92 ■ Transportation Research Record 1853

Paper No. 03- 2223

Three-Dimensional Finite Element

Analysis of Jointed Plain Concrete
Pavement with EverFE2.2
William G. Davids, Zongmu Wang, George Turkiyyah, Joe P. Mahoney, and David Bush

The features and concepts underlying EverFE2.2, a freely available specifically for analyzing rigid pavements (6, 7). EverFE1.02, which
three-dimensional finite element program for the analysis of jointed was first made available in 1998 (7), addressed these difficulties
plain concrete pavements, are detailed. The functionality of EverFE has through the use of an interactive graphical user interface for easy model
been greatly extended since its original release: multiple tied slab or definition and visualization of results, specialized techniques for mod-
shoulder units can be modeled, dowel misalignment or mislocation can eling both dowel and aggregate interlock joint load transfer (2, 8), and
be specified per dowel, nonlinear thermal or shrinkage gradients can be fast iterative solution strategies for inclusion of inequality constraints
treated, and nonlinear horizontal shear stress transfer between the slabs for modeling slab–base separation and material nonlinearity (9).
and base can be simulated. Improvements have been made to the user Recently, EverFE2.2 has been developed, which retains the orig-
interface, including easier load creation, user-specified mesh refinement, inal capabilities of EverFE1.02 while incorporating the following
and expanded visualization capabilities. These new features are detailed, features that substantially extend its usefulness:
and the concepts behind the implementation of EverFE2.2 are explained.
In addition, the results of two parametric studies are reported. The first • The ability to model tied adjacent slabs and shoulders. Multi-
study considers the effects of dowel locking and slab–base shear transfer ple slab or shoulder systems can be modeled, and transverse tie bars
and demonstrates that these factors can significantly affect the stresses are explicitly incorporated.
in slabs subjected to both uniform shrinkage and thermal gradients. The • Extended dowel modeling capabilities. Dowel–slab interaction
second study examines transverse joint mislocation and dowel looseness can be captured via either the specification of dowel looseness or
on joint load transfer. As expected, joint load transfer is greatly reduced springs sandwiched between the dowels and slabs, and the effect of
by dowel looseness. However, while transverse joint mislocation can sig- dowel misalignment or mislocation can be simulated.
nificantly reduce peak dowel shears, it has relatively little effect on total • Modeling of nonlinear thermal gradients. Bilinear or trilinear
load transferred across the joint for the models considered. thermal gradients through the pavement thickness can be specified.
• Simulation of slab–base interaction. Separation of the base and
slab under tension is handled via inequality constraints, and interme-
The use of three-dimensional finite element (FE) methods for ana- diate degrees of horizontal slab–base shear transfer can be captured.
lyzing rigid pavements subjected to mechanical and environmental • Expanded postprocessing capabilities. In addition to visualizing
loadings has grown significantly in the last decade. The increased use slab stresses and displacements—as well as retrieving precise stress
of three-dimensional FE analysis has given pavement researchers and displacement values at specific coordinates—the user can view
and designers a better understanding of critical aspects of pavement shears and moments in individual dowels.
response that cannot be captured with analytical solutions, such as • Expanded library of axle loads. Loads ranging from single wheels
joint load transfer (1, 2), the effect of slab support on stresses (3), to dual-wheel, tandem axles can be quickly created, positioned, and
and pavement response under dynamic loads (4, 5). deleted, as shown in Figure 1a.
However, many aspects of rigid pavement behavior have not been
thoroughly studied with three-dimensional FE analysis. This can be This manuscript details the features of EverFE2.2 and the con-
attributed to several factors, including the complexity of concrete cepts underlying implementation, with a primary focus on model-
pavement structures (especially joint load transfer mechanisms), ing of the dowels and ties, treatment of nonlinear thermal gradients,
the need to consider both environmental and mechanical load effects, and simulation of slab–base interaction. The results of parametric
the difficulty of model generation and result interpretation, and the studies that consider the effects of dowel locking, slab–base shear
relatively long solution times required for large three-dimensional FE transfer, and transverse joint mislocation on pavement response are
analyses. These factors become especially challenging for the analyst reported to illustrate the flexibility and modeling capabilities of
when general-purpose FE programs are used. To circumvent these EverFE2.2.
issues, three-dimensional FE analysis packages have been developed

