- used in the CRREL Shallow Snow Model (SSM 1.0)
though non-dilatant flow can be assumed for the (Blaisdell et al. 1990).

Figure 22. Calculated c' and φ' using measurements

from the CRREL Instrumented Vehicle (CIV) and the
ring shear device and predicted using the Shallow
Snow Model, SSM 1.0. (From Blaisdell et al. 1990.)

14 to contents
 Using these graphs, for a snow density of 200 Fig. 23). The cap eccentricity parameter R was cho-
kg/m3, values ranged from c' = 2.14 kPa and φ' = sen based on typical values for earth materials having
8.98° for the ring shear device to c' = 43 kPa and φ' = a very steep compaction cap (R = 0.02 to 0.0001).
14° using the CIV. Because of the nature of the test, The transition surface radius α was set to 0.0 for no
these values may be more representative of the inter- transition. The initial cap yield surface position ε plvol |
0

face shear. However, similar values of cohesion (ap- was arbitrarily set to 0.001 to allow initial softening.
proximately 20–30 kPa) were reported in Shapiro et For the foam model the initial yield stress in uni-
al. (1997). These values yield a Drucker–Prager β axial compression σ0 was based on data in Shapiro et
ranging from 15.2° to 22.5° and a d ranging from 3.7 al. (1997). The initial yield surface position in hydro-
to 72 kPa. static compression pc|0 (for exponential hardening)
Parameters describing the shape of the cap were was chosen to allow for initial softening or hardening
adjusted based on model response. The flow stress (about half the value of σ0). The hydrostatic tensile
ratio K (ratio of tensile strength to compressive strength pt was selected as 10% of pc|0 as a general
strength) was held at 1.0. This agrees with data pre- rule for foam materials. M is calculated from the
sented in Mellor (1975) and Shapiro et al. (1997), other input parameters.
indicating nearly equal values of compressive and These material parameters were refined based on
tensile strength for low-density snow (data set M in model response and are presented in Table 1.

Table 1. Initial material model parameters.

Drucker–Prager cap Foam Elastic
d = 5,000 to 30,000 Pa λ = 0.35 κ = 0.005
β = 22.53° pc|0 = 113.76 Pa ν = 0.3
– ε plvol |0 + ε plvol | = 0.0005
+
R = 2.2E-2 to 1.1E-4 pelt = 11.376 Pa
ε plvol |0 = 0.001 pc|0 + pt = 113.76 Pa E = 13.79 MPa
α = 0.0 pt = 11.376 Pa
K = 1.0 σ0 = 275.79 Pa
K = 1.0

Hardening. Two methods were used to handle the

hardening data: 1) using an exponential hardening
function and 2) using tabular data. For exponential
hardening, the slope of pure hydrostatic compression
behavior λ is obtained by plotting the void ratio
against ln[(p + pt)/pt], as shown in Figure 14, where p
is the hydrostatic pressure and pt is the strength in
hydrostatic tension. For the snow pressure–volume
relationship, void ratio was plotted against ln(p) since
pt was not precisely known. For snow, pt is a very
small quantity and only the slope is required, which
is minimally affected when p>>pt. These plots were
generated using the compression test data published
in Abele and Gow (1975).
Values for the logarithmic hardening modulus
were calculated using three different compression
tests of similar snow density, along with values cal-
culated from the final pressure–density pairs of an
additional 21 tests, all from Abele and Gow (1975).
A = Strain rates greater than 5 × 10–4 s–1
Since the measurements are from one-dimensional
B = In-situ tensile strength
compression or consolidation (oedometer) tests by
M = Uniaxial compressive and tensile strength
application of a vertical load to a sample within a
under rapid loading
rigid cylinder (i.e. uniaxial strain), the calculation of
Figure 23. Compilation of uniaxial strength data. the mean hydrostatic pressure p is not straightfor-
[After Mellor (1975) and Shapiro et al. (1997).] ward. One approximation calculates the horizontal

15 to contents
stresses σh based on earth pressure theory assuming p = 0.413σ1 . (24)
that they are related to the vertical stress σv by a fac-
tor of K0, known in soil mechanics as the earth pres- Using this expression to calculate the mean stress
sure coefficient, where from individual uniaxial consolidation tests with a
density of 200–390 kg/m3 (Abele and Gow 1975,
σh = K 0σ v . (21) tests 3, 10, and 29) results in a range of λ between
0.223 and 0.454.
The earth pressure factor can be calculated from
An additional data set was obtained using the
the friction angle using
combined data from 21 consolidation (oedometer)
K 0 = 1 − sin ϕ . (22) tests on snow with initial densities of 200–220 kg/m3
and temperatures of 0° to –3°C. These data are shown
Based on Alger’s shear measurements (Alger in Figure 24 and fall within the expected range of
1988, Alger and Osborne 1989), the average K0 value stress–density data for snow shown in Figure 25.
is approximately 0.12. Assuming that the maximum Selected points from the best-fit line of these data
principal stress is vertical (as would be expected in were used to calculate the values in Table 2 and are
uniaxial compression tests on undisturbed snow) and plotted in Figure 26 to obtain an average λ of 0.33.
using equations 8 and 21: The hardening table data were taken from an aver-
age of Abele and Gow tests on warm, 200-kg/m3
p= 1
3 ( σ1 + σ2 + σ3 ) = 13 (1 + 2 K 0 ) σ1 (23)
snow (tests 10 and 29) and are given in Table 3.
which, for K0 = 0.12, reduces to

4
10

ρ0 = 200 to 220
σ1, Major Principal Stress (MPa)

3
10

2
10

1
10
200 300 400 500 600 700 800 900

ρ, Density (kg m–3 )

Figure 24. Compression of natural snow at 0° to –3°C. (After Abele

and Gow 1975.)

Table 2. Pressure–volume data calculated from the

3
Abele and Gow (1975) data set for 200- to 220-kg/m
snow at 0° to –3°C.
ρ σ1 p = 0.413 σ1 Void ratio,
(kg/m3) (MPa) (MPa) ln p e
900 2.8 1.16 0.15 0.00
770 2 0.83 0.19 0.17
620 1 0.41 0.88 0.45
540 0.5 0.21 –1.58 0.67
470 0.2 0.08 –2.49 0.91
410 0.1 0.04 –3.19 1.20
380 0.05 0.02 –3.88 1.37

16 to contents
 A = Natural densification of snow at –1° to –48°C
B = Slow natural compression of dense firn and
porous ice (from polar caps)
C = Slow compression of solid ice
E = Calculated values from plane wave impact at
20–40 m/s
F = Hugoniot data for explosively generated shock
waves with impact velocity of 1–12 m/s at tem-
peratures of –7° to –18°C
–4 –1
J = Compression at strain rate 10 s at –7° to
–18°C
K = Compression in uniaxial strain with incre-
mental loading at –2° to –3°C
Solid Lines = Abele and Gow data, T = 0° to –34°C,
ρ0 = 90 to 270 kg/m3
Figure 25. Compression of undisturbed snow and ice. (After Abele and Gow 1975.)

1.6

1.2
λ = 0.33

0.8
e

0.4

0.0
–4 –3 –2 –1 0 1

ln(p)
Figure 26. Pressure–volume data from compression tests of Abele and Gow (1975).

17
to contents
 Table 3. Hardening table for fresh snow with a density of
3
200 kg/m .
p = 1/3 σ1 (Pa) p = 0.413 σ1 (Pa) p = σ1 (Pa) ε plvol
37.92 46.98 113.76 0
0.017E6 0.021E6 0.05E6 0.593
0.033E6 0.041E6 0.1E6 0.669
0.067E6 0.083E6 0.2E6 0.806
0.167E6 0.207E6 0.5E6 0.944
0.333E6 0.413E6 1.0E6 1.083
0.667E6 0.826E6 2.0E6 1.299
0.933E6 1.156E6 2.8E6 1.455
1.083E6 1.342E6 3.25E6 1.475
2.0E6 2.478E6 6.0E6 1.50
2.0E7 2.478E7 6.0E7 1.514

Figure 27. Comparison of model and experimental data for uniaxial

compression tests on low-density snow.

Model calibration against laboratory strength Two different plate sinkage tests were used in the
tests. A finite element model simulating the uniaxial comparisons. The field test consisted of pushing a 20-
confined compression tests was used to refine the cm- (8-in.-) diameter rigid plate, originally at the
material properties shown in Tables 1 and 3. A single surface, into 40 cm (16 in.) of fresh snow with aver-
element model was determined to be adequate after age density of 200 kg/m3 (Alger and Osborne 1989).
multi-element models gave the same results. The The laboratory test pushed a 23-cm- (9-in.-) diameter
uniaxial compressive test data used for calibration rigid plate into a cube of snow 60 × 60 × 50 cm deep
were from the best-fit line for fresh snow of density with an average initial density just under 200 kg/m3.
200–220 kg/m3 and temperatures of 0° to –3°C from Three different models were used to simulate the
Figure 24. A comparison between the experimental plate sinkage tests: 1) a Drucker–Prager cap material,
data and model results is given in Figure 27. The 2) a Drucker–Prager cap material with vertical shear
agreement between modeled and measured material surfaces along the snow–plate interface, and 3) a
behavior is good for both the Drucker–Prager cap crushable foam material model with shear surfaces
model and the crushable foam model. along snow–plate interface. The vertical frictional
surfaces allowed these elements to slide, reducing
Material model validation with element distortion, which more closely mimicked the
plate sinkage tests shear surfaces observed in the field. The slide planes
The initial model parameters were applied to a were modeled using a contact surface with Coulomb
plate sinkage test for snow of similar age and density. friction of µ = 0.3.

18 to contents
 plate and snow. The deformed meshes for each of the
models of the field test are shown in Figure 28. Fig-
ure 28b clearly shows the effect of the shear surface
on the sides of the plate, with the model deforming
similarly to field observations, compared to the unde-
sirable element distortion along the sides of the plate
when these elements are not allowed to slide (Figure
28a). In future models, adaptive meshing, where the
mesh adjusts to accommodate large distortions,
should help alleviate the problem of excessive distor-
tion without the use of the shear contact surface. Also
clearly seen in Figure 28 is the lateral deformation of
the snow beneath the plate in the Drucker–Prager cap
a. Drucker–Prager cap material.
models (Fig. 28a and b), but no lateral deformation is
seen in the crushable foam model (Fig. 28c).
The force–displacement curves for the plate mov-
ing into the snow are compared to the experimental
measurements in the laboratory and field in Figure
29. The model matches the controlled laboratory test
well, particularly for the Drucker–Prager cap model.
In the field test the hardening behavior was difficult
to simulate. The laboratory test was not run to maxi-
mum resistance and would be similar to a “deep
snow” condition, while the plate in the field test was
pushed to ultimate resistance, causing compaction
equivalent to a “shallow” snow condition.
Snow density can be calculated from the modeled
volumetric plastic strain using the hardening tables
b. Drucker–Prager cap with vertical shear contact and the pressure–volume relationships in Tables 2
surface. and 3. The snow density values from the models are
shown in Figure 30. These densities can be compared
to snow densities measured under the plate, as shown
in Figure 31. The maximum volumetric inelastic
strain in the plate sinkage model occurred in isolated
locations adjacent to the plate at values of approxi-
mately 0.5, which is equivalent to a snow density of
420 kg/m3, whereas very dense snow (440–520
kg/m3) was measured up to 5 cm below the plate
(Fig. 31). These values correspond to a strain of 0.5
m/m or greater (assuming primarily vertical deforma-
tion) on the uniaxial stress–strain model calibration
curve (Fig. 27). The divergence of the models and
data at strains greater than 0.5 m/m (Fig. 26) would
account for density discrepancies in the plate sinkage
c. Crushable foam with shear contact surface. measurements and models in high strain areas. On the
whole, however, most of the strain occurs at levels of
Figure 28. Deformed meshes for field plate sink- 0.4 or less, so the model provides a good fit to the
age simulations. overall plate sinkage behavior.
The laboratory tests were also used to evaluate the
lateral deformation of the snow, since field observa-
The plate was lowered by constraining the surface tions show lateral deformation under a vehicle load.
nodes radially while displacing them into the snow, On closer examination of the literature, however, few
effectively creating a no-slip contact between the examples were available where the snow had been

19 to contents
 a. Laboratory tests.

b. Field tests.
Figure 29. Comparison of modeled and measured plate sinkage test for
laboratory and field tests.

marked in a way that would enable the observation of the deformation (Shoop and Alger 1998). Fukue
lateral movement. (1979) reported similar observations in light snow
In addition, the nature of loading from vehicles is (120 kg/m3) as a function of penetration speed, as
not strictly vertical. Therefore, an experiment was shown in Figure 32. This substantiates the use of the
devised to mark several snow samples both vertically crushable foam model to represent fresh snow. How-
and horizontally and subject them to purely vertical ever, dense or bonded snow does show lateral defor-
loading. The snow in Figure 31 was one such ex- mation (Shoop and Alger 1998), as does snow sub-
periment, clearly indicating no lateral deformation, jected to slow loading (Fukue 1979) (Fig. 32a).
since the vertical marks remained vertical throughout

20 to contents
 a. Laboratory test simulation. b. Field test simulation.

