Abstract
The work presented in this thesis is aimed at improving the accuracy of one of the components of a flexible
pavement design procedure commonly used in South Africa, namely the South African Mechanistic-
Empirical Design Method. This is achieved through the development of a new design approach and
permanent deformation model for the pavement subgrade. The new distress model for the pavement
subgrade was developed from a comprehensive Accelerated Pavement Testing (APT) database on
subgrade behaviour and permanent deformation that was generated by a fleet of Heavy Vehicle
Simulators (HVSs) over 20 years of testing in South Africa.
A literature review of the origin of the current subgrade design model that is used by the South African
Mechanistic-Empirical Design Method revealed that that model is based on very little actual subgrade
performance data. The model was also developed from the AASHO road test data and adjusted for South
African conditions, based on general observations of subgrade behaviour without any calibration.
Previous researchers have illustrated the potential of using Heavy Vehicle Simulator data to develop
structural pavement design models and it was decided to apply a similar process to the permanent
deformation of the pavement subgrade.
The present investigation consisted of two components, namely, the evaluation of the resilient and the
permanent deformation response of the pavement subgrade, the emphasis in this thesis being more on
the permanent deformation response. A general, multi-dimensional empirical model was formulated for
the permanent subgrade deformation and the characteristics of the model investigated based on
previously published permanent deformation data and mathematical assessment.
A set of 35 HVS tests for which suitable data were available was identified and additional field and
laboratory tests were done on selected sites to improve the classification of the subgrade materials at
these sites. Standard procedures were developed to present the pavement, instrumentation and load
sequence data of each HVS test. A process for doing the initial analysis of the deflection and permanent
MDD displacement data and presenting the data was also developed.
In terms of the resilient response of the subgrade, it was shown that the vertical depth deflection and
vertical strain could be modelled accurately if an appropriate set of resilient modulus values was selected
for the pavement layers. A detailed investigation of the resilient response of selected HVS sections did,
however, clearly illustrate the stress-dependent behaviour of subgrade material, resulting in resilient
modulus values being determined, which were well outside the range that would normally be expected for
natural gravel subgrade material. This research needs to be continued to enable the development of a
comprehensive set of stress-dependent resilient modulus models for South African subgrade materials.
The selection of an appropriate critical parameter that can be used as a predictor of permanent subgrade
deformation was done by an investigation of the relationship between potential critical parameters and
several permanent deformation parameters. It was found that the subgrade elastic deflection showed the
best correlation with the subgrade bearing capacity (the number of load repetitions that can be sustained
before a terminal rut condition is reached). The vertical subgrade strain that is currently used in the South
African Mechanistic-Empirical Design Method in fact correlates poorly with subgrade bearing capacity and
has to be replaced with subgrade elastic deflection.
A set of subgrade bearing capacity or design models was developed for different levels of permanent
subgrade deformation. These models are referred to as S-N models and form contour lines on the general
permanent deformation model that was formulated. The subgrade design model accommodates loading
conditions ranging from a 40 kN dual wheel load to a 100 kN dual wheel load as well as subgrade
materials ranging from a material quality one class better than that which would normally be used for a
subgrade, to the lowest possible material class. The model is therefore very flexible in terms of its
application.
Prof. P. Pretorius