FEATURES OF EverFE2.2
W. G. Davids and Z. Wang, Department of Civil and Environmental Engineering,
University of Maine, 5711 Boardman Hall, Orono, ME 04469-5711. EverFE2.2 employs several element types to discretize concrete
G. Turkiyyah and J. P. Mahoney, Department of Civil and Environmental
Engineering, University of Washington, Box 352700, Seattle, WA 98195-2700.
pavement systems that have from one to nine slab or shoulder
D. Bush, Dynatest Consulting, Inc., 165 South Chestnut Street, Ventura, units. Up to three elastic base layers can be specified below the slab,
CA 93001. and the subgrade is idealized as either a tensionless or a tension-
Davids et al. Paper No. 03- 2223 93

(a)

(b)

FIGURE 1 Model generation with EverFE2.2: (a) axle and thermal load specification,
(b) typical model illustrating discretization and element types.

supporting dense liquid foundation. Twenty-noded quadratic hexa- longitudinal joints via transverse tie bars can also be modeled. Fig-
hedral elements are used to discretize the slabs and elastic base ure 1b is a screen shot of the EverFE2.2 meshing panel, showing
layers (10), and the dense liquid foundation is incorporated via many of the basic elements. (The user can selectively refine the
numerically integrated, eight-noded quadratic elements that are number of elements used to discretize the slabs and base or subgrade
meshed with the bottommost layer of solid elements. Linear or non- layers.) The remainder of this section highlights significant fea-
linear aggregate interlock joint load transfer as well as dowel load tures that are new to EverFE2.2; detailed discussions of the basic
transfer can be modeled at transverse joints. Load transfer across components, including the nonlinear aggregate interlock modeling
94 Paper No. 03- 2223 Transportation Research Record 1853

capabilities, are available in Davids and Mahoney (2), Davids (8), Nonlinear Thermal Gradients
and Davids and Turkiyyah (11).
Previous studies noted that thermal gradients through the depth
of concrete pavements are often nonlinear (12, 13). EverFE2.2
Dowel and Transverse Tie-Bar Modeling allows the consideration of this important effect by the specifi-
cation of a bilinear or trilinear approximation to a nonlinear gra-
EverFE2.2 models dowels and transverse tie bars explicitly with dient, which is easily defined in the loading panel (Figure 1a).
embedded flexural finite elements (8, 11), which has the advan- The temperature changes are converted to equivalent element pre-
tage of allowing the dowels and tie bars to be precisely located strains via the slab coefficient of thermal expansion, and these
irrespective of the slab mesh lines, as shown in Figure 1b. This strains are numerically integrated over the element volume to
embedded element formulation also permits significant savings in generate equivalent nodal forces (10). The 20-noded quadratic
computation time by allowing a range of load transfer efficiencies element employed by EverFE2.2 can accurately capture strains
to be simulated without requiring a highly refined mesh at the that vary linearly over its volume. This implies that multiple ele-
joints. Dowel–slab interaction can be captured either by specify- ments through the pavement thickness should be used to accu-
ing a length and magnitude of gap between the dowels and the rately model bilinear or trilinear thermal gradients. The effect of
slabs or by specifying dowel support moduli in the dowel local uniform or nonuniform shrinkage strains can be simulated through
coordinates, which translate into springs sandwiched between the their conversion to equivalent temperature changes for input to
dowels and slabs (Figure 2a). The latter approach was not avail- EverFE2.2.
able in EverFE1.02, and it permits varying degrees of dowel–slab
interaction to be modeled while avoiding the contact nonlinearity
inherent in the modeling of dowel looseness. However, this Simulation of Slab–Base Interaction
approach is a simplification of a complex phenomenon (8). The
localized stresses in the concrete surrounding the dowels may not Modeling interaction of the slab and base is crucial for accurately
be accurately predicted when the embedded element formulation predicting pavement response to axle loads near joints and ther-
is used. Tie–slab interaction is captured via user-specified tie-bar mal or shrinkage gradients. EverFE2.2 allows the specification of
support moduli in the tie-bar local coordinates. either perfect bond between the slab and base (no slip and no sep-
Once the dowels have been located within the model, the user can aration) or free separation of the slab and base under tension. In
specify four misalignment or mislocation parameters (∆ x, ∆z, α, β) both cases, the slab and base do not share nodes, and constraints
that shift an individual dowel along the x- and z-axes and define its are used to satisfy the required contact conditions (Figure 3). The
angular misalignment in the horizontal and vertical planes (see Fig- solution algorithm relies on a perturbed Lagrangian formulation
ure 2b and 2c). The dowel support moduli coincide with the local and a constraint updating scheme based on the current normal
dowel coordinate axes (q, r, s), which are rotated from the global stress between the slab and base.
(x, y, z) axes by the angles α and β. The meshing algorithm precisely Shear transfer between the slab and base can be important when
locates individual flexural elements within the mesh of solid elements analyzing pavements subject to uniform thermal expansion or con-
by first solving for the intersection of each dowel with solid element traction or shrinkage strains. Rasmussen and Rozycki (14) over-
faces and then subdividing each dowel into at least 20 individual viewed the factors governing slab–base shear transfer, noting that
quadratic embedded flexural elements. both friction and interlock between the slab and base play a role.