3
Figure 30. Modeled snow density (kg/m ) for plate sinkage test for Drucker–Prager cap material.

22.86 cm

50

Maximum Load
180 kg/m 3 1867 N 180

40

180 190
Vertical Distance (cm)

Plate (Diameter=20.32 cm)

30
520
440
210 290 200

20 290

290
190 290 190

10 310

300
200 190
300
290
0
0 10 20 30 40 50 60

Horizontal Distance (cm)

Figure 31. Measured displacement and snow density in laboratory
plate sinkage test. (From Shoop and Alger 1998.)

21 to contents
 a. At 0.097 mm/s. b. At 0.98 mm/s.
Figure 32. Deformation of fresh snow under a plate. (From Fukue 1979.)

Soil model and wet. However, the compacted McCormick Ranch

The material model used for the wheel-on-soil sand model is useful for simulating a vehicle on an
simulation was also a Drucker–Prager cap plasticity unpaved secondary road under good (summer) road
model. Initially a very soft thawing soil was to be conditions.
simulated, but the testing of the soil parameters be- The mechanical behavior of the McCormick
came quite extensive and will be completed in the Ranch sand was documented by Mazanti and Holland
future. A compacted sand material called McCormick (1970a, 1970b), and material parameters for a modi-
Ranch sand was used for initial modeling. fied Drucker–Prager cap model were developed in
McCormick Ranch sand is a clayey sand, classified DiMaggio and Sandler (1971) and implemented in
as SC, from Albuquerque, New Mexico, and is simi- ABAQUS (HKS 1996). The material parameters are
lar in grain size distribution to the Lebanon sand used given in Table 4, and the cap hardening–softening
in mobility experiments on frozen and thawing soils behavior is defined using the piecewise linear ap-
at CRREL (Fig. 33). The McCormick Ranch sand proach given in Table 5. The material model yields
was tested in a compacted state, whereas soils that excellent comparison to uniaxial strain test results
have been subjected to freeze–thaw are usually loose (Fig. 34).
0.75 in.

0.38 in.
1.5 in.

0.5 in.

#140
#200
3 in.
2 in.

1 in.
6 in.

#10

#20

#40

#60
#4

100

80
Percent Finer

60

40

20
Lebanon Sand (SM)
McCormick Sand (SC)

0
200 100 10 1 0.1 0.01 0.001
Grain Size (mm)
Figure 33. Grain size curves for McCormick Ranch sand (Mazanti and Hol-
land 1970a) and Lebanon sand (Shoop 1990, 1993a).

22 to contents
 Table 5. Hardening model for
McCormick Ranch sand (HKS
1996).
Hydrostatic
stress, Pb Volumetric plas-
pl
(MPa) tic strain, ε vol
0.138 0

Table 4. Material constants for Drucker–Prager cap model of 0.172 0.005

McCormick Ranch sand (HKS 1996). 0.434 0.01
Material parameter Value 0.896 0.02
Young’s modulus, E 689 Mpa (100000 psi) 1.655 0.03
Poisson’s ratio, ν 0.25 2.758 0.04
Drucker-Prager material angle of friction, β 14.56°
Drucker-Prager material cohesion, d 1.194 MPa (173.2 psi) 4.137 0.05
Cap eccentricity parameter, R 0.1 6.895 0.06
Initial value of volumetric plastic strain, ε plvol |0 0.001
13.447 0.10
Flow stress ratio, K 1.0

Figure 34. Validation of McCormick Ranch sand

Drucker–Prager cap model with uniaxial strain test data
from DiMaggio and Sandler (1976). (After HKS 1996.)

23 to contents
TIRE MODELS FOR A DEFORMABLE
SUBSTRATE

Tire model construction

Background
To apply a tire model to deformable terrain, a
model is needed that is efficient yet accurately por-
trays the tire structural behavior. Specifically, an ac-
curate model of the contact patch is critical for simu-
lating the impact of deformable terrain on tire
performance. Models commonly used for tire design
purposes very accurately predict deformation of the
complete tire, including the interaction of the internal Figure 35. National Tire Modeling
components. However, since our concern is only the Program (NTMP) model of a ra-
deformation as it relates to the contact region and its dial tire. (After Danielson and
ability to roll across a surface, simpler models can be Noor 1997.)
employed for better computational efficiency. To this
purpose, several tire models were evaluated for effi- using methodology developed by Darnell at the Uni-
ciency and comparison to measured tire mechanical versity of Michigan for use in vehicle dynamics, 3) a
response, particularly deformation and stress related tire model of the type used for harmonic vibration
to contact. modal analysis, with a smooth tread, and 4) a tire
Modern tires are structurally complex, consisting model similar to model 3 except with a straight (lon-
of layers of belts, plies, and bead steel imbedded in gitudinal) ribbed tread. The modal analysis models
rubber (Fig. 7). Materials are often anisotropic, and were provided by Goodyear through a Cooperative
rubber compounds vary through the tire structure. Research and Development Agreement and required
Models developed for tire design are extremely de- only minor modifications to accommodate surface
tailed, modeling each material within the tire (belts, contact and rolling. All of the models were built to
plies, rubber) for internal tire stress analysis, wear, represent the tires used in the experimental test pro-
and vibration modal analysis. For these types of gram for comparison to measured tire behavior in
models, the tire may consist of shell, continuum, and terms of deflection, contact area, deflected sidewall
rebar elements. The simulation consists of mounting profile, contact stress distribution, and performance
the tire on the rim, inflating it, lowering the tire onto on deformable terrain (snow and soil).
a rigid surface, and applying various loading condi-
tions. Such models are extremely large and take Tires simulated
many hours to run. The tire properties and associated Three tires were involved in the data collection
modeling details for tire models used by industry are phase of the project. The first tire was a Michelin
extremely difficult to obtain. Among the more nota- XCH4 235/75 R15. This tire was used in several ex-
ble published contributions to tire modeling are nu- periments on snow and soil under various moisture,
merous publications by Joseph Padovan at the Uni- compaction, and temperature conditions. The second
versity of Akron (Padovan 1977, Padovan et. al 1992, tire, a Goodyear Wrangler AT 235/75 R15, was cho-
Padovan and Padovan 1993, 1994a, 1994b) and mod- sen to be similar to the first in construction and be-
els of agricultural tires by Hu and Abeels (1994). An havior. The third was a highway tire with a ribbed
example of such a model is the National Tire Model- tread pattern, the Goodyear Wrangler HT 235/75
ing Program (NTMP) radial automobile tire shown in R15. The three tires were similar in size, construc-
Figure 35. The NTMP is a generic model used for tion, deflection, and contact area characteristics, and
calibrating modeling programs across the tire indus- data from all three were used in the following com-
try. Prior to 1997 this type of model required over 15 parisons. Although the tread patterns and compounds
CPU hours on a SUN SPARCstation 10 for inflation varied (Fig. 36), which would strongly influence trac-
and no-slip contact on a rigid surface (i.e. no rolling). tion, this study focused on the tire deformation and
In this research project, four types of tire models rolling resistance, so the tires can be considered
were evaluated for suitability to rolling on deform- equivalent. The dimensions of the test tires are given
able terrain: 1) a rigid tire, 2) a simplified tire model in Table 6.

24 to contents
 Figure 36. Tires used in the experiments: Goodyear Wrangler HT,
Michelin XCH4, and Goodyear Wrangler AT.

Table 6. Tire characteristics for the test tires at an average

vertical load of 6227 N (1400 lb). The XCH4 tire data were
used to represent the average values.
Inflation pressure Contact Deflection Max. deflected
(kPa) (psi) area (cm2) (cm) (%) section width (cm)
103 8 509.7 6.3 38 27.9
138 15 419.4 5.4 33 25.9
179 26 364.5 4.6 28 26.0
241 35 281.7 3.1 19 25.1
Unloaded tire diameter: 73.7 cm
Unloaded section width: 22.9 cm
Unloaded section height: 16.4 cm
Tread width: 15.7 cm
Ratio of voids to tread contact area: 0.44
Inflated overall section width: 24.9 cm
Inflated diameter: 74.7 to 75.3 cm

All three tires were evaluated for their deforma- tween the different tires and models. Based on the
tion characteristics on a rigid surface. A comparison good match in the measured tire deformations on a
of the deflection and contact area measurements col- rigid surface (deflection and contact area), the tires
lected from various sources is given in Figures 37 were assumed to be sufficiently similar for inter-
and 38. All of the data were collected after a tire changeable use in comparing to model results. The
break-in period of operation of at least 160 km. Con- normal operating pressure for the tires is 241 kPa (35
tact area is calculated from a static contact “perime- psi), and that is where the similarity between models
ter” area without accounting for voids due to tread and data is best.
pattern so that comparisons can easily be made be-

25 to contents
Figure 37. Deflection measurements for the tires used in the experi-
ments at inflation pressures of 241 kPa (35 psi) and 179 kPa (26 psi).

26 to contents
 Figure 38. Comparison of measured contact area for the three test
tires at 241 kPa (35 psi).

