Road Materials and Pavement Design, 2015

Vol. 16, No. 2, 392–404, http://dx.doi.org/10.1080/14680629.2015.1013053

Finite element modelling for prediction of permanent strains in

fine-grained subgrade soils
G.Y. Yesuf ∗ and I. Hoff

Department of Civil and Transport Engineering, Norwegian University of Science and Technology,
Høgskoleringen 7A, 7491, Trondheim, Norway

(Received 14 February 2014; accepted 24 January 2015 )

Roads are designed with good reliability to fulfil the long-term performance during the design
period. In order to achieve an effective design, the long-term performance of pavements should
be optimized during the design phase. One of the strategies is to develop models which are
capable of predicting the long-term performance. This paper focuses on one of the most critical
distress modes in flexible pavements, namely, rutting. The plastic strain of subgrade soils
is modelled to quantify the amount of rut contributed from the subgrade. The deformation
from the first loading cycle is incorporated in the elastoplastic theoretical framework based on
the Drucker–Prager yield criterion. The proximity of deviator stress to the static failure limit
of subgrade soils is considered to predict the amount of incremental plastic strains at each
loading cycle. The model is implemented in the User Material Subroutine in the finite element
program ABAQUS™. The prediction of plastic strains using the proposed model provides a
good agreement with laboratory test with deviator stress level up to 50% of the static strength
of the soil. The mobilisation of subgrade soils in pavements is normally low due to reduced
stress magnitude at the subgrade level, which makes the proposed model in this study more
appealing to understand the development of plastic strains in subgrade soils.
Keywords: elastoplasticity; finite element method; pavement; rutting; subgrade

1. Introduction
Pavement design is carried out in consideration with the success of short-term and long-term
performances. The long-term performance of flexible pavements is evaluated based on threshold
values of road performance indicators and pavement distresses. Common modes of distresses
in pavements associated with the long-term performance are accumulated permanent deforma-
tion (rutting), surface cracking and roughness. Rutting is the term used to describe the surface
depression in the wheel-path. It is considered as one of the most critical parameters that strongly
affect the pavement performance. Other distress modes, such as cracking and roughness, are
often associated with rutting. Rutting is also the most common cause of resurfacing in road
maintenance.
Some of the consequences of rutting are high risk of skidding when ice or water covers the sur-
face, increasing fuel consumption and reducing the drainage performance of the asphalt surface.
In cold climate regions, the difficulty to remove the snow or ice from ruts results in an increased
cost for winter maintenance. In some cases, waterway collect in a buried rut in the subgrade
and results in reduced load-bearing capacity of the granular layers (Dawson & Kolisoja, 2006).

*Corresponding author. Email: girum.yimer.yesuf@vegvesen.no

© 2015 Taylor & Francis

 Road Materials and Pavement Design 393

Field observations showed that once the permanent deformation exceeds a certain limit, there
is an increasing probability of cracking in the wheel track in the asphalt layer. It was reported
in Croney and Croney (1997) that heavy traffic loads exceeding the design limit would lead to
cracking as a result of large elastic deformations. Water entering the cracks is then likely to accel-
erate failure, leading to break up of the pavement surface and give rise to potholing. Then after,
weakening of subgrade soils leads to appreciable permanent deformation.
The accumulated knowledge of permanent deformation in pavements follows the hierarchy
of pavement layers from the top to the bottom. Particularly for subgrade soils, the emphasis
has been very low due to the notion that the magnitude of stress is reduced at the subgrade
level. However, the variation of ground condition is immense and quality of subgrade soils can
neither be fully investigated nor be ensured during design and construction of roads, unlike the
overlying pavement layers. Besides, the strength of subgrade soils is relatively low that even a
small magnitude of stress can cause plastic deformation. In Pavement Management System for
road maintenance planning, the amount of rut depth is measured on the surface of the asphalt
layer. Although the contribution of rutting in the subgrade to the total rut is small, the explicit
quantification is important to outline different maintenance strategies.
An adverse effect of permanent deformation at the subgrade level is often observed in low-
traffic volume roads. These types of roads are designed with a thin asphalt layer and the
anticipated stress at the subgrade level is high. In addition, the expected performance of pavement
layers in low-traffic volume roads may not be obtained due to drainage problems and limited
maintenance budget. In cold climate regions, inadequate frost protection layers lead to frozen
subgrade in the winter period and the soil becomes saturated following melting of the segre-
gated ice-layers during the thawing period. In such conditions, plastic shear strains are induced
upon high axle loads. Moreover, the permanent deformation from re-consolidation of saturated
subgrade soils aggravates cracking and roughness at the pavement surface. Field observations
have shown that subgrade soils in low-traffic volume roads are highly prone to structural and
functional failures.
The accumulation of plastic strains in pavement layers cannot be avoided, but it can be sub-
stantially reduced if the pavement layers have good resistance against rutting. This is, however,
not the best method since the selection of geo-materials for pavement layers must consider both
economical and practical aspects. In practice, when the rut depth at the surface of the pavement
reaches to a maximum allowable limit, a road maintenance task is carried out. An example of the
Norwegian practice (NPRA, 2003) is that a maximum deformation of 25 mm in the wheel tracks
at road section measurements is defined as a failure condition. The measured rut depth on the
surface of the pavement is a result of the wear of the road surface and the cumulative permanent
deformation in each layer. Hence, the long-term deformation properties of each pavement layer
must be quantified explicitly for efficient design and maintenance of roads.
In mechanistic pavement design method, the vertical elastic strain on the top of the subgrade
is used as a limiting criterion against the excessive deformation in the pavement layers. This con-
cept is explained by Huang (2004) that the plastic strains are proportional to the elastic strains
in road materials and by limiting the elastic strains on the subgrade, the elastic strains in other
components above the subgrade will also be controlled. Consequently, the magnitude of perma-
nent deformation on the pavement surface will be controlled. The other procedure to limit rutting
in flexible pavements is to limit the total accumulated permanent deformation on the pavement
surface based on the permanent deformation properties of each individual layers. The permanent
deformation from one load repetition is very small. However, the gradual accumulation of plas-
tic strains for a large number of load repetitions could result in excessive rutting, a consequent
failure of the pavement structure.
394 G.Y. Yesuf and I. Hoff