x
∆x

Original
position
α
Misaligned r
position q

Gap between Plan View

dowel and slab
Slab Original
x
C.L. position

z
∆z

β
s
Misaligned q
position
Dowel-slab springs
(a) Elevation
(b)

FIGURE 2 (a) Dowel–slab interaction and (b) dowel misalignment parameters.

Davids et al. Paper No. 03- 2223 95

Pairs of nodes vertically constrained

if compression at interface
τ Slab element
τo δzz

kSB

δx or δy
δo Interface Base element δx or δy
elements transfer
(a) shear stress
(b)

FIGURE 3 Modeling of (a) slab–base interaction and (b) interface shear transfer.

In addition, a bilinear, elastic-plastic shear transfer model was cal- EFFECT OF DOWEL LOCKING AND SLAB–BASE
ibrated on the basis of push tests of slabs on various bases. The study SHEAR TRANSFER ON THERMAL STRESSES
concluded that the effect of slab–base shear transfer should be in-
corporated in three-dimensional analyses of pavement systems. A The potential detrimental effects of dowel locking—where the dow-
study by Zhang and Li (15) focused on developing a one-dimensional els become effectively bonded to the slabs—on pavement response to
analytical model for predicting shrinkage-induced stresses in con- thermal loads are well recognized. Dowel locking is commonly attrib-
crete pavements that accounts for slab–base shear transfer. Like uted to dowel misalignment, which can cause flexure of the dowels
the model developed by Rasmussen and Rozycki, that model ulti- and large frictional forces to develop at locations of dowel–slab con-
mately relied on a bilinear, elastic-plastic shear transfer model. tact, or corrosion of the dowels, which can result in bond between
Zhang and Li concluded that the type of supporting base—and thus the dowels and slabs. In addition, one study suggested that friction
the degree to which it restrains slab shrinkage—significantly affects between properly aligned dowels and slabs can provide significant
slab stresses. axial restraint and increased stresses in slabs that are simultaneously
To capture slab–base shear transfer, EverFE2.2 employs a subjected to a uniform temperature change and a negative thermal
16-noded, zero-thickness quadratic interface element that is meshed gradient (1). Other studies (14, 15) also concluded that shear trans-
between the slab and the base (Figure 3). The element constitutive fer at the slab–base interface can significantly affect slab stresses.
relationship is based on that given by Rasmussen and Rozycki (14) Here, EverFE2.2 is used to simulate the effect of dowel locking on
and Zhang and Li (15). The bilinear constitutive relationship, defin- a rigid pavement system subjected to a variety of thermal and self-
ing the relationship between the shear stress (τ) and the relative slip weight loadings. The degree of slab–base interaction also is varied
between the slab and base is shown in Figure 3. This relationship is to study the effect of this important parameter on response.
characterized by an initial distributed stiffness kSB (MPa/mm) and
slip displacement δ0. (While kSB has the same units as the well-
known modulus of subgrade reaction, kSB is a distributed stiffness Model Description
in the horizontal direction, and the shear stresses developed at the
A three-slab system was modeled to capture the effect of the re-
slab–base interface depend on the relative horizontal displace-
straint provided by adjacent slabs. The 250-mm-thick slabs were
ments between the slab and the base layer.) This constitutive rela-
4,600 mm long and 3,600 mm wide, with a modulus of elasticity E
tionship is assumed to apply independently in both the x and y of 28,000 MPa, a Poisson ratio ν of 0.20, a coefficient of thermal
directions if the slab and base remain in contact, which implies that expansion of 1.1 × 10−5 per °C, and a density of 2,400 kg/m3. The
a compressive normal stress exists at the slab–base interface. That slabs were founded on a 150-mm-thick asphalt-treated base with E
there will be little or no shear transfer when slab–base separation of 3,500 MPa, ν of 0.20, and density of 2,000 kg/m3. The dense liq-
occurs is accommodated by setting the interface stiffness and shear uid foundation was assumed to have a modulus of subgrade reaction
stress to zero whenever δz > 0. Modeling this loss of shear trans- of 0.03 MPa/mm. Each transverse joint had 11 dowels 32 mm in
fer with loss of slab–base contact is important, especially when diameter and 450 mm long, spaced at 300 mm on center. The FE
thermal gradients are simulated. The interface element stiffness mesh, shown in Figure 4, had 3,024 solid elements. The center slab
matrix and nodal force vector are computed numerically via 3 × 3 was meshed with 18 × 18 elements in plan, and the outer slabs were
Gauss point integration. meshed more coarsely, as they are of secondary interest.
For very large values of kSB, this model approaches Coulomb fric- The analyses considered dowels that were both locked and un-
tion with a very large friction coefficient, and for very small values bonded (free slip). In all cases, the locked and unbonded dowels
of kSB, it is equivalent to a frictionless interface. An advantage of were assumed to have no looseness (i.e., they provided maximum
this modeling scheme is that the symmetry of the system stiffness vertical joint load transfer). No tensile bond stresses were allowed
equations is maintained, which allows the use of the existing, between the slab and the base, but three levels of slab–base shear
highly efficient preconditioned conjugate-gradient solver. Ideal- transfer were considered in the analyses to capture the effect of this
izing slab–base interaction with conventional Coulomb friction important parameter. The low degree of slab–base interaction corre-
would destroy this symmetry, requiring the use of more complex sponded to a slab–base interface shear stiffness kSB of 0.0001 MPa/mm,
(and likely less efficient) solution techniques. which is the minimum value used by EverFE2.2; this value might
96 Paper No. 03- 2223 Transportation Research Record 1853