Rigid tire model ric analysis and refinement of the code (Alverez Sanz
If the deformable material is very soft, as in fresh 1999).
snow, the tire can be approximated as a rigid wheel. The objectives of the Darnell tire model are simi-
The rigid wheel is modeled as a rigid analytical sur- lar to those needed for this study: an efficient, accu-
face 0.74 m in diameter and 0.272 m wide (for a full rate, three-dimensional model to predict spindle
wheel), with a 0.051-m radius of curvature on the tire (axle) forces but not the internal stresses in the tire.
shoulder. An unsprung mass of 636 kg is placed at The contact forces are calculated in the process, so
the axle to simulate the weight of the tire and the load the model is also suitable for efficient modeling of
of the vehicle on the tire (6240 N); the rotational the tire–terrain interaction, where accurate modeling
inertia of the wheel was set to 2.15 kg m2. of the contact stress distribution is imperative.
The tire is simulated with three modeling ele-
Darnell tire model ments: the tread elements are six-degree-of-freedom
To adequately model the dynamics of a rolling shell elements (S4R); the sidewall elements are three-
tire, a very efficient tire model is needed. Thus, the node, user-defined elements based on pre-calculated
tire model developed by Darnell and Hulbert at the results from a sidewall model; and the contact is
University of Michigan is a promising alternative. modeled using a “hard” contact, penalty method
The Darnell tire model (Darnell et al. 1997, Darnell, (HKS 1998). The roadway is modeled as a rigid sur-
in progress) is an extension of a two-dimensional tire face.
modeling concept developed by Mousseau (Mous- The sidewall model is a special-purpose finite
seau and Hulbert 1996) for interfacing with vehicle element code that generates a lookup table of side-
dynamics simulations. The tire is modeled using ma- wall forces (and geometry) using ten equal-length,
terial properties for the composite material rather non-extensible beam elements with inflation pressure
than the individual components. The composite prop- as a follower load. The sidewall model is generated
erties were obtained by sectioning the tire and per- from the geometry of the uninflated sidewall, the
forming laboratory tests on each of the major section bending and shear stiffness for each element, and the
components (tread, sidewall, and shoulder). A similar tire inflation pressure. The results of the sidewall
approach was also developed at the University of model are implemented in the tire model through the
Birmingham in the UK for tire modal analysis (Burke user-defined element, which attaches to the tread
and Olatunbosun 1997a, 1997b, 1997c). Additional section with rotational springs (Fig. 39).
work on the Darnell model has been toward paramet-

27 to contents
 d2
Tread
θ tread

Tread
Spring Tread Reference
Angle (γ)
Sidewall
Length
Side
Radius of
Wall
Curvature

Rim s
Spring

d1
Bead Reference
Angle (θ )

Figure 39. Construction elements of the Darnell tire model. (After Darnell et al. 1997.)

The Darnell tire model was validated for a Ford

Taurus tire for lateral forces and deflection on a
smooth surface and while rolling over an angled ver-
tical step (Darnell et al. 1997) and for slip angle and
aligning moment.* The model results sufficiently
mimic tire behavior while being computationally
efficient, indicating its potential for simulations of a
tire on deformable terrain. Some alteration in the
sidewall model may be necessary to account for the
interaction of the sidewall with the deformable terrain
if sinkage is substantial.
The same tire used to gather motion resistance
data on various terrain (snow, freezing and thawing
ground) was cut into four sections: sidewall, lateral
tread, longitudinal tread, and shoulder (Fig. 40). The Figure 40. Cut tire sections used to determine
sidewall was further dissected into smaller sections composite tire properties.
for future work expanding the sidewall model. The
sections were tested in the laboratory to determine I = moment of inertia
the beam stiffness (EI) by clamping an end of the test l = beam length.
section onto a workbench to form a cantilever beam.
Deflections were measured for a range of loads, and Young’s modulus (E) is calculated from EI using
stiffness was calculated using the equation for a can- I = wbeam tbeam
3
/12 , where wbeam and tbeam are the width
tilever beam with a point load on the free end: and thickness of the tire samples, respectively. The
test results are summarized in Table 7.
The sidewall was divided into ten equal-length
1 Wl 3
y max = (25) beam elements in a special-purpose finite element
3 EI code. This code generates the sidewall force and ge-
ometry table for the user-defined element represent-
where ymax = maximum deflection at the free end ing the sidewall in the tire model. The stiffness and
of the beam curvature of the sidewall model are given in Table 8
W = applied load and illustrated in Figure 41.
E = Young’s modulus of the beam Additional measurements needed for the model
are sidewall curvature, tread hoop stiffness, rotational
* Personal communication with I. Darnell, University spring constants, reference angles, tire and rim geo-
of Michigan, 1997.

28 to contents
 Table 7. Bending stiffness of tire sections.
Width Length Thickness EI
(cm) (cm) (cm) (N m2)
Sidewall section A1 5.2 3.3 1.3 0.1564
Sidewall section A2 5.2 2.0 1.3 0.0287
Sidewall section A1+A2 5.2 5.3 1.3 0.1521
Sidewall section B 5.8 6.1 0.6 0.0459
Sidewall section C 7.0 2.1 1.1 0.0172
Lateral tread section 7.0 16.5* 1.7 0.1550
(0.9 w/o tread)
Longitudinal tread section 12.5 27.2 1.7 0.5137
(0.9 w/o tread)
Shoulder 33.5 cm
along tread
* 23.4 cm including shoulders

Table 8. Properties of sidewall elements and

tire model components.
Curvature Stiffness per unit
Element (1/m) width (N m)
1 (Bead end) 0.01 2 (rigid)
2 6.76 1.5
3 7.63 0.15
4 9.95 0.04
5 8.62 0.04
6 9.71 0.04
7 9.01 0.04
8 9.09 0.15
9 14.49 0.5
10 (Tread end) 17.54 1.0
Total length of sidewall = 16.1 cm, each element L = 1.6
cm
Tread attachment angle = 133°
Rim attachment angle = 23°
Tread diameter to end of last sidewall element = 72.5 cm
Tread diameter to center of belts = 69.5 cm
Rim diameter to first sidewall element = 41.6 cm
Tread longitudinal shell flexural rigidity for unit section
width, E1I = 4.1261 Nm2
For 1.11-mm shell thickness, E1 = 3.62 × 1010 N/m2
Tread lateral shell flexural rigidity for a unit section width,
E2I = 2.259 Nm2
For 1.11-mm shell thickness, E2 = 1.98 × 1010 N/m2

metry, and sidewall and shoulder stiffness. These are

listed at the bottom of Table 8.
For the elements representing the tread, the shell
stiffness was calculated with E1 being along the tread
in the longitudinal direction and E2 being across the
tread in the lateral direction. An equivalent shell
thickness of 1.11 mm was used in the model since
this was found to yield a reasonable hoop stiffness
based on measured changes in circumference with
inflation pressure.*

* Personal communication with I. Darnell, University Figure 41. Tire geometric parameters for
of Michigan, 2000. the sidewall model.

29 to contents
 a. Coarse (3 × 36) mesh.

Figure 42. Shoop–Darnell tire model (user elements not shown).

The Shoop–Darnell model (the model of the test then a plasticity model for the terrain material. Two
tire created with the Darnell methodology) was con- Shoop–Darnell mesh configurations were used in the
structed using ABAQUS for a quasi-static simulation. analysis: a coarse and a fine mesh (Fig. 42). Since the
This adequately simulates low-speed vehicle travel sidewalls are modeled with user-defined elements,
and is appropriate for 5 mph, which is the standard they do not appear on the standard finite element
test speed for off-road traction and motion resistance visualization.
testing. The tire model is first subjected to internal
pressure (inflated), then lowered onto the surface and Modal analysis tire model: Smooth tread
allowed to equilibrate with the applied vertical load The smooth tread tire model is typical of the type
before rolling. The Darnell methodology for the tire of models used for tire vibration modal analysis. The
model was validated for the test tire on a hard surface carcass is composed of a single layer of four-node,
and then used to simulate the three-dimensional con- reduced-integration shell elements (S4R) with mate-
tact between two deformable bodies on a homogene- rial properties representing the composite behavior
ous terrain, first using an elastic terrain model and through the carcass thickness. The treadcap is con-

30 to contents
 b. Fine (6 × 72) mesh.

Figure 42 (cont.). Shoop-Darnell tire model.

structed of linear, hybrid continuum elements The analysis is done in three steps: inflating the
(C3D8H), with constant pressure (simulating the tire to the desired pressure, vertically loading it based
nearly incompressible nature of rubber). The result is on the vehicle’s weight, and rolling it along the road
a reasonable approximation of both the structural by translating its centerpoint while allowing free rota-
behavior and the contact patch. The model is con- tion about its axis. The creation and seating of the
structed using the cross section shown in Figure 43, bead are accomplished by multi-point constraint and
repeated every three degrees around the axis of the fixed displacement of the first four rings of nodes
hub (creating 120 cross sections). For most of the closest to the wheel rim. During the inflation and
simulations the tire model was cut in half using a loading steps, no friction is included on the contact
vertical symmetry plane along the longitudinal axis. surface. Variations of the analysis include a static
The wheel mass is represented by one mass element rolling step, a dynamic rolling step, and changes in
and one rotary inertia element located at the wheel’s the definition of the shell element local coordinate
centerpoint on the symmetry plane. system.

31 to contents
Figure 43. Modal analysis tire model with smooth tread.

32 to contents
Modal analysis tire model: Straight ribbed tread wheel’s centerpoint while allowing the tire to freely
The ribbed tread tire model is also a modal analy- rotate about its axis due to friction, simulating a
sis tire model. Construction is similar to the smooth towed wheel. The contact is frictionless during the
tread model, with the carcass composed of four-node loading process, and friction is added during the roll-
reduced-integration shell elements and the treadcap ing step. Variations of the analysis include a static or
constructed of linear, hybrid continuum elements dynamic analysis procedure during rolling and
with constant pressure. The major difference is the changes in the convergence controls.
four grooves across the width of the tire created by The towing phase of the dynamic analysis is bro-
omitting sections of the tread continuum elements ken into two steps. The first is to accelerate the tire
(Fig. 44). The wheel is represented by one mass ele- from rest to 1 m/s at a constant acceleration of 1 m/s2
ment and one rotary inertia element located at the for one second (traversing 500 mm). The second step
centerpoint. Details for the smooth and ribbed tread continues to roll the tire at a constant velocity of 1
models are given in Table 9. m/s for two more seconds (2000 mm). For all cases
The analysis is completed in eleven steps, com- the analyses converge for a while but eventually
prising inflation, bead seating, lowering to the road reach a point where the tire stops moving by taking
surface, and vertical loading. The loading occurs in increasingly smaller time steps due to convergence
eight steps of alternating loose and tight convergence difficulties. The problem occurs when the shell is
control. Inflation is handled similarly. The creation subjected to simultaneous distortion and rotation as
and seating of the bead are accomplished using multi- the rotation angle approaches 360°. This is a bug in
point constraints on the first five rings of nodes clos- the ABAQUS shell element that will be corrected in
est to the wheel rim. An additional step translates the future versions (Bug Report v58_2449).

Figure 44. Modal analysis tire model with ribbed tread.

33 to contents
 Figure 44 (cont.). Modal analysis tire model with ribbed tread.

Table 9. Features of the modal analysis models with smooth and straight
ribbed tread.

Smooth tread Straight ribbed tread

Half carcass model Half carcass model
Carcass elements (number and type) 3360 S4R 6240 S4R
Treadcap elements (number and type) 1440 C3D8H 2400 C3D8H
120 C3D6H 120 C3D6H
Tread rubber: neo-Hookean, C10 1.109596 2.408340
Tread density (Ns2/mm4) 0.814575E-9 6.6255E-10
Tread gaps (width x depth, mm) None 5.1 × 16.6
5.3 × 16.2
Total no. elements 6602 13562
Total model variables 32172 63396

34 to contents
Evaluation of tire models on a rigid surface 241 kPa (35 psi), 179 kPa (26 psi), and 103 kPa (15
The tire models were applied to a rigid surface psi). The suggested inflation pressure for these tires is
through a range of pressures and loads for compari- 241 kPa. Lower inflation pressures are sometimes
son to measured deflections, contact patch areas, di- used when driving in off-road and marginal road
mensions, and stresses. conditions (snow and soft soil) and for minimizing
damage to unpaved travel surfaces. Therefore, the
Deflection two lower inflation pressures were also evaluated.
A complete data set for deflection was gathered Performance at a range of inflation pressures is also
for the Wrangler AT (Fig. 45) for use in comparisons of interest to industries using vehicles with Central
with model results from the Shoop–Darnell model, Tire Inflations Systems (CTIS) (i.e. military, forestry,
the smooth tread model, and the ribbed tread model. and agriculture).
Each model was loaded at a range of vertical loads A compilation of the data and model results for
(from 0 to 8000 N) at three tire inflation pressures: each inflation pressure is shown in Figure 46. The
6

4
Deflection (cm)

103 kPa
1 179 kPa
241 kPa

0
0 2,000 4,000 6,000 8,000 10,000 12,000

Vertical Load (N)

Figure 45. Measured deflection for three tire pressures.

