You are on page 1of 9

Advanced Materials Research Online: 2013-08-16

ISSN: 1662-8985, Vol. 723, pp 196-203


doi:10.4028/www.scientific.net/AMR.723.196
© 2013 Trans Tech Publications, Switzerland

Static and Dynamic Backcalculation Analyses of an inverted pavement


structure

James Maina1, a, Wynand JvdM Steyn2,b , Emile B van Wyk2,c


and Frans le Roux2,d
1
CSIR Built Environment, Pretoria, South Africa
2
Department of Civil Engineering, University of Pretoria, Lynnwood road, Hatfield, South Africa
a
jmaina@csir.co.za, bwynand.steyn@up.ac.za, cvanwyk.eb@gmail.com, dfwleroux@hotmail.com

Keywords: backcalculation, inverted pavement, FWD.

Abstract. A crucial part of any maintenance strategy is an intricate understanding of the material
characteristics of the pavement, so that the current level of damage may be accurately assessed and an
appropriate plan implemented. Advances in the precision to which these parameters can be
determined, as well as improvements in how these results are interpreted under varying conditions of
measurement and analysis, are essential in the effective execution of a maintenance strategy. Results
from Falling Weight Deflectometer (FWD), which is a Non-Destructive Testing (NDT) device, can
be used to predict elastic modulus of any layer by comparing measured deflection data to calculated
values through an iterative process referred to as back-calculation. This paper presents a comparison
between static and dynamic back-calculation procedures, specifically with regard to typical South
African inverted pavements. The analysis indicates a dynamic analysis provides results of greater
accuracy than a static analysis, although the effect of the difference requires further investigation.

Introduction
Maintaining serviceable conditions across any given road network is a fundamentally important
aspect of any society, as the efficient movement of goods and services along routes composing a
network results in positive growth of infrastructure that serves to create a better quality of life for all.
The development and implementation of a maintenance strategy is, however, a costly undertaking,
with countries all over the world requiring larger portions of their budgets to deal with the incipient
deterioration across networks as traffic volumes increase.
In order to accurately assess the structural condition of a pavement, individual layer characteristics
need to be obtained, preferably without disruption of the structure. The primary parameter of interest
is the stiffness or elastic modulus of each layer from which stresses and strains at any point in the
structure can be determined for a given load. The problem of obtaining layer moduli is complex,
requiring the use of an inverse mapping strategy to find the unknown parameters from initial estimates
or seed values in an iterative process known as back-calculation. There are a number of different
approaches for the back-calculation that often yield significantly digressive results, and it is for this
reason, as well as for the fact that only recently the computational power to solve such complex
problems in a realistic time has been available, that there is a substantial amount of work that can be
done in order to refine the mathematical strategy and to determine how the results apply to specific
contexts. In South Africa, considerable research continues to be conducted through the CSIR and it is
in this context that this research project was undertaken.

Background
Inverted Pavements. The inverted pavement system is a fairly unique concept and highly regarded as
a feasible method of pavement construction, both in terms of cost efficiency and structural
performance (1). According to (2), “in an inverted pavement, the stiffness of one of the lower
supporting layers is greater than the stiffness of the upper structural layers”. A notable feature of the

All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of Trans
Tech Publications, www.ttp.net. (ID: 130.203.136.75, Pennsylvania State University, University Park, USA-18/05/16,01:51:56)
Advanced Materials Research Vol. 723 197

