Professional Documents
Culture Documents
Finite Element Studies of Asphalt Concrete Pavement Reinforced With Geogrid
Finite Element Studies of Asphalt Concrete Pavement Reinforced With Geogrid
Abstract: Many geotechnical applications are becoming more sophisticated and solutions derived from simplistic procedure are no
Downloaded from ascelibrary.org by SUN YAT-SEN UNIVERSITY on 03/19/14. Copyright ASCE. For personal use only; all rights reserved.
longer reasonable or solutions do not exist. This paper describes two-dimensional finite element studies that analyzed the behavior of
reinforced asphalt pavement under plane strain conditions and subject to monotonic loading. The asphalt material and soils were expressed
using triangular elements of elastoplastic behavior that obeys Mohr–Coulomb criteria with associated and nonassociated flow rules. The
geogrid was modeled using a one-dimensional linear elastic bar element. The finite element procedure was validated by comparing the
results of analysis with the results obtained from a series of model tests. The load–settlement relationships, settlement profile, and strains
in the geogrid were compared. The failure load obtained by assuming subgrade foundation with nonassociated flow rule was smaller than
that of associated flow rule. There was only minor difference between the results obtained from the associated and nonassociated plastic
models. The finite element procedure was capable of determining most measured quantities satisfactorily except the tensile strain in the
geogrid, which was assumed linear elastic. The effects of the stiffness of geogrid reinforcement, thickness of asphalt layer, and strength
of subgrade foundation were also investigated. The finite element procedure is a versatile tool for enhanced design of reinforced pavement
systems.
DOI: 10.1061/共ASCE兲0733-9399共2003兲129:7共801兲
CE Database subject headings: Finite elements; Asphalt pavements; Geogrids; Bearing capacity; Settlement; Plasticity.
Introduction sons between the theory and field test results, and the theory now
forms a standard methodology for the design and analysis of
Pavements are classified as rigid or flexible depending on the pavement systems 共e.g., Huang 1993兲. An overview of the com-
surfacing layer, which can be asphalt or portland concrete. The puter codes that use elastic layer theory was given by Haas and
surface layer of asphalt or portland concrete helps to redistribute Rodway 共2001兲.
the wheel load to the base and subgrade. After the Second World Burmister’s theory was formulated under idealized conditions,
War, many highways were designed based on the California bear- especially with regard to the elastic material properties. It is well
ing ratio 共CBR兲, which is a semiempirical procedure. The settle-
known that soil and asphalt materials are inelastic such that the
ment could at best be estimated using Boussinesq’s solution,
approximation can be oversimplified. The theory under the semi-
which is valid for an elastic and homogeneous medium with an
infinite elastic half space is applicable to the simple boundary and
infinite width and depth. The theory is oversimplified for a
loading conditions that preclude complicated loading conditions
multilayer system. In the 1940s, theoretical developments had
of multi-wheel loads of trucks and aircrafts. Thus, the numerical
enabled elastic solution to be obtained for layered system such as
pavements 共Burmister 1943, 1945兲. The theory assumes continu- procedures, such as finite element methods, were attempted for
ity of vertical and shear stresses as well as displacements at the the pavement systems. Duncan et al. 共1968兲 was among the first
interfaces of two- and three-layer systems. The results were com- to apply finite element method to the pavement analysis. The
piled as design chart so that the influence values are obtained for material was modeled as nonlinear elastic that depended on the
the deflection and vertical stress. The solution indicated that an stress levels. Nonlinear properties of the base, subbase, and sub-
increase in the stiffness of the top layer may be interpreted as grade were obtained from the triaxial compression tests. The re-
equivalent to a reduction in the required thickness of the layer in sults of analysis compared favorably well with the measured field
design. Burmister 共1958, 1962兲 demonstrated successful compari- deflection. Note that the methodology proposed by Duncan et al.
共1968兲 is for prefailure analysis and was not suitable to detect
1
Associate Professor, Dept. of Civil Eng. and Eng. Mechanics, failure.
