See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/277205104

Interim revision of the South African Mechanistic-Empirical pavement design

method for flexible pavements

Article · September 2011

CITATIONS READS

6 3,862

4 authors, including:

Hechter Theyse Morris De Beer

GHD Pty Ltd Council for Scientific and Industrial Research, South Africa
29 PUBLICATIONS   293 CITATIONS    57 PUBLICATIONS   630 CITATIONS   

SEE PROFILE SEE PROFILE

James Maina
University of Pretoria
82 PUBLICATIONS   350 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Advanced numerical modeling View project

Revison of the South African Pavement Design Method View project

All content following this page was uploaded by Morris De Beer on 21 July 2015.

The user has requested enhancement of the downloaded file.

 10th CONFERENCE ON ASPHALT PAVEMENTS FOR SOUTHERN AFRICA

INTERIM REVISION OF THE SOUTH AFRICAN MECHANISTIC-EMPIRICAL PAVEMENT

DESIGN METHOD FOR FLEXIBLE PAVEMENTS

Authors name(s): H L Theyse1, M de Beer2, J W Maina2, L Kannemeyer3

First author's address:1. Pavement Modelling Corporation

1397 Starkey Avenue
Waverley, 0186
Co-authors' addresses: 2. CSIR Built Environment
P O Box 395
Pretoria, 0001
3. South African National Roads Agency Limited
P O Box 415
Pretoria, 0001

Abstract

Pavement design methods, in combination with network level management systems must
enable road authorities to develop reliable long-term financial plans based on the
estimated structural capacity of the road network. Inaccurate design models at the core
of such a design system could, however, result in significant design risk and inappropriate
financial planning. The design model for unbound granular layers contained in the 1996
version of the South African Mechanistic-Empirical Design Method (SAMDM) for flexible
pavements was shown to be overly sensitive to minor changes in certain input variables
at the Conference on Asphalt Pavements for Southern Africa held in 2004 and the general
accuracy of this model was challenged.

This paper presents an interim revision of the SAMDM recommending increased contact
stress values at the tyre-pavement interface; updated unbound material characterisation
parameters; effective stress analysis for unbound material; and revised damage models
for estimating the structural capacity of unbound granular layers including the pavement
subgrade. The internationally accepted subgrade vertical strain damage model is
replaced with a model calibrated for local material and environmental conditions. The
Factor of Safety (FoS) model unbound, granular base and subbase layers is replaced with
a Stress Ratio (SR) model which is explicitly calibrated for the effect of material density
and saturation levels. The revised models are shown to be less sensitive to variation in
the resilient input properties of unbound pavement layers and better suited to simulating
the permanent deformation of granular pavement layers under repeated traffic loading.

-1-
 10th CONFERENCE ON ASPHALT PAVEMENTS FOR SOUTHERN AFRICA

1. INTRODUCTION

1.1. Background

Pavement design is essentially an endurance problem and not a stability problem such as
the design challenges in structural and geotechnical engineering. Pavement failures are
definitely undesirable but generally the consequences of pavement failures are not as
catastrophic as the collapse of a bridge or a building. Instead, deteriorating road conditions
and ultimate pavement failures have a long-term negative economic impact and financial
implications that are not immediately apparent. Pavement design methods, in combination
with network level management systems must therefore enable road authorities to develop
reliable long-term financial plans based on the estimated structural capacity of the road
network. Certainly, the pavement design method should also guard against premature
failure as far as possible but often these premature failures are not caused by a design error
but rather attributed to:

• insufficient information collected during the design investigation;

• incorrect interpretation of the information and more specifically oversight of risk

factors, and

• violation of design assumptions during construction as a result of poor construction

quality.

The vision for future pavement design systems in South Africa is to incorporate the
mechanistic-empirical models in a holistic design system that also addresses the interaction
with network level management systems; that guides the design investigation process and
ultimately base design decisions on economic considerations to alleviate some of the above
aspects. However, inaccurate design models contained at the core of such a design system
will result in significant, unknown design risk and it cannot be denied that the 1996 version
of the South African Mechanistic Design Method (SAMDM) (Theyse et al., 1996) for flexible
pavements contained such inaccurate design models.

1.2. Historical development

The development of mechanistic-empirical (ME) based flexible pavement design procedures

have long been pursued in South Africa (Van Vuuren et al., 1974, Walker et al., 1977,
Paterson and Maree, 1978, Maree and Freeme, 1981, Jordaan, 1993). The overview of the
method published in 1996 (Theyse et al., 1996) was based largely on the content of these
earlier publications. The implementation of this 1996 version of the method in software
packages resulted in a much larger user group having access to the method and using the
method for pavement design. The wide-scale use of the method highlighted certain of the
problems associated with the method and culminated in a paper at the CAPSA 2004
conference (Jooste, 2004) questioning the value of the method, especially for the design of
unbound, granular pavement layers.

Some of the problems raised at the CAPSA 2004 conference were known to the research
fraternity and in certain cases research at the CSIR was already aimed at developing
solutions to these problems. This research effort was significantly increased with the

-2-
 10th CONFERENCE ON ASPHALT PAVEMENTS FOR SOUTHERN AFRICA

initiation of a project by the South African National Roads Agency Limited (SANRAL) to
revise the South African pavement design method for flexible and rigid pavements (SAPDM)
(www.SAPDM.co.za). A framework for the revision of the mechanistic-empirical design
method for flexible pavements was presented at CAPSA 2007 and although it is not possible
to present the final revised design method yet, significant improvements to the design of
unbound, granular layers are already possible.

1.3. Proposed interim improvements

This paper presents an interim revision of the mechanistic-empirical design method for
flexible pavements, addressing the following aspects:
1. Load characterisation
a. Recommendations on increased tyre contact stress levels for design;
2. Material characterisation
a. Updated resilient response characterisation of the pavement subgrade;
b. Updated strength parameters for unbound granular layers;
3. Primary pavement response analysis
a. The introduction of effective stress analysis incorporating suction pressure
and residual compaction stress;
4. Revised damage models
a. Interim models for the permanent deformation of unbound pavement layers
including the pavement subgrade and structural layers.

