near-surface stress states in flexible pavements using

Upload: arthur-nnang

Post on 10-Jan-2016

219 views

Category:

Documents


0 download

DESCRIPTION

Near-surface Stress States in Flexible Pavements Using

TRANSCRIPT

  • Near-surface stress states in exible pavements usingmeasured radial tire contact stresses and ADINA

    Marc Novak, Bjorn Birgisson *, Reynaldo Roque

    Department of Civil and Coastal Engineering, University of Florida, P.O. Box 116580, Gainesville, FL 32611-6580, USA

    Abstract

    The nite element code ADINA was used to identify the three-dimensional stress states in a typical exible pavement

    conguration, resulting from measured radial tire contact stresses. The predictions show that measured radial tire

    contact stresses result in stress states being both larger in magnitude and more focused near the surface than those

    obtained from traditional uniform vertical loading conditions. In terms of eects of possible pavement damage

    mechanisms, predicted high near-surface shear stresses may be a part of an explanation for near-surface rutting failure

    modes, as supported by near-surface slip planes seen in the eld.

    2003 Elsevier Science Ltd. All rights reserved.

    Keywords: Tire contact stresses; Contact surfaces; Instability rutting; Pavement analysis

    1. Introduction

    Instability rutting in asphalt pavements occurs within

    wheel paths and is due to the lateral displacement of

    material within the pavement layer. It occurs when the

    structural properties of the compacted pavement are

    inadequate to resist the stresses imposed upon it. Despite

    instability rutting being the predominant mode of pre-

    mature rutting failures in modern exible pavements,

    current pavement structural design approaches do not

    deal with rutting in the asphalt concrete layer [18].

    Recent studies have shown that instability rutting is

    primarily a near-surface phenomenon, aecting only the

    top 13 in. of the asphalt concrete layer, with visible slip

    surfaces associated with the rutting failure [9,10]. Simi-

    larly, studies by Myers et al. [11] and de Beer et al. [12]

    have shown that tire contact stresses in this near-surface

    region are greatly inuenced by the structural charac-

    teristics and design of radial truck tires. Measured radial

    tire contact stresses are both distributed highly non-

    uniformly over the tire footprint and larger in magni-

    tude than the traditional uniform circular load used in

    pavement design. Dealing appropriately with instability

    rutting for both the design of asphalt mixtures and for

    the structural design of pavements, will require a com-

    plete understanding of the mechanisms that induce rut-

    ting within the surface layer and the identication of the

    key factors that aect these mechanisms.

    Currently, more than 98% of trucks use radial tires,

    because of the associated fuel savings, and higher reli-

    ability of newer tire structures [13,14]. Radial tire con-

    tact stresses are highly complex, with non-uniform

    vertical stresses throughout the tire contact area, as well

    as large lateral contact stress components in both the

    transverse and longitudinal (along wheel path) direc-

    tions. The eects of these complex tire contact stresses

    have not been widely analyzed. Myers et al. [15] and

    Roque et al. [16] present the results of a series of

    two-dimensional nite element analyses where a two-

    dimensional cross-section of measured three-dimensional

    tire contact stresses was applied on a layered half plane.

    The two-dimensional nite element analysis results show

    that the predicted stress state is signicantly dierent

    from that of a uniform vertical strip load. Similar results

    were also reported based on three-dimensional layered

    *Corresponding author. Tel.: +1-352-392-9537x1462; fax:

    +1-352-392-3394.

    E-mail address: [email protected] (B. Birgisson).

    0045-7949/03/$ - see front matter 2003 Elsevier Science Ltd. All rights reserved.doi:10.1016/S0045-7949(02)00413-3

    Computers and Structures 81 (2003) 859870

    www.elsevier.com/locate/compstruc

  • elastic theory solutions. However, because of the limi-

    tations of the semi-analytical layered elastic solution

    approach, a number of simplifying assumptions were

    made regarding the application and distribution of

    measured tire loads. These results were based on a tire

    contact measurement system developed by Pottinger

    [17], which was especially developed for tire research,

    and consists of 1200 distinct measurement points, which

    register contact stresses in the x, y and z direction. Todate, no detailed three-dimensional nite element anal-

    ysis of the eects of tire contact stresses has been per-

    formed, using the results from Pottingers tire contactmeasurement system [17]. A part of the challenge in

    modeling actual tire stresses is that the radial tire contact

    area is rather small (4060 square inches (in.2)) and

    highly non-uniform. Typical exible pavement struc-

    tures also consist of a relatively thin layer (48 in.) of

    asphalt concrete, overlying a granular base course (812

    in. thick), which rests on a semi-innite foundation.