Columbia Univ., 500 West 120th Street, New York, NY 10027. E-mail: A failure analysis, such as bearing capacity for shallow foun-
Ling@civil.columbia.edu dation, may be captured by using elastoplastic models 共e.g., Grif-
2
Graduate Research Assistant, Dept. of Civil Engineering and fiths 1982; de Borst and Vermeer 1984兲. In the analysis by Grif-
Engineering Mechanics, Columbia Univ., New York, NY 10027. fiths 共1982兲, Mohr–Coulomb failure criteria was used with a
Note. Associate Editor: Victor N. Kaliakin. Discussion open until De- nonassociated flow rule. An initial strain solution technique using
cember 1, 2003. Separate discussions must be submitted for individual viscoplastic algorithm was used with iteration until convergence.
papers. To extend the closing date by one month, a written request must
The failure load was determined when the solution failed to con-
be filed with the ASCE Managing Editor. The manuscript for this paper
was submitted for review and possible publication on December 11, verge. The analysis used eight-node isoparametric elements with
2001; approved on December 4, 2002. This paper is part of the Journal reduced integration. Collapse load was well captured, but the con-
of Engineering Mechanics, Vol. 129, No. 7, July 1, 2003. ©ASCE, ISSN vergence was affected by the increase in the angle of internal
0733-9399/2003/7-801– 811/$18.00. friction, with a limit of ⫽35%. de Borst and Vermeer 共1984兲
Fig. 1. 共a兲 Laboratory loading test setup 共after Ling and Liu 2001兲 Zheng 共1993兲 employed elastoplastic models for the asphalt ma-
and 共b兲 finite element mesh terial and soils. The model developed by Weissman and Sousa
共1993兲 required a number of tests in obtaining the parameters for
asphalt materials. Wright and Zheng 共1993兲 indicated that the
used 15-node triangular elements, with full integration, and dem- plastic properties of asphalt materials are significant in determin-
onstrated the solution for associated and nonassociated flow rules ing the permanent deformation of the flexible pavements.
up to ⫽40°. In the U.S., the Federal Aviation Administration 共FAA兲 has
The use of finite element method for pavement analysis has concentrated on the three-dimensional 共3D兲 finite element analy-
gained increasing attention only in recent years. In the Science sis of airport pavement using NIKE3D 共Maker 1995兲. Full-scale
Program of the European Community–DG XII, Jouve and Gu- testings were also conducted so that the results of analysis and
ezouli 共1993a兲 compared several finite element programs for the measurements can be compared. The study was conducted by
analysis of pavements. The programs that were compared are assuming the materials to be linear elastic 共Brill 1998兲. It is ex-
FENLAP of the University of Nottingham, DIANA of the Delft pected that future design of airport pavement will be conducted
University of Technology, NOEL of the University of Nantes, and using linear 3D finite element procedures. However, little empha-
CESAR of the Laboratoire Central des Ponts et Chaussees sis has been directed toward the nonlinearity of material proper-
共LCPC兲. Jouve and Guezouli 共1993b兲 also compared the effects of ties. It may be argued that the load levels in the pavement are very
using different soil models, including linear to complicated non- low, typically, in the elastic range, but failure of pavement due to
linear elastic models, and elastoplastic model. Jouve 共1993兲 com- overstressing are not uncommon. While three-dimensional finite
pared the test results of LCPC with the finite element analysis element analysis is costly, its success has been demonstrated by
conducted using four computer programs. In the analysis, the soil several studies, such as Hjelmstad et al. 共1997兲, Kuo et al. 共1995兲,
foundation was assumed linear elastic, whereas the granular ma- Uddin 共1998兲, and Zaghloul and White 共1993兲.
