Professional Documents
Culture Documents
net/publication/273023432
CITATIONS READS
13 349
2 authors, including:
Mesbah Ahmed
Gemini Technologies, Inc.
34 PUBLICATIONS 77 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Field and Laboratory Evaluation of Warm-Mix Asphalt (Phase II) View project
All content following this page was uploaded by Mesbah Ahmed on 04 December 2017.
Abstract: The falling weight deflectometer (FWD) testing develops a deflection basin on the pavement surface. Depths of this deflection basin
from the center of the falling weight are measured at different radial offsets. These deflections are used for the backcalculation of the pavement
layer moduli. Most of the available backcalculation software uses the layered elastic theory and static load to calculate moduli from known
pavement surface deflections. However, the FWD test load is dynamic, and layer materials may show nonelastic behavior. Layered elastic the-
ory in these types of software cannot characterize dynamic response of the pavement. Also, elastic theory is unable to accurately predict the
surface deflection whenever stress developed in any pavement layer exceeds the yield point. For this reason, this study has performed a fi-
nite-element analysis of the airport pavement under the FWD test considering the dynamic load and materials plasticity. The analysis presented
here includes elastoplastic behavior of pavement layer materials. Both axisymmetric and quarter cube models have been developed in ABAQUS.
Time-deflection histories are simulated to match the FWD test data. A comparison is made between the dynamic, static, and field deflection
basins. Contours of vertical deflection and strain are also plotted to observe their distribution on both the axisymmetric and quarter cube models.
Analysis results show that the time-deflection histories are in close agreement with the field data. The axisymmetric model yields better results
than the quarter cube model. Deflections from the static analysis are greater than the dynamic analysis for an identical set of the layer modulus of
elasticity. A uniform distribution of strain is observed from the static analysis in both of the geometries. However, the dynamic analysis does not
show similar distribution because of the time-dependent response. DOI: 10.1061/(ASCE)GM.1943-5622.0000305. © 2014 American Society
of Civil Engineers.
Author keywords: Falling weight deflectometer; Backcalculation; Modulus; Deflection; Stress; Strain; Finite-element method (FEM);
Flexible pavements; Nondestructive tests.
(Chen and Mizuno 1990). Therefore, elastodynamic analysis of the FEM Model
pavement under the FWD test may not produce accurate results. A
3D finite-element model was developed by Schwartz (2002) to Development of a finite-element model of multilayered flexible
observe the effect of stress dependency of the base layer on the pavement structure under the FWD test involves selection of model
pavement response. This model was simulated for the static geometry, assignment of layer (material) property, meshing of the
loading condition. In addition, analysis results from KENLAYER model, imposing the boundary condition, and dynamic load as-
(KENPAVE 2004) were used to validate the model instead of any signment similar to the FWD test. Each step of this finite-element
field data. Dong et al. (2001) developed dynamic finite-element analysis is subsequently described.
analysis to perform the time domain backcalculation to determine
the pavement layer properties. Backcalculated layer moduli from
this model showed close agreement with the other methods. In this Model Geometry
model, the FWD loading area was considered as a rectangle. In Clayton Airport pavement has mainly three different layers. These
reality, the FWD loading area is a circle with a radius of 300–450 mm layers are the surface, base, and subgrade. Table 1 shows that the
(12–18 in.). In addition, model validation was not shown by the samples are collected at 304.8 m (1,000 ft) from the east end of
comparison between the FEM simulated and field time-deflection Runway 12/30. The thickness of the surface, base, and subgrade are
history data. To this end, this study develops a finite-element model 5.08, 40.64, and 106.68 cm (2, 16, and 42 in.) according to drilling
of flexible pavement under the FWD test assuming dynamic load and information. The surface course consists of asphalt concrete, whereas
material plasticity. In particular, the load is considered dynamic with the base and subgrade consist of sand and silty clay, respectively. The
a very short duration, and the material of the base and the subgrade FWD test load is applied on the surface by a steel plate of 15.24-cm
are assumed to be elastoplastic. For validation, the dynamic response (6-in.) radius resting on the pavement surface. Sensors at different
of the pavement is compared with the field test data. In addition to the radial locations record vertical movement of the surface as shown in
dynamic analysis, an analysis is performed considering the load as Fig. 1. At a certain radius, the surface deflection is assumed to be the
static, and the results from the static analysis are compared with same in all radial directions. Therefore, it forms a deflection basin/
dynamic analysis results. influence zone.