TABLE 1 Maximum Principal Stresses Caused by Temperature

Curling or Shrinkage or Both

Degree of Slab-Base Interaction

Dowel Type Load Case Low Intermediate High

DL – T 0 0.159 (B) 0.594 (B)

DL + ∆ T 0.871 (B) 0.886 (B) 0.945 (B)
Locked DL + ∆ T – T 0.870 (B) 0.973 (B) 1.18 (B)
DL – ∆ T 0.689 (T) 0.705 (T) 0.815 (T)
DL – ∆ T – T 0.688 (T) 0.818 (T) 0.991 (T)
FIGURE 4 Finite element mesh used in parametric study on
DL – T 0 0.118 (B) 0.591 (B)
dowel locking.
DL + ∆ T 0.871 (B) 0.880 (B) 0.938 (B)
Free
DL + ∆ T – T 0.872 (B) 0.906 (B) 1.51 (B)
Slip
DL – ∆ T 0.689 (T) 0.703 (T) 0.785 (T)
DL – ∆ T – T 0.689 (T) 0.669 (T) 0.547 (T)
be expected when a bond-breaker such as polyethylene sheeting is
placed on the base before the slab pour. (Note that kSB cannot be *All values in MPa; letter in parentheses indicates either top (T) or bottom (B) of
taken as zero because the slabs would be horizontally unrestrained, slab.
giving an unstable model.) The intermediate slab–base shear trans-
fer parameters were kSB = 0.035 MPa/mm and δ0 = 0.60 mm, which
correspond to an asphalt-treated base (15). The high slab–base shear Parametric Study Results and Significance
transfer parameters of kSB = 0.416 MPa/mm and δ0 = 0.25 mm were
reported by Zhang and Li for a hot-mix asphalt concrete base (15). Table 1 shows the maximum principal stresses predicted in the cen-
Five load cases were considered: ter slab for all parameter combinations and loadings. These stresses
occurred at either the top center or the bottom center of the middle
• A uniform temperature change of −10°C (DL − T); slab. Note that when there is full bond between the dowels and slabs,
• A positive thermal gradient of 0.032°C/mm (DL + ∆T); the model may predict higher tensile stresses around the dowels;
• A negative thermal gradient of −0.032°C/mm (DL − ∆T); however, these stresses are not reliable because of insufficient mesh
• A positive thermal gradient of 0.032°C/mm plus a uniform refinement at the joints. Figure 5 shows a colormap of principal
temperature drop of −10°C (DL + ∆T − T); and stresses and the deformed shape of the system under DL − ∆T − T
• A negative thermal gradient of −0.032°C/mm plus a uniform assuming high slab–base shear transfer.
temperature change of −10°C (DL − ∆T − T). When an intermediate level of slab–base shear transfer is
assumed, dowel locking increases stresses due to a uniform shrink-
The term DL refers to model self-weight. The uniform temperature age load (DL − T ) by 35%. When high slab–base shear transfer
change is equivalent to a uniform slab shrinkage of 110 µ. exists, dowel locking has a much less marked effect for this load