Figure 46. Comparison of measured and modeled deflection for the

Shoop–Darnell, smooth tread, and ribbed tread models for three infla-
tion pressures.

35 to contents
 Figure 47. Comparison of all measured and modeled deflections at
241 kPa (35 psi) inflation pressure.
600

103 kPa
500 179 kPa
241 kPa
Contact Area (cm2 )

400

300

200

100

0
0 2,000 4,000 6,000 8,000 10,000

Vertical Load (N)

Figure 48. Measured contact areas at three inflation pressures.
model predictions follow the same general trend as contact areas for the three inflation pressures are
the data for all tire pressures, with the largest differ- given in Figure 48. All contact areas are based on the
ence occurring at the lower inflation pressure. This is perimeter of the contact, without accounting for voids
understandable since undesirable tread and sidewall within the area due to tread design. Comparisons of
behavior can occur when tires are underinflated, and the model results (the Shoop–Darnell model with
these pressures and behaviors are neither within the both coarse and fine mesh, the smooth tread model,
design range nor accurately accounted for in the and the ribbed tread model) with measured data are
models. On the other hand, the agreement between shown in Figure 49.
data and models at the standard inflation pressure is In general, the agreement between the models and
exceptionally good, as illustrated in more detail in data is reasonable. Surprisingly the Darnell gives the
Figure 47, showing all measured and modeled deflec- best fit, but the model mesh is coarse, which makes
tions at 241 kPa. the results erratic since the contact area changes
abruptly as each shell element makes contact. The
Contact area refined mesh on the Darnell model is only a slight
Comparisons of contact area were done much the improvement, and the results are worse at the higher
same as the deflection comparisons. The measured loads where the contact nodes pop off the surface just

36 to contents
Figure 49. Comparison of measured and modeled contact areas at three
inflation pressures.

37 to contents
 Figure 50. Comparison of Shoop–Darnell model results with measured
contact areas (for coarse and fine meshes).

Figure 51. Comparison of the smooth and ribbed modal analysis tire models
with measured contact areas.

inside of the tire shoulder due to buckling of the car- contact area is slightly less than the smooth model,
cass (as discussed later). This behavior strongly af- even after being corrected for the tread voids.
fects the contact stress and area since adjacent nodes
may alternate between high loads and no contact. Contact stress distribution
Figure 50 compares the Shoop–Darnell models to the Contact stress measurements for the Wrangler HT
measured data. were provided by Goodyear. For comparison, the
Comparisons of the smooth and ribbed tread mo- modeled contact stresses were displayed at the same
dal analysis models to measured data are shown in contour levels and color scheme, as shown in Figure
Figure 51. Both models provide contact areas less 52. Both models and measurements show high stress
than the measured data at the high loads, indicating values at the tire shoulder and, to a lesser extent,
that the models are slightly stiff. The ribbed model along the tire centerline. Agreement with the meas-
ured contact stress distribution is excellent.

38 to contents
 Measured Smooth tread model Ribbed tread model

7 69 131 193 262 324 386 448 510 572 641 703 765 kPa
Figure 52. Measured and modeled contact stress distribution for half carcass on a hard surface
(207 kPa inflation and 6627 N load).

Figure 53. Irregular stress contours generated by Figure 54. Close-up view of the contact for the
the Shoop–Darnell model on a rigid surface (full ribbed tire model, showing buckling just inside
carcass width). the shoulder area (half tire carcass).

During contact on a rigid surface, the Shoop– gressively finer meshes fail to converge [similar be-
Darnell model suffered from buckling just inside the havior was noted and explored by Alvarez Sanz
tire shoulder area, which strongly affected the contact (1999)]. The buckling behavior also occurs to a lim-
stress. Although a finer mesh would normally smooth ited extent in the physical tire and in the modal
the contact stress contours, the buckling causes alter- analysis tire models (seen in the contours in Figure
nate nodes to lift off the surface, resulting in the 52). The buckling is worse at low inflation pressures
stress pattern shown in Figure 53. Eventually, pro- and high loads (Fig. 54). The impact is minor for the

39 to contents
 Figure 55. Measured and modeled sidewall profiles for the Shoop–
Darnell tire model.

actual tires and the models where the rubber tread Run times
elements conform to the surface, softening the con- Model efficiency is of primary importance for use
tact. For our purposes, this behavior disappears when in rolling a tire on a deformable material, so the mod-
the tire model is applied to a terrain surface, which els were run with various options and using different
softens the contact through terrain compliance. computers in order to assess the effects on run time.
The primary variables considered were the effect of
Sidewall profile the tread (smooth vs. ribbed), static or dynamic
Measurements of the deflected shape of the side- analysis, convergence controls, optimum use of sym-
wall under various loads and inflation pressures were metry, and computing system. The results of the
collected to check the calibration of the Shoop– many runs are included in the following discussions
Darnell tire sidewall model. Selected measurements and qualitatively summarized in Table 10. Examples
of the profiles, along with comparisons to the mod- of some of the run times are given at the bottom of
eled geometry, are given in Figure 55. The modeled the table. Consideration was also given to how far the
sidewall geometry closely duplicates the measured tire would roll in the model run. A roll distance of
geometry. The modeled sidewall is slightly longer between one third and two revolutions, depending on
than the measured profiles since the measurements the tire, is needed to remove abnormal stresses built
did not include the sidewall portion covered by the up in the tire structure due to lowering it onto the
wheel rim. surface. One revolution was considered the minimum
requirement. All comparison simulations were done
Hard-surface rolling resistance using the implicit solver.
Ideally forces generated from rolling the model Static vs. dynamic. The static models were run in a
should be comparable to the measured hard-surface steady-state type of analysis. The same models were
rolling resistance. However, rolling resistance is modified for a dynamic analysis where the inertia of
largely due to viscoelastic behavior in the rubber the system is considered. In general, the static models
compounds, and these models omit the viscoelastic roll over twenty times farther in less than twice the
material properties for increased efficiency. There- CPU time than the dynamic models. However, the
fore, the modeled hard-surface rolling resistance dynamic models used less memory. Ultimately, a
forces reflect only resistance due to interface shear. dynamic model is the desired choice to include mass,
For an interface friction coefficient of 0.825, similar velocity, and acceleration effects and for operating
to asphalt pavement, the rolling forces modeled are the combined model using an explicit finite element
less than 10 N at an inflation pressure of 241 kPa, code.
while measured values are on the order of 25 N [at Full tire vs. half-tire. The full-tire version of the
241 kPa and low speed (8 kph)]. ribbed model is prohibitively expensive in terms of

40 to contents
 Table 10. Qualitative summary of runtime analysis.
Model parameters compared Change in performance
Static vs. dynamic analysis Static models rolled 2 m vs. 0.08 m before conver-
gence problems
Half tire carcass vs. full tire carcass Half tire runs three times faster
Smooth vs. ribbed tread model Smooth model runs four times faster
SGI Octane vs. Cray J932 Octane is two to three times faster
Tight vs. loose convergence controls Loose controls run faster but tight controls let the tire
roll further before convergence problems
Example run times
Ribbed tread, full tire, tight convergence Rolls 25 mm in 488 hr on J932
Ribbed tread, half tire, dynamic analysis Rolls 46 mm in 252 hr on J932
Smooth tread, half tire, static analysis Rolls 2 m in 22 hr on Octane
Shoop–Darnell model (full tire) Roll 20 m in 11 min on Octane

run time. It takes almost three times as long as the COMBINED TIRE–TERRAIN MODEL
half-tire model to go the same distance. No full-tire
model has stopped due to non-convergence, but Modeling the tire–terrain contact interface
rather due to system errors or time limits. In the fu-
ture, if side forces, cornering forces, or steering reac- Contact
tions are investigated, then a full-tire model will be To merge the tire and terrain models, the interface
required. between the two meshes must interact. This is ac-
Smooth tread vs. ribbed tread. The model with the complished by defining where contact is allowed,
smooth tread has fewer nodes and elements, is less how the contact occurs, and how forces are transmit-
geometrically complex, and uses default convergence ted. For two meshes to come into contact, a surface
controls. It runs more than four times faster than the must be placed on each material. For a deformable
ribbed tread model during the static analysis and tire on snow or soil, one surface covers the outside of
more than eleven times faster for the dynamic analy- the tire and another surface lies on the top of the ter-
sis. In most instances the accuracy of the smooth rain layer. The contact is defined by the proximity of
model is as good as or better than the ribbed model the surfaces. When both material meshes are deform-
(i.e. the model results are closer to the experimental able, one surface is chosen as a “master” surface and
data). In the dynamic analyses the smooth tread the other as a “slave” surface. The more finely
model went almost twice as far as the ribbed tread meshed surface is usually chosen as the slave since
model. the nodes on the slave surface are not allowed to
Cray J932 vs. SGI octane. The models are being penetrate the master surface.
run on an SGI Octane (Intrepid, 120 megaflop) and a Contact can be enforced using either “penalty” or
Cray J932 (Chilkoot, 12 CPU at 2.4 gigaflop each, “kinematic” contact methods. Both methods were
200 total if optimized for this platform). Generally used. Penalty contact was implemented for pseudo-
the same models took two to three times longer to run static analyses, where inertia forces were not consid-
on the Cray J932 than on the SGI Octane. This is ered. In penalty contact the penetration of the sur-
likely because HKS (the ABAQUS software devel- faces into each other is resisted by linear spring
oper) uses the SGI as a development platform, which forces with values proportional to the distance of
results in increased efficiency on that equipment. penetration. These forces pull the surfaces into an
Convergence controls. For the static analysis equilibrium position with no penetration. Kinematic
ribbed tire model, the tire is maneuvered in several contact was used in some of the dynamic simulations
steps, with the convergence controls adjusted at each (using the explicit code). Kinematic contact considers
step to increase efficiency. The loose convergence the inertia forces of the material when calculating the
controls will allow the model to run faster, but con- forces relative to the surface positions. Although the
vergence stops after rolling only a short distance. For overall model results were the same with both contact
dynamic rolling the model using the default conver- methods, the kinematic contact was slightly more
gence controls runs faster, and the tire rolls three representative of field observations of the snow de-
times farther, than when convergences controls are formation beneath the wheel (Haehnel 2000).
looser or tighter than default values.

