Page 230 - MODELING OF ASPHALT CONCRETE

208 Cha pte r Ei g h t

model such as the layered elastic theory and FE procedure that includes elastic, nonlinear
elastic [e.g., resilient modulus (RM) model] or elastoplastic models. The last may involve
such classical plasticity models as von Mises, Mohr-Coulomb, and hardening or
continuous yielding (Desai 2001; Schofield and Wroth 1968; Vermeer 1982). The stresses
and strains, usually under total or incremental application of the wheel load, are
computed and used in empirical formulas for calculation of rutting, damage, cracking
under mechanical and temperature load, and cycles to failure. Very often, uniaxial
quantities such as the tensile strain, e , at the bottom of the asphalt layer, vertical
t
compressive strain e at the top of the subgrade layer, vertical stress s under the wheel
c y
load, and tensile stress s at the bottom of the asphalt layer, are used to compute various
t
distresses with the empirical formulas (Huang 1993).
The M-E approach can lead to improved design compared to the empirical approach.
However, it does not allow for realistic material behavior as affected by elastic, plastic,
and creep responses under mechanical and environmental loading. Furthermore,
evaluation of distresses based on the uniaxial quantities in empirical formulas may not
provide accurate predictions of the distresses as affected by the multidimensional
geometry, nonhomogeneities, anisotropy, and nonlinear material response, which is
dependent on stress, strain, time, and load repetitions.
The full mechanistic (M) approach allows for geometry, nonhomogeneities,
anisotropy, and nonlinear material properties of all layers in a unified manner. As a
result, the distresses are evaluated as a part of the solution (e.g., finite element)
procedure, without the need of the empirical formulas.
The AASHTO Design Guides (e.g., 1986, 1993) are often used for pavement design.
The most recent Design Guide (NCHRP 2004) includes the M-E approach. The Strategic
Highway Research Program (SHRP) (Lytton et al. 1993) and ongoing Superpave research
attempt to develop the general and unified material models. However, such unified
models are very often based on combinations of models for specific material properties
such as linear elastic creep, viscoplastic creep, damage and fracture (Kim et al. 1997;
Rowe and Brown 1997; Schapery 1965, 1990, 1999; Secor and Monismith 1962). Although
such models have been used commonly in the pavement engineering area, often ad hoc
combination may not be suitable for the realistic behavior of materials in which the
elastic, plastic, creep, damage, fracture, and healing occur simultaneously under the
applied loading. The combination of models suffers from some limitations: The component
models may not be consistently integrated, they can be relatively complex, and the
material parameters involved can be large. Some of the parameters do not have physical
meanings, they may not relate to the specific states during deformation, and hence, they
need to be determined by using mainly curve fitting and least square procedures.
On the other hand, unified and concise models that can overcome many of the above
limitations are available in other engineering areas such as mechanics, geomechanics,
and mechanical engineering. In the early 1980s, the author conducted a research project
(Desai et al. 1983) where constitutive models available at that time were developed and
implemented in mechanistic 2-D and 3-D finite element procedures for track support
and pavement systems. The models were calibrated by using comprehensive material
tests and the computer codes were verified with respect to field observations. Indeed,
there exists a need for advanced and unified mechanistic models. One such model based
on the hierarchical single surface (HISS) plasticity model by Desai and coworkers (Desai
et al. 1986; Desai et al. 1993; Desai 2001) has been also used for pavement analysis by
Scarpas et al. (1997). The HISS model has been used successfully for unbound materials