Professional Documents
Culture Documents
99-0105 81
The joint load transfer modeling capabilities of EverFE, a recently lem of shear transfer across discrete cracks in concrete has shown
developed rigid pavement three-dimensional finite element analysis the mechanics of aggregate interlock shear transfer to involve a
tool, are verified through comparisons with available experimental complex interaction between two deformable, rough surfaces. In
data. Dowel joint load transfer is examined via comparison of dis-
addition to contact between the rough joint surfaces, there typically
placements predicted by EverFE with results from laboratory tests of
two small-scale doweled pavement systems, and dowel looseness is is localized crushing, the amount of which depends on the joint
shown to be a probable cause for experimentally observed differential opening, normal restraint of the joint, the strength of the concrete,
joint displacements. Results of finite element analyses using EverFE’s and the size and distribution of the aggregate particles (12). Despite
nonlinear, two-phase aggregate interlock constitutive model are shown these complexities, most finite element models of rigid pavements
to agree well with available experimental data. A parametric study is idealize aggregate interlock with linear springs spanning the joint
performed that examines the effect of joint opening on aggregate inter-
(2,13,14) or Coulomb friction with contact (1,3). Although the use
lock load transfer and illustrates the importance of considering non-
linearities in joint load transfer when predicting pavement response. of linear springs may be reasonable for an examination of the effect
Recommendations for future research on joint load transfer modeling of aggregate interlock shear transfer effectiveness on the global slab
are also discussed. response (13), the rational choice of a spring stiffness may be diffi-
cult, if not impossible. Coulomb friction, although playing an impor-
tant role in aggregate interlock load transfer, does not wholly account
The realistic modeling of both aggregate interlock and dowel load for the complexities inherent in aggregate interlock because of the
transfer across rigid pavement joints is necessary for the accurate idealization of the joint surfaces as smooth.
prediction of pavement response to applied axle and thermal load- EverFE, a recently developed program for the 3D finite element
ings using the finite element method. Unfortunately, the complexi- analysis of rigid pavements (15), incorporates joint load transfer
ties inherent in the mechanics of joint load transfer make achieving modeling strategies that consider important aspects of dowel and
this requirement difficult. Although nonlinear three-dimensional aggregate interlock load transfer not previously addressed. In par-
(3D) finite element modeling is rapidly becoming a standard ana- ticular, EverFE implements an embedded dowel element that per-
lytical tool in research settings (1–3), methods for modeling both mits the explicit and rigorous consideration of dowel looseness (16)
aggregate interlock and dowel joint load transfer still warrant further and a nonlinear aggregate interlock constitutive model employing a
investigation. two-phase idealization of the slab concrete that accounts for aggre-
Dowel-slab interaction plays an important role in dowel load gate size, joint opening, and concrete strength (12,17). Two studies
transfer; in particular, dowel looseness has been shown experi- verifying these joint load transfer models via comparisons with
mentally (4,5) and analytically (6,7 ) to have potentially detrimen- available laboratory data (11,18) are presented. The importance of
tal effects on joint load transfer. Most two-dimensional (2D) finite incorporating load transfer nonlinearities is demonstrated through a
element analyses account for dowel-slab interaction by incorpo- parametric study, and conclusions and recommendations for future
rating a Winkler foundation between the dowel and surrounding research are also addressed.
slab (8,9). More sophisticated models recently have been devel-
oped that account for the finite embedded length of the dowel (10);
however, these still rely on the common assumption of a Winkler
MODELING STRATEGIES
foundation sandwiched between the dowels and surrounding slab.
Although the Winkler foundation modulus can be reduced to account
Discretization of Slab and Base
for dowel looseness, determination of the proper foundation mod-
ulus can be accomplished only through backcalculation using
The finite element models employed in this study are similar to that
measured data. Further, a backcalculated modulus is effectively a
shown in Figure 1. The slabs and base are linearly elastic and are
secant stiffness, valid only for one loading, geometry, and set of
discretized with 20-noded quadratic hexahedra. A Winkler founda-
material properties.
tion is used to represent the subgrade below the bottommost base
Colley and Humphrey (11) demonstrated that aggregate interlock
layer. Loss of contact between the slab and the upper base layer is
load transfer effectiveness and endurance depend on many factors,
explicitly considered by using a nodal contact approach, illustrated
including slab thickness, load magnitude, foundation type, subgrade
in Figure 1 (15,19). The model generation capabilities and solution
modulus, and aggregate shape. Research into the more general prob-
efficiency of EverFE are central in permitting the simulation of such
systems with relative ease on desktop computers. In addition, EverFE
W. G. Davids, Department of Civil and Environmental Engineering,
University of Maine, 5711 Boardman Hall, Orono, ME 04469-5711. J. P.
permits the consideration of various wheel and axle load configu-
Mahoney, Department of Civil and Environmental Engineering, University rations as well as linear temperature gradients through the slab
of Washington, Box 352700, Seattle, WA 98195-2700. thickness (not considered in the present study).