Uzan (2004) suggested two approaches to incorporate rut prediction models in pavements. The
first approach is computing the permanent strains under the traffic load at different locations and
summing up the contribution of all layers. This approach is used when the stress state dominates
the permanent deformation. It is a typical phenomenon for soils where the deformation is signifi-
cant in the first loading cycles and stabilizes afterwards. In the second approach, the incremental
rutting at each load application is computed and is integrated over the design life. The latter one
incorporates the effect of the stress state as well as the number of load repetitions. This has been
observed in laboratory tests of fine-grained soils (Brown, Lashine, & Hyde, 1975; Chai & Miura,
2002; Li & Selig, 1996).

2. Mechanism of permanent deformation

There are different mechanisms that contribute for the deformation of unbound layers in flex-
ible pavements. Most research developments are mainly for coarse aggregates based on the
notion that these materials are prone to high stress from the traffic. In low-traffic volume roads,
high stress is expected on both fine-grained and coarse-grained subgrade soils which can cause
appreciable permanent deformation. However, the mechanisms are not entirely the same for
coarse-grained and fine-grained soils. The deformation mechanisms of coarse-grained soils are
governed by rearrangement and/or crushing of particles, and shear deformations. The extent
of deformation is affected by the type of compaction, grain size and distribution, grain shape,
moisture content and amount of fines (Hoff, 1999; Lekarp & Dawson, 1998).
In fine-grained soils such as clay and silt, consolidation and shear deformation are dominant
mode of deformation and their behaviour is affected by the stress history, moisture content and
extent of compaction (Brown et al., 1975; Elliott, Dennis, & Qiu, 1999; Horvli, 1979; Pup-
pala, Saride, & Chomtid, 2009). The accumulation of plastic strains on both fine-grained and
coarse-grained soils is affected by load-related factors such as the stress level, confining pressure,
number of load applications and rotation of principal stresses. The major primary factors that
affect the deformation properties of fine-grained soils are moisture content, magnitude of devia-
tor stress and freezing–thawing cycle. Other factors such as density, duration and frequency of
the deviator stress have less effect on permanent deformation as compared to the primary factors.
The accumulation of plastic strain is highly dependent on the deviator stress level, almost for
all types of subgrade soils. For the case of pavements, the deviator stress is the magnitude of
additional stress from the traffic. For fine-grained soils, there exists a critical level of repeated
deviator stress above which the soil plastic deformation increases rapidly as the number of load
repetition increases. This critical level (dynamic strength of soil) is usually smaller than the soil
static strength determined under monotonic loading. For stresses below the critical level, a non-
failure equilibrium state is reached after many cycles of repeated loading. The critical stress level
generally varies depending on the soil type. For instance, it was estimated to be 50% of the static
shear strength for Ottawa area Champlain Sea clay soil (Mitchell & King, 1976), and 83% for
Mexico City clay soils (Diaz-Rodriguez, 1989).
The level of confining pressure also controls the accumulated plastic strains though its effect
is lesser compared to deviator stress. For saturated clay soils, Brown et al. (1975) showed that
for a constant deviator stress, the accumulation of plastic strains strongly depends on the Over-
Consolidation Ratio and the confining stress level. Elliott, Dennis, and Qiu (1998) presented the
significant effect of confining pressure on compacted subgrade soils. The effect of principal stress
rotation on the permanent deformation of pavement materials is not well understood yet, due to
complexity of testing methods for continuous directional change of the principal stresses. Results
from large-scale rutting tests (Chan, 1990) showed that the permanent vertical deformation of a
 Road Materials and Pavement Design 395