(a)

(b)

FIGURE 5 (a) Tensile stresses on top of slabs and (b) displaced shape (scale factor 
500) (DL  T  T, high slab–base shear transfer).
Davids et al. Paper No. 03- 2223 97

case because of the significant reduction in overall slab shortening. 1.6

Dowel locking increases stresses for the DL − ∆T − T load case for
both intermediate slab–base shear transfer (16% increase) and high 1.5 Intermediate Slab-Base Shear
High Slab-Base Shear

Maximum Principal Stress (MPa)

slab–base shear transfer (81% increase). However, stress increases
caused by dowel locking are only 7% for the DL + ∆T − T load case 1.4
with intermediate slab–base shear transfer (although the overall
stresses are higher than for DL + ∆T − T ). This difference can be
1.3
attributed to the fact that under DL + ∆T − T, the bottom of the slab
is shrinking under a net temperature change of −14°C. This causes
shear stresses at the slab–base interface that are concentrated near 1.2
the edges of the slab and act away from the slab center, which tends
to increase tensile stresses in the bottom of the slab significantly 1.1
more than dowel locking alone. In contrast, under DL − ∆T − T, the
net temperature change at the bottom of the slab is only −6°C, and 1
the resulting shear stresses, which are concentrated near the center
of the slab, tend to reduce the peak tensile stress that occurs at the top 0.9
0 200 400 600 800 1000 1200 1400 1600 1800 2000
of the slab. Further, the ends of the slab are not in contact with the
base because of slab liftoff (see Figure 5). As a result, the increase in Dowel Axial Restraint Modulus (MPa)
tensile stress arising from the restraining effect of the dowels is more FIGURE 6 Variation in peak slab stress caused by dowel axial
pronounced. restraint (DL  T  T ).
A counterintuitive result is the 28% decrease in slab stresses
caused by dowel locking under DL + ∆T − T with a high degree of
slab–base shear transfer. This can be explained by the fact that dowel crease results from the increased shear stresses between the slab
locking tends to prevent contraction of the bottom of the slab, reduc- and the base under uniform temperature shrinkage that tend to
ing the relative displacements between the slab and base and thus the reduce the peak tensile stress at the top of the slab.
shear at the slab–base interface near the transverse joints. In fact, the In general, the results of the simulations indicate that there is a
maximum relative x-direction displacement between the central slab complex interaction among dowel locking, slab–base interaction,
and the base predicted by EverFE, which occurs at the slab ends, is and thermal loading. The need for three-dimensional analysis
0.078 mm when dowel locking exists, giving τ = 0.032 MPa. In con- (instead of one- or two-dimensional) when simulating both ther-
trast, when there is no dowel locking, the x-direction relative dis- mal gradients and shrinkage is evident: even under uniform shrink-
placements at the slab ends are 0.25 mm, implying that the peak age, the slab–base shear stresses acting at the bottom of the slab
value of τ = τ0 = 0.104 MPa. As discussed, this reduction in slab result in slab–base separation and a nonuniform distribution of
stress with dowel locking was not observed for the intermediate stresses over the slab thickness because of the eccentricity of the
degree of slab–base interaction, where the reduced stiffness kSB of shear stress respective to the center of gravity of the slab. How-
the slab–base interface allows a relative x-direction displacement of ever, creep of both slab and base, which is not considered by
0.208 mm when the dowels are locked and 0.308 mm when the dow- EverFE2.2, will mitigate these stress increases.
els are unbonded. These values result in relatively low slab–base
interface shear stresses of 0.0073 MPa and 0.011 MPa, respectively.
This explanation was further verified by running of simulations in EFFECT OF TRANSVERSE JOINT LOCATION ON
which the effect of the degree of bond between the dowels and slabs JOINT LOAD TRANSFER
on slab stress was simulated by varying the dowel–slab axial re-
straint modulus. Figure 6 shows the results of these analyses for Poor construction can lead to effective misalignment or mislocation
models with both high and intermediate degrees of slab–base shear of dowels at transverse contraction joints (16 ). While dowel mis-
transfer subjected to DL + ∆T − T. Note the increase in slab stresses alignment or mislocation is suspected to decrease load transfer (16),
with increasing dowel–slab restraint modulus for the case of inter- this topic has not been extensively studied either experimentally or
mediate slab–base shear transfer. Conversely, slab stresses decrease numerically. Here, the effect of transverse joint location is examined
with increasing dowel–slab restraint modulus, assuming a high by using the dowel mislocation–misalignment feature of EverFE2.2.
degree of slab–base shear transfer. For both degrees of slab–base
shear transfer, the limiting stresses given in Table 1 bound the results
shown in Figure 6. Model Description
Increasing slab–base shear transfer tends to increase slab stresses
significantly for most loadings. As expected, when kSB = 0.0001 MPa/ The FE model used in this parametric study has the same slab dimen-
mm, there are no slab stresses for the DL − T load case, as shrink- sions and material properties assumed in the previous parametric
age is effectively unrestrained; however, significant tensile stresses study. However, only two slabs are modeled, because the focus is joint
are observed for DL − T loading with intermediate slab–base shear load transfer, and the slab–base shear transfer parameters were fixed
transfer for both locked and unlocked dowels. The effect of increas- at kSB = 0.035 MPa/mm and δ0 = 0.60 mm. The only load case con-
ing slab–base shear transfer is also dramatic for the model with sidered is an 80-kN dual-wheel axle located at the joint and centered
locked dowels subjected to DL −∆T − T, where slab stresses increase transversely on the left-hand slab combined with a negative thermal
44% as slab–base shear transfer increases from low to high. Only gradient of −0.032°C/mm. Each slab was discretized with 18 × 18 ele-
the DL − ∆T − T loading with unbonded dowels shows a decrease ments in plan, and the slab and base each had two elements through
in slab stresses with increasing slab–base shear transfer. This de- their thickness.
98 Paper No. 03- 2223 Transportation Research Record 1853