82 Paper No. 99-0105 TRANSPORTATION RESEARCH RECORD 1684
U de = [ ]
Ud
Ue
(1)
The new displacement vector for the embedded dowel element can
be transformed back to Ud as follows:
U d = TUde (2)
Kde = T T K d T (3)
τ = σ pu ( Ay + µ Ax) (5)
MODEL VALIDATION: DOWEL LOAD TRANSFER slipping plastic drinking straws over their ends. The model was cast
in a steel reaction box, and the ends of the slabs were clamped
This section presents a comparison of predicted pavement response between two angles to prevent displacement and increase their
with measured laboratory data (18). In particular, dowel looseness is effective length. For further details on all tests, see Hammons (18).
shown to be a potential cause for the observed differential vertical
joint displacements of a scale model jointed plain concrete pavement
system subjected to corner loading. Finite Element Model
Dowel Modeling
Two different finite element analyses were run for comparison with
test LSM-2: the first assumed no dowel looseness; the second run
assumed a uniform gap, γ, of 0.08 mm between the dowel and the
unloaded slab. This value of the gap was not measured, but it was
backcalculated to give a relative vertical displacement between the
two loaded and unloaded slabs that closely matches the experimen-
tally measured values. To place this value in perspective, the mea-
FIGURE 4 Experimental model and finite element mesh:
(a) model dimensions and instrumentation layout; and
sured wall thickness of a typical drinking straw is approximately
(b) finite element mesh for LSM-5. 0.13 mm—nearly 60 percent greater than γ. The vertical displace-
ment of the top surface of the slabs was interpolated along the line
y = 432 mm to give the analytical deflection profile corresponding
to that measured experimentally, as shown in Figure 6(a).
liquid, and the slab and treated base are modeled as elastic con-
tinua. Figure 4(b) shows the finite element mesh used to model test
LSM-5; an identical model without the base layer was used for
LSM-2.
The material properties used in the finite element model were
those determined experimentally by Hammons. The slab and base
layer moduli, E, were taken as 27 600 MPa and 1410 MPa, respec-
tively; the corresponding values of Poisson’s ratio, ν, were 0.18
and 0.20. For the dowels, E = 200 000 MPa and ν = 0.30. The rub-
ber pad had a modulus, k, of 0.09 MPa/mm as backcalculated with
a 3D finite element model developed by Hammons to give good
deflection profile comparisons with tests of a single undoweled
slab founded directly on the rubber pad and subjected to corner
loading.
The model boundary conditions were chosen to reflect those of
the laboratory tests. The displacements measured at transducers D7
through D10 near the slab corners [see Figure 4(a)] were maintained
at the slab ends. Although the slabs were cast against the reaction
box walls (which were coated with a form release agent), the sides
of the slabs and base layers (± y faces) were not restrained. There
were three reasons for this: (a) the slabs were restrained only when
displacing in the +y direction on the +y face and in the −y direction
on the −y face; (b) even if contact between the slabs and box walls
occurred, the restraint provided by the box walls was not that of a
rigid support; and (c) shrinkage of the slabs could have resulted in a
gap between the slabs and box walls before loading, which had to
be overcome before any contact could occur. FIGURE 5 Assumed profile of gap around dowels.
Davids and Mahoney Paper No. 99-0105 85
2 du
JTE = ( 7)
dl + du
du JTE
JTE = = (8)
dl 2 − JTE
Colley and Humphrey reported values of joint efficiency, JTE, for FIGURE 8 Comparison of measured and computed LTE:
various values of joint opening, w, and number of load cycles. Joint (a) results for 178-mm-thick slab; and (b) results for 229-mm-
efficiency is defined according to thick slab.
Davids and Mahoney Paper No. 99-0105 87
accuracy may be achieved by backcalculating the parameters σpu ment was used for the upper layer of interface elements [Figure 9(c)].
and µ to fit these particular test results. It must be noted that although EverFE implements the two-phase
nonlinear aggregate interlock model, some manual modification of
the input files was necessary to include these zero-stiffness elements.