The internationally accepted vertical strain damage model for the pavement subgrade is
challenged by the paper based on the available experimental data. The plastic strain
damage model for the upper unbound pavement layers also shows that the density and
saturation of the material dominates the permanent deformation of the material in
combination with the imposed stress condition.

The application of the interim solutions presented in the paper are tested against limited
Accelerated Pavement Testing (APT) results and compared to results from the 1996 method,
showing significant improvement in the modelling ability of the revised models. The design
cases used in the critical review by Jooste (2004) are also re-analysed, showing a significant
reduction in the disproportionate sensitivity of the method to variation in certain input
parameters.

2. MECHANISTIC-EMPIRICAL PAVEMENT MODELLING CONCEPTS

Figure 1 illustrates the dissipation of the stress resulting from an external wheel-load
through a layered pavement system. Mechanistic-Empirical pavement design methods
attempt to model the resilient response of the pavement and more importantly, the
damage caused by the external wheel-load throughout the pavement system. All
mechanistic-empirical pavement design methods that are intended for routine pavement
design separate the resilient response modelling of the pavement from the damage
modelling similar to the diagram shown in Figure 2. Models such as plasticity theory that
combines the elastic and plastic response of materials in a single constitutive material
model are unlikely to be implemented in routine design methods in the near future as they

-3-
 10th CONFERENCE ON ASPHALT PAVEMENTS FOR SOUTHERN AFRICA

are computationally intensive and still do not cater for all the possible responses of the
materials used in pavement engineering.

Half-axle tyre loads

Wearing course Surfacing

Base layer Structural layers

•High shear stresses
•Large strains
Subbase layer

Upper imported
subgrade
Lower imported Subgrade
subgrade •Low shear stresses
•Small strains
In situ subgrade

Figure 1. Dissipation of externally applied stress in a layered pavement system

(from Theyse, 2009)

Input
In service ME model

Geometry

Axle load Fixed

Computer
solution

Traffic histogram load

Damage models
data Contact stress Fixed
histogram contact stress
Thermal cracking

Fatigue
HMA

Resilient
Grading response models Plastic strain/
permanent deformation
Binder
content Top-down cracking
Mr = Constant
HMA

Binder
Plastic strain/
Unbound

properties Primary permanent deformation

pavement
Temperature response model Critical
Mr = f (Temp, Other
parameters
frequency) Structural
Other or capacity
Crushing
estimate
Resilient
response
Stabilized

Grading Mr = f (Dens, Stiffness reduction

analysis
saturation)
Material data

Atterberg Plastic strain/

limits permanent deformation
Unbound

Moisture Mr = f (Bulk Other

content and shear
stress) Plastic strain/
Subgrade

Density
permanent deformation

Other Other
Mr = f (Strain)

UCS

Stress and
Stabilized

Poisson’s
strain at break ratio
Time/previous
loading

Other

Figure 2. Schematic representation of the main components of a mechanistic-

empirical pavement model (from Theyse, 2008)
-4-
 10th CONFERENCE ON ASPHALT PAVEMENTS FOR SOUTHERN AFRICA

Mechanistic-empirical design methods may be classified according to the manner in which

the damage modelling is done. Typically, when new pavements are opened to traffic there
is an initial rapid accumulation of distress or damage (bedding-in), followed by a period of
linear accumulation of damage and finally an accelerated increase in the damage as
illustrated by paths
and  in Figure 3. The damage models in classical mechanistic-
empirical methods are calibrated to estimate the number of load repetitions required to
progress from a condition of zero initial distress or damage to the terminal condition which
is merely a predefined, unacceptable condition and does not imply complete failure of the
facility. Given that the damage models of classical ME-design methods do not retain any
information regarding the progression of damage, a linear progression following path  in
Figure 3 is implied. Damage models that relate a critical stress or strain parameter (S) to the
number of repetitions (N) that can be sustained before the terminal condition is reached are
referred to as S-N damage models in this paper.

Another characteristic of classical ME-design methods is the use of a design standard load.
The structural capacity of the design pavement is expressed in terms of the number of
design standard loads required to reach the terminal structural condition.

Structural
distress/damage

Terminal structural
condition

1
2

Facility Traffic/Time
opened to Structural capacity
traff ic

Figure 3. Illustration of various hypothetical damage progression paths for

pavements (from Theyse, 2009)

More recently, ME-design methods have evolved to accommodate mixed traffic and time-
dependent input variables by utilising recursive modelling. These recursive methods may in
turn be classified as linear recursive methods, which utilise the damage models from
classical ME design methods in combination with Miner’s law; or non-linear recursive
methods that are based on incremental damage models and the strain-hardening approach.

The interim solutions proposed in this paper are formulated in the context of a classical ME
design method which may also be applied in a linear recursive modelling process.
Ultimately the revised South African Pavement Design Method will be based on non-linear
recursive modelling. However, for the interim models to be immediately useful, they have
to be implemented within the current classical ME-design framework used in South Africa.
This interim version of the classical South African Mechanistic-Empirical Design Method for
flexible pavements is, therefore, referred to as SAMDM2011 to distinguish from the 1996
version.

-5-
 10th CONFERENCE ON ASPHALT PAVEMENTS FOR SOUTHERN AFRICA

3. REVISED ME-DESIGN MODELS FOR SAPDM2011

3.1. Load characterisation

The fact that the SAPDM2011 remains a classical ME-design method has the implication that
the method relies on the concept of a standard design load to quantify the structural
capacity of the pavement. The 80 kN single axle standard design load with a dual-wheel
configuration is therefore retained in SAMDM2011 similar to the standard design load in the
1996 version. However, the revised method deviates from the 1996 version regarding the
tyre-pavement contact stress that is used for design analysis. Stress-In-Motion (SIM)
measurements results have shown the value of 520 kPa recommended by the 1996 version
to be too low (De Beer et al., 1999) and SAMDM2011, therefore, makes provision for using
increased tyre-pavement contact stress values.