For finite-element modeling of the multilayered pavement
structure, different geometries are used in different studies (Uddin
Objectives and Scope and Garza 2010; Xia 2010; Kim et al. 2009; Hjelmstad et al. 1997).
These model geometries can be two-dimensional (2D) plane strain,
The primary objective of this study is to evaluate the deflection basin axisymmetric, or 3D cube. The 2D plane strain model is used by
based on the dynamic characteristic of the load applied in the FWD several researchers to characterize the behavior of the pavement
test and the elastoplastic behavior of the pavement layer. For this (Van Metzinger and McCullough 1991; Lytton et al. 1993). It
reason, the finite-element model of the flexible pavement under the requires little computation time and memory storage for analysis;
FWD test has been developed in the commercial general purpose however, it can represent actual traffic loading, and the footprint of
finite-element program, ABAQUS (ABAQUS 2010). The specific the loading is typically elliptical. It can be represented by two
objectives of this study are as follows. semicircles and a rectangle. The plane strain model only uses the
• To determine the time-deflection history and peak deflection of line load (Cho et al. 1997).
the sensors from the dynamic analysis. These results will then be The axisymmetric model is in a geometry developed in two
compared with the field FWD test data to validate the model. dimensions and then is revolved around with reference to a vertical
• To determine the deflection basin from the static analysis and to axis to form a cylinder (Thompson 1982; Nam 1994). The advantage
compare this with the deflection basin from both the dynamic of this model is that it can solve the problem of a 3D structure with
analysis and field FWD test. 2D formulation using cylindrical coordinates. The main limitation of
of the plate-load diameter. The FWD test shows that deflection at the
Fig. 1. Zone of influence during the FWD test
loading point is the maximum. It approaches a very small value at the
sensor farthest from the loading point. Surface deflection, that is,
deflection at the loading point, is contributed by the sum of the
this model is that only the circular load can be applied on this model. deflection of all layers. As the depth increases, the load spreads away
The load assignment from the dual tire is complicated. Moreover, more that results decreases in stress. Therefore, vertical stress de-
this model is unable to consider interface shear; although, a special veloped in the subgrade is small. Deflection at the sensor farthest
case of all-round radial shear can be analyzed. It cannot address from the loading point is negligible because of this lower stress
pavement discontinuities or shoulder conditions. Therefore, this contributed from the subgrade (Huang 2004). Though previous
model can be used only for the region of pavement far from any literature has suggested the depth to be 4, 10, and 50 times the plate-
cracks or shoulder (Cho et al. 1997). load diameter, respectively (Logan 2007; Yamada 1970; Duncan
In this study, a 3D FEM model is used. A 3D FEM can address et al. 1968), the vertical length of the model is considered to be 508
different issues related to pavement, such as multiple wheel loading, cm (200 in.), which is 33.33 times the loading radius. During de-
nonlinear behavior of base materials, and distresses in pavements velopment of this model, geometries with varying vertical length
(Kuo and Chou 2004). It can precisely analyze the structural re- were considered. At the end, this dimension was found to be suf-
sponse of pavement-subgrade systems subjected to static and dy- ficient for negligible defection and stress at the bottom. Considering
namic loads for new pavements and pavement with joints, cracks, computational time and memory storage, the vertical length of the
and discontinuities (Uddin et al. 1995). The traditional 2D finite- entire pavement model is assigned to be 508 cm (200 in.). Therefore,
element model is not able to address these problems (Uddin 2008). the thickness of the subgrade is selected to 462.28 cm (182 in.),
However, it requires a long time and large storage capacity, espe- which is greater than 10 times the diameter of the loading plate.
cially with material nonlinearity (Hjelmstad et al. 1997; Cho et al.
1997). 3D Quarter Cube
This study performs finite-element analysis using both axisym- This model has two axes of symmetry: saving time and storage
metric and quarter cube geometries of flexible pavement to represent capacity. The 3D quarter cube model herein is shown in Fig. 3.
the actual field condition and loading configuration of the FWD test. Similar to the axisymmetric model, the thicknesses of the surface and
The details of the geometries are subsequently mentioned. base of this model are 5.08 and 40.64 cm (2 and 16 in.), respectively.