inverted pavement is the use of high quality granular materials as a base layer, placed on a lightly
cemented subbase layer (3). The base layer is referred to as an Unbound Aggregate Base (UAB) layer
and, as an overlay on top of a cemented subbase, can prevent the spread of shrinkage cracks in the
subbase to the surface (4).
Non-Destructive Testing. Non-Destructive Testing (NDT) is a tool utilized to determine
properties of the various layers of a pavement system without altering the in-situ properties of the
materials in question (5). These tests can be broadly categorized into devices which measure
deflections, and those which measure surface wave propagation (6). The Falling Weight
Deflectometer (FWD) (Fig. 1) falls into the former category. Devices in the latter category are based
on the generation and detection of stress waves in the pavement structure, including devices such as
the Portable Seismic Pavement Analyzer (PSPA) (5). First introduced in France in the 1960s, the
FWD has since gained widespread acceptance as one of the fastest and most effective devices for the
evaluation of flexible roads (7). Although no stranger to criticism, the FWD is currently the most
widely used device for the determination of layer properties by deflection measurement, and also the
most studied form of non-destructive pavement evaluation (9).

FIGURE 1: Schematic diagram of FWD (8).

The FWD test is a form of plate bearing test, whereby an impulse load is delivered to the pavement
on a circular plate at the desired location and the resulting deflections are measured and recorded. The
magnitude of the applied force is recorded by a load cell, and the pavement response recorded by
sensors (geophones) such as seismometers, velocity transducers or accelerometers (10). Readings
from these sensors can be used to compute the resulting vertical deflections (11). The sensors are set
to record either the resulting maximum deflections at positions radially outward from the location of
the applied load, or alternatively to record time series deflections (5). This choice of deflections
measurement is determined by whether a static or dynamic analysis is to be performed on the
pavement structure.

Methodology
Back-calculation. Back-calculation can be defined as “an optimization process performed to obtain
inverse mapping of a known relation established by discrete or continuous data points.” (6).
Back-calculation problems thus consist of a forward calculation, whereby a mathematical model is
used to theoretically determine strains and displacements using initially assumed moduli, and
comparing these with measured displacements to back-calculate actual layer moduli through a series
of iterations (5). The back-calculation procedure can broadly be performed in three different ways,
varying in complexity as simplified models are adapted to closer conform to real conditions.
198 Innovation and Sustainable Technology in Road and Airfield Pavement

The first, an adaptive approach, only makes use of peak deflections at selected points without
directly employing a pavement response model (6). The measured deflections are matched with
similar deflections in a database making this method purely empirical, and the lack of a sound
theoretical basis does not leave much room for development or improvement (12). The remaining two
methods are static and dynamic back-calculation strategies, which are each covered in more detail in
the following sections. Fig. 2 depicts various back-calculation strategies and their respective
application constraints. Finite element models have also been developed to solve back-calculation
problems for multi-layer systems, and have been shown (13) to be able to verify results of
back-calculation programs based on multi-layer elastic theory. Although research into such methods
continues, the tediousness of their use due to the required computational effort currently makes them
unattractive for use in general pavement analysis (6).

FIGURE 2: A schematic illustration of available back-calculation techniques (6).


Static Back-calculation. The second available method to determine layer properties from FWD
data is by performing a static back-calculation. This method for determining layer moduli by
simplification of the model to a statically applied loading, is the most popular approach employed
(13). Despite its widespread application, it has been shown to produce unsatisfactory results due to the
assumptions inherent in its application (9). In the static back-calculation, only the peak deflections are
utilized and considered as “pseudo-static” deflections. These deflections are iteratively matched to
computed deflections to obtain layer moduli (12). Deflections, whether peak or time series, depend on
the thickness and the elastic modulus of each layer (14), and hence if the layer thicknesses are known,
the moduli can be back-calculated.
Fig. 1 shows how the load is assumed to be distributed throughout the pavement system, and
roughly depicts the deflection basin obtained with the FWD test. This distribution of the load is often
assumed to spread at 45º into the pavement structure. Only deflections measured with geophones at
their maximum are used in the static back-calculation.
Dynamic Back-calculation. The nature of the applied load delivered by the FWD is transient and
impulsive, which is suitable in closely simulating the effect of a moving wheel on the pavement
surface (6). A dynamic analysis of FWD data therefore logically yields results which are more
accurate than those obtained from a static approach, in which the theoretical model is considerably
more simplified and thus effectively more ‘unrealistic’ (15). The following parameters determine the
nature of the dynamic response of a pavement subjected to an applied impulse load (6):
Advanced Materials Research Vol. 723 199

 Elastic moduli;
 Layer thicknesses;
 Poisson’s ratios;
 Mass densities, and
 Damping ratios.