Although the concept of a standard design load is retained in SAPDM2011,

recommendations are provided for the axle mass and contact stress on other axle groups
and wheel configurations to allow assessment of the effect that these load configurations
have on the structural capacity of a pavement. Table 1 provides a summary of typical axle
and wheel-load configurations with recommended contact stress values for design analysis.
The static axle loads listed in Table 1 are based on the current legal axle loads for these axle
groups and wheel configurations.

3.2. Material characterisation

As indicated previously, the mechanistic-empirical method consists of two main modelling

components, the primary pavement response model that calculates the elastic response of
the pavement to loading and the damage models that quantify the damage in all the
pavement layers given certain elastic response parameters. The primary pavement
response model used in most mechanistic-empirical design methods is a multi-layer, linear-
elastic continuum mechanics model that requires Young’s modulus and Poisson’s ratio to
characterise the resilient response of the materials found in each of the pavement layers.
Young’s modulus and Poisson’s ratio are, however, theoretical concepts that apply to
perfectly elastic materials. The Young’s modulus that represents the “stiffness” of materials
can only be approximated from experimental results and is most often approximated by the
resilient modulus for unbound granular material. The resilient modulus is a measure of the
elastic recovery of a specimen of material given the repeated application and removal of an
axial load under compressive stress conditions. Figure 4 shows a simplified case
representing a single cycle of load application and removal. The resilient modulus therefore
represents the secant slope through the two extremes of a load-unload, stress-strain
hysteresis loop in repeated load tri-axial testing.

Most road-building materials also exhibit stress-dependent and apparent anisotropic

behaviour. The magnitude of the resilient modulus therefore depends on the level of
confinement of the material and differs under tensile and compressive stress conditions, not
to mention the effect of density and saturation. Strictly speaking, there is no single resilient
modulus value for a given unbound, granular material but rather an infinite range of
possible values. The resilient modulus values recommended for design in this paper,
therefore, represents typical stiffness values in a compressive stress region underneath the
external wheel-load.

-6-
 10th CONFERENCE ON ASPHALT PAVEMENTS FOR SOUTHERN AFRICA

Table 1 Typical axle and wheel-load configurations for pavement design

Vehicle Axle group Wheel Typical half-axle configuration for Static Recommended
type configuration configuration analysis * axle tyre contact
group stress (kPa) *
load
(kN)
All Standard Dual 40 kN
R = 98.97 mm
80 650
design load
350 mm

Contact area

Truck Steering axle Single 38,5 kN 77 900

R = 116.69 mm

Single axle Single 40 kN 80 900

R = 118.94 mm

Dual 45 kN
R = 101.15 mm
90 700

350 mm

Tandem axle Single 40 kN 40 kN 160 900

R = 118.94 mm R = 118.94 mm

1350 mm

Dual 45 kN
R = 101.15 mm
45kN
101.15 mm
180 700

350 mm

1350 mm

Tridem axle Single 40 kN 40 kN 40 kN 240 900

R = 118.94 mm R = 118.94 mm R = 118.94 mm

1350 mm 1350 mm

Dual 40 kN 40 kN 40 kN 240 650

R = 98.97 mm R = 98.97 mm R = 98.97 mm

350 mm

1350 mm 1350 mm

Bus Steering axle Single 38,5 kN 77 900

R = 116.69 mm

Single axle Dual 51 kN

R = 104.03 mm 102 750

350 mm

* Load radii (R) and tyre contact stresses estimated from SIM measurements on 12R22.5
and 315/80 R22.5 tyres.

-7-
 10th CONFERENCE ON ASPHALT PAVEMENTS FOR SOUTHERN AFRICA

At rest Load Unload Applied stress = Load/area

= P/A
Load, P Loading strain = reduction in length/length
= ∆l/l

Reduction in Reversible strain = length recovered/length

Length
length, ∆l = ∆le/l
recovered, ∆le
Material specimen Stress, σ

specimen

specimen
Materialspecimen
specimen
Length, l

Applied Load
stress Mr

Material

Material
1 Unload

Material Reversible strain

Strain ε

Applied stress
Resilient Modulus (M r ) =
Reversible strain
Cross sectional
Pl
area, A =
A∆le
Figure 4. Approximation of Young’s modulus by the resilient modulus (from
Theyse, 2009)

3.2.1 Resilient response characterisation of the pavement subgrade

The resilient modulus values recommended for the pavement subgrade are based on back-
calculation from multi-depth deflectometer (MDD) deflection profiles recorded during
Heavy Vehicle Simulator (HVS) testing in South Africa. The subgrade was modelled as a
semi-infinite half-space in the back-calculation process although the MDD system is
normally anchored at a depth of 2,5 to 3 metres. Modelling the subgrade as a semi-infinite
half-space results in an over-estimation of the resilient modulus of the subgrade by roughly
20 to 25 %. However, when this “incorrect” modulus is used in a forward design calculation
in combination with a semi-infinite half-space, the subgrade deflection which is the
proposed new critical parameter for subgrade design is calculated correctly. Using the over-
estimated subgrade modulus values with a vertical subgrade strain damage model is not
recommended as the structural capacity of the subgrade will be over-estimated.

Table 2 provides a summary of average, long-term resilient modulus values for different
subgrade types and moisture conditions for a 40 kN dual wheel-load. A Poisson’s ratio of
0,35 is recommended for subgrade materials.

Table 2 Typical resilient modulus values for subgrade material

Subgrade Typical subgrade material Grading Plasticity Moisture Resilient modulus
type Modulus Index condition (MPa)
(GM) (PI) (MC)
Gravel Sandstone conglomerate 2,2 Non- Dry 200 – 300
(Gauteng) plastic Wet 100 – 150
Ferricrete (Gauteng) 1,2 10 Dry 100 – 200
Wet 60 – 100
Decomposed granite 2,1 14 Dry 300 – 500
(Limpopo) Wet 100 – 150
Calcrete (Karoo) 1,4 – 1,6 Non- Dry 200 – 300
plastic Wet 100 – 150
Sand Deep sand (Western Cape and 1,0 – 1,2 Non- Dry 100 – 150
KwaZulu-Natal plastic Wet 75 – 125
-8-
 10th CONFERENCE ON ASPHALT PAVEMENTS FOR SOUTHERN AFRICA

In general, course material tends to have a higher modulus and the higher the plasticity of
the material the bigger the difference between the dry and wet modulus of the material.