granular base subjected to a moving wheel load was at least three times higher than the one
subjected to a repeated vertical load with the same magnitude.

3. Development of rutting models

3.1. Existing models
An abundant experience is obtained from laboratory testing of permanent deformation of
subgrade soils. However, there is no unique constitutively defined model for estimating the per-
manent deformation. Different researchers have used different approaches. Most models relate
the plastic strain to the number of cycles, stress level or both. The most widely known empirical
model is the power model, proposed by Monismith, Ogawa, and Freeme (1975) which is shown
in Equation (1). The elegance of the power model is its simplicity

εp1 = AN b , (1)

where εp1 is the cumulative axial plastic strain (in %), N is the number of load repetition, A and
b are parameters depending on the stress state and soil properties. An explicit form of the effect
of the stress state is not included in the model. Alternatively, Li and Selig (1996) developed
an improved model with a reasonable quantification of the influence of the stress state. The
extended power model is shown in Equation (2) where the effect of the stress is included. Li and
Selig reported that the model is capable of predicting cumulative plastic strains for unconfined
triaxial tests
 b
σd
εp1(N ) = a N c, (2)
σs
where σd is the deviatoric stress, σs is the soil static strength, N is the number of load repetitions,
and a, b, c are material constants depending on the soil type. Another model proposed by Chai
and Miura (2002) is presented in the following equation:
 b  
σd σs0 b c
εp1(N ) = a 1+ N , (3)
σs σs
where σs0 is the initial static deviator stress and the remaining parameters are similar to the
terms defined in Equation (2). This model is able to incorporate the static stress history. A four-
parameter permanent deformation model was proposed by Puppala et al. (2009) as shown in
Equation (4). This model considers the effect of both confining and deviator stresses
   
σoct α3 τoct α4
εp1 = α1 N α2 , (4)
Pref Pref
where α1 , α2 , α3 and α4 are model constants determined from laboratory tests, σoct and τoct are
the octahedral normal and shear stresses, respectively, and Pref is the reference stress. For triaxial
loading conditions, the octahedral and shear stresses are defined in the following equation:
√
1 2
σoct = (σ1 + 2σ3 ), τoct = (σ1 − σ3 ). (5)
3 3
All models proposed earlier appeared to fit the experimental results that they are derived from.
The model proposed by Puppala et al. (2009) includes the effect of mean stress and shear stress
which are fundamental stress components for the development of plastic strains. These stress
components are used to determine the resilient modulus of soils and aggregates in the current
mechanistic-empirical pavement design methods (NCHRP, 2004).
396 G.Y. Yesuf and I. Hoff

