Farhad Salour Doctoral Thesis

Figure 13.
Schematic overview of a Falling Weight Deflectometer and its components 
(Doré and Zubeck, 2008). 
The magnitude and the shape of the deflection basin under a given load pulse can be 
used as a representative for the material properties of pavement system layers and their 
overall stiffness. It is generally accepted that deflections at some distance from the plate 
load centre may be correlated to the deformation at a corresponding depth beneath the 
road surface (Mork, 1990). This denotes that deflections close to the centre of the load 
plate reflect the material properties of the pavement layers, while deflections sufficiently 
away from the load centre only reflect the subgrade properties. 
Generally, the data from FWD measurements is analysed at two different levels: the 
first level of analysis includes deflection basin analysis or deflection basin shape 
indicators. These indices are good indicators for a fast and preliminary assessment of 
the pavement structure and to sort out the weak sections along the road. The seasonal 
climatic effects on pavement overall response and material stiffness can generally be 
perceived using the deflection bowl and deflection basin indices. For instance, in 
pavements exposed to freeze and thaw cycles, depending on the thawing progression in 
the pavement structure, the deflection basin changes in both the shape and the size 
(Simonsen, 1999). In early spring, the pavement is in a fully frozen state, the deflection 
basin is narrow and the measured deflections are very small. As thawing penetrates the 
pavement structure, the deflection basin widens and increases in depth. The maximum 
basin deflection is observed at the end of the spring-thaw period or within a few days 
after the thawing is completed, in which the subgrade is usually in a very wet condition. 
The second level of FWD data analysis includes estimation of the stiffness or resilient 
modulus of the pavement layers through backcalculation procedures. This is a more 
advanced procedure that is usually used in the mechanistic based evaluation of 
pavement structures. This technique applies the multi-layer elastic theory and models. 

---

26 
In backcalculation, the FWD impact load, pavement structure layer thicknesses, material 
Poisson’s ratio and the measured surface deflections are assigned as the inputs of the 
procedure. Using the multi-layer elastic theory, a set of pavement layer moduli that 
would theoretically produce the measured deflection basin with a certain level of errors 
is determined. The backcalculation is an iterative procedure from the simple equivalent 
thickness approach to more sophisticated approaches that apply least-square 
optimization methods. 
It should be noted that special concerns should be taken into account when performing 
backcalculation on pavements that are exposed to significant environmental variations 
such as moisture content variations due to groundwater fluctuations or pavements that 
are exposed to frost-thaw actions. Conduction quality backcalculation regarding these 
pavements usually requires substantial information on depth to the frozen layers, the 
temperature gradient of the asphalt concrete layer, moisture distribution within the 
unbound layers and depth the groundwater and/or bedrock. For this purpose, 
pavement instrumentation can be taken as collecting data on the climatic condition of 
pavement structures and their induced stresses. Pavement instrumentation and intensive 
data collection can be a valuable enhancement for a better understanding of factors 
influencing pavement behaviour and response and both their short term and long term 
impacts. However, intensive data collection, management and analyses are usually 
expensive and time consuming and therefore are mainly limited to research purposes. 
Stress dependent behaviour of pavement unbound material from field measurements 
The nonlinear stress dependent behaviour of pavement unbound materials has 
traditionally been determined from laboratory-based studies using RLT testing due to its 
manageability and relative simplicity as well as time and cost efficiency. Even though 
RLT testing is designed to simulate the internal in situ loading and environmental 
condition of pavement materials and subgrade soils, it might still not be fully capable of 
reproducing the internal structure, overburden pressure and traffic induced stress states 
(Karasahin et al., 1993; Ke et al., 2000). 
Non-destructive testing such as FWD and Seismic Pavement Analyser (SPA) 
measurements are probably the most effective methods to capture the in situ behaviour 
and response of pavement materials and structures. FWD measurements and 
backcalculation of pavement layer moduli are in particular cost efficient and 
well-recognised methods for structural evaluation of pavement systems that are widely 
used by the road authorities. The ability of the FWD to stimulate traffic loading, its 
mobility and capacity to collect large amounts of data at network level have gained 
interest among the pavement community for more advanced analysis of FWD data. 
This is along the path towards development of mechanistic-empirical design framework 
for pavement systems that requires improved knowledge about the materials, climatic 
factors, geometry and traffic on pavements structural response and performance. 

---

27 
Even though most pavement unbound materials exhibit nonlinear stress dependent 
behaviour, they are traditionally treated as linear elastic materials in backcalculation of 
FWD data. A majority of the backcalculation programs apply the simplified multilayer 
linear elastic theory to backcalculate pavement layer stiffness. However, realistic 
analyses of deflection data may require more advanced backcalculation techniques that 
account for material nonlinearity. 
Over the years, several studies have been conducted to evaluate the nonlinearity of 
pavement material from deflection basins. Uzan (2004) applied both linear and 
nonlinear procedures to analyse the FWD data obtained from an instrumented test site 
in Hanover, New Hampshire. In his study, the backcalculation of the data exhibited 
superior fit to the deflection bowl when the nonlinear approach was implemented in 
comparison with the linear approach. This was in agreement with the stress and strain 
measurements from the site instrumentation that also confirmed the nonlinear response 
of the pavement material. 
Meshkani et al. (2003) investigated the feasibility of backcalculating flexible pavement 
layer nonlinear parameters from FWD and SPA testing. They used four hypothetical 
pavement sections with nonlinear base and subgrade materials to study the feasibility 
and accuracy of backcalculating the nonlinear parameters from deflection bowls. 
Although nonlinear parameters for thinner pavement structures could be estimated, 
they concluded that in many cases the deflection bowl did not provide sufficient 
information to reliably backcalculate the nonlinear parameters. 
Steven et al. (2007) modelled a thin-surfaced flexible pavement structure that was 
instrumented with strain gauges and pressure cells using the Finite Element (FE) 
method in ABAQUS computer code. In their FE model, both the granular layer 
(anisotropic) and subgrade (isotropic) were modelled using the generalized nonlinear 
constitutive model as expressed in Equation 3. The material model parameters used in 
the FE model were obtained from RLT testing. They calibrated their model so that the 
measured and calculated stresses and strains levels agreed within the acceptable 
tolerance. The computed surface deflection from the FE model and the surface 
deflections measured FWD at the test section showed an excellent match for all the 
FWD load levels. They concluded that a fully nonlinear pavement model can realistically 
represent the response of the pavement structure during the FWD tests. A similar 
conclusion was also made in a study conducted by Uzan (1994). 
In 2005, Li and Baus conducted a full-scale static and cyclic plate load test to investigate 
the mechanical properties of unbound granular materials and the effect of moisture 
content. Based on the plate load measurements, they developed a procedure to 
backcalculate the material nonlinear parameters of the 


k
model that was 
implemented in the algorithm. 

---

28