    Hence, the combination of a small, highly non-uni-

    formly loaded contact area and relatively thin surface

    layers, connecting to a semi-innite half plane requires a

    large number of elements.

    In this study, the nite element code ADINA [18] is

    used in modeling the three-dimensional eects of mea-

    sured tire contact stresses in a typical pavement cong-

    uration. Although exible pavement materials are

    generally non-linear in nature, a rm understanding of

    the linear elastic stress states should precede any further

    analysis. All pavement layers are assumed to be linear

    elastic, and dynamic eects are ignored in favor of

    promoting a basic understanding of static stress states

    before complicating the analysis with dynamic eects.

    Due to the complicated nature of the measured radial

    tire contact stresses, contact surfaces were used exten-

    sively to control the size of the problem.

    The results presented show that the predicted stress

    states obtained with measured radial tire loads are dif-

    ferent from those resulting from circular uniform verti-

    cal loads. In particular, high transverse near-surface

    shear stresses are observed in the pavement, which are

    not present in the traditional approach, which may

    partly explain the near-surface instability rutting failure

    modes recently reported in the literature [9,10].

    In the following, an overview of the tire contact

    measurement system is provided, followed by the de-

    scription of the nite element modeling, and the pre-

    sentation of the modeling results.

    2. Measurement of radial tire contact stresses

    Based on contact stress measurements, Pottinger [17]

    has identied two distinct types of contact stress eects

    that exist under truck tires. These are generally referred

    to as the pneumatic eect and Poissons eect [13]. The

    overriding eect induced under radial truck tires is the

    Poissons eect. This is a direct result of tire construc-tion. Radial tires are constructed to have sti treads and

    exible sidewalls, to minimize the deformation of the tire

    during rolling. Thus, the lateral stresses induced on the

    road by the radial truck tire will tend to push out from

    the center of the tire ribs. In contrast, bias-ply tires tend

    to have high wall stiness and a exible tread, resulting

    in smaller lateral contact stresses. Modern radial truck

    tires also can withstand higher ination pressures and

    higher loads than bias-ply tires, resulting in higher

    contact stresses [13].

    By using triaxial load pin transducers inserted onto a

    at steel test track, Pottinger [17] was able to measure

    tire-interface forces and displacements for vertical, lon-

    gitudinal, and transverse axes. The experimental setup

    used was also capable of determining the rolling tire

    footprint shape. Fig. 1 shows the test track conguration

    that was used. The experimental setup consisted of a

    rolling steel treadmill device in which the tire was held in

    one location, while the bed was moved longitudinally,

    causing the tire to roll over a row of 16 transducers.

    Stresses and displacements were recorded every 0.20 in.

    longitudinally (parallel to wheel path) and every 0.15 in.

    transversely (perpendicular to wheel path) by varying

    the transverse position of the sensors resulting in over

    1200 contact points, at which contact stresses in the x, y,and z directions were recorded. This resulted in 3600distinct stress measurements for a radial tire with ve

    ribs and a gross contact area of 47 in.2 with an ination

    pressure of 115 psi. The measurements provided a high

    denition of actual tire contact stresses.

    The other available measurement system, developed

    by de Beer et al. [12], also measures contact stresses in

    the x, y, and z directions, but uses only 13 triaxial straingauge steel pins, mounted on a steel plate and xed ush

    with the road surface.

    Because of the complexities involved in measur-

    ing contact stresses under tires, it is not possible to ob-

    tain these measurements directly on real pavements. In

    particular, the question arises as to whether stresses

    measured under a tire on a rigid foundation with em-

    bedded sensors are similar to the contact stresses that a

    exible pavement will experience. Roque et al. [16] pre-

    sented a two-dimensional nite element model of a ra-

    dial tire, using the nite element program ABAQUS

    [19]. The results showed that contact stresses were nearly

    identical whether the contact surface was rigid or had

    properties similar to an asphalt pavement. Eectively,

    the stiness of the tire is so much less than the pavement,

    that the resulting contact stresses are similar to those

    obtained from a rigid steel bed. It was concluded that

    contact stress measuring devices, such as the one used in

    this study, with rigid foundations are suitable for the

    prediction of response of exible highway pavements

    [16].

    860 M. Novak et al. / Computers and Structures 81 (2003) 859870

  • 3. Pavement structure and loading conditions

    A typical three-layer (asphalt concrete, base, and

    foundation) pavement structure was used in this analy-

    sis, used previously by Myers et al. [15] and Drakos et al.