terial was assumed to be nonlinear, predominantly nonlinear elas- Geosynthetic material 共Koerner 1998兲 has been used to tensile
tic. The asphalt layer was assumed to be linear elastic. Chen et al. reinforce earth structures and unpaved roads 共e.g., Giroud and
共1995兲 assessed several computer programs for analysis of flex- Noiray 1981; Holtz and Sivakugan 1987; Love et al. 1987兲. Geo-
ible pavement system. Additional finite element simulations of synthetic has been used to inhibit reflective cracking by acting as
pavement systems were reported, for example, by Almedia an interlayer between the old pavement and the overlay 共e.g.,
共1993兲, Balay and Kabre 共1993兲, Cho et al. 共1996兲, Guezouli Rigo et al. 1993; Roschen 1997兲. Many highway engineers have
et al. 共1993兲, Schelt et al. 共1993兲, and Uddin 共1998兲. The analyses not realized possible tensile contributions of geosynthetic in in-
were incapable of capturing the failure and it was found that creasing the bearing capacity. The study has been initiated by
assumption of a linear elastic property for the asphalt materials Brown et al. 共1985兲 and Haas 共1985兲 to look at the possibility of
was oversimplified. Weissman and Sousa 共1993兲 and Wright and reinforcing asphalt concrete layer by comparing the behavior of
Fig. 3. Settlement profile of asphalt concrete surface: 共a兲 Associated Fig. 4. Normalized load–settlement relationships of geogrid-
flow rule and 共b兲 nonassociated flow rule reinforced pavement: 共a兲 Associated flow rule and 共b兲 nonassociated
flow rule
Fig. 5. Settlement profile of reinforced asphalt concrete layer with Fig. 6. Settlement profile of reinforced asphalt concrete layer with
geogrid stiffness J⫽1160 kN/m: 共a兲 associated flow rule and 共b兲 non- geogrid stiffness J⫽400 kN/m: 共a兲 associated flow rule and 共b兲 non-
associated flow rule associated flow rule
Fig. 7. Strain distribution of geogrid in asphalt concrete layer with Fig. 8. Strain distribution of geogrid in asphalt concrete layer with
geogrid stiffness J⫽1,160 kN/m: 共a兲 associated flow rule and 共b兲 geogrid stiffness J⫽400 kN/m: 共a兲 associated flow rule and 共b兲 non-
nonassociated flow rule associated flow rule
Fig. 11. Effect of geogrid thickness on normalized load–settlement Fig. 12. Effect of asphalt concrete stiffness on normalized load–
relationships: 共a兲 associated flow rule and 共b兲 nonassociated flow rule settlement relationships: 共a兲 associated flow rule and 共b兲 nonassoci-
ated flow rule
Fig. 13. Effect of strength of subgrade foundation on normalized Fig. 14. Effect of strength of subgrade foundation on normalized
load–settlement relationships with associated flow rule: 共a兲 re- load–settlement relationships with nonassociated flow rule: 共a兲 re-
inforced pavement and 共b兲 unreinforced pavement inforced pavement and 共b兲 unreinforced pavement
gave a better prediction of strain distribution. The same trend of ness is 4 cm. Note that for thickness greater than 4 cm, the bear-
results is seen for the strain versus displacement relationships ing capacity of the pavement system was determined by the as-
throughout the loading test for the associated and nonassociated phalt layer. Yielding occurred in the asphalt concrete before the
flow rules 共Figs. 9 and 10兲. sand foundation.
The results of simulation for the reinforced pavement seem Fig. 12 shows the effects of geogrid stiffness on the response
satisfactory for both associated and nonassociated flow rules, but of the pavement. It is seen that the bearing capacity of the pave-
the nonassociated case seems produced slightly better agreement. ment improved as the stiffness of geogrid is increased. There is
The strains predicted in the geogrid were less satisfactory when also an upper limit of stiffness that may render improvement in
the stress level increased. The results were due to the limitation of the bearing capacity. The stiffness affected displacement up to
a linear elastic model in expressing the nonlinear behavior of about 20% the footing width. For presented parametric studies,
geogrid. J⫽1,000 kN/m gave practically the same results as that of J
⫽400 kN/m.
Figs. 13 and 14 show the normalized load and settlement re-
Parametric Studies lationships of reinforced and unreinforced asphalt pavement as-
suming associated and nonassociated flow rules for the subgrade
The analytical procedures had been validated with well-controlled foundation, respectively. The angle of internal friction of sub-
tests of sand foundation, unreinforced and reinforced pavements. grade foundation was investigated for the range of values ⫽25–
Therefore, the effects of several design parameters, such as the 45° for associated flow rule and ⫽33– 45° for nonassociated
thickness of asphalt layer, stiffness of geogrid, and strength of flow rule. Note that Griffiths 共1982兲 and de Borst and Vermeer
foundation were investigated. The input parameters and geom- 共1984兲 had indicated that the Young’s modulus does not affect the
etries used for reinforced pavement, as presented earlier, were failure load. In using nonassociated flow rule, the angle of dila-
used. tion was decided following Eq. 共2兲. The analyses resembled the
Fig. 11 shows the effects of the thickness of asphalt layer on cases where the foundation was prepared with the same sand at
the load–displacement relationship where the stiffness of geogrid different densities, and thus different friction angles are obtained.
was selected as 400 kN/m. The analysis indicated that as the The results of analysis show that the unreinforced pavement gave
thickness of asphalt layer increases, the bearing capacity of as- lower failure load than the reinforced case. The failure load ob-
phalt pavement is also increased. However, there exists an upper tained from nonassociated flow rule was smaller than correspond-
limit related to the thickness of asphalt layer, upon which no ing results of associated flow rule. The results from nonassociated
improvement in bearing capacity was observed. The ideal thick- flow rule were less stable than those of associated flow rule.