Two primary parameters are considered in the analyses: dowel 10

mislocation (simulated through specification of ∆x, as shown in Fig-
ure 2a) and dowel looseness. Values of ∆x ranged from −100 mm to 9
100 mm, where ∆x = 0 corresponds to a perfectly located sawed joint; 8
note that a negative value of ∆x corresponds to a joint sawed too far
to the right (i.e., more of the dowel is located in the loaded slab than 7 Gap = 0 mm, ∆x = 0

Dowel Shear (kN)

Gap = 0.10 mm, ∆x = 0
in the unloaded slab). Dowel looseness was simulated by explicitly 6 Gap = 0.10 mm, ∆x = 100 mm
modeling gaps of 0 to 0.2 mm between the dowels and slabs, which Gap = 0.10 mm, ∆x = -100 mm
can have significant effects on joint load transfer (8, 17, 18). The 5
gaps were assumed to vary parabolically along the embedded por- 4
tions of each dowel, with no gap at the dowel start or end and with
the maximum gap at the joint. To ensure sufficient potential points 3
of nodal contact between the dowels and slabs, 24 three-noded 2
flexural elements were used to discretize each dowel.
1

0
Parametric Study Results and Significance -1500 -1000 -500 0 500 1000 1500
Dowel Location Across Joint (mm)
Figure 7 shows the variation in peak dowel shear and total shear
transferred across the joint with gap for ∆x = −100 mm, 0 mm, and FIGURE 8 Variation in dowel shear across joint with x
100 mm. The peak dowel shear occurs at the third dowel in from the (gap fixed at 0.10 mm).
pavement edge, which is centered between two wheels on one side
of the axle. As expected, both peak and total shear decrease rapidly
with increasing dowel looseness; when ∆ x = 0, total shear trans- However, while this equalization of dowel shear can be expected to
ferred across the joint decreases by 73% as the gap increases from lead to lower peak dowel–slab bearing stresses, it cannot be con-
0 mm to 0.2 mm. In addition, the effect of transverse joint location cluded from this that dowel mislocation is beneficial. This equal-
on peak dowel shear is pronounced for intermediate values of dowel ization of dowel shear implies less effective dowel load transfer and
looseness (0.05 to 0.10 mm). However, joint location has a small higher slab stresses caused by edge loading. In fact, as dowel loose-
effect on total load transferred across the joint. This can be explained ness increases from 0 to 0.2 mm with no joint mislocation, the joint-
by the equalization of shear between dowels that grows both with displacement load transfer computed between the two wheels on
increasing gaps and with increasing ∆x. Figure 8 shows the varia- each side of the axle decreases from 99% to 45%, and the peak ten-
tion in dowel shear across the joint for selected values of dowel sile stress on the slab bottom under the wheel load increases from
looseness and ∆ x, highlighting this equalization of dowel shear. 0.401 to 0.522 MPa.
The results of the analyses indicate that fairly small shifts in joint
location can have a large effect on peak dowel shears. Further,
dowel looseness has a large effect on joint load transfer. However,
10 total shear transferred across the joint remains relatively constant
Peak Dowel Shear (kN)