    [10]. The thickness of the asphalt concrete layer was 8

    in., overlying a 12-in. thick granular base course. The

    foundation was assumed to consist of a 52-in. thick

    layer. The properties of each layer were assumed to be

    isotropic, homogenous, and linear elastic. Table 1 lists

    the elastic moduli, Poissons ratio and the layer thicknessfor each of the pavement layers. The asphalt concrete

    layer modulus corresponds to that experienced on a

    warm summer day on a fairly new pavement, and the

    base and foundation values were chosen based on typi-

    cal measured values in the State of Florida [10,15].

    4. ADINA nite element model

    A part of the challenge with three-dimensional

    modeling of a pavement system with measured tire

    contact stresses is that the resulting nite element model

    needs to combine a small, highly non-uniformly loaded

    contact area with relatively thin surface layers, con-

    necting to a semi-innite half space, resulting in a large

    number of elements. The measured contact area under

    the radial tire used in this study is 47 in.2. The radial tire

    used was the same one used by Myers [15] and Drakos

    [10] and has been determined to be typical for radial

    truck tires currently used in the Unites States of Amer-

    ica.

    An initial assessment of the grid size requirements

    based on using a uniform stress distribution demon-

    strated that the three-dimensional model should be at

    least 72 in. deep and extend laterally at least 60 in. in

    each direction from the center of the tire contact load to

    adequately represent the semi-innite half space condi-

    tions associated with pavement problems. Because of the

    size of this model, the renement needed near the tire

    contact area, and the desire to remain within the 3:1

    element length to width ratio, the resulting memory re-

    quirements for the Silicon Graphics Multiprocessor

    computer available for this analysis exceeded the 1300

    MB RAM memory available. To overcome the limita-

    tions associated with building a traditional mesh, con-

    tact surfaces were introduced, where a ne graded mesh

    representing the loaded surface was attached (glued)

    onto a coarse-graded mesh. This allowed for the intro-

    duction of coarse meshes at distances further away from

    the loaded area where the change in stress was more

    gradual, and far eld stresses dominated the response.

    The use of contact surfaces was further justied based

    on the primary area of interest being the near-surface

    area under and immediately surrounding the loaded tire,

    thus negating any possible negative numerical eects of

    far away contact surfaces.

    The measured tire contact measurements reported

    tire contact stresses as uniform stresses acting over areas

    of 0.03 in.2 in size. Three dierent stresses were pro-

    vided, namely vertical normal stresses, transverse shear

    stresses, and longitudinal shear stresses. Ideally, each

    Fig. 1. Schematic of system developed by Pottinger [17] used to measure tire contact stresses.

    Table 1

    Material properties and layer thicknesses of pavement structure

    used in the nite element analysis

    Layer Modulus

    (psi)

    Poissonsratio

    Thickness

    (in.)

    Asphalt concrete 100,000 0.45 8

    Base 40,000 0.45 12

    Foundation 15,000 0.45 52

    M. Novak et al. / Computers and Structures 81 (2003) 859870 861

  • uniform stress should be applied to a single element.

    Unfortunately, the number of elements needed would

    have again exceeded the memory of the Silicon Graphics

    multiprocessor computer available for the analysis.

    Thus, the use of fewer elements was required under the

    contact area, which subsequently required the determi-

    nation of the equivalent nodal forces to be applied to

    each node. The appropriate nodal forces for each ele-

    ment were determined by converting each uniform stress

    into an equivalent concentrated force. These forces were

    then applied to each element and distributed over the

    element nodes according to the rules described below.

    If it is assumed that the concentrated forces ff g hascomponents fx, fy , and fz, then the element load vector,freg, acting on a surface of an element, is dened by:freg

    ZSe

    N TfUgdS 1

    where N T is the transpose of the shape function matrix,and fUg is the surface traction vector. The contributionof ff g to freg can be determined by viewing the con-centrated force as a large traction, fUg, acting over asmall area, dS. Subsequently, the concentrated forcevector ff g can be denoted as:ff g fUg dS 2The integral of N TfUgdS thus becomes N Tff g, re-sulting in Eq. (1), with n concentrated forces, becoming:

    freg Xni1

    N Ti ff gi 3

    where N i is the value of N at the location of ff gi.The nal three-dimensional mesh consisted of

    204,185 nodes with three degrees of freedom per node,

    resulting in a total of 612,555 degrees of freedom. The

    elements under the radial contact area had uniform di-

    mensions of 0.30 by 0.40 in. Contact surfaces were used

    for the transition from the asphalt layer to the base,

    from the base to the foundation, and from the ne mesh

    near the tire contact area to the peripheral areas. Fig. 2

    illustrates the nal three-dimensional mesh, with Fig. 3

    showing a plan view of the contact area of the three-

    dimensional mesh.