3.2.2 Resilient response characterisation of unbound granular material

The resilient modulus values recommended for unbound granular base and subbase layers
remain the same as for SAMDM1996 and are summarised in Table 3. The recommended
Poisson’s ratio for these materials is 0,35.

Table 3 Recommended resilient modulus values for unbound granular base and
subbase layers
Moisture condition Dry Wet
Support condition Cemented Over granular Cemented Over granular
Material Code Material layer in slab or equivalent layer in slab or equivalent
(CSRA, 1985) Description state granular layer state granular layer
G1 High quality 250 - 1000 150 - 600 50 - 250 40 - 200
crushed stone (450) (300) (250) (200)
G2 Crushed stone 200 - 800 100 - 400 50 - 250 40 - 200
(400) (250) (250) (200)
G3 Crushed stone 200 - 800 100 - 350 50 - 200 40 - 150
(350) (230) (200) (150)
G4 Natural gravel (base 100 - 600 75 - 350 50 - 200 30 - 150
quality) (300) (225) (200) (150)
G5 Natural gravel 50 - 400 40 - 300 30 - 150 20 - 120
(250) (200) (150) (120)
G6 Natural gravel (sub- 50 - 200 30 - 200 20 - 150 20 - 120
base quality) (200) (150) (150) (120)

Although the resilient modulus ranges in Table 3 are wide, the sensitivity of the design
model for unbound material to stiffness variation is countered by the introduction of
suction pressure and residual compaction stress that act as beneficial pre-stressing of the
unbound layers.

3.3. Primary pavement response modelling

The relationship between the ratio of applied stress to the shear strength (Stress Ratio or
Factor of Safety) and the plastic strain of unbound material has been confirmed by local
(Maree, 1978 and Theyse, 2008) as well as international research (Huurman, 1997 and van
Niekerk et al, 1998). One of the major problems associated with the SAPDM1996 is the
calculation of inadmissible stress conditions (FoS < 1, SR > 1) for unbound layers. These
inadmissible stresses are partly the result of not considering the effective stress condition
and the behaviour of unbound granular material depends largely on the effective stress
condition to which the material is subjected.

SAMDM1996 only considered the stress resulting from the application of the external
wheel-load in the primary response analysis of unbound granular layers. While the damage
(primarily permanent deformation) of unbound granular layers is the direct result of the
stress caused by the external wheel-load, the amount of damage is determined by the
effective stress regime in the unbound granular layers. The introduction of effective stress,
especially the residual compaction stress component, causes the unbound pavement layers

-9-
 10th CONFERENCE ON ASPHALT PAVEMENTS FOR SOUTHERN AFRICA

to be constrained similar to pre-stressed concrete thereby preventing the development of

tensile stress in the unbound material.

SAMDM2011 deviates from the past approach by introducing effective stress analysis in the
primary pavement response model for unbound granular base and subbase layers by
considering the following stress components:

• The vertical overburden stress in combination with residual compaction stress;

• Internal suction pressure resulting from the partial saturation of the material; and

• The stress caused by the external wheel-load.

The development of residual compaction stress behind retaining walls have long been
studied and documented in geotechnical engineering. Similarly, residual compaction stress
occurs in pavement layers and Uzan (1985) formulated theoretical models for the residual
compaction stress in pavement layers based on static equilibrium conditions. Dehlen (1958)
provides experimental proof of residual compaction stress. A simplified residual
compaction stress model was developed for SAPDM2011 using the theory formulated by
Uzan, the shear strength properties of a selection of unbound granular road-building
materials and the experimental results from Dehlen.

The concepts of matric suction and suction pressure are also well accepted in geotechnical
engineering (Fredlund, 1985: 465 - 472, Vanapalli et al, 1996a: 259 – 268, Vanapalli et al,
1996b: 379 – 392, Vanapalli and Fredlund, 1999: 93 – 96 and Vanapalli and Fredlund, 2000:
195 - 209) and Heath (Heath et al, 2002 and Heath, 2002) introduced suction pressure in the
analysis of granular pavement materials in California. Locally, a suction pressure model has
been developed based on a linear approximation of the soil-water characteristic curve for a
range of unbound granular material (Theyse, 2008) and the suction pressure model
parameters were found to be largely determined by the grading of the material. This model
is incorporated in SAPDM2011. Research is continuing to develop a general suction
pressure model for design application as part of the ongoing SANRAL research project.

The stress caused by the external wheel-load is calculated using a multi-layer, linear elastic
software programme such as GAMES (Maina and Matsui, 2004). The combination of the
residual compaction stress, suction pressure and external load stress in an effective stress
analysis has been coded into a software package for mechanistic-empirical pavement design
that will be released during 2012.

3.4. Damage models

SAMDM2011 provides revised damage models for the pavement subgrade and unbound
granular base and subbase layers.

3.4.1 Subgrade permanent deformation damage models

Similar to the 1996 version, SAMDM2011 makes provision for two levels of subgrade
permanent deformation, namely 10 and 20 mm. Using multi-depth deflection data from
HVS tests, Theyse (2001) showed that the vertical subgrade strain at the top of the subgrade
is not a good predictor of subgrade permanent deformation. The subgrade deflection was
found to have a better correlation with the subgrade permanent deformation. This does
not imply that the subgrade plastic strain is not correlated to the subgrade vertical elastic
-10-
 10th CONFERENCE ON ASPHALT PAVEMENTS FOR SOUTHERN AFRICA

strain. In fact, the integration of the subgrade vertical elastic and plastic strain over the full
depth of the subgrade result in the subgrade deflection and subgrade permanent
deformation respectively. Such an approach, however, requires that the exact subgrade
stiffness (and hence vertical strain) profile is known. This detailed information is rarely
available for calibration data sets generated from controlled testing, let alone pavement
design cases.

Subgrade deflection is therefore used as the critical parameter for subgrade permanent
deformation. The resilient modulus values listed in Table 2 should be used in combination
with a semi-infinite subgrade to calculate the subgrade deflection between wheel-loads for
design purposes.