3.2. Finite element modelling in this study

In the numerical modelling of permanent deformation in subgrade soils, the concept of plasticity
models must be applied. Conventional elastoplastic models are well known and are widely used
in the analysis of geotechnical problems. Most models are efficient for static loading condition:
for static load application, incremental loading or constant strain rate loading. Nonetheless, the
conventional elastoplastic constitutive models based on a strain hardening mechanism are able to
simulate the plastic strains under monotonic loading. Due to this, the models do not account for
the plastic strains, after the first load repetitions unless the load level is increased. New modelling
concepts based on isotropic and kinematic hardening mechanisms have been emerged to simulate
cyclic loading applications in geo-materials (Manzari & Prachathananukit, 2001; Rouainia &
Wood, 2001). Such models are, however, appealing to few numbers of loading cycles and very
large computation time is required to simulate the repeated traffic load applications in pavements.
Besides, extensive tests are required to determine the model parameters. Therefore, simplified
and efficient models are yet to be developed for simulation of a large number of load repetitions
which can be applied in a routine pavement design.
The fundamental principles of elastoplastic models are available in the literature (Chen & Han,
1988; de-Souza-Neto, Peric, & Owen, 2008). When soil models are considered to be employed
for the prediction of permanent deformation, the hardening rule is considered to control the
accumulation of plastic strains. Two distinct types of hardening mechanisms are available in
elastoplastic soil models, namely, isotropic and kinematic hardening. Isotropic hardening is a
reasonable assumption for monotonic loading. It has been observed in experiments that some
materials exhibit translation of the yield surface upon stress reversal during cyclic loading; and in
this case, consideration of kinematic hardening provides a proper hardening mechanism. Traffic
loading on pavements is not a complete cyclic loading, that is, there is no deviator stress reversal.
Traffic loading is usually termed as “repeated loading” rather than “cyclic loading”. Due to this,
only isotropic hardening is assumed in this study.
The premise for the hypothesis of the model presented in this paper is based on the following
facts.
Bonaquist and Witczak (1996) presented incremental permanent deformations normalized
by the first cycle strain for subgrade and base materials. The accumulation of plastic strain is
strongly related to the magnitude of plastic strain on the first cycle of loading.
The accumulation of plastic strains depends on the confining pressure (Elliott et al., 1998;
Puppala et al., 2009). In order to include this effect in the long-term prediction in this study,
the permanent deformation from the first loading cycle is obtained based on the Drucker–Prager
yield criterion.
The static stress limit is related to the accumulation of plastic strains (Chai & Miura, 2002;
Li & Selig, 1996). In the semi-analytical method described in Li and Selig (1996), the static
strength of the soil is used to include the effect of the physical state of the soil. The proximity
to static failure is very important in the development of plastic deformation. The rate at which
plastic strain accumulates was found to relate quite well to the proximity of the applied stress
state to the stress conditions which are needed to cause failure under a single load (Dawson &
Kolisoja, 2006; Muhanna, Rahman, & Lambe, 1998).
The proposed model in this study is presented in the following equations:

εp1,N = N m(1−(σd /qs ) εp1,0 , (6)


N
εp1,total = (N m(1−(σd /qs )) )εp1,0 , (7)
N =1
 Road Materials and Pavement Design 397

where εp1,N is the axial plastic strain for load cycle N, εp1,total is the accumulated (total) axial
plastic strain, N is the number of load repetition, εp1,0 is the accumulated axial plastic strain at
the first loading cycle, σd is the deviatoric stress (loading from the traffic), qs is the static strength
of the soil and m is a constant to fit laboratory tests. The ratio of deviator stress to static strength
of the soil (σd /qs ) represents the proximity of deviator stress level and it is directly related to the
stability of the soil relative to the apparent shakedown limit.
It is noted in Equations (6) and (7) that the rate of accumulation of plastic strains increases as
the ratio of deviator stress to the static strength of the soil approaches to unity. When the level of
deviator stress increases, the accumulation of plastic strains also increases. The plastic strain from
the first cycle load is obtained from an equivalent static load in the first cycle based on the theory
of plasticity. The Drucker–Prager yield criterion is used to account for the effect of confining
pressure. The Drucker–Prager yield criterion first proposed by Drucker and Prager (1952) is a
smooth approximation to the Mohr–Coulomb law. The model is a modification of the von-Mises
criterion where extra term is included to introduce pressure sensitivity. The main advantage of
the Drucker–Prager yield surface is that it is easy for numerical implementation. However, it
assumes the same yield limit for triaxial compression and extension unlike of the Mohr–Coulomb
yield surface. Nonetheless, at particular points of triaxial compression, the Drucker–Prager yield
surface fits well with the Mohr–Coulomb model. In subgrade soils in pavements, there will not be
extension loading condition. Additionally, the principal stress rotation is minimal at the subgrade
level that the stress state is fully compressive. So, the approximation of the deformation from the
first cycle loading using the Drucker–Prager model is this paper is consistent with the Mohr–
Coulomb model approximation.
The Drucker–Prager criterion states that plastic yielding begins when the second deviator
stress and the mean stress reach a critical combination. The outset equation for plastic yielding
is given in the following equation:

f (σij , c, ϕ) = J2 + αp − κc, (8)