41 to contents
Friction The simulation of driving and braking traction
Forces parallel to the interface are transmitted would be an important extension of this research.
based on a friction law. Friction is a complex phe- More-elaborate friction models could be generated
nomenon involving stress and strain distributions; based on the multitude of experimental traction data
heat transfer; hydrodynamic and elastohydrodynamic obtained by CRREL. However, this need for a trac-
fluid flow; material structure; chemical interactions tion curve a priori seems to negate the usefulness of
between surfaces, surface coatings, and lubricants; the model for predicting traction (i.e. the traction
and phase change. Current models of friction are em- input is needed to get a valid traction output). Thus,
pirical or semi-empirical formulations in which sev- the current modeling effort focused on the tire forces
eral variables and combinations of variables are developed due to deformation of the snow rather than
modified with coefficients and exponents based on due to the interaction at the interface (i.e. a model of
experimental observations. These equations apply rolling resistance rather than traction). Therefore, for
only to the system for which they were developed, this project the simple Coulomb friction model was
though they may be used to estimate the behavior of used:
similar systems. They are not applicable to other ex-
perimental configurations or test conditions and can- µ= τ . (27)
not be extrapolated outside of their bounds with any σ
degree of confidence. Thus, an accurate model of
friction is currently non-existent and is not likely to Rigid wheel in snow
be developed in the immediate future. The only rea-
sonable way to predict the behavior of most frictional Model construction
systems today is to test them (Ludema 1996a, 1996b, The snow–wheel model simulates one wheel of
Barber 1991). the CRREL Instrumented Vehicle moving through a
The experimental measurement of tire–terrain fric- range of shallow (20 cm) to deep (50 cm) fresh snow
tion (including pavement) is typically performed as a with a density of 200 kg/m3. The wheel is modeled as
vehicle traction or braking test [a discussion of winter a rigid surface 0.74 m in diameter and 0.272 m wide
traction test methods is available in Shoop et al. (for a full wheel) with a 0.051-m radius of curvature
(1994)]. Test data are generally reported as a traction on the tire shoulder. An unsprung mass of 636 kg is
(or friction) coefficient (a peak value or an averaged placed at the hub to simulate the weight of the vehi-
value over a specified range) or as a traction curve cle and tire; the rotational inertia of the wheel was set
with respect to wheel slip. Traction coefficients and to 2.15 kg m2. The snow was modeled with eight-
curve shapes are functions of both the tire and the node continuum elements near the wheel and four-
surface. Curves of traction data with wheel slip can node infinite elements for the far-field snow cover.
be implemented in ABAQUS or other finite element The Drucker–Prager cap material model with linear
analysis codes by specifying friction as a function of elasticity was used for the snow, and a linear elastic
relative velocity. Liu and Wong (1996) took this ap- model was used for the infinite elements. Coulomb
proach using a friction law of the following form and friction was applied at the tire–snow interface (µ =
implementing it in a tire–terrain model using the gen- 0.3). The deep snow was modeled using Adaptive
eral-purpose finite element program called MARC: Lagrangian–Eularian (ALE) meshing to accommo-
date the large displacements, so slide planes were not
−j k needed along the sides of the tire. The shallow snow
τ = µσ(1 − e ) (26)
model had 11,520 continuum elements and 1,418
infinite elements; the deep snow model had 14,400
where µ =friction coefficient
continuum and 1,962 infinite elements. In all models
τ =shear stress
the snow rested on a rigid surface. The shallow snow
σ =normal stress
model is shown in Figure 56.
j =relative slip distance between the
The rolling resistance test is simulated by first
wheel and the terrain
lowering the wheel into the snow by gravity, accelerat-
k = constant having the same units as j.
ing it to the desired speed, and then translating it at a
This equation is of the type commonly used in constant velocity. The wheel is moved longitudinally
conjunction with soil shear data. However, Liu and by displacing the axle node. This simulates a towed
Wong were not convinced that this equation ade- wheel and duplicates the procedure of a rolling resis-
quately described traction. tance test using the CRREL Instrumented Vehicle.

42 to contents
 These models were configured for both implicit mesh, therefore, does not directly reflect the deforma-
dynamic analysis and explicit dynamic analysis. The tion of each element, as in a standard finite element
explicit models used ALE meshing, where the mesh mesh, because the nodes and elements have moved
automatically adjusts during the deformation by al- relative to the snow mass. However, the final surface
lowing mass transfer between elements. The final shape is correctly represented, and tracer nodes can
be used to track the mass movement. The ALE mesh-
ing significantly reduces model run time (from over
48 hours to approximately 24 hours on an SGI Ori-
gin), eliminates severe mesh distortion, and smooths
the resultant load response, but it does not affect the
numerical results. Figure 57 shows the deformed
mesh for the deep snow model with and without the
ALE meshing.
The snow material properties were adjusted to
match field measurements of sinkage and motion
resistance. This was done primarily through changes
to the cap hardening and consequent adjustments
to the Drucker–Prager parameters. Since snow com-
Figure 56. Model of a wheel rolling on 20 cm of paction is the primary mechanism contributing to
fresh snow (infinite elements not shown). motion resistance, the pressure–volume relationship for

Figure 57. Comparison of standard mesh (top) and ALE mesh (bot-
tom). (From Shoop et al. 1999.)

43 to contents
snow, reflected in the hardening table, has the great-
est influence on resulting tire forces and sinkage.
Changes to the acceleration, velocity, and friction
coefficient have little effect on model results, partly
because of the low velocities and accelerations in-
volved in standard snow rolling resistance tests.

Motion resistance forces and sinkage

The tire–snow model was compared to field
measurement of tire forces made with an instru-
mented vehicle. Model results were also compared to
rolling resistance and sinkage predictions made from
established algorithms for snow (as adopted by the
NATO Reference Mobility Model).
Experimental measurements. Vehicle performance
was measured using the CRREL Instrumented Vehi-
cle (CIV), which is instrumented to measure vertical,
longitudinal, and lateral forces at the tire–terrain in-
terface; wheel speed at each wheel; true vehicle
speed; and steering angles. The CIV is designed as a
research tool to perform various mobility tests (trac-
tion, resistance, and maneuverability) using different
Figure 59. Configuration of speed
tires, traction aids, and vehicle configurations (of
sensors and axle-mounted load
braking and driving wheels) on a range of terrain cells on the CIV (From Shoop et al.
surfaces including dry, wet, snow, ice, and freezing 1994.)
or thawing ground. The data obtained from these tests
are used for model validation and for developing al- on the vehicle. The difference between the wheel
gorithms to predict vehicle performance on cold- speed vw and the vehicle speed vv is termed the dif-
weather terrain. ferential interface velocity (DIV), also called the
The vehicle, originally a 1977 American Motors “slip speed” or “longitudinal slip velocity” (SAE
Corporation Jeep Cherokee, is shown in Figure 58. 1992). This is often normalized by the speed of the
Each wheel is instrumented with a three-component vehicle and called slip i (ISTVS 1977):
load cell (Fig. 59), designed and calibrated to respond vw − vv
to the forces at the terrain interface. The wheel speed i= ×100% . (28)
vv
is measured using a proximity sensor with a set of
100 measurement pegs set into the brake rotor. Wheel Using this convention, driving slip is positive and
speed is compared to the true vehicle speed measured braking slip is negative.
using a fifth wheel or a sonic speed sensor mounted Rolling resistance or motion resistance is the sum
of the forces resisting vehicle motion. These are
forces due to the internal friction of the moving parts
of the vehicle, the air drag, the internal resistance of
the tires due to flexing of the belts and plies and vis-
cous properties of the rubber compounds, and the
added resistance due to the deformation of the sur-
face. The CIV measures motion resistance forces
directly at the wheel, eliminating the forces acting on
the vehicle body. Motion resistance is tested with the
CIV by rolling the vehicle through snow at a constant
speed. The vehicle is driven with the rear wheels
only, and the front brakes are released so that the
only forces on the front wheels are due to the tires
and snow. The internal resistance of the tire is meas-
ured by rolling it on a hard surface. This value is sub-
Figure 58. The CRREL Instrumented Vehicle (CIV). tracted from the snow motion resistance measure-

44 to contents
ment to obtain the resistance due only to the deform- study. Resistance in shallow snow was studied fur-
able snow (or other terrain). The experimental values ther in an experimental program in 1993, also in
of rolling resistance presented here are only the por- Houghton, Michigan, documented in Richmond
tion of the resistance due to the deformation of the (1995). The data from these studies are given in Ta-
terrain (snow). ble 11. The majority of these data were collected us-
CRREL has a large database of vehicle measure- ing Michelin XCH4 tires except where a different
ments in snow. For comparison to model data, most radius is noted. This database was used to develop a
of the field measurements were taken from a study of semi-empirical performance model for mobility on
wheeled and tracked vehicle performance in snow snow, which was incorporated into the NATO Refer-
that was performed in Houghton, Michigan, in 1988 ence Mobility Model (NRMM II) (Richmond et al.
and 1989 and documented in Blaisdell et al. (1990) 1995, Ahlvin and Shoop 1995). Both the original data
and Green and Blaisdell (1991). Motion resistance and the predictive algorithms were compared to the
measurements formed a significant portion of this tire–snow finite element model.

Table 11. Measured sinkage and resistance in snow. All

tires are the same (radius = 38 cm), except for the last four
measurements.
Snow depth Sinkage Resistance Snow density
(m) (m) coefficient (kg/m3)
0.13 0.10 0.05 160
0.16 0.12 0.06 160
0.15 0.12 0.05 170
0.19 0.15 0.08 170
0.16 0.13 0.07 160
0.17 0.14 0.06 160
0.18 0.15 0.06 170
0.19 0.16 0.07 170
0.21 0.17 0.07 220
0.21 0.17 0.10 220
0.20 0.16 0.07 220
0.21 0.17 0.12 220
0.20 0.16 0.07 220
0.21 0.16 0.09 220
0.18 0.14 0.07 220
0.19 0.16 0.10 220
0.21 0.16 0.06 220
0.22 0.17 0.07 220
0.21 0.15 0.07 220
0.18 0.13 0.05 220
0.18 0.13 0.06 220
0.20 0.15 0.07 220
0.21 0.16 0.07 220
0.22 0.17 0.08 220
0.22 0.17 0.06 220
0.22 0.17 0.05 220
0.14 0.10 0.02 110
0.13 0.90 0.03 110
0.17 0.14 0.03 110
0.16 0.12 0.04 110
0.18 0.15 0.03 110
0.16 0.12 0.06 110
0.16 0.13 0.14 230
0.38 0.33 0.19 220
0.38 0.31 0.19 160
0.36 0.31 0.21 160
0.32 (r = 35 cm) 0.26 0.17 190
0.28 (r = 33 cm) 0.25 0.10 120
0.26 (r = 33 cm) 0.22 0.16 170
0.25 (r = 33 cm) 0.23 0.12 120

45 to contents
 Figure 60. Parameters used to predict motion resistance us-
ing the NRMM algorithm. (From Richmond 1995.)

NRMM prediction of motion resistance in snow. snow, since a driven wheel would not likely have
The following method was proposed by CRREL for enough traction to overcome the large resistance and
predicting the motion resistance in snow. It was would become immobile.
adopted into the NATO Reference Mobility Model as
part of the overall vehicle performance predictions For ρ0 > 150 kg/m3 and 2/3 r < z < r, (32)
scheme. Only the portions of the algorithms related to Factor = 1.5
motion resistance in snow are stated here. Based on
Richmond (1995), the motion resistance for the lead- For ρ0 > 150 kg/m3 and z > r, (33)
ing tire in shallow snow Rs is based on the vehicle Factor = 2.5.
sinkage z according to
Comparison of measurements, NRMM, and finite
Rs = 13.6041(ρ0 wa) 1.26
(29) element results. In the FEA model the longitudinal
reaction force at the wheel hub represents the motion
resistance force on the wheel due to deformation of
a = r arcos [(r-z)/r] (30) the snow and is directly comparable to the motion
resistance measured with the CRREL Instrumented
Vehicle. Similarly the vertical displacement of the
 ρ0 
z = h 1 −  (31) hub node is equivalent to the sinkage of the wheel
 0.519 + 0.0023 p max  into the snow (since the wheel itself is not deforming
in this simulation). Figure 61 shows the resistance
where Rs = motion resistance (leading tire only) (N) coefficient (the longitudinal load divided by the verti-
ρ0 = initial snow density (kg/m3) cal load on the hub) and vertical displacement for one
w = maximum tire width (m) of the snow model runs. For this model the vertical
a = arc length in contact with snow (m) loading occurs from 0 to 1.5 s, the horizontal accelera-
r = tire radius (m) tion occurs from 1.5 to 2.0 s, and then constant speed
z = sinkage (m) translation is maintained from 2.5 to 5 s at a speed of 8
h = snow depth (m) kph (5 mph). The sinkage and motion resistance val-
pmax = maximum tire contact pressure (kPa). ues are chosen as the average of the values occurring
during the constant speed portion of the simulation.
These parameters are illustrated in Figure 60. Comparisons of the measured and modeled sink-
Very little data is available for wheeled vehicles in age and motion resistance are shown in Figure 62 as
deep snow. However, based on numerous observa- a function of snow depth. The model overestimates
tions of snow vehicle movement, the following fac- motion resistance, but the trends are consistent with
tors are applied to the shallow snow equations to es- the measured data and the model falls directly along
timate the additional motion resistance in deep snow. the upper edge of the measurements, constituting a
The factors represent an engineering estimate of the conservative prediction. The NRMM predictions of
additional forces due to plowing of the vertical face motion resistance are discontinuous based on the
of the wheel and undercarriage drag (Richmond et al. snow depth. For this tire, motion resistance for snow
1995). The deep snow modifiers would be applicable depths less than 25 cm are calculated using NRMM
to wheels being dragged or towed through deep eq 29, for depths from 25 to 38 cm using eq 31, and for

46 to contents
Figure 61. Modeled resistance force in coefficient form (longitudi-
nal/vertical) and vertical displacement of the wheel hub for the FEM
wheel rolling through 20-cm snow.