A range of S-N type subgrade permanent deformation models were calibrated as part of the
SANRAL SAPDM project, providing for subgrade deformation levels from 1 to 20 mm. It was
also found that a separation had to be made between the models for deep sand subgrades
typical of KwaZulu-Natal and the Western Cape and gravel subgrades. The S-N models for
10 and 20 mm subgrade permanent deformation included in SAMDM2011 are illustrated in
Figure 5.

The scatter in the data plotted in Figure 5 is wide but such variation is an inherent
characteristic of pavement performance. The distinction between sandy and gravel
subgrades as well as subgrade capacities of a particular subgrade type at different levels of
subgrade deflection are, however, clear for a given level of subgrade permanent
deformation. Although there is significant overlap in the data ranges for 10 and 20 mm
permanent deformation, the risk profiles are different for the two levels of permanent
deformation for a given traffic demand.

S-N data S-N data

10 mm permanent def ormation 20 mm permanent def ormation

1.0E+09 1.0E+09
Subgrade capacity (repetitions)

Subgrade capacity (repetitions)

1.0E+08 1.0E+08

1.0E+07 Gravel 1.0E+07 Gravel

Sand Sand

1.0E+06 1.0E+06

1.0E+05 1.0E+05
0 0.5 1 1.5 0 0.5 1 1.5

MDD subgrade deflection (mm) MDD subgrade def lection (mm)

Figure 5. Subgrade permanent deformation damage models

3.4.2 Plastic strain damage models for unbound granular layers

Plastic strain data from repeated load tri-axial tests by Theyse (2008) were used for the
development of a series of S-N type plastic strain damage models under the SANRAL SAPDM
project. It was found the volumetric density and degree of saturation of the material had to
be incorporated in the formulation of the critical parameter in addition to the stress ratio.
The stress ratio is based on effective stress according to Equation (1). A better correlation
was found between the plastic strain laboratory results and the stress ratio formulated in
terms of major principal stress than the stress ratio formulated in terms of deviator stress.
-11-
 10th CONFERENCE ON ASPHALT PAVEMENTS FOR SOUTHERN AFRICA

Implementation of the stress ratio in a pavement system requires that the stress ratio be
calculated from the effective vertical and horizontal stress under the wheel instead of the
major and minor principal stress. The reason for this being that the overburden pressure
and residual compaction stress are uniform stress fields that affect the major and minor
principal stress but do not contribute to the vertical permanent deformation of the
pavement layer, only the vertical stress component of the external wheel-load does.

σ v′
SR =
σ h′ tan 2  45° + φ 2  + 2C tan  45° + φ 2 
Eq. (1)
   
where SR = stress ratio
σ’v = effective vertical stress
σ’h = effective horizontal stress
φ = friction angle of the material
C = cohesion of the material (kPa)

Recommended Mohr-Coulomb shear strength parameters for unbound granular base and
subbase materials are summarised in Table 4 for preliminary design. It is strongly
recommended that shear strength tests be performed on the materials selected for the
construction of a road and that the design should be finalised using project specific material
properties.

Table 4 Recommended shear strength properties for unbound granular material to

be used during preliminary design
Application Material Saturation level Recommended shear strength
parameters
Cohesion (kPa) Friction angle (º)
Base G1 20 % (dry) 90 – 130 53 – 57
50 % (moderate) 75 – 100 51 – 55
80 % (wet) 50 – 75 50 – 53
G2 20 % (dry) 100 – 125 54
50 % (moderate) 50 52
80 % (wet) 45 50
G3 20 % (dry) 75 51
50 % (moderate) 40 51
80 % (wet) 20 50
G4 20 % (dry) 75 51
50 % (moderate) 40 47
80 % (wet) 20 45
Subbase - coarse G5/6 20 % (dry) 100 – 125 45 – 49
material (BLS x P425 < 50 % (moderate) 50 – 100 41 – 45
170) 80 % (wet) 10 – 50 39 – 42
G5/6 20 % (dry) 225 – 275 45 – 49
(BLS x P425 > 50 % (moderate) 50 – 100 41 – 44
170) 80 % (wet) 25 – 35 31 – 33
Subbase - fine G5/6 20 % (dry) 125 – 250 43 – 45
material (BLS x P425 < 50 % (moderate) 40 – 50 43 – 45
100) 80 % (wet) 10 – 25 40 – 43
Notes: Course material – grading modulus (GM) 1,7 to 2,3; maximum particle size 26,5 to 37,5 mm;
Fine material – grading modulus (GM) 1,5 to 1,6; maximum particle size < 13,2 mm;
BLS – Bar Linear Shrinkage;
P425 – Percentage passing 0,425 mm sieve

-12-
 10th CONFERENCE ON ASPHALT PAVEMENTS FOR SOUTHERN AFRICA

The damage models for unbound base and subbase layers makes provision for different
levels of plastic strain ranging from 1 to 19 % plastic strain. The appropriate level of plastic
strain depends on the terminal deformation level selected by the designer i.e. a terminal
condition of 20 mm permanent deformation equates to 13 % plastic strain for a 150 mm
thick layer and 20 % plastic strain for a 100 mm thick layer. Figure 6 illustrates the S-N
damage model for 19 % plastic strain as a function of the Stress Ratio (SR), saturation level
(S) and volumetric density (VD) of the material.

S-N model
19 % plastic strain
9.0

7.0
log (N)

Observed

Predicted

5.0

3.0
0 0.2 0.4 0.6 0.8 1
SR*S/VD

Figure 6. Plastic strain damage model for 19 % plastic strain of unbound base
and subbase layers

4. VALIDATION OF SAMDM2011

The effective stress analysis and damage models presented in the preceding sections were
coded into a software package for mechanistic-empirical pavement design (me-PADS)
developed at the CSIR. As indicated earlier, the revised models are implemented in a
classical ME-design approach meaning that an estimate of the structural capacity of each
pavement layer is made independently but with due consideration of the interaction
between layers in the pavement system. The revised models were evaluated at the hand of
the sensitivity of the models to changes in the resilient response characteristics of the
pavement materials and the ability of the models to simulate permanent deformation
damage recorded during HVS testing.