To maintain the similarity in dimensions, the length, width, and
Axisymmetric breadth of the 3D quarter cube model are kept at 508, 508, and 508
Fig. 2 shows the axisymmetric structure of the pavement under the cm (200, 200, and 200 in.), respectively.
FWD test. This model is symmetric in all the radial directions with
reference to the vertical axis. It has three layers (surface, base, Layer Property
subgrade) of finite thickness individually. The thicknesses of the Each model has three layers as previously mentioned. The surface
surface and base layers are 5.08 and 40.64 cm (2 and 16 in.), re- layer consists of asphalt concrete. The layer underneath the surface
spectively. The vertical and horizontal lengths of the model are course is the base course with compacted coarse-graded gravelly
selected in such a manner that the influence of the loading area soil. The bottom layer is the subgrade with natural soil. Material
becomes insignificant in the far-field region. According to the properties of the different layers are subsequently discussed.
studies by Yamada (1970), Koswara (1983), and Dunlop et al.
(1968), consideration of the vertical and horizontal lengths of the Surface Course
entire model, equal to 4–10 times the width of the loading area, The surface course is an asphalt concrete layer, which consists of
ensures that the effect of the load will be insignificant in the far-field viscoelastic material. The effect of viscoelasticity is pronounced at
region. In the dynamic finite-element model by Dong et al. (2001), a higher temperature and at load with slower rates. The FWD load is
the ratio of the vertical length and the FWD loading radius was 37.6. instantaneous; therefore, the strain in this test is recoverable. Strain
Duncan et al. (1968) considered the depth of analysis to be 50 times disappears just after the removal of the load (Haddad 1995).
Base Course
The base course consists of granular material, which is assumed
linear elastic up to the yield stress. It begins to show plasticity
whenever stress because of the applied load reaches the yield point.
Soils and geomaterials usually do not have a readily available yield
point. However, in this study, the yield point is defined by 482.63
kPa (70 psi) stress when the base materials show plasticity. The
triaxial test data are collected from the study performed by Garg and
Thompson (1997) to determine the yield stress for the assignment of
plasticity in the base layer. In that study, several triaxial tests have
been conducted with the rapid load rate, which is more appropriate
for addressing the rate of FWD loading.
The stress-strain plot of the granular material for the base course
is shown in Fig. 4. The yield point of this stress-strain distribution is
considered 482.63 kPa (70 psi) because part of the curve before this
yield stress can reasonably be assumed as linear elastic. Based on the Fig. 4. Stress-strain distribution of the granular soil in the base course
classification of the soil (particle distribution and soil index property (data from Garg and Thompson 1997)
test conducted in the laboratory) in Clayton Airport, the initial
axial test data have been assumed as accurate representation of zone, whereas the mesh is coarser at the far end of loading. A cubic
stress-strain distribution for the subgrade. Several researchers per- model with identical length, breadth, and height is used. This geom-
formed triaxial tests on different types of soil (El-Badawy and etry leads to uniform meshing of the model; however, it increases
Witczak 2007; Kim 2004). Subgrade soil in Clayton Airport has analysis time and needs more memory storage. Coarse mesh is
been identified as silty clay from the soil index property. For the avoided because of maintaining an aspect ratio between 1 and 2. The
assignment of subgrade stress-strain distribution, triaxial data were aspect ratio is important for accuracy (Logan 2007).
collected from the U.S. Army Corps of Engineers (2003). Fig. 5 The mesh for both the axisymmetric and quarter cube models was
shows the stress-strain distribution of the subgrade. The yield stress selected after a number of iterations with a varying number of ele-
for the subgrade materials is assumed to be 143.69 kPa (20.84 psi). It ments. The goal was to determine the optimum number of elements
would be appropriate if triaxial tests could have been performed on that can produce reliable analysis results within a shorter analysis
soils and aggregate, which requires laboratory capacity and other time. After a number of iterations with a varying number of ele-
resources. On the hand, the approach of this study is to perform soil ments, the optimum number of elements for the axisymmetric and
classification (e.g., grading, Atterberg limit) and use that information quarter cube models was 4,440 and 46,274, respectively.
to determine the material property from the available literature. Table 3 shows the parameters considered during the mesh as-
Material parameters for the layers (i.e., modulus of elasticity, signment on the pavement models. The number of nodes, element
Poisson’s ratio, and density) are summarized in Table 2. type, and number of elements for both of the geometries are mentioned
briefly. The table shows that the quarter cube model has more than 10
times the number of elements than the axisymmetric model.