    Finally, Fig. 4 shows a plan view of the radial tire

    contact stresses applied as nodal forces onto the pave-

    ment model surface at locations consistent with ve tire

    ribs. Fig. 5 shows a cross-sectional prole of the tire

    contact stresses. It is apparent from Figs. 4 and 5 that

    the measured tire contact stresses are highly non-sym-

    metric, thus forcing the modeling of the whole tire

    contact structure.

    4.1. Solution process

    There are four dierent types of solution schemes

    available in ADINA, namely, (1) direct solver, (2) sparse

    solver for very large problems, (3) iterative solver, for

    non-linear problems, and (4) multigrid solver, for par-

    allel processing solutions. The direct solver requires a

    large amount of storage and is not recommended for

    large three-dimensional models [17]. The multigrid sol-

    ver is intended for large three-dimensional problems

    with very large systems of equations, but when contact

    Fig. 2. Three-dimensional nite element mesh used in the

    pavement response analysis.

    Fig. 3. Plan view of the contact area of the three-dimensional

    mesh used in the pavement response analysis.

    862 M. Novak et al. / Computers and Structures 81 (2003) 859870

  • surfaces are used, ADINA does not allow for the use of

    the multigrid solver. The best solver for large memory

    limited problems is the sparse solver. Unfortunately, for

    the number of equations anticipated, the system requires

    a 64-bit version solution solver, with only the 32-bit

    version currently available for this research. Hence,

    since the iterative solver is also recommended for large

    problems, it was used by default.

    The equilibrium equations to be solved in a non-

    linear static analysis of a nite element model with

    contact surfaces in ADINA are:

    tDtR tDtF 0 4

    where tDtR is the vector of the external nodal loads andtDtF is the force vector equivalent to the element stressesat time t Dt [18]. In non-linear analysis three iterationmethods/schemes are available in ADINA, namely, (1)

    full Newton method, with or without line searches, (2)

    modied Newton method, with and without line sear-

    ches, and (3) the BroydenFlectherGoldfarbShanno

    (BFGS) matrix update method. In this study, the full

    Newton method with lines searches was employed,

    based on its ability to converge and obtain accurate

    solutions [18]. The iterative solution process required

    1295 MB of RAM memory, with a resulting solution

    time of 9000 s, using a single processor.

    5. Predicted stress states

    In the following, a comparison will be made between

    predicted near-surface stress states predicted from: (1)

    an axisymmetric uniform vertical load, to simulate cur-

    rent practice, (2) a three-dimensional uniform vertical

    load applied over measured gross tire contact area, to

    evaluate the eects of the tire footprint shape, and (3) a

    three-dimensional radial tire load with vertical, tangen-

    tial and longitudinal contact stresses, to represent full

    measured radial truck tire loading eects. Near-surface

    vertical stresses, horizontal stresses, shear stresses, and

    conning stresses are presented and compared between

    the three dierent surface loading conditions used.

    5.1. Axisymmetric model

    Current practice in pavement engineering uses solu-

    tions based on three-dimensional layered elastic theory,

    in which the tire load is modeled as a circular uniform

    vertical load. Typical pavement engineering analysis and

    design programs that use this loading conguration in-

    clude BISAR [20], ELSYM5 [21], and WESLEA [22]. A

    two-dimensional axisymmetric model was generated in

    ADINA [18] to provide a comparison between the

    stresses induced by a circular uniform vertical load and

    the more complicated radial tire loading eects. Fig. 6

    displays the two-dimensional axisymmetric nite ele-

    ment mesh used in the analysis. Because of the sym-

    metric nature of the problem, only one half of the loaded

    area is modeled. The nite element model is 72 in. tall

    and 30 in. wide, with a uniform vertical surface load of

    115 psi, distributed over a radius of 4 in. The elements

    used consist of eight-noded isoparametric elements, with

    72 vertical rows of elements, each containing 99 ele-

    ments, for a total of 7128 elements. The layer thicknesses

    and elastic properties are the same as in the three-

    dimensional nite element model, discussed previously.

    To evaluate the results from the axisymmetric nite

    element model, a comparison was performed between

    predicted shear stresses at the edge of the loaded area and

    those obtained from a semi-analytic layered-elastic the-

    ory solution, using the program BISAR [20]. Fig. 7

    shows that the shear stress predictions obtained with

    ADINA [18] and BISAR [20] are very similar for the

    exact same loading conditions, meaning that the axi-

    symmetric model adequately captures the loading re-

    sponse due to the circular uniformly loaded vertical load.