4.1. Assessment of the sensitivity of the method to input variation

The critical review of SAMDM1996 by Jooste (2004) focused on the sensitivity of the
method, specifically the models for unbound materials, to changes in the resilient response
parameters of the unbound layers and their supporting layers. Jooste established a base
case for analysis consisting of the pavement structure summarised in Table 5. Jooste (2004)
analysed four variations of the base case of which two related to variation in the resilient
properties of the pavement layers and two related to changes in the shear strength
properties of the G1 base layer. The modified resilient response parameters are highlighted
in Table 5 for the first two variations analysed by Jooste.
-13-
 10th CONFERENCE ON ASPHALT PAVEMENTS FOR SOUTHERN AFRICA

The base case and first two variations were recalculated using the proposed new design
models and the same resilient and shear strength properties used by Jooste (2004). The
pavement was analysed using a 20 kN dual-wheel load at 750 kPa contact pressure similar
to Jooste. Table 6 summarises the base layer capacity reported by Jooste and the results
from the revised models. The revised method is far less sensitive to variation in the resilient
input parameters than the 1996 method for the particular case investigated.

Table 5 Pavement structure used by Jooste (2004) to perform a sensitivity analysis of

SAMDM1996
Layer Material Base case Variation 1 Variation 2
thickness Description Resilient Poisson’s Resilient Poisson’s Resilient Poisson’s
modulus ratio modulus ratio modulus ratio
(MPa) (MPa) (MPa)
40 mm Asphalt overlay 2500 0,40 2500 0,40 2500 0,40
150 mm G1 overlay 650 0,35 650 0,35 650 0,38
250 mm Old cemented layer 450 0,35 400 0,35 450 0,35
150 mm Selected subgrade 140 0,35 140 0,35 140 0,35
Semi-infinite Sandy subgrade 90 0,35 90 0,35 90 0,35

Table 6 Comparison of the G1 base layer capacity according to SAMDM1996 and

SAMDM2011
Method Parameter Base case Variation 1 Variation 2
SAMDM1996 Base capacity 10,7 x 106 4,2 x 106 54,5 x 106
(from Jooste, 2004) Deviation from base 0% -25 % + 168 %
SAMDM2011 Base capacity 26 x 106 25 x 106 27 x 106
(me-PADS) Deviation from base 0% -3 % +5 %

4.2. Simulation of limited Accelerated Pavement Testing (APT) results

A full depth granular pavement consisting of a crushed stone base on top of a crushed stone
subbase with a deep sandy subgrade was tested on the National Road 7 (N7) near Cape
Town during 2001-2002 with the HVS. Figure 7 shows the pavement structure on the fast
lane of the southbound carriageway that was tested while Table 7 summarises the test
programmes for two HVS tests done on this pavement.

50 mm
Novachip and HMA
15 mm old seals
200 mm G2
Crushed stone
base
150 mm G3
(crushed stone)
subbase

Sand subgrade

Figure 7. Pavement structure tested by the HVS on the N7 near Cape Town

-14-
 10th CONFERENCE ON ASPHALT PAVEMENTS FOR SOUTHERN AFRICA

Table 7 HVS test programme for the crushed stone pavement on the N7
Trafficking wheel-load and tyre pressure
Dry Conditions Wet conditions
40 kN, 620 kPa 60 kN, 620 kPa 80 kN, 850 kPa 40 kN, 620 kPa
Test 417A5 0 – 55 000 55 000 – 147 444 147 444 – 553 644 -
Section 418A5 0 – 616 806 - - 616 806 – 1 143 423

Deflections and permanent deformation data were recorded at three locations (MDDs 4, 8
and 12) on the HVS tests. The resilient moduli back-calculated for each of the pavement
layers using the depth-deflections recorded at regular intervals during the HVS tests were
used as common input to SAMDM1996 and SAMDM2011 to estimate the structural capacity
of each pavement layer to a 20 mm terminal rut condition. The evolution of permanent
deformation during the two HVS tests was subsequently modelled using a linear recursive
method based on Miner’s law.

Figure 8 shows examples of the observed and modelled base layer permanent deformation
for HVS tests 417A5 and 418A5 using SAMDM1996 and SAMDM2011. In the case of test
417A5 during which the wheel-load was increased from 40 to 60 and then 80 kN the
SAMDM1996 models did not respond well and the base layer permanent deformation only
started to increase slightly under the 80 kN load. Although the SAMDM2011 models do not
simulate the initial bedding-in that occurred, the rate of permanent deformation increased
for both the 60 and 80 kN load portions of test 417A5. In the case of test 418A5 during
which the section was soaked with water from about 600 000 repetitions onwards, both
models responded equally and realistically to the increase in saturation levels of the base
layer.

HVS test 417A5 - MDD 12 HVS test 417A5 - MDD 12

2 2
1.8 1.8
Permanent Deformation (mm)

Permanent Deformation (mm)

1.6 1.6
1.4 1.4
1.2 1.2 HVS 417A5
HVS417A5
1 1
0.8 0.8 SAPDM2011
SAPDM1996
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 200000 400000 600000 0 200000 400000 600000

Repetitions Repetitions

(a) SAMDM1996 – 417A5 (b) SAMDM2011 – 417A5

HVS test 418A5 - MDD 4 HVS test 418A5 - MDD 4
7 8

6 7
Permanent Deformation (mm)

Permanent Deformation (mm)

6
5
5
4 HVS 418A5
HVS 418A5
4
3
SAPDM1996 3 SAPDM2011
2
2
1
1

0 0
0 200000 400000 600000 800000 1000000 1200000 0 200000 400000 600000 800000 1000000 1200000
Repetitions Repetitions

(c) SAMDM1996 – 418A5 (d) SAMDM2011 – 418A5

Figure 8. Base layer permanent deformation for HVS tests 417A5 and 418A5
-15-
 10th CONFERENCE ON ASPHALT PAVEMENTS FOR SOUTHERN AFRICA

The observed and modelled permanent deformation at the top of the base (20 mm),
subbase (255 mm) and subgrade (405 mm) are shown in Figure 9 for the two HVS tests. In
addition to the poor modelling accuracy of SAMDM1996 for the base layer of test 417A5,
the subgrade permanent deformation is also under-estimated significantly resulting in
almost no permanent deformation being modelled for the duration of the test. Although
the pavement structure permanent deformation is over-estimated by SAMDM2011 for test
417A5, the model response to increasing wheel-load is far more realistic than that of
SAMDM1996. In the case of test 418A5 no permanent deformation other than that
originating from the base layer during the wet portion of the test is predicted by
SAMDM1996. SAMDM2011 shows an accumulation of permanent deformation at all levels
in the pavement structure during the dry and wet portions of the tests that agrees with the
observed permanent deformation behaviour.