Meshing of the Model
Axisymmetric Model Boundary Condition
The aspect ratio of the mesh elements is kept between 1 and 2 all Axisymmetric
through the model. The aspect ratio is defined as the ratio of the Interfaces at the two adjacent layers are fully bonded, that is, slip is
longest dimension to the shortest dimension (Logan 2007). The not allowed. Roller supports are assigned at the vertical left end to
axisymmetric model of the multilayered flexible pavement structure allow vertical movement and restraining horizontal movement as
is meshed finely near the loading region. The region closest to the shown in Fig. 2. The right end of the bottom block (subgrade) is
restrained to move in the horizontal direction, and rollers are
assigned as the support. The presence of rigid bedrock is assumed at
the bottom subgrade that leads to no deflection at the interface of the
rigid/stiff layer and subgrade. For this reason, the bottom of the
model is assigned with hinge supports to restrain the translational
movement in all directions.
Quarter Cube
The interface of the two adjacent layers is assumed to be fully
bonded, that is, the slip between the two adjacent layers is not
allowed. In response to the FWD test load, the deflection becomes
zero at some depth, and it indicates the presence of rigid layers at this
point. To address this phenomenon, hinge supports are assigned at
the bottom; therefore, they constrain the bottom from moving along
both the vertical and horizontal directions. Two intersected edges
with the loading point (along two axes of symmetry) are free to move
in the vertical direction and are restrained in the horizontal direction
Fig. 5. Stress-strain distribution of the subgrade soil from the triaxial from the adjacent material. In this model, these edges are assigned
test (data from U.S. Army Corps of Engineers 2003) with roller supports to ensure this behavior of the structure. There is
no movement in the horizontal direction for some distance after the
loading pattern is assigned according to the field test. A pressure of Dynamic Time-Deflection History
548.82 kPa (79.6 psi) is applied on the loading area; then, this
pressure is integrated with a field time-amplitude variation to match Time-deflection histories determined from the dynamic finite-
with the load-time history under a field test. element analysis have been plotted in Fig. 7(a). This plot shows
Static Load
In static analysis, the load is assumed constant with time. Only the
peak of the load-time history is used as a static load in this analysis.
The magnitude of the peak is 40.03 kN (9 kips), and it develops
a pressure of 548.82 kPa (79.6 psi) on the loading plate of a
15.24-cm (6-in.) radius.
Finite-Element Analysis
a closed loop. These loops are formed because of the time lag gradually diminishes radial outward. This is because of the higher stress
between the load and deflection. Every loop initiates from the origin near the loading region. Stress reduces both vertically and horizontally.
and terminates at the origin because of the elasticity. These loops are Consequently, vertical deflection decreases significantly.
getting thinner with an increase in the geophone distance because the
deflections are decreasing. All these observations are in very close Quarter Cube
agreement with the field test data. Therefore, this dynamic finite- Fig. 9(b) shows the contour for the vertical deflection distribution
element analysis is producing reasonable values of time-deflection over the quarter cube model. The top corner of this model is sub-
histories. Although the base and subgrade layers were assigned jected to the FWD impulse load. Therefore, this part of the model
elastoplastic materials models, stress because of a small FWD load undergoes maximum deflection. With an increase in distance along
did not exceed the limiting stresses or yield stress. Therefore, both of both the vertical and horizontal directions, the vertical deflection
the layers show only elastic behavior; essentially, they could have decreases. The color spectrum transition from black to gray indicates
been assigned only elastic materials (E, n) as ABAQUS inputs.