∆x = -100 mm with joint location, even at shifts in joint location approaching half
8
∆x = 0 the embedded length of the dowel. The results of this study cannot
6 ∆x = 100 mm
be considered conclusive, because only a single load case, system
4 geometry, and set of material properties were considered. Further,
2 dowel mislocation may produce high, localized stresses in the con-
crete surrounding the dowels that the models employed here cannot
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 capture. However, the need for three-dimensional analysis with
Magnitude of Gap (mm) which to simulate these effects is evident.
(a)

SUMMARY AND CONCLUSIONS

40
Total Shear (kN)

∆x = -100 mm
30 ∆x = 0 This paper highlighted the features of the program EverFE2.2, which
∆x = 100 mm was developed specifically for the three-dimensional FE analysis of
20 jointed plain concrete pavements. EverFE2.2 allows the modeling of
10
one to nine slab–shoulder units with tied adjacent slabs and shoulders
and the rigorous treatment of joint load transfer via dowels, aggregate
0 interlock, and transverse tie bars. Dowel misalignment or mislocation
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
can be specified per dowel. In addition, nonlinear thermal and shrink-
Magnitude of Gap (mm)
age gradients can be treated, and slab–base interaction—including
(b)
separation and horizontal shear stress transfer between the slab and
FIGURE 7 Variation in dowel shear with joint location and dowel base—can be incorporated in the analyses. The interactive, user-
looseness: (a) peak dowel shear and (b) total shear transferred friendly interface of EverFE2.2 eases model generation and result
across joint. interpretation through simple creation or deletion of a variety of axle
Davids et al. Paper No. 03- 2223 99