HVS test 417A5 - MDD 12 HVS test 417A5 - MDD 12

7 7
Permanent Deformation (mm)

Permanent Deformation (mm)

6 6

5 MDD 20 mm MDD 20 mm
5
MDD 255 mm MDD 255 mm
4 4
MDD 405 mm MDD 405 mm
3 3
Model 20 mm Model 20 mm
2 Model 255 mm 2 Model 255 mm
1 Model 405 mm
1 Model 405 mm

0
0
0 200000 400000 600000
0 200000 400000 600000
Repetitions
Repetitions

(a) SAMDM1996 – 417A5 (b) SAMDM2011 – 417A5

HVS test 418A5 - MDD 4 HVS test 418A5 - MDD 4
12 12

10 10
Permanent Deformation (mm)

Permanent Deformation (mm)

8 MDD 20 mm 8 MDD 20 mm

MDD 255 mm MDD 255 mm

6 6
MDD 405 mm MDD 405 mm

4 Model 20 mm Model 20 mm
4
Model 255 mm Model 255 mm
2 Model 405 mm 2 Model 405 mm

0 0
0 500000 1000000 1500000 0 500000 1000000 1500000
Repetitions Repetitions

(c) SAMDM1996 – 418A5 (d) SAMDM2011 – 418A5

Figure 9. Pavement structure permanent deformation evolution for HVS tests
417A5 and 418A5

Table 8 provides a summary of the final deformation results recorded at three locations
(MDD4, MDD8 and MDD12) on HVS test sections 417A5 and 418A5. In general,
SAMDM1996 underestimates the base and total pavement deformation with the error in
some cases exceeding 80 % and approaching 100 %. In all cases, the SAMDM1996 predicts
almost no deformation from the subgrade which does not correspond with the observed
data. Given this underestimation of permanent deformation, the structural capacity of the
pavement will be overestimated significantly resulting in very high design risk. The
sensitivity of SAMDM1996 to variations in the input variables is highlighted by the results
for HVS test 417A5 with the base deformation being overestimated at MDD4 and
underestimated at the other locations. Although the base and pavement deformation
-16-
 10th CONFERENCE ON ASPHALT PAVEMENTS FOR SOUTHERN AFRICA

estimated by SAMDM2011 is not very accurate, the results from SAMDM2011 are
consistent and more accurate than the SAMDM1996 results.

Table 8 Summary of results for the crushed stone pavement on the N7

Deformation at the end of the each test
HVS SAMDM 1996 SAMDM 2011
Test Base MDD4 1.6 2.2 (+37.5%) 0.9 (-43.8%)
417A5 deformation MDD8 0.9 0.1 (-88.9%) 0.8 (-11.1 %)
MDD12 1.1 0.2 (-81.8%) 0.8 (-27.3%)
Pavement MDD4 5.4 2.2 (-59.3%) 4.8 (-11.1%)
deformation MDD8 5.9 0.2 (-96.6%) 9.2 (+55.9%)
MDD12 5.4 0.4 (-92.6%) 8.0 (+48.1%)
Test Base MDD4 6.2 4.8 (-22.6%) 7.5 (+21.0%)
418A5 deformation MDD8 3.9 3.2 (-17.9%) 6.6 (+69.2%)
MDD12 5.2 4.4 (-15.4%) 7.5 (+44.2%)
Pavement MDD4 11.7 4.8 (-58.9%) 14.8 (+26.5%)
deformation MDD8 8.0 3.3 (-58.8%) 13.8 (+72.5%)
MDD12 9.6 4.5 (-53.1%) 14.0 (+45.8%)

5. CONCLUSIONS AND RECOMMENDATIONS

In addition to safeguarding against inappropriate (over- and under-designed) pavement

designs, pavement design methods and specifically mechanistic-empirical design methods
must provide realistic pavement performance information for long-term financial planning.
In the process of modelling the pavement performance, the design method should not be
overly sensitive to variation in secondary design inputs but should respond realistically to
changes in the primary inputs that are known to govern the performance of pavements.

Jooste (2004) illustrated that the 1996 version of the South African Mechanistic-Empirical
Design Method for flexible pavements (SAMDM1996) was overly sensitive to small changes
in the resilient response parameters of pavement layers. As shown in this paper, the
simulation of Heavy Vehicle Simulator (HVS) permanent deformation data for full-depth
granular pavements also revealed that the models for unbound granular layers were not
sufficiently sensitive to changes in loading conditions.

The paper presents revised permanent deformation damage models for unbound granular
pavement layers. The internationally accepted vertical subgrade strain design model was
found to be inadequate and replaced with a model based on the total subgrade deflection.
The factor of safety model for unbound granular base and subbase layers was replaced with
a stress ratio model with the distinction that the stress ratio is calculated from effective
stress conditions, not only the stress associated with the external wheel-load. The effective
stress includes the vertical overburden pressure, residual compaction stress, suction
pressure and the stress resulting from the external wheel-load.

The revised method shows considerably less sensitivity to changes in the resilient response
parameters of pavement layers compared to the SAMDM1996 models with appropriate
responsiveness to changes in input variables such as wheel-load, density and saturation
levels as well as material shear strength. Using the revised models, it was possible to

-17-
 10th CONFERENCE ON ASPHALT PAVEMENTS FOR SOUTHERN AFRICA

simulate the development of the permanent deformation of a full-depth granular pavement

with reasonable accuracy.