where J2 is the second deviator stress invariant, p is the mean stress, α and κ are defined in
Equation (9) and (10), respectively, c is cohesion and ϕ is the friction angle. The isotropic hard-
ening mechanism of soils is controlled by cohesion or friction angle depending on the type of
soils (Nordal, 2010; de-Souza-Neto et al., 2008). Cohesion hardening is considered for cohesive
subgrade soils, while hardening governed by friction angle is used for non-cohesive soils such as
sand and aggregates. Approximation techniques are obtained by making the yield surfaces of the
Drucker–Prager and Mohr–Coulomb criteria coincident at the outer or inner edges of the Mohr–
Coulomb surface. For the case of compression, the approximation is provided in the following
equations:
6 ∗ sin ϕ
α(ϕ) = √ , (9)
3(3 − sin ϕ)
6 ∗ cos ϕ
κ(ϕ) = √ . (10)
3(3 − sin ϕ)
In the theory of plasticity, the fundamental property of plastic strain increments is postulated
in the following equation:
p ∂g(σij , q∗ )
dεij = dλ · , (11)
∂σij
where dλ is a scalar quantity that quantifies the amount of plastic strain increment, g is the
plastic potential which controls the direction of plastic flow, σij is the stress and q∗ represents
398 G.Y. Yesuf and I. Hoff

Figure 1. Determination of model parameters from triaxial testing from cohesion-controlled hardening
yield surface.

internal variables. Equations (12) and (13) define the plastic potential for the Drucker–Prager
yield criterion.

g(σij , q∗ ) = J2 + α(ψ)p, (12)
6 ∗ sin ψ
α(ψ) = √ , (13)
3(3 − sin ψ)

where ψ is the dilation angle. When ψ = ϕ, the plastic flow follows an associated flow rule
which is a valid assumption for stable, work hardening materials. If ψ = ϕ, the plastic flow
follows a non-associative flow rule. The plastic strain increment for non-associated flow rule is
normal to the plastic potential which is different from the yield surface. The consequence of using
a associative flow rule for materials possessing non-associative behaviour (such as aggregates) is
that plastic volume change is overestimated. In this study, it is assumed that the extent of dilation
in cohesive soils is minimal, and the associated flow rule is used.

3.3. Determination of parameters

The material parameters in the proposed model are obtained from triaxial testing. The deviator
static stress is the difference between the axial stress and the confining (radial) pressure at failure.
The friction angle (ϕ) and cohesion (c) are obtained from the effective mean stress (p) and devi-
atoric stress (q) plot. Figure 1 shows the incremental failure for cohesion hardening mechanism.
It is noted that the soil gradually hardens (consequently the yield surface is expanded) before it
finally reaches the ultimate failure state.

4. Numerical implementation
The basic platforms of elastoplastic constitutive model are decomposition of total strain into elas-
tic and plastic components, a relationship that governs the elastic contribution and conditions for
the development of plastic strains. The three ingredients necessary to control the plastic con-
tribution in plasticity theory are yield criterion, flow rule and hardening rule. The stress update
mechanisms based on the aforementioned preconditions are available in the literature (Simo &
Hughes, 1998; de-Souza-Neto et al., 2008).
 Road Materials and Pavement Design 399

The integration schemes for stress updates are generally categorized as an explicit and implicit
scheme. In the explicit integration process, the stress increment is obtained with the knowledge
of the total strain increment and the stress. Although this method is simple and straight forward,
it is conditionally stable, that is, the solution may diverge. The accuracy of integration also
depends on the increment size. In the implicit integration scheme, a trial stress increment is
chosen which takes the updated stress outside the yield surface. The stress is then updated with
a plastic correction to bring it back onto the yield surface. The implicit elastic predictor-return
mapping algorithm for the Drucker–Prager model described in de-Souza-Neto et al. (2008) is
used for the numerical implementation in this study. The implicit scheme, in this case, insures a
converged solution of the plastic strains in a few load increments.
If the applied load exceeds the yield limit of the soil, plastic strains develop and are obtained
based on Equation (11). Since plastic strains do not increase after the first load increment unless
the load level is increased, we implement a semi-analytical model based on Equations (6) and (7).
A flowchart for the numerical algorithm is presented in Figure 2. In the numerical scheme, each
time increment after the solution is converged for the applied load, is considered as equivalent
loading repetitions. As it is shown in Figure 3, the first loading cycle is obtained after few time
increments using a numerical implicit stress update scheme. The consistent tangent modulus
for Drucker–Prager model formulated by de-Souza-Neto et al. (2008, p. 338–339) is used. It is
further noted that the main essence of the model is that the incremental plastic strains are defined
as some percentage of the initial strains as it is observed in experimental tests where the cyclic
hardening behaviour can be approximated by the power model (Bonaquist & Witczak, 1996).

Figure 2. Schematic presentation of the proposed constitutive model.

400 G.Y. Yesuf and I. Hoff

Figure 3. Concept of repeated load application assumed in this study.