The latter three parameters are usually assumed to be known as their influence on the dynamic
response is limited (6). Common practice is to use empirically based values for Poisson’s ratios and
mass densities, while damping ratios are either incorporated in the analysis as coefficients in a
complex modulus or as fractions of layer moduli (12).

FIGURE 3: Typical deflection graph obtained with a FWD (6).

Fig. 3 illustrates how deflections are measured at the geophone sensors over the time of the
loading. Of importance is that the maximum deflections at each sensor occur at different times, an
incidence which is not taken into account in the static procedure.

Data analysis
Optimization. The comparison of measured and computed deflections by adaptation of parameter
values is referred to as a non-linear least square or non-linear minimization problem (5). The
evaluation function of such a problem as related to back-calculation by the matching of time series
deflections (Eq. 1) (12):

J
1 L N K (l )
 
2 l 1 i 1 k 1
 2
wi (t k )  z i(l ) ( X , t k ) . (1)

where:
= measured response at point i and time tk for the lth data set
= computed response at point i and time tk for the lth data set
X = vector of layer modulus and damping coefficient (unknown parameters)
K = number of time steps
N = number of deflection sensors
L = number of data sets
Computed and measured deflections can be compared to within a predetermined tolerance, rate of
change of calculated layer modulus or number of iterations. Minimizing the evaluation function
furthermore requires the use of an optimization strategy such as Truncated Singular Value
Decomposition (TSVD), Ridge Regression, Limited memory quasi-Newton or Genetic Algorithms,
because of the inherently unstable nature of the back-calculation process (5).
200 Innovation and Sustainable Technology in Road and Airfield Pavement

Data collection
Site Selection. As part of on-going research by the CSIR to evaluate different back-calculation
procedures, deflections from FWD tests were analyzed. The selection of a suitable site for the
implementation of NDT was based on the following:
 The pavement is of a typically “inverted” South African design (thin asphalt surfacing, unbound
granular base, stabilized subbase, upper selected layer, and in-situ subgrade), which makes it ideal
as being representative of the local conditions in which the software will be implemented (Fig. 4),
and
 A variety of stages of deterioration are present on the stretch of road so that the effect of surface
cracking and damage of the lower layers on the analysis may also be determined.

Also of note is that conditions of the pavement structure under consideration may vary over the
course of testing as loading induced by testing devices may result in traffic associated deterioration
due to the unbalanced nature of the pavement structure.

FIGURE 4: Cross section of N4 national freeway.

Data Collection. The FWD testing procedure consisted of the following:


 120 drops were performed at each test point and the time-load history and time-deflection
history from each drop was measured;
 Each set of 120 drops consists of 40 drops of differing heights, attempting to obtain peak loads
of 40 , 50 and 70 kN, and
 These drops were performed respectively three at a time, i.e., a 40 kN drop followed by a
50 kN drop and then a 70 kN drop, repeating this 40 times.

Pavement Modelling. In both static and dynamic backcalculation, the pavement is modeled (Table
1(a) and 1(b)) as a 3-layer system as suggested by ASTM, with layer 1 comprised of the surfacing and
base, layer 2 the two subbase layers, and layer 3 the selected and subgrade layer extending infinitely
downwards. A notable difference in the static analysis is the approach to seed values for elastic
Advanced Materials Research Vol. 723 201

TABLE 1: Pavement input parameters used the analysis.