Fig. 7(c) is plotted to validate the analyzed time-deflection history
by the field test data. This figure shows the time-deflection history at
the geophone. The solid line represents the time-deflection history
from the finite-element simulation, whereas the dotted line repre-
sents the data collected from the field test. It is evident that the peak
deflections from both the finite-element simulation and the field test
are very close, that is, 0.875 mm (0.03445 in.) from the finite-element
simulation and 0.89 mm (0.035 in.) from the field test. Time lag
between these peak deflections is very small (i.e., ∼ 1:5 ms). These
two time-deflection histories show some noticeable deviation before
attaining their peak values. The deviation becomes insignificant just
after the peak deflections. The resulting time-deflection history from
the finite-element simulation is almost similar to that from the field
FWD test data.
Deflection Basin
Fig. 8 shows the deflection basins calculated from both the dynamic
and static finite-element simulations. These deflection basins have
also been compared with the field FWD deflection basin. Both of
these deflection basins are near to the field deflection basin. The
RMS of the error is then calculated to determine the nearness of
Fig. 8. Deflection basin from static and dynamic FEM analyses Fig. 9. Contour of vertical deflection
Acknowledgments
The authors thank the NMDOT, Aviation Division for funding this
study. Special thanks go to the NMDOT field exploration team for
their sincere effort behind the asphalt coring, soil sampling, and
FWD data collection from Runway 12/30 of Clayton Airport.
Fig. 10. Contour of true strain over the domain of the axisymmetric
model
References
that the vertical deflection is compressive (i.e., sink in) around the ABAQUS. (2010). ABAQUS 6-10 EF-2 documentation, Dassault Systèmes
loading zone. It gradually turns to extensive (i.e., pile up) farther Simulia, Providence, RI.
away from the loading zone. Ameri, M., Yavari, N., and Scullion, T. (2009). “Comparison of static and
dynamic backcalculation of flexible pavement layers moduli, using four
softwares.” Asian J. of Applied Sciences, 2(3), 197–210.
Contour of Strain Chen, W. F., and Mizuno, E. (1990). “Nonlinear analysis in soil mechanics:
Theory and implementation.” Developments in geotechnical engineering,
Axisymmetric Vol. 53, Elsevier, Amsterdam, Netherlands.
Vertical strain distribution over the axisymmetric model is shown in Cho, Y. H., McCullough, B. F., and Weissmann, J. (1997). “Considerations
Fig. 10(a). This contour shows that the magnitude of the vertical on finite element method application in pavement structural analysis.”
strain is higher near the loading region and is compressive in nature. Transportation Research Record 1539, Transportation Research Board,
Compressive strain decreases gradually farther away from the Washington, DC, 96–101.
loading zone. Vertical tensile strain is observed on the surface layer Dong, Q., Matsui, K., and Yamamoto, K. (2001). “Time domain back-
at a distance a little farther from the loading zone. This is because of calculation of pavement structure material properties using 3D FEM with
the pile effect resulting during FWD load application. Ritz vectors.” Int. J. Geomech., 10.1061/(ASCE)1532-3641(2001)1:3
(325), 325–336.
Duncan, J. M., Monismith, C. L., and Wilson, E. L. (1968). “Finite element
Quarter Cube analysis of pavements.” Highway Research Record No. 228, Highway
Fig. 10(b) shows the contour of the vertical strain distribution over the Research Board, Washington, DC.
Dunlop, P., Duncan, J. M., and Seed, H. B. (1968). “Finite element
quarter cube model. The top corner region of this model undergoes
analysis of slopes in soil.” Rep. TE 68-3, Univ. of California, Berkeley,
compressive strain because of the application of the load directly on CA.
this region. With an increase in distance, the compressive strain El-Badawy, S. M., and Witczak, M. W. (2007). “Development of a universal
reduces and turns into tensile strain. It indicates that the part of the permanent strain model for the subgrade pavement materials.” Proc.,
model near to the loading radius shows the sink in effect, whereas the Transportation Research Board 86th Annual Meeting (CD-ROM),
rest of model shows the pile up effect. Transportation Research Board, Washington, DC.
Upper Saddle River, NJ. Thompson, M. R. (1982). ILLI-PAVE, user’s manual, Transportation Re-
Irwin, L. H. (2002). “Backcalculation: An overview and perspective.” search Facilities, Dept. of Civil Engineering, Univ. of Illinois, Urbana,
Æhttps://docs.google.com/file/d/0BwRtUOIxzlF4eGVXRzFFZVY0QWM/ Champaign, IL.
edit?pli=1æ (Nov. 29, 2009). Uddin, W. (2008). “3D-FE dynamic analysis of pavements subjected to
KENPAVE. (2004). “KENPAVE: A computer package for pavement nondestructive impact testing.” NED University Journal of Research,
analysis and design.” Univ. of Kentucky, Lexington, KY. 5(1), 1–15.