    5.2. Comparison of vertical stress states

    Figs. 8 and 9 show the near-surface vertical stress

    contours obtained from the measured radial truck tire

    Fig. 4. Illustration of the contact area and the radial tire nodal

    forces used in the pavement response analysis.

    Fig. 5. Cross-sectional view of applied radial tire nodal forces used in the pavement response analysis.

    M. Novak et al. / Computers and Structures 81 (2003) 859870 863

  • loads and the corresponding three-dimensional uniform

    vertical load, respectively. The contour plots are ob-

    tained under the left-most rib of the three-dimensional

    radial tire, and the corresponding location for the uni-

    form vertical load. The radial tire produces new surface

    vertical stresses that are higher in magnitude than those

    produced by the uniform vertical load. Fig. 10 shows a

    comparison between predicted vertical stresses at the

    edge of the radial tire and those obtained from the three-

    dimensional vertical load and the axisymmetric loading

    case. Below 1.0 in., the radial tire loading case ap-

    proaches the uniform vertical and the axisymmetric

    loading cases.

    Fig. 6. A cross-sectional view of the axisymmetric nite ele-

    ment mesh used for comparison purposes.

    0.0

    0.5

    1.0

    1.5

    2.0

    2.5

    3.0

    3.5

    0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0Shear Stress (psi)

    Dep

    th (i

    nche

    s)

    BISAR

    ADINA

    Fig. 7. Comparison of shear stresses under the edge of a two-dimensional axisymmetric circular uniform vertical load predicted with

    BISAR and ADINA.

    Fig. 8. Predicted vertical stress contours under left edge of

    three-dimensional radial tire load.

    864 M. Novak et al. / Computers and Structures 81 (2003) 859870

  • The three-dimensional uniform vertical and two-

    dimensional axisymmetric results show very similar

    vertical stress proles at the edge of the tire, implying

    that for uniform loading conditions, the circular uni-

    form vertical loading condition is suciently detailed in

    many cases, negating the need for three-dimensional

    modeling. However, the addition of lateral surface

    tractions, as in the radial tire case, signicantly changes

    the near-surface vertical stress proles, thus requiring a

    full three-dimensional analysis of loading eects.

    5.3. Comparison of horizontal stress states

    Figs. 11 and 12 show the near-surface horizontal

    stress contours obtained under the left-most rib of the

    three-dimensional radial tire, and the corresponding

    location for the uniform vertical load. The radial tire

    contact stresses produce horizontal compressive stresses

    that are both higher in magnitude and more intense in

    the top 0.25 in. than the corresponding uniform vertical

    load. These high very near-surface stress states may be

    partly due to the discrete nature of the tire contact

    Fig. 9. Predicted vertical stress contours under left edge of

    three-dimensional uniform vertical tire load acting over mea-

    sured tire contact surface.

    0.0

    0.5

    1.0

    1.5

    2.0

    2.5

    3.0

    3.5

    0.0 20.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0

    Stress (psi)

    Dep

    th (i

    nche

    s)

    AxisymmetricRadialUniform

    Fig. 10. A comparison of predicted vertical stress with depth at the left edge of loaded area.

    Fig. 11. Predicted horizontal stress contours under left edge of

    three-dimensional radial tire load.

    M. Novak et al. / Computers and Structures 81 (2003) 859870 865

  • measurement system and the constraints associated with

    applying the measured tire contact stresses as discrete

    nodal forces in the nite element model. However, the

    general pattern of stresses should be representative,

    discounting the higher local peaks in stress under dis-

    crete nodal locations. Fig. 13 shows a comparison of

    horizontal stress with depth at the edge of the tire load.

    The results show that the high horizontal stresses dissi-

    pate very quickly over the top 0.25 in. of the pavement,

    below which they become similar in magnitude to those

    produced by the uniform loading conditions. These re-

    sults imply that the contribution from bending may

    dominate the response at greater depths for this partic-

    ular pavement system, overwhelming the eects from the

    lateral tire contact stresses.

    5.4. Comparison of shear stress states

    Figs. 14 and 15 show the near-surface shear stress

    contours obtained under the left-most rib of the three-

    dimensional radial tire, and the corresponding location

    Fig. 12. Predicted horizontal stress contours under left edge of

    three-dimensional uniform vertical tire load acting over mea-

    sured tire contact surface.