While the failure load was attained in 20% the displacement ratio
or less in unreinforced pavement, the reinforced pavement
reached much larger displacement up to 60% before failure load of sand foundation, unreinforced and reinforced pavement sys-
was reached. Figs. 15 and 16 show the ratio of strength mobilized tems. The procedures were able to simulate the improved per-
for the reinforced and unreinforced pavements at a displacement formance of geogrid-reinforced asphalt pavement over unrein-
ratio of 5%, 10%, and 20% of footing width. It is natural to forced pavement system.
anticipate that as the strength of subgrade foundation increases, • The difference in results of using nonassociated and associated
the reinforcement effect reduces. The reinforcement effect in- flow rules were small for the pavement systems. The settle-
creases with increasing displacement, that is, when the strain in ment profiles were well predicted. The failure load obtained
the reinforcement was mobilized. The ratio of ultimate load was from nonassociated flow rules was slightly smaller than that of
in the range of 1.9–2.5 for the values of soil strength investigated. associated flow rule. The results showed that nonassociated
It is of interest to note that in the associated flow, when the angle flow rule is more relevant for the simple elastoplastic analysis
of internal friction exceeded 40°, the reinforcement ratio started of asphalt pavements that are reinforced with geogrid.
to reduce. The result was due to the failure of the asphalt concrete • An increased thickness of asphaltic concrete layer improved
layer instead of the subgrade foundation. The strength ratio was the performance of asphalt pavement system. However, when
greater for nonassociated flow and the ratio also increased with an upper limit of thickness is attained, the failure occurred in
the increase in the strength of subgrade foundation. the asphalt layer that determined the failure load of the pave-
ment system.
• The stiffness of geogrid affected the load–settlement relation-
Summary and Conclusions ships. There is an upper limit of stiffness that led to the im-
provement. However, different values of stiffness gave rise to
This paper outlined a two-dimensional finite element procedure different strain distributions in the geogrid. A linear elastic
for analyzing geogrid-reinforced pavement system. The proce- model of geogrid predicted poorly the strain distributions in
dures were validated with several well-controlled laboratory load- the geogrid layer. Thus, PLAXIS should be incorporated with
ing tests and then extended for a series of parametric studies. The a nonlinear geogrid model for future analysis of reinforced
procedure proved very useful for designing unconventional geo- pavement system.
technical system where solutions do not exist. The following con- • The reinforcement effect was more pronounced for weaker
clusions were drawn from this study: subgrade foundation. The reinforcement effect increased as the
• The finite element procedures employed were able to predict strain level was increased. These effects were related to the
with reasonable accuracy the load–displacement relationships tensile strain mobilized in the geogrid reinforcement.
grid, mastic, and membrane interlayers on asphalt concrete overlays.’’ law for asphalt aggregate mixtures.’’ Flexible pavements, Balkema,
Proc., Geosynth. 97, Industrial Fabrics Association International, Rotterdam, The Netherlands, 207–214.
725–747. Wright, P. J., and Zheng, L. 共1993兲. ‘‘Calculation of the permanent de-
Schelt, W., Vos, E., Galjaard, P. J., and Hendriks, M. A. N. 共1993兲. ‘‘The formation in a flexible pavement.’’ Flexible pavements, Balkema, Rot-
nonlinear elastic model of Boyce: An evaluation of the model in the terdam, The Netherlands, 215–221.
FEM program DIANA.’’ Flexible pavements, Balkema, Rotterdam, Zaghloul, S., and White, T., 共1993兲. ‘‘Use of a three-dimensional, dy-
The Netherlands, 165–171. namic finite element program for analysis of flexible pavement.’’
Tanaka, T., and Kawamoto, O. 共1989兲. ‘‘Plastic collapse analysis of strain Transp. Res. Rec. No. 1388, National Research Council, Washington,
softening materials.’’ Numerical models in geomechanics, S. Pi- D.C., 60– 69.