ANALYTICAL STUDY The 150-mm-thick base layer was assumed to be a compacted
gravel with E = 150 MPa and ν = 0.20. Below the base layer, 300 mm
This section illustrates the importance of modeling nonlinearities in of natural soil was modeled as a linearly elastic continuum having
joint shear transfer via a parametric study on aggregate interlock E = 75 MPa and ν = 0.20. A dense liquid was used to represent the
shear transfer. In particular, the effect of joint opening, w, on system remaining natural soil with a modulus of 0.054 MPa/mm. The elas-
response is examined. The importance of accounting for aggregate tic material properties of the slab were E = 28 000 MPa and ν = 0.25,
interlock nonlinearity is emphasized by comparing results of models and the mass density, ρ, was 2400 kg/m3. To develop the two-phase
assuming equivalent linear joint springs with those employing the aggregate interlock model, µ = 0.4 and σpu = 45 N/mm2 were taken as
two-phase model for varying magnitudes of axle loads. typical; pk was fixed at 0.75 and dmax set to 20 mm.
The parametric studies employ a system similar to that shown previ- To study the effect of joint opening, a single, transversely centered
ously in Figure 1; a plan view of the model showing boundary con- 80-kN axle load was applied at the joint. As the joint opening, w,
ditions and loading is given in Figure 9(a). The slabs were 230 mm varies from 0.10 mm to 2 mm (a typical seasonal maximum for 3-
thick; for simplicity, the joint was not skewed, and the boundaries of to 5-m joint spacings), LTE decreases nearly 60 percent, as shown
the subgrade were not extended beyond the edges of the slabs. Axle in Figure 10(a).
loadings were applied at the joint through four wheels, each acting Contact conditions at the joint were monitored during the solu-
over a rectangular area 180 mm long by 180 mm wide (selected to tion, and the maximum relative normal displacement of the joint
give a realistic uniform pressure of approximately 0.62 MPa for an surfaces was approximately the minimum joint opening of 0.10 mm.
80-kN axle). The finite element mesh used in the analyses had approx-
Commensurate with the change in LTE is a rapid decrease in shear
imately 29,000 degrees of freedom and is shown in Figure 9(b). The
stresses on the joint surfaces with increasing w. The maximum shear
sawcut used at the joint was assumed to extend through one-third of
stress, corresponding to w = 0.10 mm, is 204 kPa; this is reduced
the pavement thickness; to reflect this detail, a zero-stiffness joint ele-
more than 50 percent for w = 0.40 mm and to only 11 kPa for
w = 1.60 mm. The maximum principal stresses occur at the bottom
of the loaded slab under the wheel loads and range from 704 kPa for
w = 0.10 mm to 1063 kPa for w = 2.0 mm, an increase of 51 percent;
as w increases, the loading approaches that of an unprotected edge.
in the present study, nonlinearities due to dowel looseness will study on the effect of joint load opening on aggregate interlock load
present similar difficulties. transfer was conducted.
In general, the validation studies indicated that the aggregate
interlock and dowel load transfer modeling capabilities of EverFE
CONCLUSIONS are reasonably accurate and viable. Further, the parametric study
results showed that joint load transfer nonlinearities can have sig-
This study presented experimental verification of the nonlinear joint nificant effects on pavement response, and a backcalculated linear
load transfer modeling strategies employed by EverFE. Dowel loose- spring stiffness should be used with caution when different loadings
ness was shown to be a probable cause for observed differential ver- are considered.
tical joint displacements in the scale model rigid pavement tests of Although the modeling techniques verified here represent sig-
Hammons (18), and results predicted by the two-phase aggregate nificant advances in quantifying aggregate interlock and dowel
interlock model compared reasonably well with the data of Colley joint load transfer, several issues should be addressed in future
and Humphrey (11). In addition to model validation, a parametric research:
Davids and Mahoney Paper No. 99-0105 89
• Both the loss of aggregate interlock shear transfer efficiency TRB, National Research Council, Washington, D.C., 1994, pp.
and the formation of dowel looseness under cyclic loading should 123–133.
4. Snyder, M. B. Cyclic Shear Load Testing of Dowels in PCC Pavement
be quantified. This likely would require significant extension to and Repairs. In Transportation Research Record 1215, TRB, National
refinement of the existing capabilities of EverFE. Research Council, Washington, D.C., 1989, pp. 246–257.
• More controlled laboratory studies under the high-cycle, low- 5. Buch, N., and D. G. Zollinger. Development of Dowel Looseness Pre-
stress conditions common to pavements should be conducted to diction Model for Jointed Concrete Pavements. In Transportation
Research Record 1525, TRB, National Research Council, Washington,
permit construction and calibration of more refined models. D.C., 1996, pp. 21–27.
• Field verification is necessary to provide a higher degree of 6. Zaman, M., and A. Alvappillai. Contact-Element Model for Dynamic
confidence in the model predictions. Analysis of Jointed Concrete Pavements. Journal of Transportation
Engineering, Vol. 121, No. 5, 1995, pp. 425– 433.