Although some of the known problems associated with the design models for unbound
granular layers have been alleviated by the proposed revised models, the models for other
material types are still under revision as part of the SANRAL SAPDM project to ensure that
all materials are assessed according to true performance potential in the South African
Mechanistic-Empirical Pavement Design Method.

ACKNOWLEDGMENT

The revision of the South African Mechanistic-Empirical Design Method for flexible
pavements presented in this paper and the work that preceded this revision was funded
predominantly by the CSIR and SANRAL. Both organisations are acknowledged for their
support.

REFERENCES

CSRA (Committee of State Road Authorities). 1985. “TRH14: 1985 Guidelines for Road
Construction Materials”. Department of Transport, Pretoria, South Africa.

De Beer, M., Kannemeyer, L., Fisher, C., 1999. “Towards Improved Mechanistic Design of
Thin Asphalt Layer Surfacings Based on Actual Tyre/Pavement Contact Stress - In -
Motion (Sim) Data in South Africa.” Proceedings of the 7th Conference on Asphalt
Pavements for Southern Africa. Victoria Falls, Zimbabwe.

Dehlen G. L., 1958. “The Residual Lateral Stress in a Soil and its Effect on the Load-Bearing
Capacity of the Soil”. M Sc (Eng) thesis. University of the Witwatersrand, Johannesburg.

Fredlund D. G., 1985. “Soil mechanics principles that embrace unsaturated soils”.
Proceedings of the 11th International Conference on Soil Mechanics and Foundation
Engineering. San Francisco.

Huurman M. 1997. Permanent deformation in concrete block pavements. Delft: Technical

University of Delft. Ph D thesis.

Jooste F., 2004. “Re-Evaluation of Some Aspects of the Mechanistic-Empirical Design

Approach.” Proceedings of the 8th Conference on Asphalt Pavements for Southern Africa.
Sun City, South Africa. [CD]

Jordaan G. J., 1993. Users Manual for the South African Mechanistic Pavement
Rehabilitation Design Method. Department of Transport, Pretoria. (Interim Report
IR 91/242/2)

Maina, J. W. and K . Matsui. 2004. Developing Software for Elastic Analysis of Pavement
Structure Responses to Vertical and Horizontal Surface Loadings. Transportation Research
Records, No. 1896, pp. 107-118.

Maree J. H. 1978. Ontwerpparameters vir klipslag in plaveisels (Design parameters for

crushed stone in pavements). Pretoria: University of Pretoria. M Eng thesis.
-18-
 10th CONFERENCE ON ASPHALT PAVEMENTS FOR SOUTHERN AFRICA

Maree J. H and Freeme C. R., 1981. The Mechanistic Design Method Used to Evaluate the
Pavement Structures in the Catalogue of the Draft TRH4 1980. Technical Report RP/2/81,
National Institute for Transport and Road Research, CSIR, South Africa.

Paterson W. D. O. and Maree J. H., 1978. An Interim Mechanistic Procedure for the
Structural Design of Asphalt Pavements. Technical Report RP/5/78, National Institute for
Transport and Road Research, CSIR, South Africa.

South African Pavement Design Method. [Online]. Available: www.SAPDM.co.za [7 March

2011].

Theyse H. L., 2001. “The Development of Mechanistic-Empirical Permanent Subgrade

Deformation Models from Heavy Vehicle Simulator Data”. M Eng thesis. Rand Afrikaans
University, Johannesburg, South Africa.

Theyse H. L., 2008. “A Mechanistic-Empirical Design Model for Unbound Granular Pavement
Layers”. PhD thesis. University of Johannesburg, Johannesburg, South Africa.

Theyse H. L., 2009. “The Classical South African Mechanistic-Empirical Design Method”.
Flexible Pavement Design Course. Pavement Modelling Corporation, Pretoria, South Africa.

Theyse H. L., de Beer M and Rust F. C., 1996. “Overview of the South African Mechanistic
Pavement Design Method”. Transportation Research Record 1539. Transportation
Research Board, Washington D.C.

Uzan J., 1985. “Characterization of Granular Material.” Transportation Research Record

1022. Transportation Research Board. Washington D. C. pp 54.

Van Vuuren D. J, Otte E and Paterson W. D. O., 1974. “The Structural Design of Flexible
Pavements in South Africa”. Proceedings of the 2nd Conference on Asphalt Pavements in
South Africa. Durban, South Africa.

Vanapalli S. K., Fredlund, D. G. and Pufahl D. E., 1996a. “The relationship between the soil-
water characteristic curve and the unsaturated shear strength of a compacted glacial till”.
Geotechnical Testing Journal.

Vanapalli S. K., Fredlund, D. G. and Pufahl D. E., 1996b. “Model for the prediction of shear
strength with respect to soil suction”. Canadian Geotechnical Journal No 33.

Vanapalli, S. K. and Fredlund, D. G., 1999. “Empirical procedures to predict the shear
strength of unsaturated soils”. Proceedings of the 11th Asian Regional Conference on Soils
Mechanics and Geotechnical Engineering. Balkema, Rotterdam.

Vanapalli, S. K. and Fredlund, D. G., 2000. “Comparison of empirical procedures to predict

the shear strength of unsaturated soils using the soil-water characteristic curve”. American
Society of Civil Engineers, Special Publication, No. 99.
-19-
 10th CONFERENCE ON ASPHALT PAVEMENTS FOR SOUTHERN AFRICA

Van Niekerk A. A., Houben L. J. M. and Molenaar A. A. A. 1998. “Estimation of the

mechanical behaviour of unbound road building materials from physical material
properties”. Proceedings of the 5th International Conference on the Bearing Capacity of
Roads and Airfields. Trondheim, Norway.

Walker R. N., Paterson W. D. O., Freeme C. R. and Marais C. P., 1977. “The South African
Mechanistic Pavement Design Procedure”. Proceedings of the 4th International Conference
on the Structural Design of Asphalt Pavements. University of Michigan, Ann Arbor,
Michigan.

KEY WORDS

Mechanistic-Empirical design, unbound granular layers, effective stress analysis, damage

models, linear recursive simulation.

-20-

View publication stats