Figure 62. Finite element model, measured data, and NRMM

predictions for sinkage and motion resistance in fresh snow.

47 to contents
snow depths greater than 38 cm using eq 32. The with zero slip and with unrestricted slip. The NRMM
finite element model falls between the NRMM mo- deep snow resistance predictions are close to the fi-
tion resistance prediction equations for deep snow, as nite element results for a free slipping wheel, as
does the measured data. The bottom part of Figure 62 would be expected for a wheel being dragged through
shows the corresponding sinkage, which the finite deep snow, while shallow snow results are best repre-
element model (with zero slip) underestimates, sented by the simulation with zero slip.
particularly for the deeper snow. The bottom of Figure 63 shows the measured
The NRMM predictions are also a function of the sinkage data falling between the zero slip and unre-
initial snow density. Although the material model stricted slip models. Measured slip in shallow snow is
was designed to simulate 200-kg/m3 snow, NRMM near zero, but higher slip measurements are not un-
predictions based on 240-kg/m3 snow yield an excel- common as the snow gets deeper. Realistically the
lent match to the finite element resistance, as seen in model should behave somewhere between these two
the top of Figure 63. This would indicate that the extremes. This part of the physics of the interaction is
FEA model is more representative of a slightly not accurately reproduced in the model, partly be-
denser snow. cause the tread, which is designed to resist slippage,
Figure 63 also shows model results for a wheel is not yet accurately modeled.

Figure 63. Finite element simulations at zero slip and at unlimited

3
slip and NRMM motion resistance predictions for 240-kg/m snow.

48 to contents
Snow deformation beneath the wheel snow deformation in cross sections in the direction of
Snow deformation under a vehicle can be meas- travel (longitudinal) and transverse to the direction of
ured by excavating a cross section of the snow as travel are shown in Figures 65 and 66. The deviatoric
shown earlier in Figure 3. Additional detail can be stress contours shown in the longitudinal cross sec-
obtained by marking the snow prior to deformation tion reveal the location where a new shearing surface
by backfilling a small-diameter, vertical hole with begins to develop ahead of the advancing wheel. The
dark-colored chalk dust. The vehicle is then driven failure arc, even for snow, clearly advances down-
into the snow, and the changes in the lines can be ward and away from the wheel, just as shown ex-
seen when the cross section is excavated. This proc- perimentally for sand in Figure 1. The deformation
ess is illustrated in Figure 64. predicted in the model is slightly less than the ob-
Once the snow cross section has been excavated, served deformation, partly because the model is for
snow density can be measured by collecting and deeper (20 cm) and denser (200 kg/m3) snow than
weighing samples of a known volume. This was done occurred in the field. The general shape of the dis-
in several experiments (Richmond 1995). Modeled placement is the same, however.
deformation of 20-cm, 200-kg/m3 snow was com- The measured density and the modeled density are
pared to measured snow deformation beneath the in good agreement, as shown in the transverse cross
CIV in similar snow conditions. Because the model section in Figure 67. The modeled density directly
used ALE meshing, the displacement of the snow beneath the wheel (460 kg/m3) is only 10% different
was documented by placing tracer particles in the from the measured density (510 kg/m3). Near the tread
model, spaced at the same distance as the markings in shoulder, the modeled density was 290–380 kg/m3,
the field. Comparisons of the measured and modeled with measured values ranging from 280 to 300 kg/m3.

Figure 64. Marking the snow to observe snow de-

formation after vehicle passage.

49 to contents
 Figure 65. Comparison of measured displacement (14.5-cm snow) to
modeled displacement and deviatoric stress (20-cm snow) in the lon-
gitudinal direction.

Figure 66. Comparison of measured (19-cm snow) and modeled (20-cm snow) displacement in a cross
section transverse to the direction of travel.

50 to contents
 Figure 67. Comparison of measured (19-cm snow) and
modeled (20-cm snow) densities in a cross section trans-
verse to the direction of travel.

Deformable tire on soil The same tire was then rolled on a Drucker–Prager
On harder terrain it is important to model the de- cap plasticity model representing the McCormick
formation of the tire in addition to the deformation of Ranch sand (discussed earlier). The results of this
the terrain. Because of the large deformation and run simulation are given in Figure 69, showing the maxi-
times involved, it is important that the tire model be mum principal plastic strain and the principal stress
as efficient as possible while still maintaining an ac- directions. The maximum principal stress is compres-
curate contact patch. Thus, the tire model used for sive beneath and slightly forward of the tire. The
rolling on soil was based on the formulation proposed minimum principal stress is tensile and is largest just
by Darnell et al. (1997). below and to the sides of the tire, oriented away from
The Shoop–Darnell tire model was modified to the tire. The difference between these stresses indi-
roll across deformable material, such as soil. This tire cates the magnitude of shear stress in the soil, which
was first placed on an elastic terrain material with the is largest beneath and directly in front of the tire. This
properties of compacted sand. The simulation was corresponds to the general shear zones beneath a
conducted in four steps: the tire was first inflated, towed rigid wheel observed by Wong and Reece
then lowered into the soil, allowed to come to equi- (1967), shown in Figure 1.
librium, and then rolled to the left, as seen in Figure The tire-soil models are prototypes requiring re-
68. Only the tread and soil elements are shown in the finement and development. They will eventually be
figure; the user-defined sidewall elements are not compared to field measurements of tire forces, de-
displayed. Figure 68d is an oblique view of the three- flection, and contact area, including sinkage and
dimensional model showing the contact pressure. deformation of the soil surface.

51 to contents
 a. Tire inflated.

b. Tire lowered onto soil.

c. Tire rolled to the left.

Figure 68. Rolling tire on an elastic material (sand).

52 to contents
 d. Contact pressure.
Figure 68 (cont.).

a. Plastic strain.

b. Principal stress directions as the tire rolls to the left.

Figure 69. Rolling Shoop–Darnell tire on a Drucker–Prager

cap model of the McCormick Ranch sand.

53 to contents
 Figure 70. Deformed mesh of the modal analysis tire model in
snow after 8 cm of sinkage in snow.

Figure 71. Close-up of a tire after sinking 8 cm into the

snow at the beginning of roll. The tire shows very little de-
formation (the deflection is 3 mm), indicating that a rigid
wheel may be a good approximation.

Deformable tire on snow seen in Figure 71, revealing that the rigid wheel
Simulations of the modal analysis tire on snow are model for deep fresh snow may be a valid assumption
in progress.* Although preliminary runs are very (although these results are only preliminary).
slow and have not progressed to completion, inter-
mediate results are available. Figure 70 shows the
simulation after the tire has been lowered onto the CONCLUSIONS
snow and begins to roll. The sinkage is approxi-
mately 8 cm, which is about the same as the modeled Summary
sinkage for the rigid wheel on 20-cm snow, as seen in The interaction between a tire and deformable ter-
Figure 61. Of significant note, however, is that the rain is a complicated dynamic process that involves
deformable tire has undergone very little deformation the deformation of both the tire and the terrain in
(the deflection is 3 mm, which is less than 2%), as three dimensions. The tire is a complex structure
made of many materials, and the terrain undergoes
* Personal communication with K. Kestler, private large deformations in an inelastic manner, so neither
consultant, 2000. component is easily modeled. The objective of this

54 to contents
research was to produce a three-dimensional finite tion resistance forces but underestimates sinkage. The
element model of tire–terrain interaction that can be measured deformation patterns are duplicated in the
used to explore the effects of tire and terrain variables model, but the modeled displacements are less than
on vehicle mobility. Such a model can be used for what is measured in the field. Model results were also
tire design and specification for off-road vehicles, for compared to snow mobility predictions made using
vehicle performance prediction, and for terrain dam- the largely empirical NATO Reference Mobility
age prediction and reduction of the environmental Model, with good agreement for forces but underes-
impact of off-road travel. timation of sinkage. The amount of wheel slip had a
The details of the tire–soil modeling problem are major impact on the modeled sinkage.
divided into three topic areas: 1) material models for Simulations of a tire rolling on soil and snow us-
the terrain material, 2) tire models for use on a de- ing the Shoop–Darnell tire and the modal analysis
formable substrate, and 3) the combined tire–terrain tire model are operational but have not been vali-
model and the treatment of the interface. The terrain dated. Of significant note, however, is that a simula-
materials simulated in this study were fresh snow and tion using the modal analysis tire on fresh snow pre-
compacted sandy soil. The two material models used dicts very little tire deformation (3 mm, or less than
were a modified Drucker–Prager cap plasticity model 2% deflection). This suggests that the assumption of
and a critical-state, crushable foam model. Both a rigid tire may be used for soft terrain such as deep,
models were considered suitable for capturing the fresh snow without loss of model accuracy.
highly compressible behavior of fresh, natural snow
(initial density of 200–250 kg/m3 at temperatures of Significant findings
–10° to –1°C). Model parameters were generated The following summarizes the significant
from field test data and from the literature, matching achievements of this work:
this snow type as closely as possible. The snow 1. A material model was developed for fresh snow
model was validated using plate sinkage test data for and validated with plate sinkage tests in the lab
snow of similar age and density. The soil model and field. Good agreement of measured and mod-
represents sand similar to that used during vehicle eled forces, displacements, and changes in mate-
mobility experiments at CRREL. The material was rial density were achieved.
modeled using a Drucker–Prager cap plasticity model 2. Evaluations of several finite element tire models
with input parameters from the literature. suitable for rolling on a deformable substrate in-
To apply a tire model to deformable terrain, the dicate that the Darnell model yielded accurate re-
model must be efficient yet accurately portray the tire sults and was computationally efficient. The mo-
structural behavior. Four tire models were evaluated dal analysis type of tire model is also suitable but
for suitability to rolling on deformable terrain: 1) a is more computationally intensive.
rigid tire model, 2) a simplified tire model using 3. Combined tire–terrain models utilizing the
methodology developed by Darnell at the University Shoop–Darnell tire and the modal analysis tire
of Michigan for use in vehicle dynamics simulations, model are operational but have not yet been vali-
3) a tire model of the type used for harmonic vibra- dated. Preliminary results of the modal analysis
tion modal analysis, with a smooth tread, and 4) a tire tire model on snow show very little deformation
model similar to model 3 except with a straight in the tire, indicating that the rigid wheel simplifi-
ribbed (longitudinal) tread. All of the models were cation may be valid for soft terrain.
built to represent tires used in the experimental test 4. A model of a rigid wheel on fresh snow, validated
program for comparison to measured tire behavior in experimentally, shows good agreement with
terms of deflection, contact area, deflected sidewall measured motion resistance forces, snow dis-
profile, contact stress distribution, and rolling resis- placement, and snow compaction and agrees with
tance forces on deformable terrain (snow and soil). results predicted by the NATO Reference Mobil-
Three models of the combined tire and terrain ity Model.
were developed. The first is a rigid tire on fresh 5. The rigid wheel on snow model does not capture
snow. The second is the Shoop–Darnell tire model the tread effects of the tire–terrain interaction (as
rolling on a soil, and the last is the modal analysis tire tread patterns were not expressly modeled). For a
on snow. Model results simulating a rigid tire rolling free-rolling wheel, the amount of slip and sinkage
on snow were compared to tire forces measured using is overpredicted. When the model wheel slip is set
an instrumented vehicle. The measured snow defor- to zero, as is measured during shallow snow resis-
mation under the wheel was also compared to model tance tests in the field, sinkage is underpredicted.
results. The model exhibits good agreement with mo- The most realistic results lie between these two

55 to contents
 cases and can possibly be modeled with “rough” rioration, loading effects on soil water (pumping
friction attributes. A full understanding of this ef- up through pavement layers and cracks), and the
fect will be critical for traction and braking stud- function of drainage materials, including geosyn-
ies. thetics.
4. The impact of temperature-dependent material
Recommended applications and future properties, such as frictional heating and melting
research needs of snow, ice, or frozen ground.