7. Davids, W. G., G. M. Turkiyyah, and J. Mahoney. Modeling of Rigid
Achieving these future research objectives would lead to more Pavements: Joint Shear Transfer Mechanisms and Finite Element Solu-
reliable and accurate quantification of aggregate interlock and tion Strategies. Report WA-RD 455.1. Washington State Department of
dowel load transfer. It must also be noted that the effects of tem- Transportation, Olympia, 1998.
8. Tabatabaie, A. M., and E. J. Barenberg. Structural Analysis of Concrete
perature and moisture-induced slab curling should be considered in Pavements. Journal of the Transportation Division, ASCE, Vol. 106,
future studies; EverFE currently permits the inclusion of a linear tem- No. TE5, 1980, pp. 832–849.
perature gradient. Integration of such models with a reliable, effi- 9. Tayabji, S. D., and B. E. Colley. Analysis of Jointed Concrete Pave-
cient, and user-friendly 3D finite element package such as EverFE ments. Report FHWS-RD-86-041. FHWA, U.S. Department of Trans-
portation, 1986.
and their use in conjunction with appropriate nondestructive joint 10. Guo, H., J. A. Sherwood, and M. B. Snyder. Component Dowel-Bar
evaluation techniques would aid researchers, designers, and planners Model for Load-Transfer Systems in PCC Pavements. Journal of
in determining pavement retrofit and maintenance schedules. Transportation Engineering, Vol. 121, No. 3, pp. 289–298, 1995.
11. Colley, B. E., and H. M. Humphrey. Aggregate Interlock at Joints in
Concrete Pavements. Bulletin 189, HRB, National Research Council,
Washington, D.C., 1967, pp. 1–18.
ACKNOWLEDGMENTS 12. Walraven, J. C. Fundamental Analysis of Aggregate Interlock. Journal of
the Structural Division, ASCE, Vol. 107, No. ST11, 1981, pp. 2245–2270.
13. Ioannides, A. M., and G. T. Korovesis. Aggregate Interlock: A Pure-
This work was supported in part by WSDOT under contract T9903-54 Shear Load Transfer Mechanism. In Transportation Research Record
and by fellowships from the Valle Program and the Osberg Foun- 1286, TRB, National Research Council, Washington, D.C., 1990, pp.
dation at the University of Washington. Michael Hammons of Applied 14–24.
14. Brill, D. R., G. F. Hayhoe, and X. Lee. Three-Dimensional Finite Ele-
Research Associates, Inc., is gratefully acknowledged for providing ment Modeling of Rigid Pavement Structures. In Aircraft/Pavement
the doweled slab laboratory data referenced in this manuscript. Technology: In the Midst of Change, ASCE, 1997, pp. 151–165.
15. Davids, W., G. Turkiyyah, and J. Mahoney. EverFE—Rigid Pavement
Three-Dimensional Finite Element Analysis Tool. In Transportation
Research Record 1629, TRB, National Research Council, Washington,
REFERENCES D.C., 1998, pp. 41–49.
16. Davids,W., and G. Turkiyyah. Development of Embedded Bending
Member to Model Dowel Action. Journal of Structural Engineering,
1. Uddin, W., R. M. Hackett, A. Joseph, Z. Pan, and A. B. Crowley. Three- Vol. 123, No. 10, 1997, pp. 1312–1320.
Dimensional Finite-Element Analysis of Jointed Concrete Pavement 17. Walraven, J. C. Rough Cracks Subjected to Earthquake Loading. Jour-
with Discontinuities. In Transportation Research Record 1482, TRB, nal of Structural Engineering, Vol. 120, No. 5, 1994, pp. 1510–1524.
National Research Council, Washington, D.C., 1995, pp. 26–32. 18. Hammons, M. I. Development of an Analysis System for Discontinuities
2. Kuo, C.-M., K. T. Hall, and M. I. Darter. Three-Dimensional Finite Ele- in Rigid Airfield Pavements. Report GL-97-3. U.S. Army Corps of Engi-
ment Model for Analysis of Concrete Pavement Support. In Trans- neers, 1997.
portation Research Record 1505, TRB, National Research Council, 19. Davids, W., and G. Turkiyyah. Multigrid Preconditioner for Unstruc-
Washington, D.C., 1995, pp. 119–127. tured Nonlinear 3D FE Models. Journal of Engineering Mechanics,
3. Zaghloul, S. M., T. D. White, and T. Kuczek. Evaluation of Heavy Load Vol. 125, No. 2, 1999, pp. 186–196.
Damage Effect on Concrete Pavements Using Three-Dimensional, Non-
linear Dynamic Analysis. In Transportation Research Record 1449, Publication of this paper sponsored by Committee on Rigid Pavement Design.