Layer E-Modulus (seed – MPa) F – Damping (MPa) v ρ (kg/m3)
(a) DYNAMIC
1 1,000 10 0.35 2,000
2 5,000 50 0.2 1,800
3 200 2 0.35 1,800
(b) STATIC
1 10 – 1,000 0.35
2 10 – 5,000 Not applicable 0.2 Not applicable
3 10 - 200 0.35

modulus. In this analysis, seed values are randomly generated between a specified range and the best
fit results chosen from the resulting range of solutions. This range was set between 10 and 1,000 MPa
for the base, 10 and 5,000 MPa for the subbase and 10 and 500 MPa for the subgrade. Thus, the upper
limits for the seed values are approximately equal to the seed values used in the dynamic analysis.

Discussion of results
Comparison between static and dynamic analyses is shown in Fig. 5 for the three repeats. The first
notable observation is that the disparities between moduli calculated from the static and dynamic
analyses intensify as the applied load increases. For the static analysis, the higher loads, and thus
higher strains, place more pressure on the simplification of assumed static loading conditions as the
difference between actual dynamic strain and assumed static strain increases. This leads to an
overestimation of the elastic moduli derived from the static analysis.
The same decreasing trend in subbase modulus throughout the progression of testing is observed in
both analyses, with pronounced changes in results from approximately drop 27 onwards. In terms of
computational speed, the static analysis takes place virtually instantaneously in comparison to the
dynamic analysis, where one iteration takes approximately one second depending on the available
computational power.
For clarity, in Fig. 6 the elastic moduli for the surface and subgrade layers are plotted without
inclusion of the results for the base layer. For the surfacing layer, results from static and dynamic
analyses are virtually indistinguishable, apart from larger variation in the dynamic results. In the
subgrade the difference is slightly more pronounced. However, with differences consistently only in
the region of approximately 50 MPa, the extra computational effort required for the dynamic analysis
is not necessarily warranted.

FIGURE 5: Comparison of static and dynamic analyses - Repeat 1.


202 Innovation and Sustainable Technology in Road and Airfield Pavement

FIGURE 6: Static and dynamic analyses compared for layer 1 and 3 only - Repeat 1.

Of prime interest is the difference in calculated layer modulus of the cemented subbase, the
structural backbone of the inverted pavement. With the maximum difference in the region of
1,900 MPa, the difference between the two approaches becomes clear. One suggestion could be to
add a factor of safety for elastic modulus calculated with assumed static conditions, although the
variability in the differences lead one to suspect that this may be over-conservative in many cases. In
order to fully understand the significance of these differences, it would be necessary to determine how
the remaining life prediction differs from the static and dynamic inputs, although this falls outside the
scope of this project.

Conclusion and recommendations


Dynamic back-calculation is a complex problem, but with a shortage of the data needed to adequately
maintain an ever deteriorating road network, it is necessary to constantly enhance our ability to
accurately and efficiently determine pavement structural condition.
In this paper, it has been shown that application of the Visco-BackGAMES software to dynamic
back-calculation of layer moduli with FWD measurements yields reasonable results, which can be
used to effectively assess the structural condition of an inverted pavement without altering its in situ
state. The results are validated by the ability of the program to closely match measured and computed
deflections within a specified error range.
A gradual decline in subbase modulus at the selected testing point was observed during the course
of testing. For the 40 kN target load, subbase stiffness varied from approximately 3,000 MPa to
approximately 1,000 MPa over the 120 drops of the falling weight, which suggested an accelerated
transition into traffic associated deterioration.
Finally, it was determined that layer moduli determined with a dynamic analysis compare
favorably with those obtained by a static analysis. It was found that the largest discrepancy between
the two approaches lies in the calculated elastic moduli of the cemented subbase layer, where this
parameter is overestimated in the static approach. The largest difference, in the region of 1,900 MPa,
could lead to problems in a maintenance strategy where a static analysis is employed if the results of
such an analysis are not interpreted with gumption.
The significance of the difference in statically and dynamically derived elastic moduli requires
further research into the effect of the difference on remaining pavement life predictions. With the
relative speed of the static back-calculation, it is recommended that the development of dynamic
methods continues, and that this development is aimed at enhancing the efficiency of the more
complex approach.
Advanced Materials Research Vol. 723 203