Parametric design analysis Additional terrain material models

The primary applications of a tire–terrain model Future work should also include extending the ma-
are 1) all-season and off-road performance prediction terial models to other terrain materials, such as wet,
for tires, 2) tire design, optimizing geometry, materi- trafficked snow or slush, and soft, loose soils. Traf-
als, and tread patterns, and 3) specification of the ficked snow is typical of much of the snow on roads
correct tires based on expected operating conditions. before clearing and therefore is of primary interest to
Naturally, design use of the model could include a snow maintenance operations, tire companies for
parametric analysis of the impact of tire and terrain improving design of snow tires, automotive compa-
parameters on tire performance. Such a study, evalu- nies for optimizing vehicle handling, and for the roll-
ating the effects of tire geometry and loading on roll- ing resistance of the snow and its effect on fuel econ-
ing resistance in snow, may now be undertaken using omy. The thawing soils material represents a critical
the results of this research. case for roads subjected to freeze–thaw, where soil
moisture is pulled to the surface during freezing (ex-
Interface friction and interlocking of panding soil pores or forming ice lenses), resulting in
tread and terrain a wet and loose soil during seasonal, intermittent, or
The processes occurring at the interface of the tire even daily thawing events. Some of the difficulties in
and terrain are also of great interest. Advances in modeling this unique material may be simulated us-
driving and braking traction mechanics through so- ing the more flexible Multi-Mechanical Model pro-
phisticated study of interface friction, and the imple- posed by Peters* and Smith (2000) or through hydro-
mentation of this in a numerical model, are of major dynamic modeling. Additional behaviors specific to
significance because of their large impact on tire and the interaction of snow grains during loading and
vehicle performance and safety. Since the tread pat- resulting sintering may be better approached using a
tern was not modeled in this study, the details of the discrete element model as proposed by Johnson and
interface friction and frictional interlocking, which Hopkins.**
affected the resulting slip of the tire, were not spe-
cifically modeled. Issues include the interaction of
terrain material and tread blocks and the impacts of a LITERATURE CITED
spinning tire on traction through frictional heating
and mass shearing. Modeling challenges include cou- Abele, G., and A. Gow (1975) Compressibility char-
pling with other models, such as submodels of tread acteristics of undisturbed snow. Research Report 336,
blocks, and discrete element or hydrodynamic models U.S. Army Cold Regions Research and Engineering
of terrain material. The practical application of this is Laboratory, Hanover, NH.
readily apparent for tread design. Ahlvin, R.B., and P.W. Haley (1992) NATO refer-
ence mobility model edition II, NRMM II user’s
Deformation of terrain materials guide. WES Technical Report GL-92-19, U.S. Army
subjected to vehicle loading Waterways Experiment Station, Vicksburg, MS.
Research is also needed in areas relating to the de- Ahlvin, R., and S.A. Shoop (1995) Methodology for
formation of the materials beneath the wheel, such as: predicting for winter conditions in the NATO Refer-
1. Soil deformation under vehicles and implications ence Mobility Model. In Proceedings, 5th North
for terrain damage through soil compaction, rut
formation, root tearing, etc.
2. Washboard formation on secondary roads, includ-
ing the effects of suspension parameters, vehicle * Personal communication with J.F. Peters, GSL,
speed, and road material. ERDC, 1999.
3. Dynamic models of vehicles on pavements, in- ** Personal communication with J. Johnson and M.
cluding impacts on pavement structure and dete- Hopkins, CRREL, 2000.

56 to contents
American Conference of the ISTVS, Saskatoon, Sas- MSC/NASTRAN. International Journal of Vehicle
katchewan, Canada, May 1995, p. 320–334. Design, 18(2): 203–212.
Alger, R. (1988) Effect of snow characteristics on Burke, A., and O.A. Olatunbosun (1997c) Contact
shear strength. Contract Report DACA89-88-K-004, modelling of the tyre/road interface. International
Keweenaw Research Center, Michigan Technological Journal of Vehicle Design, 18(2): 194–202.
University, Houghton, Michigan. Chi, L., and S. Tessier (1995) Finite element analy-
Alger, R., and M. Osborne (1989) Snow characteri- sis of soil compaction reduction with high flotation
zation field data collection results. Final Report tires. In Proceedings, 5th North American Confer-
ACA89-85-K-002, Keweenaw Research Center, ence of the ISTVS, Saskatoon, Saskatchewan, Can-
Michigan Technological University, Houghton, ada, p. 167–176.
Michigan. Chiroux, R.C., W.A. Foster, Jr., C.E. Johnson,
Alkire, B.D. (1992) Seasonal soil strength by spec- S.A. Shoop, and R.L. Raper (1997) Three-
tral analysis of surface waves, Journal of Cold Re- dimensional finite element analysis of soil interaction
gions Engineering, 6(1): 2–38. with a rigid wheel. In Proceedings, 1997 ASAE An-
Alvarez Sanz, M.B. (1999) A parametric study of a nual International Meeting, Minneapolis, Minnesota.
three-dimensional tire finite element model. Master’s ASAE paper no. 971028, St. Joseph, Michigan.
thesis, Department of Mechanical Engineering and Clark, S.K. (1981) Mechanics of Pneumatic Tires.
Applied Mechanics, University of Michigan. U.S. Department of Transportation, National High-
Aubel, T. (1993) FEM simulation of the interaction way Traffic Safety Administration.
between elastic tyre and soft soil. In Proceedings, Danielson, K.T., and A.K. Noor (1997) Finite ele-
11th International Conference of the ISTVS, Lake ments developed in cylindrical coordinates for three-
Tahoe, Nevada, vol. 2, p. S791–802. dimensional tire analysis. Tire Science and Technol-
Aubel, T. (1994) The interaction between the rolling ogy, 25(1): 2–28.
tyres and the soft soil FEM simulation by VENUS Darnell, I. (In progress) Efficient three-dimensional
and validation. In Proceedings, 6th European Con- tire models. Ph.D. Dissertation, Department of Me-
ference of the ISTVS, Vienna, vol. 1, p. 169–188. chanical Engineering and Computational Mechanics,
Bailey, A.D., and C.E. Johnson (1989) A soil com- University of Michigan.
paction model for cylindrical stress states. Transac- Darnell, I., G.M. Hulbert, and C.W. Mousseau
tions of the ASAE, 32(3): 822–825. (1997) An efficient three-dimensional tire model for
Barber, J.R. (1991) Is modeling in tribology a useful vehicle dynamics simulations. Mechanical Structures
activity? In Tribological Modeling for Mechanical and Machinery, 25(1): 1–19.
Designers, ASTM STP 1105. DiMaggio, F.L., and I.S. Sandler (1971) Material
Bekker, G. (1956) Theory of Land Locomotion. Ann model for granular soils. Journal of Engineering Me-
Arbor, Mich.: The University of Michigan Press (2nd chanics Division, ASCE, 97(EM3): 935–950.
Edition, 1962). DiMaggio, F.L., and I.S. Sandler (1976) General-
Bekker, G. (1960) Off-Road Locomotion. Ann Ar- ized cap model for geological materials. Journal of
bor, Mich.: The University of Michigan Press Geotechnical Engineering Division, ASCE,
Bekker, G. (1969) Introduction to Terrain–Vehicle 102(GT7): 683–699.
Systems. Ann Arbor, Mich.: The University of Fervers, C.W. (1994) FE simulations of tyre-profile
Michigan Press. effects on traction on soft soil. In Proceedings, 6th
Blaisdell, G.L., P.W. Richmond, S.A. Shoop, C.E. European Conference of the ISTVS, Vienna, Austria,
Green, and R.G. Alger (1990) Wheels and tracks in p. 618–633.
snow: Validation study of the CRREL shallow snow Fervers, C.W. (1997) Effects of traction and slip,
mobility model. CRREL Report 90-9, U.S. Army investigations with FEM. In Proceedings, 7th Euro-
Cold Regions Research and Engineering Laboratory, pean Conference of the ISTVS, Ferrara, Italy, p.
Hanover, NH. 117–124.
Bowden, F.P. and D. Tabor (1964) The Friction Fervers, C.W. (1999a) Phenomena of tyre-profile on
and Lubrication of Solids. London: Oxford Univer- different soils. In Proceedings, 13th International
sity Press. Conference of the International Society of Terrain–
Burke, A., and O.A. Olatunbosun (1997a) Static Vehicle Systems, Munich, Germany, p. 337–344.
tyre/road interaction modelling. Meccanica, 32(5): Fervers, C.W. (1999b) Phenomena of air-filled tires
473–479. and soft terrain. Investigations with FEM. Doctoral
Burke, A., and O.A. Olatunbosun (1997b) New Thesis, University of Federal Armed Forces, Ham-
techniques in tyre modal analysis using burg, Germany (in German).