The prediction of the plastic strain from the first loading cycle is based on plasticity theory and
is algorithmically consistent. However, the subsequent plastic strains are hypothesized based on
experimental observations.
Two methods of load increment are available in ABAQUS™ (Standard). In the first method,
the load is applied proportional to the increments for a single step (ramp). The second method
is an instantaneous loading increment in which the entire load is applied instantly in the first
few time increments. The STEP option must be chosen in proposed flowchart to ensure that
the required load is fully applied in the first loading cycle. The stress updates in the subsequent
steps are used to evaluate the proximity of the deviator stress to compute the accumulated plastic
strains.
The proposed model is implemented in the Users-Material Subroutine (UMAT) in
ABAQUS™ for axisymmetric elements. Axisymmetric elements are used for numerical sim-
ulation of triaxial testing. They can also be used for multi-layer analysis with the assumption of
circular contact area between the vehicle tire and the pavement surface.

5. Results and discussions

Numerical simulations are carried out based on the proposed model in this study to compare
the results with laboratory tests obtained from Townsend and Chisolm (as cited in Li & Selig,
1996). The physical properties of the soil were dry density, ρd = 1440 kg m−3 , and water content,
w = 28.7 %. The maximum static strength obtained from uniaxial laboratory testing was qs =
159 kPa. The following input parameters are calibrated for the same maximum static strength,
that is, 159 kPa, using finite element analysis (strain-controlled uniaxial test simulation): E =
100 MPa, υ = 0.3, ϕ = 15◦ , ψ = 15◦ . Similarly, cohesion hardening variables are calibrated as
(c0 = 10 kPa, ε̄p,0 = 0) and (c1 = 61 kPa, ε̄p,1 = 0.001275). The laboratory test results reported
in Li and Selig (1996) did not explicitly present the values of friction angle and cohesion, and a
trial-and-error method is used to obtain the friction angle and cohesion.
Three sets of deviator stress are considered similar to the laboratory tests: σd = 52.5, 73.1 kPa
and 81.1 kPa. The material parameter m in Equations (6) and (7) is set to −1.49. It is important
to note that the test results presented in Li and Selig (1996) consists of two independent tests at
different water contents. One, unconfined compression test to obtain the static failure limit of the
soil, and the other test is cyclic loading test. Hence, the stiffness, friction angle and hardening
 Road Materials and Pavement Design 401

Figure 4. Comparison of the proposed model in this study with test results reported in Li and Selig (1996)
for soil type ρd = 1440 kg m−3 , w = 28.7 % and qs = 159 kPa.

parameters are calibrated from the static test. The back-calculated parameters are then used to
validate the proposed model against another test results from the same soil type (mainly repeated
loading). Figure 4 shows the comparison of the numerical simulation with laboratory tests at
different deviator stress levels.
The proposed model is also validated for the same soil type but with different water con-
tents where the test data are obtained from Townsend and Chisolm (as cited in Li & Selig,
1996). Ying, Chengcheng, Hongjun, and Wei (2011) showed that the cohesion does not change
appreciably at high water content, whereas the friction angle changes substantially. Hence, in
the calibration of material parameters in this study, the cohesion hardening parameters are kept
the same, and only the friction angle is changed to obtain a static strength of qs = 193.0 kPa
for the soil with dry density, ρd = 1470 kg m−3 , and water content, w = 27.6 %. Consequently,
ϕ = 26◦ is obtained. The calibrated input parameters for the numerical model are E = 120 MPa,
υ = 0.3, ϕ = 26◦ , ψ = 26◦ . The cohesion hardening variables are (c0 = 10 kPa, ε̄p,0 = 0) and
(c1 = 61 kPa, ε̄p,1 = 0.001275). The results from finite element analysis in comparison with
laboratory tests are presented in Figure 5.
It is shown in Figures 4 and 5 that the numerical model proposed in this paper gives a very
good agreement with test results up to 50% of the static strength of the soil. For higher deviator
stresses, the results do not match well. At low deviator stress level, the mobilisation of soil
grains is low and the structural bond of the soils is affected to a low extent. So, the accumulation
of plastic strains depends on the previous history of the soil, for example, on the deformation
from first cycle loading. However, when the deviator stress level approaches to the proximity
of the static strength of the soil (in other words, at high deviator stress level), the soil will be
mobilised to a large or full extent – and hence the structure in soil grains is partially or fully
lost. Consequently, the rate of accumulation of plastic strains increases highly. This phenomenon
is commonly observed in laboratory tests. The physical process in soils at this phase is difficult
to describe it quantitatively. Robust numerical models in this context are yet to be developed.
402 G.Y. Yesuf and I. Hoff

Figure 5. Comparison of the proposed model in this study with test results reported in Li and Selig (1996)
for soil type ρd = 1470 kg m−3 , w = 27.6 % and qs = 193 kPa.