Kim, D. G. (2004). “Development of a constitutive model for resilient Uddin, W., and Garza, S. (2010). “3D-FE simulation study of structural
modulus of cohesive soils.” Ph.D. thesis, Dept. of Civil Engineering, response analysis for pavement-subgrade systems subjected to dynamic
Ohio State Univ., Columbus, OH. loads.” Proc., Pavements and materials: Testing and modeling in mul-
Kim, M., Tutumluer, E., and Kwon, J. (2009). “Nonlinear pavement tiple length scales, ASCE, Reston, VA, 170–181.
foundation modeling for three-dimensional finite-element analysis of Uddin, W., Hackett, R. M., Joseph, A. P., and Algred, B. (1995). “Three-
flexible pavements.” Int. J. Geomech., 10.1061/(ASCE)1532-3641 dimensional finite-element analysis of jointed concrete pavement with
(2009)9:5(195), 195–208. discontinuities.” Transportation Research Record 1482, Transportation
Koswara, H. (1983). “A finite element analysis of underground shelter Research Board, Washington, DC, 26–32.
subjected to ground shock load.” M.S. thesis, Rose-Hulman Institute of Uddin, W., Hackette, R. M., Noppakunwijai, P., and Pan, Z. (1996). “Three-
Technology, Terre Haute, IN. dimensional finite element simulation of FWD loading on pavement
Kuo, C. M., and Chou, F. J. (2004). “Development of 3-D finite element model systems.” Meeting the Challenge: Proc., 24th Int. Air Transportation
for flexible pavements.” J. Chin. Inst. Chem. Eng,., 27(5), 707–717. Conf., ASCE, Reston, VA, 284–294.
Logan, D. L. (2007). A first course in the finite element method, 4th Ed., U.S. Army Corps of Engineers. (2003). “Engineering and design—Slope
Thomson, Toronto. stability.” Rep. EM 1110-2-1902, Washington, DC.
Lukanen, E. O. (1993). “Effects of buffer on falling weight deflectometer Van Metzinger, W. A., and McCullough, B. F. (1991). “An empirical-
loadings and deflections.” Transportation Research Record 1355, mechanistic design method using bonded concrete overlays for the re-
Transportation Research Board, Washington, DC, 37–51. habilitation of pavements.” Rep. No. 1205-1, Center for Transportation
Lytton, R. L., et al. (1993). “Development and validation of performance Research, Univ. of Texas at Austin, Austin, TX.
prediction models and specifications for asphalt binders and paving mixes.” Xia, K. (2010). “Finite element modeling of dynamic tire/pavement in-
Rep. SHRP A-357, Transportation Research Board, Washington, DC. teraction.” Proc., Pavements and materials: Testing and modeling in mul-
Meier, R. W., and Rix, G. J. (1998). “Backcalculation of flexible pavement tiple length scales, ASCE, Reston, VA, 204–214.
moduli from falling weight deflectometer data using artificial neural Yamada, Y. (1970). “Dynamic analysis of civil engineering structures.”
networks.” Chapter 7, ASCE manuals and reports on engineering Recent advances in matrix methods of structural analysis and design,
practice, ASCE, Reston, VA, 162–190. University of Alabama Press, Tuscaloosa, AL, 487–512.
Nam, D. (1994). “Effect on slab flexibility on the vertical stiffness of circular Yin, D., and Mrawira, D. (2009). “Comparison between laboratory in-
foundations.” M.S. thesis, Univ. of Texas at Austin, Austin, TX. vestigation and non-destructive testing methods for mechanistic char-
National Cooperative Highway Research Program (NCHRP). (2008). acterization of asphalt pavement.” Proc., 88th Annual Meeting of
Mechanistic-empirical pavement design guide, Interim Ed., Trans- Transportation Research Board, Transportation Research Board,
portation Research Board, Washington, DC. Washington, DC.