References
[1] Terrel, R.G., Cox, B. R., Stokoe, K. H., Allen, J. J. and Lewis, D. Field Evaluation of the Stiffness
of Unbound Aggregate Base Layers in Inverted Flexible Pavements. In Transportation Research
Board (TRB) Annual Meeting, USA. 2003.
[2] South African National Roads Agency. High Performance Crushed Stone Bases (G1) Used on
South African Highways - A General Overview. 1998.
[3] De Beer, M., Kannemeyer, L. and Fisher, C. Towards Improved Mechanistic Design of Thin
Asphalt Layer Surfacings Based On Actual Tyre/Pavement Contact Stress in Motion (SIM) Data in
South Africa. In 7th Conference on Asphalt Pavements for Southern Africa. 1999.
[4] Committee of Land Transport Officials (COLTO). Draft TRH4: Structural Design of Flexible
Pavements for Interurban and Rural Roads. Pretoria. 1996.
[5] Maina, J.W., van Rensburg, Y. and de Beer, M. Pavement Analysis Based on Linear Elastic
Theory. Revision of South African Pavement Design Method. CSIR, South Africa. 2010.
[6] Goktepe, A.B., Agar, E. and Lav, A.H. Advances in Back-Calculating the Mechanical Properties
of Flexible Pavements. Advances in Engineering Software, 37, pp. 421-431. 2006.
[7] Karadelis, J.N., A Numerical Model for the Computation of Concrete Pavement Moduli: a
Non-Destructive Testing and Assessment Method. NDT & E International, 33, pp. 77-84. 2000.
[8] Das, A. Interpretation of Falling Weight Deflectometer data. Seloflex World Conference. 2010.
[9] Hadidi, R. and Gucunski, N. Comparative Study of Static and Dynamic Falling Weight
Deflectometer Back-Calculations Using Probabilistic Approach. Journal of Transportation
Engineering, 136(3), pp. 196-204. 2010.
[10] ASTM D4694. Standard Test Method for Deflections with a Falling-Weight-Type Impulse Load
Device. West Conshohocken, PA: ASTM International. 2003.
[11] Sun, L., Luo, F. and Gu, W. Elastodynamic Inversion of Multilayered Elastic Media via Surface
Waves - Part 1: Methodologies. Journal of Applied Mechanics, 78, pp. 041004-1 - 041004-6. 2011.
[12] Matsui, K., Dong, Q., Hachiya, Y. and Takahashi, O. Difference in Estimated Layer Moduli from
Static and Dynamic Back-Calculations. In Federal Aviation Administration Airport Technology
Transfer Conference. 2002.
[13] Loizos, A. and Scarpas, A.T. Verification of Falling Weight Deflectometer Backanalysis Using a
Dynamic Finite Elements Simulation. The International Journal of Pavement Engineering, 6(2), pp.
115-213. 2005.
[14] Seo, J.-W., Kim, S.-I., Choi, J.-S. and Park, D.-W. Evaluation of Layer Properties of Flexible
Pavements Using a Pseudo-static Analysis Procedure of Falling Weight Deflectometer. Construction
and Building Materials, 23, pp. 3206-13. 2009.
[15] Picoux, B., El Ayadi, A. and Petit, C. Dynamic Response of a Flexible Pavement Submitted by
Impulsive Loading. Soil Dynamics and Earthquake Engineering, 29, pp. 845-54. 2009.
Innovation and Sustainable Technology in Road and Airfield Pavement
10.4028/www.scientific.net/AMR.723

Static and Dynamic Backcalculation Analyses of an Inverted Pavement Structure


10.4028/www.scientific.net/AMR.723.196

You might also like