Figure 6. The effect of confining stress on the accumulation of plastic strain at a deviator stress level of
73.1 kPa.

Fortunately, the degree of mobilisation of subgrade soils in pavements is low since the additional
load from the traffic is reduced at the subgrade level.
The Drucker–Prager failure criterion used to obtain the plastic strains in the first loading
cycle provides the capability of the proposed model to predict the effect of confining pres-
sure on the accumulated permanent deformation. For the input parameters of E = 100 MPa,
υ = 0.3, ϕ = 15◦ , ψ = 15◦ , and cohesion hardening parameters (c0 = 10 kPa, ε̄p,0 = 0) and
(c1 = 61 kPa, ε̄p,1 = 0.001275), the deviator stress is kept the same at σd = 73.1 kPa, and
 Road Materials and Pavement Design 403

different confining stresses are considered. In addition, a loading cycle up to 10,000 repetitions
is simulated. The results are presented in Figure 6.
The effect of confining pressure on the accumulation of plastic strains is clearly shown. As
the confining pressure increases, the permanent deformation is reduced. When the confining
pressure is doubled (from 15 to 30 kPa) the accumulation of plastic strains is reduced by 29%.
This implies that confining stress is an important component in the prediction and quantification
of plastic strains.
The nonlinear elastic stiffness (often called resilient modulus) affects the point at which the
soil starts yielding. However, the study presented in this paper does not consider the nonlinear
behaviour of the soil stiffness. Nonetheless, it is important to note that different stiffness values
are chosen by the user for different soil types and conditions, but that values remain constant
for that particular test or simulation. In principle, the nonlinear behaviour should be included
in the constitutive model and this can be one of the limitations of the proposed model in this
study. It will be possible to update the stress in the elastic range using “consistent elastic tangent
modulus” which requires additional task in the numerical implementation.

6. Conclusion
The property of subgrade soils is commonly quantified by its cohesion and friction angle. These
two parameters are not unique quantities; rather, they describe the property of the soil at a par-
ticular state or condition such as moisture content and degree of compaction. This implies that
the accumulation of plastic strains for different soil conditions can be linked to the cohesion and
friction angle. The advantage of the model presented in this paper over the existing empirical and
analytical models is that the effect of stress is fully incorporated in the analysis, and the basic soil
parameters, that is, soil cohesion and friction angle, are used. Moreover, parameters of the model
are obtained from conventional triaxial (strain controlled) tests which are commonly carried out
in routine geotechnical investigation. The prediction of plastic strains using the proposed model
in this study provides a good agreement with laboratory test with deviator stress level up to 50%
of the static strength of the soil. The mobilisation of subgrade soils in pavements is normally low
due to reduced stress magnitude at the subgrade level, which makes the proposed model in this
study more appealing to understand the development of plastic strains in subgrade soils.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
This research is done during the Ph.D. study of the first author at the Norwegian University of Science
and Technology [Project number: 25100500]. The research fund support from the Norwegian Public Roads
Administration is greatly acknowledged.

References
Bonaquist, R., & Witczak, M. W. (1996). Plasticity modeling applied to the permanent deformation response
of granular materials in flexible pavement systems. Transportation Research Record: Journal of the
Transportation Research Board, 1540, 7–14. doi:10.3141/1540-02
Brown, S. F., Lashine, A. F., & Hyde, A. L. (1975). Repeated load triaxial testing of a silty clay.
Geotechnique, 25(1), 95–114. doi:10.1680/geot.1975.25.1.95
Chai, J.-C., & Miura, N. (2002). Traffic-load-induced permanent deformation of road on soft subsoil. Jour-
nal of Geotechnical and Geoenvironmental Engineering, 128(11), 907–916. doi:10.1061/(ASCE)1090-
0241(2002)128:11(907)
404 G.Y. Yesuf and I. Hoff