57 to contents
Foster, W.A., C.E. Johnson, R.L. Raper, and S.A. Kestler, M.A., S.A. Shoop, K.S. Henry, J.A. Stark,
Shoop (1995) Soil deformation and stress analysis and R.T. Affleck (1999) Rapid stabilization of thaw-
under a rolling wheel. In Proceedings, North Ameri- ing soils for enhanced vehicle mobility: A field dem-
can Conference of the ISTVS, Saskatoon, Saskatche- onstration project. CRREL Report 99-3, U.S. Army
wan, Canada, p. 194–203. Cold Regions Research and Engineering Laboratory,
Fukue, M. (1979) Mechanical Performance of Snow Hanover, NH.
under Loading. Tokyo: Tokai University Press (based Ladanyi, B. (1997) Mechanical properties data base
on author’s thesis, McGill University, 1977). for ground freezing applications. Ground Freezing
Gill, W.R., and G.E. Vanden Berg (1967) Soil Dy- 97, p. 43–52.
namics in Tillage and Traction. Agriculture Hand- Lade, P.V., and M.K. Kim (1995) Single hardening
book No. 316, Agricultural Research Service, USDA, constitutive model for soil, rock and concrete. Inter-
U.S. Government Printing Office. national Journal of Solids Structures, 32(14): 1963–
Goodsell, D. (1995) Dictionary of Automotive Engi- 1978.
neering. Warrendale, Pennsylvania: Society of Liu, C. (1994) Traction mechanisms of automobile
Automotive Engineers. tires on snow. Ph.D. dissertation, Vienna University
Green, C.E., and G.L. Blaisdell (1991) U.S. Army of Technology, Vienna, Austria.
wheeled versus tracked vehicle mobility performance Liu, C.H., and J.Y. Wong (1996) Numerical simula-
test program. Report no. 2: Mobility in shallow snow. tions of tire-soil interaction based on critical state soil
Technical Report GL-91-7, U.S. Army Waterways mechanics. Journal of Terramechanics, 33(5): 209–
Experiment Station, Vicksburg, MS. 222.
Haehnel, R. (2000) Summary of model changes for a Ludema, K.C. (1996a) Friction, Lubrication and
wheel rolling on a snow surface: FY2000. Unpub- Wear: A Textbook in Tribology. New York: CRC
lished memo. Press.
Harrison, W.L. (1975) Vehicle performance over Ludema, K.C. (1996b) Mechanism-based modeling
snow: Math-model validation study. U.S. Army Cold of friction and wear. Wear, 200(1-2): 1–7.
Regions Research and Engineering Laboratory, Mazanti, B.B., and C.N. Holland (1970a) Study of
Technical Report 268. soil behavior under high pressure. Defense Technical
HKS (Hibbitt, Karlsson, and Sorensen, Inc.) Information Center, Contract Report, report 1, vol. 1.
(1996) Analysis of geotechnical problems with AD709315.
ABAQUS. ABAQUS Course Notes, Pawtucket, Mazanti, B.B., and C.N. Holland (1970b) Study of
Rhode Island. soil behavior under high pressure. Defense Technical
HKS (Hibbitt, Karlsson, and Sorensen, Inc.) Information Center, Contract Report, report 1, vol. 2.
(1998) ABAQUS Theory and User’s Manuals. Paw- AD756124.
tucket, Rhode Island. Mellor, M. (1975) A review of basic snow mechan-
Hu, Y.K., and P.F.J. Abeels (1994) The deforma- ics. In Proceedings of the International Symposium
tion of agricultural tire on rigid surface by FEM. on Snow Mechanics, Grindelwald, Switzerland, April
American Society of Agricultural Engineers Interna- 1–4, 1974. IAHS-AISH Publication 114, p. 251–291.
tional Summer Meeting, Kansas City, Missouri, June Meschke, G., C. Liu, and H.A. Mang (1996) Large
19–22, 1993, ASAE Paper No. 941043. strain finite-element analysis of snow. Journal of
ISTVS (International Society of Terrain–Vehicle Engineering Mechanics, July: 581–591.
Systems) (1977) ISTVS standards. Journal of Ter- Mousseau, C.W., and G. M. Hulbert (1996) An
ramechanics, 14(3): 153–182. efficient tire model for the analysis of spindle forces
Jaeger, J.C., and N.G.W. Cook (1969) Fundamen- produced by a tire impacting large obstacles. Com-
tals of Rock Mechanics. New York: John Wiley and puter Methods in Applied Mechanics and Engineer-
Sons, Inc. ing, 135(1-2): 15–34.
Johnson, J., J. Brown, E. Gaffney, G. Blaisdell, Mundl, R., G. Meschke, and W. Liederer (1997)
and D. Solie (1992) Shock response of snow. Friction mechanism of tread blocks on snow surfaces.
CRREL Report 92-12, U.S. Army Cold Regions Re- Tire Science and Technology, 25(4): 245–264.
search and Engineering Laboratory, Hanover, NH. Pi, W.S. (1988) Dynamic tire/soil contact surface
Karafiath, L.L., and E.A. Nowatzki (1978) Soil interaction model for aircraft ground operations.
mechanics for off-road vehicle engineering. In Series Journal of Aircraft, 25(11): 1038–1044.
on Rock and Soil Mechanics. Clausthal, Germany Padovan, J. (1977) On standing waves in “tires.”
Trans Tech Publications, 2(1974/77), no. 5. Tire Science and Technology, 5(2): 83–101.

58 to contents
Padovan, J., and P. Padovan (1993) Spinup wear in Shoop, S.A. (1993) Thawing soil strength measure-
aircraft “tyres.” Tire Science and Technology, 21(3): ments for predicting vehicle performance. Journal of
138–162. Terramechanics, 30(6): 405–418.
Padovan, J. and P. Padovan (1994a) Modelling Shoop, S., and R. Alger (1998) Snow deformation
wear at intermittently slipping high speed interfaces. beneath a vertically loaded plate. In Proceedings,
Computers and Structures, 52(4): 795–812. ASCE Cold Regions Specialty Conference, Duluth,
Padovan, J., and P. Padovan (1994b) Modelling Minnesota, September 1998.
“tyre” performance during antilock braking. Tire Sci- Shoop, S., B. Young, R. Alger, and J. Davis (1994)
ence and Technology, 22(3): 182–204. Winter traction testing. Automotive Engineering,
Padovan, J., A. Kazempour, F. Tabaddor, and B. 102(1): 75-78, SAE paper 940110.
Brockman (1992) Alternative formulations of rolling Shoop, S.A., R. Haehnel, K. Kestler, K. Stebbings,
contact problems. Finite Elements in Analysis and and R. Alger (1999) Finite element analysis of a
Design, 11(4): 275–284. wheel rolling in snow. In Proceedings of the 10th
Peters, J.F. (1991) Computational aspects of endo- International Conference on Cold Regions Engineer-
chronic plasticity. In Proceedings, 7th International ing, Lincoln, New Hampshire, August, 1999, p. 519–
Conference on Computer Methods and Advances in 530.
Geomechanics, Cairns, Australia. Smith, D.M. (2000) Response of granular layers in
Pi, W.S. (1988) Dynamic tire/soil contact surface flexible pavements subjected to aircraft loading.
interaction model for aircraft ground operations. ERDC/GL TR-00-3, U.S. Army Corps of Engineers,
Journal of Aircraft, 25(11): 1038–1044. Engineer Research and Development Center, Vicks-
Richmond, P.W. (1995) Motion resistance of burg, MS.
wheeled vehicles in snow. CRREL Report 95-7, U.S. Timoshenko, S.P., and J.N Goodier (1970) Theory
Army Cold Regions Research and Engineering Labo- of Elasticity. New York: McGraw-Hill Book Com-
ratory, Hanover, NH. pany, Third Edition.
Richmond, P.W., S.A. Shoop, and G.L. Blaisdell Tordesillas, A. (1996) A contact mechanics ap-
(1995) Cold regions mobility models. CRREL Report proach to the soil-tire interaction problem. In
95-7, U.S. Army Cold Regions Research and Engi- Proceedings, Second North American Workshop on
neering Laboratory, Hanover, NH. Modeling the Mechanics of Off-Road Mobility,
SAE (Society of Automotive Engineers) (1992) Vicksburg, Mississippi.
SAE Glossary of Automotive Terms. Warrendale, Wong, J.Y. (1989) Terramechanics and Off-Road
Pennsylvania: Society of Automotive Engineers. Vehicles. New York: Elsevier.
Saliba, J.E. (1990) Elastic-viscoplastic finite- Wong, J.R., and A.R. Reece (1966) Soil failure be-
element program for modeling tire-soil interaction. neath rigid wheels. In Proceedings, 2nd International
Journal of Aircraft, 27(4): 350–357. Conference of the International Society for Terrain–
Schmid, I.C. (1995) Interaction of vehicle and ter- Vehicle Systems, p. 425–445.
rain: Results from 10 years research at IKK. Journal Wong, J.R., and A.R. Reece (1967) Prediction of
of Terramechanics, 32(1): 3–26. rigid wheel performance based on the analysis of
Schofield, A.N., and C.P. Wroth (1968) Critical soil-wheel stresses. Journal of Terramechanics, 4(2):
State Soil Mechanics. London: McGraw-Hill. 7–25.
Scott, R.F. (1985) Plasticity and constitutive rela- Wood, D.M. (1990) Soil Behavior and Critical State
tions in soil mechanics. Journal of Geotechnical En- Soil Mechanics. New York: Cambridge University
gineering, 11(5): 563–605. Press.
Shapiro, L., J. Johnson, M. Sturm, and G. Blais- Yong, R.N., and E.A. Fattah (1976) Prediction of
dell (1997) Snow mechanics: Review of the state of wheel-soil interaction and performance using finite
knowledge and applications. CRREL Report 97-3, element method. Journal of Terramechanics, 13(4):
U.S. Army Cold Regions Research and Engineering 227–240.
Laboratory, Hanover, NH. Yong, R.N., E.A. Fattah, and P. Boosinsuk (1978)
Shen, J., and R.L. Kushwaha (1998) Soil-Machine Analysis and prediction of tyre-soil interaction and
Interactions. New York: Marcel Dekker, Inc. performance using finite elements. Journal of Terra-
Shoop, S.A. (1990) Mechanisms controlling vehicle mechanics, 15(1): 43–63.
mobility on a thawing soil. In Proceedings of the Yong, R.N., E.Z. Fattah and N. Skiadas (1984)
10th International Conference of the International Vehicle Terrain Mechanics. New York: Elsevier.
Society of Terrain Vehicle Systems, Kobe, Japan,
August 1990, vol. I, p. 301–312.

59 to contents
 Form Approved
REPORT DOCUMENTATION PAGE OMB No. 0704-0188
Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the
data needed, and completing and reviewing this collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing
this burden to Department of Defense, Washington Headquarters Services, Directorate for Information Operations and Reports (0704-0188), 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302.
Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to any penalty for failing to comply with a collection of information if it does not display a currently valid
OMB control number. PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ADDRESS.
1. REPORT DATE (DD-MM-YY) 2. REPORT TYPE 3. DATES COVERED (From - To)
November 2001 Technical Report
4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER

5b. GRANT NUMBER

Finite Element Modeling of Tire–Terrain Interaction

5c. PROGRAM ELEMENT NUMBER

6. AUTHOR(S) 5d. PROJECT NUMBER

5e. TASK NUMBER

Sally A. Shoop
5f. WORK UNIT NUMBER

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION REPORT

NUMBER
U.S. Army Engineer Research and Development Center
Cold Regions Research and Engineering Laboratory
72 Lyme Road ERDC/CRREL TR-01-16
Hanover, NH 03755-1290

9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR / MONITOR’S ACRONYM(S)

Office of the Chief of Engineers

11. SPONSOR / MONITOR’S REPORT
Washington, D.C. 20314-1000 NUMBER(S)

12. DISTRIBUTION / AVAILABILITY STATEMENT

Approved for public release; distribution is unlimited.

Available from NTIS, Springfield, Virginia 22161.

13. SUPPLEMENTARY NOTES

14. ABSTRACT
The desire to incorporate theoretical mechanics into off-road vehicle performance prediction has generated great interest in applying
numerical modeling techniques to simulate the interaction of the tire and terrain. Therefore, a full three-dimensional model simulating a tire
rolling over deformable terrain was developed. Tires were simulated using a rigid wheel, a deformable tire simplified with user-defined
sidewall elements, and modal analysis tire models. Model comparisons with measured, hard-surface tire deformation and contact stress
showed very good agreement. The simplified tire model was much more computationally efficient but the modal analysis model yielded
better contact stress distribution. Each of the tire models was then combined with rolling on deformable terrain. Fresh snow and compacted
sand surfaces were modeled using critical-state plasticity models. The rigid wheel model was validated on snow using field measurements
of tire forces and snow deformation and then compared to performance predictions using the NATO Reference Mobility Model. These
comparisons indicate excellent agreement between the model and the measurements. Preliminary results of the modal analysis tire model on
snow show very little deformation in the tire, indicating that the rigid wheel simplification may be a good approximation for soft terrain.

15. SUBJECT TERMS

Mobility Tires Tire modeling

16. SECURITY CLASSIFICATION OF: 17. LIMITATION OF 18. NUMBER 19a. NAME OF RESPONSIBLE PERSON
OF ABSTRACT OF PAGES

a. REPORT b. ABSTRACT c. THIS PAGE 19b. TELEPHONE NUMBER (include area code)

U U U U 70
Standard Form 298 (Rev. 8-98)
to contents Prescribed by ANSI Std. 239.18