types, automatic mesh generation, and efficient visualization of slab REFERENCES

stresses and displaced shapes. The use of specialized solvers targeted
to the model geometry and mechanics allows solutions to be obtained 1. William, G. W., and S. N. Shoukry. 3D Finite Element Analysis of
rapidly on modern desktop machines. Temperature-Induced Stresses in Dowel Jointed Concrete Pavements.
International Journal of Geomechanics, Vol. 1, No. 3, 2001, pp. 291–308.
Two parametric studies were completed that illustrated the fea- 2. Davids, W. G., and J. P. Mahoney. Experimental Verification of Rigid
tures of EverFE2.2. These studies examined the effect of dowel Pavement Joint Load Transfer Modeling with EverFE. In Transporta-
locking and slab–base shear transfer on pavement stresses due to tion Research Record: Journal of the Transportation Research Board,
thermal gradients and uniform slab shrinkage, as well as the effect No. 1684, TRB, National Research Council, Washington, D.C., 1999,
pp. 81–89.
of dowel misalignment and looseness on pavement response. From
3. Kuo, C.-M., K. T. Hall, and M. I. Darter. Three-Dimensional Finite Ele-
these studies, the following conclusions can be drawn: ment Model for Analysis of Concrete Pavement Support. In Trans-
portation Research Record 1505, TRB, National Research Council,
• Slab stresses can be highly affected by shear transfer between Washington, D.C., 1995, pp. 119–127.
the slab and base. In turn, the degree of slab–base shear depends on 4. Shoukry, S. N. Backcalculation of Thermally Deformed Concrete Pave-
ments. In Transportation Research Record: Journal of the Transporta-
base type and the particular environmental loading (combination of tion Research Board, No. 1716, TRB, National Research Council,
temperature gradient and uniform shrinkage) considered in an analy- Washington, D.C., 2000, pp. 64–72.
sis. The complex interaction between the effect of slab–base shear 5. Vepa, T. S., and K. P. George. Deflection Response Models for
transfer and dowel locking is best captured with three-dimensional Cracked Rigid Pavements. Journal of Transportation Engineering,
Vol. 123, No. 5, 1997, pp. 377–384.
FE analysis.
6. Brill, D. R., and I. D. Parsons. Three-Dimensional Finite Element Analy-
• Dowel locking can have an effect on pavement stresses. The sis in Airport Pavement Design. International Journal of Geomechanics,
effect of dowel locking on stresses caused by shrinkage and Vol. 1, No. 3, 2001, pp. 273–290.
by combined shrinkage and thermal gradients is significant for 7. Davids, W. G., G. M. Turkiyyah, and J. P. Mahoney. EverFE: Rigid
the range of slab–base shear transfer values considered here. The Pavement Three-Dimensional Finite Element Analysis Tool. In Trans-
portation Research Record 1629, TRB, National Research Council,
effect of dowel locking was most pronounced for a combined neg- Washington, D.C., 1998, pp. 41–49.
ative thermal gradient and shrinkage, producing an increase in 8. Davids, W. G. Effect of Dowel Looseness on Response of Jointed
peak tensile stress of 81% when there is a high degree of slab–base Concrete Pavements. Journal of Transportation Engineering, Vol. 126,
shear transfer. No. 1, 2000, pp. 50–57.
• Mislocation of transverse doweled joints can affect joint load 9. Davids, W., and G. Turkiyyah. Multigrid Preconditioner for Unstruc-
tured Nonlinear 3D FE Models. Journal of Engineering Mechanics,
transfer. When moderate degrees of dowel looseness exist (0.05 to Vol. 125, No. 2, 1999, pp. 186–196.
0.10 mm), peak dowel shears can be reduced significantly by joint 10. Zienkiewicz, O. C., and R. L. Taylor. The Finite Element Method, Vol. 1,
mislocation. However, because of equalization of dowel shears, the 4th ed. McGraw-Hill, London, 1994.
total load transferred across the joint remains relatively constant 11. Davids, W., and G. Turkiyyah. Development of Embedded Bending
Member to Model Dowel Action. Journal of Structural Engineering,
even with a mislocation of the transverse joint approaching half the Vol. 123, No. 10, 1997, pp. 1312–1320.
embedded length of the dowel. 12. Masad, E., R. Taha, and B. Muhunthan. Finite Element Analysis of
• Dowel looseness has a large effect on joint load transfer. Para- Temperature Effects in Plain-Jointed Concrete Pavements. Journal of
bolically varying gaps around the dowels as small as 0.20 mm can Transportation Engineering, Vol. 122, No. 5, 1997, pp. 388–398.
13. Pane, I., W. Hansen, and A. R. Mohamed. Three-Dimensional Finite Ele-
reduce joint load transfer by as much as 73% under a combined
ment Study on Effects of Nonlinear Temperature Gradients in Concrete
80-kN axle load and negative thermal gradient. Pavements. In Transportation Research Record 1629, TRB, National
Research Council, Washington, D.C., 1998, pp. 58–66.
These parametric studies have not fully explored the features of 14. Rasmussen, R. O., and D. K. Rozycki. Characterization and Modeling
EverFE2.2, and it is expected to be a valuable tool for a wide range of Axial Slab-Support Restraint. In Transportation Research Record:
Journal of the Transportation Research Board, No. 1778, TRB, National
of problems in the forensic analysis of pavements as well as pave- Research Council, Washington, D.C., 2001, pp. 26–32.
ment design. EverFE2.2 is freely available, and documentation and 15. Zhang, J., and V. C. Li. Influence of Supporting Base Characteristics on
details for obtaining EverFE2.2 can be found at cae4.ce.washington. Shrinkage-Induced Stresses in Concrete Pavements. Journal of Trans-
edu/everfe/. portation Engineering, Vol. 127, No. 6, 2001, pp. 455–642.
16. Guide to Developing Performance-Related Specifications. FHWA-RD-
98-155, FHWA-RD-98-156, FHWA-RD-98-171, Vol. III, Appendix C.
www.tfhrc.gov/pavement/pccp/pavespec/. Accessed March 5, 2003.
ACKNOWLEDGMENTS 17. Zaman, M., and A. Alvappillai. Contact-Element Model for Dynamic
Analysis of Jointed Concrete Pavements. Journal of Transportation
EverFE2.2 was developed with financial support from the Wash- Engineering, Vol. 121, No. 5, 1995, pp. 425–433.
ington and California Departments of Transportation. The authors 18. Guo, H., R. M. Larson, and M. B. Snyder. A Nonlinear Mechanistic
Model for Dowel Looseness in PCC Pavements. Proc., Fifth Inter-
thank Linda Pierce of the Washington State Department of Trans- national Conference on Concrete Pavement Design and Rehabilitation,
portation and John Harvey of the University of California at Davis West Lafayette, Ind., 1993.
for their valuable advice and input during the development of
EverFE2.2. Publication of this paper sponsored by Committee on Rigid Pavement Design.