Chan, F. W. K. (1990). Permanent deformation resistance of granular layers in pavements (PhD thesis).
University of Nottingham, England.
Chen, W. F., & Han, D. J. (1988). Plasticity for structural engineers. New York, NY: Springer.
Croney, D., & Croney, P. (1997). Design and performance of road pavements (3rd ed.). New York, NY:
McGraw-Hill.
Dawson, A., & Kolisoja, P. (2006). Managing rutting in low volume roads (executive sum-
mary). Roadex III Project. Retrieved from http://www.roadex.org/uploads/publications/docs-RII-
S-EN/Managing%20Rutting_English.pdf
Diaz-Rodriguez. (1989). Behavior of Mexico city clay subjected to undrained repeated loading. Canadian
Geotechnical Journal, 26(1), 159–162. doi:10.1139/t89-016
Drucker, D. C., & Prager, W. (1952). Soil mechanics and plastic analysis for limit design. Quarterly of
Applied Mathematics, 10(2), 157–165.
Elliott, R. P., Dennis, N., & Qiu, Y. (1999). Permanent deformation of subgrade soils, phase II: Repeated
load testing of four soils. Springfield, VA: National Technical Information Service.
Elliott, R. P., Dennis, N. D., & Qiu, Y. (1998). Permanent deformation of subgrade soils (a
test protocol: MBTC FR-1069). University of Arkansas, Fayetteville. Retrieved from http://ww2.
mackblackwell.org/web/research/ALL_RESEARCH_PROJECTS/1000s/1069-elliott/MBTC1069.pdf
Hoff, I. (1999). Material properties of unbound aggregates for pavement structures (PhD thesis).
Norwegian University of Science and Technology, Trondheim.
Horvli, I. (1979). Dynamisk prøving av leire for dimensionering av veger (in Norwegian) [Dynamic
testing of clay subgrades for pavement design] (PhD thesis). Norwegian Institute of Technology,
Trondheim.
Huang, Y. H. (2004). Pavement analysis and design (2nd ed.). Upper Saddle River: Pearson Prentice
Hall.
Lekarp, F., & Dawson, A. (1998). Modelling permanent deformation behaviour of unbound granular
materials. Construction and Building Materials, 12(1), 9–17. doi:10.1016/S0950-0618(97)00078-0
Li, D., & Selig, E. T. (1996). Cumulative plastic deformation for fine-grained subgrade soils. Journal of
Geotechnical Engineering, 122(2), 1006–1013. doi:10.1061/(ASCE)0733-9410(1996)122:12(1006)
Manzari, M. T., & Prachathananukit, R. (2001). On integration of a cyclic soil plasticity model.
International Journal for Numerical and Analytical Methods in Geomechanics, 25(6), 525–549.
doi:10.1002/nag.140
Mitchell, R. J., & King, R. D. (1976). Cyclic loading of an Ottawa area Champlain sea clay. Canadian
Geotechnical Journal, 14(1), 52–63. doi:10.1139/t77-004
Monismith, C. L., Ogawa, N., & Freeme, C. R. (1975). Permanent deformation characteristics of subgrade
soils due to repeated loading. Transportation Research Record, 537, 1–17.
Muhanna, A. S., Rahman, M. S., & Lambe, P. C. (1998). Resilient modulus and permanent strain
of subgrade soils. Transportation Research Record 1619, Transportation Research Board, 85–93.
doi:10.3141/1619-10
NCHRP. (2004). Guide for mechanistic-empirical design of new and rehabilitated pavement structures.
Washington, DC: Transport Research Board.
Nordal, S. (2010). Soil modeling (Lecture notes for PhD course). Trondheim: Norwegian University of
Science and Technology.
NPRA. (2003). Håndbok 111:Standard for drift og vedlikehold (in Norwegian) [Standard for operation and
maintenance of roads]. Oslo: Norwegian Public Roads Administration.
Puppala, A. J., Saride, S., & Chomtid, S. (2009). Experimental and modeling studies of permanent strains
of subgrade soils. Journal of Geotechnical and Geoenvironmental Engineering, 135(10), 1379–1389.
doi:10.1061/(ASCE)GT.1943-5606.0000163
Rouainia, M., & Wood, D. M. (2001). Implicit numerical integration for a kinematic hardening soil plas-
ticity model. International Journal for Numerical and Analytical Methods in Geomechanics, 25(15),
1305–1325. doi:10.1002/nag.179
Simo, J. C., & Hughes, T. J. R. (1998). Computational inelasticity. New York, NY: Springer.
de-Souza-Neto, E. A., Peric, D., & Owen, D. R. J. (2008). Computational methods for plasticity: Theory
and applications. Chichester: John Wiley.
Uzan, J. (2004). Permanent deformation in flexible pavements. Journal of Transportation Engineering,
130(1), 6–13. doi:10.1061/(ASCE)0733-947X(2004)130:1(6)
Ying, G., Chengcheng, Z., Hongjun, L., & Wei, S. (2011). The impact of water content and compaction of
the soil on the slope protection by vegetation. Paper presented at the third international conference on
transportation engineering, Chengdu, China.
Reproduced with permission of the copyright owner. Further reproduction prohibited without
permission.