    0.0

    0.5

    1.0

    1.5

    2.0

    2.5

    3.0

    3.5

    0.0 20.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0 180.0

    Stress (psi)

    Dep

    th (i

    nche

    s)

    AxisymmetricRadialUniform

    Fig. 13. Comparison of predicted horizontal stress with depth at the left edge of loaded area.

    Fig. 14. Predicted shear stress contours in the lateral direction

    (yz-plane) under left edge of three-dimensional radial tire load.

    866 M. Novak et al. / Computers and Structures 81 (2003) 859870

  • for the uniform vertical load, respectively. Again, the

    radial tire contact stresses produce shear stresses that are

    both higher in magnitude and more intense in the top 1

    in. than the corresponding uniform vertical load. The

    zone of maximum shear is also closer to the surface for

    the radial tire contact stress case.

    Figs. 16 and 17 show a plot of the near-surface shear

    stresses under the edge of the loaded area, as well as 0.15

    in. away from the loading. The results show high near-

    surface shear stresses (68 psi) under the edge of the

    loaded tire for the radial tire contact stress case, dissi-

    pating rapidly both vertically and horizontally away

    from the tire. In contrast, the uniform vertical loading

    cases do not show these high near-surface shear stresses.

    5.5. Comparison of hydrostatic pressure states

    Since asphalt concrete, like other granular geomate-

    rials, is pressure dependent, the eects of tire contact

    stresses on the distribution of hydrostatic pressures was

    also evaluated. The hydrostatic pressure, herein denoted

    simply as pressure, is dened as the sum of the normal

    stresses divided by three. Figs. 18 and 19 show contour

    plots of pressure obtained under the left-most rib of the

    three-dimensional radial tire, and the corresponding

    location for the uniform vertical load, respectively. The

    radial tire conguration produces much higher pressures

    near the edges of the loaded area than the corresponding

    uniformly loaded cases, implying a conning eect due

    to the radial tire contact stresses.

    Figs. 20 and 21 show proles of pressure versus depth

    under the edge of the tire and at a distance of 0.15 in.

    away from the tire. Again, the results show that pressure

    dissipates fast in both the vertical and horizontal direc-

    tions away from the loaded area, dropping from a near-

    surface high of 158 psi under the edge of the loaded area,

    down to 9 psi at a horizontal distance of 0.15 in. away

    from the tire.

    6. Summary and conclusions

    The commercial nite element code ADINA [18] was

    used to identify the stress states in a typical exible

    pavement conguration, resulting from measured radial

    tire contact stresses. Due to the complicated nature of

    the measured tire contact stresses used in this paper, the

    resulting nite element model consisted of 204,185

    nodes, with a total of 612,555 degrees of freedom. An

    Fig. 15. Predicted shear stress contours in the lateral direction

    (yz-plane) under left edge of three-dimensional uniform verticaltire load acting over measured tire contact surface.

    0.0

    0.5

    1.0

    1.5

    2.0

    2.5

    3.0

    3.5

    0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0

    Shear Stress (psi)

    Dep

    th (i

    nche

    s)

    RadialUniformAxisymmetric

    Fig. 16. Comparison of predicted shear stresses in the lateral direction (yz-plane) with depth at the left edge of loaded area.

    M. Novak et al. / Computers and Structures 81 (2003) 859870 867

  • iterative solution scheme consisting of the full Newton

    method with lines searches was employed. The iterative

    solution process required 1295 MB of RAM memory,

    with a resulting solution time of 9000 s, using a single

    processor. Due to memory limitations, contact surfaces

    were used extensively in the three-dimensional nite el-

    ement model. The results show clearly that contact

    surfaces can be employed successfully if they are su-

    ciently far away from the areas of interest, since no

    signicant numerical errors were observed in the near-

    surface stress states due to the presence of contact sur-

    faces.

    In this paper, measured tire contact stresses are ap-

    plied as nodal forces in a three-dimensional nite ele-

    ment model. The representation of measured tire contact

    stresses at discrete points is currently a function of the

    limitations of available tire contact measurement sys-

    tems [17]. The authors recognize that some predicted

    high concentrations or peaks in local near-surface stress

    states under nodal points may be partly due to the dis-

    crete nature of the tire contact measurement system and

    the constraints associated with applying the measured

    tire contact stresses as discrete nodal forces in the nite

    element model. However, the general pattern of stresses

    Fig. 18. Predicted pressure contours under left edge of three-

    dimensional radial tire load.

    0.0

    0.5

    1.0

    1.5

    2.0

    2.5

    3.0

    3.5

    0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0

    Shear Stress (psi)

    Dep

    th (i

    nche

    s)

    RadialUniformAxisymmetric

    0.0

    0.5

    1.0

    1.5

    2.0

    2.5

    3.0

    3.5

    0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0

    Shear Stress (psi)

    Dep

    th (i

    nche

    s)

    RadialUniformAxisymmetric

    Fig. 17. Comparison of predicted shear stresses in the lateral direction (yz-plane) with depth at 0.15 in. away from the left edge ofloaded area.

    Fig. 19. Predicted pressure contours under left edge of three-

    dimensional uniform vertical tire load acting over measured tire

    contact surface.

    868 M. Novak et al. / Computers and Structures 81 (2003) 859870

  • should be representative, discounting the higher local

    peaks in stress under discrete nodal locations.

    Even though it has been shown that contact stresses

    obtained with the tire contact measurement system used

    were nearly identical whether the contact surface was

    rigid or had properties similar to an asphalt pavement

    [16], it has to be recognized that the an actual tire has

    some transverse rigidity [16]. Thus in reality subsequent

    application of measured tire contact stresses at discrete

    points without incorporating the actual tire structure is

    an approximation of the conditions present in a real

    situation. However, the current approach represents the

    rst attempt to model the eects of measured tire in-

    terface stresses in three dimensions. The incorporation

    of a full tire model would be signicantly more com-

    plicated, with numerous other issues needing to be ad-

    dressed, such as the eective stiness of radial tire walls

    and treads that are reinforced with steel wiring, and the

    net eect of tire bulging and deformation on the redis-

    tribution of stresses onto the pavement.

    Within the context of the limitations of the current

    tire measurement system, the results in this paper still

    provide a good starting point for developing the needed

    level of understanding of the mechanisms that may in-

    duce instability rutting in the near-surface region of an

    asphalt concrete layer. Current pavement structural

    0.0

    0.5

    1.0

    1.5

    2.0

    2.5

    3.0

    3.5

    0.0 20.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0 180.0

    Pressure (psi)

    Dep

    th (i

    nche

    s)

    RadialUniformAxisymmetric

    Fig. 20. Comparison of predicted pressures with depth at the left edge of loaded area.

    0.0

    0.5

    1.0

    1.5

    2.0

    2.5

    3.0

    3.5

    0.0 20.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0 180.0

    Pressure (psi)

    Dep

    th (i

    nche

    s)

    RadialUniformAxisymmetric

    Fig. 21. Comparison of predicted pressures with depth at 0.15 in. away from the left edge of loaded area.

    M. Novak et al. / Computers and Structures 81 (2003) 859870 869

  • design approaches do not deal with rutting in the asphalt

    concrete layer. The results presented in this paper show

    that measured radial tire contact stresses result in near-

    surface stress states are dierent from those obtained

    using traditional circular uniform-loading conditions.

    The resulting stress states appear larger in magnitude

    and more focused near the surface than those obtained

    from uniform vertical loading conditions. In particular,

    higher transverse near-surface shear stresses are ob-

    served in the top 1 in. of the pavement, which may partly

    explain the near-surface instability rutting failure modes

    recently reported in the literature [9,10]. However, at

    depths below 12 in., bending eects tended to dominate

    the response, minimizing the dierence between the ra-

    dial tire stress states and those obtained from uniformly

    distributed loading conditions.

    In summary, further research is needed, in which more

    radial tire congurations are evaluated, along with other

    pavement geometries. The eects of variation of modulus

    within the asphalt layer and non-linear material behavior

    may also be important, and require further study.

    References

    [1] American Association of State Highway and Transporta-

    tion Ocials (AASHTO). AASHTO guide for design of

    pavement structures. Washington, DC: AASHTO; 1998.

    [2] Claessen AIM, Edwards JM, Sommer P, Uge P. Asphalt

    pavement design: The shell method. In: Proceedings of the

    Fourth International Conference on the Structural Design

    of Asphalt Pavements. University of Michigan, 1977. p. 1.

    [3] The Asphalt Institute. Research and development of the

    Asphalt Institutes thickness design manual (MS-1), 9th ed.Research report 82-2. College Park, Maryland: The

    Asphalt Institute; 1982.

    [4] Maree JH, Freeme CR. The mechanistic design method to

    evaluate the pavement structures in the catalogue of the

    draft TR H4 1980. Technical report RP/2/81. Pretoria,

    South Africa: NITRR; 1981.

    [5] Kenis WJ, Sherwood JA, McMahon RF. Verication and

    application of the VESYS structural subsystem. In: Pro-

    ceedings of the Fifth International Conference on the

    Structural Design of Asphalt Pavements, vol. 1. University

    of Michigan and Delft University of Technology, 1982.

    p. 33348.

    [6] Brown JF, Brunton JM, Pell PS. The development and

    implementation of analytical pavement design for British

    conditions. In: Proceedings of the Fifth International

    Conference on the Structural Design of Asphalt Pave-

    ments, vol. 1. University of Michigan and Delft University

    of Technology, 1982. p. 17491.

    [7] Bayomy FMA, Fawzi A, Smith RM. Mechanistically

    based exible overlay design system for Idaho. J Transport

    Res Board 1996;TRR1563:109.

    [8] Timm DH, Birgisson B, Newcomb DE. Development of

    mechanistic-empirical pavement design for Minnesota. J

    Transport Res Board 1998;TRR1629:1818.

    [9] Dawley CB, Hogenwiede BL, Anderson KO. Mitigation of

    instability rutting of asphalt concrete pavements in Leth-

    bridge, Alberta, Canada. J Assoc Asphalt Paving Technol

    1990;59:481508.

    [10] Drakos CA, Roque R, Birgisson B. Eect of measured tire

    contact stresses on near-surface rutting. J Transport Res

    Board 2001;TRR1764:5969.

    [11] Myers L, Roque R, Ruth B, Drakos C. Measurement

    of contact stresses for dierent truck tire types to evalu-

    ate their inuence on near-surface cracking and rutting.

    J Transport Res Board 1999;TRR1655:17584.

    [12] de Beer M, Fisher C, Jooste F. Determination of

    pneumatic tire/pavement interface contact stresses under

    moving loads and some eects on pavements with thin

    asphalt surfacing layers. In: Proceedings of the Eighth

    International Conference on Asphalt Pavements. Seattle,

    Washington, 1997. p. 179226.

    [13] Myers L, Roque R, Ruth BE. Mechanisms of surface-

    initiated longitudinal wheel path cracks in high-type

    bituminous pavements. J Assoc Asphalt Paving Technol

    1998;TRR1667:40232.

    [14] Roque R, Myers L, Ruth BE. Loading characteristics of

    modern truck tires and their eects on surface cracking of

    asphalt pavements. In: Proceedings of the Fifth Interna-

    tional Conference on the Bearing Capacity of Roads and

    Airelds, vol. 1(1). 1998. p. 93102.

    [15] Myers LA, Roque R, Birgisson B. Propagation mecha-

    nisms for surface initiated longitudinal wheelpath cracks. J

    Transport Res Board 2001;778:11321.

    [16] Roque R, Myers LA, Birgisson B. Evaluation of measured

    tire contact stresses for the prediction of pavement response

    and performance. J Transport Res Board 2000;TRR1716:

    7381.

    [17] Pottinger MG. The three-dimensional contact stress eld

    of solid and pneumatic tires. Tire Sci Technol 1992;20(1):

    332.

    [18] Bathe KJ. ADINA system 7.5, User manual. Watertown,

    Mississippi: ADINA R&D, Inc., 2001.

    [19] Hibbit, Karlsson, and Sorensen, Inc. ABAQUS, User

    manual. Pawtucket, Rhode Island: Hibbit, Karlsson, and

    Sorensen, Inc., 1997.

    [20] De Jong DL, Peatz GF, Korswagen AR. Computer

    program BISAR, layered systems under normal and tan-

    gential loads. External report. Amsterdam, The Nether-

    lands: Koninklijke/Shell-Laboratorium; 1973.

    [21] Alhborn G. Elastic layered system with normal loads.

    Berkeley, California: ITTE, University of California;

    1972.

    [22] Van Caulwelaert FJ, Alexander DR, White TD, Barker

    WR. Multi-layer elastic program for backcalculating layer

    moduli in pavement evaluation. In: Bush III AJ, Baladi

    GY, editors. Nondestructive testing of pavements and

    backcalculation of moduli ASTM STP 1026. Philadelphia,

    Pennsylvania: American Society for Testing and Materials;

    1989. p. 17188.

    870 M. Novak et al. / Computers and Structures 81 (2003) 859870

    Near-surface stress states in flexible pavements using measured radial tire contact stresses and ADINAIntroductionMeasurement of radial tire contact stressesPavement structure and loading conditionsADINA finite element modelSolution process

    Predicted stress statesAxisymmetric modelComparison of vertical stress statesComparison of horizontal stress statesComparison of shear stress statesComparison of hydrostatic pressure states

    Summary and conclusionsReferences