You are on page 1of 14

This article was downloaded by: [Mr Pusjatan Bandung]

On: 27 August 2015, At: 02:35


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: 5 Howick Place,
London, SW1P 1WG

International Journal of Pavement Engineering


Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/gpav20

A model for equivalent axle load factors


a a a b
Sara I.R. Amorim , Jorge C. Pais , Aline C. Vale & Manuel J.C. Minhoto
a
Territory, Environment and Construction Centre (C-TAC), University of Minho, Campus de
Azurém, 4800-058 Guimarães, Portugal
b
School of Technology and Management, Polytechnic Institute of Bragança, Campus de
Santa Apolónia, Apartado 1134, 5301-857 Bragança, Portugal
Published online: 13 Oct 2014.

Click for updates

To cite this article: Sara I.R. Amorim, Jorge C. Pais, Aline C. Vale & Manuel J.C. Minhoto (2015) A model for equivalent axle
load factors, International Journal of Pavement Engineering, 16:10, 881-893, DOI: 10.1080/10298436.2014.968570

To link to this article: http://dx.doi.org/10.1080/10298436.2014.968570

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
International Journal of Pavement Engineering, 2015
Vol. 16, No. 10, 881–893, http://dx.doi.org/10.1080/10298436.2014.968570

A model for equivalent axle load factors


Sara I.R. Amorima, Jorge C. Paisa*, Aline C. Valea and Manuel J.C. Minhotob
a
Territory, Environment and Construction Centre (C-TAC), University of Minho, Campus de Azurém, 4800-058 Guimarães,
Portugal; bSchool of Technology and Management, Polytechnic Institute of Braganc a, Campus de Santa Apolónia,
Apartado 1134, 5301-857 Braganc a, Portugal
(Received 9 October 2012; accepted 9 July 2014)

Most design methods for road pavements require the design traffic, based on the transformation of the traffic spectrum, to be
calculated into a number of equivalent passages of a standard axle using the equivalent axle load factors (EALFs).
In general, these factors only consider the type of axle (i.e. single, tandem or tridem), but they do not consider the type of
wheel on the axles, i.e. single or dual wheel. The type of wheel has an important influence on the calculation of the design
traffic. The existing design methods assume that the EALFs are valid for all pavement structures and do not consider the
thickness and stiffness of the pavement layers. This paper presents the results of the development of a model for the
calculation of the EALFs considering the type of axle, the type of wheel and the constitution of the pavement. The model
Downloaded by [Mr Pusjatan Bandung] at 02:35 27 August 2015

was developed based on the tensile strain at the bottom of the asphalt layer that is responsible for bottom-up cracking in
asphalt pavement, which is the most widely considered distress mode for flexible road pavements. The work developed in
this study also presents the influence of the type of wheel (single and dual) on pavement performance. The results of this
work allowed the conclusion that the EALFs for single wheels are approximately 10 times greater than those for a dual
wheel. This work also proposes average values for the EALFs. An artificial neural network was developed to calculate the
EALFs.
Keywords: road pavements; traffic; axle type; wheel type; equivalent single axle load; equivalent axle load factor

1. Introduction group.
Traffic data constitute one of the key elements required X
n
for the design/analysis of pavement structures. Despite the ESAL ¼ EALFi £ ni : ð1Þ
Mechanistic-Empirical Pavement Design Guide, where i¼1
traffic is defined by axle load spectra, the number of For flexible pavements, the EALF, which is defined
equivalent single axle loads (ESALs), as defined in the based on experience and on the results of the AASHTO
1993 Association of State Highway and Transportation Road Test (1962), is defined as
Officials (AASHTO) Design Guide, is a required traffic
feature for most pavement structural design procedures, W t18
which use the concept of the equivalent load to transform EALF ¼ ; ð2Þ
W tx
the expected traffic into the design traffic.
The first equivalency factors used to determine the where
number of ESAL were based on the Present Serviceability  
Index concept and are dependent on the pavement type and W tx
Log ¼ 4:79 log ð18 þ 1Þ 2 4:79 log ðLx þ L2 Þ
surface thickness of the pavement. W t18
The ESAL uses the concept of the equivalent axle load Gt Gt
þ 4:33 log L2 þ 2 ;
factor (EALF), which defines the ratio between the bx b18
damage caused by the passage of an axle on a pavement ð3Þ
 
and the damage caused by the passage of a standard axle 4:2 2 pt
Gt ¼ log ; ð4Þ
on the same pavement. This EALF is used in pavement 4:2 2 1:5
design to convert the spectrum of vehicles with different
types of axles (i.e. single, tandem and tridem) into single 0:081ðLx 2 L2 Þ3:23
bx ¼ 0:40 þ ; ð5Þ
axles with dual tires, i.e. the ESAL, by using Equation (1), ðSN þ 1Þ5:19 L3:23
2
where n is the number of axle load groups, i is the number
of the axle load group, EALFi is the EALF for ith axle load where Wt18 is the number of standard 18-kip axle
group and ni is the number of passes of the ith axle load applications, Wtx is the number of x-axle applications,

*Corresponding author. Email: jpais@civil.uminho.pt


q 2014 Taylor & Francis
882 S.I.R. Amorim et al.

Lx is the axle load (kip), L2 is the axle code (1 for single P80 is the load of the standard axle.
axle, 2 for tandem axles and 3 for tridem axles), pt is the  
terminal serviceability, SN is the structural number (in) Px
EALF ¼ : ð9Þ
and b18 is the value of bx when Lx is equal to 18 and L2 is P80
equal to 1.
In the mechanistic method, the EALF is obtained by The different methods for the calculation of the EALF
the ratio between the pavement life for the standard axle fail to consider the type of axle; they only consider the
and the pavement life for the actual load, as expressed in total load of the axles. However, for tandem and tridem
Equation (2). Using the current pavement terminology, axles, the distance between the axles in the group has an
Equation (2) can be rewritten as important effect on the state of stress and strain in the
pavement and thus has an important effect on the load
N 80 equivalency factor.
EALF ¼ ; ð6Þ
Nx This effect is considered in the French Pavement
Design Manual (LCPC 1994), where the EALF is
where N80 is the pavement life for the standard axle load, generally expressed as Equation (10), where Px is the
usually the 80-kN axle load, and Nx is the pavement life for actual axle load, P80 is the load of the standard axle, k is a
the actual axle load. coefficient that is a function of the type of axle (i.e. single,
Downloaded by [Mr Pusjatan Bandung] at 02:35 27 August 2015

By considering the pavement life given by fatigue tandem or tridem) (Table 1) and a is a coefficient function
cracking, i.e. the main distress mode that appears in of the type of pavement.
flexible road pavements, the pavement life can be
 a
expressed as a function of the asphalt mixture stiffness Px
and the tensile strain at the bottom of the asphalt layer, as EALF ¼ k : ð10Þ
P80
expressed in Equation (7), where N is the pavement life, E
is the asphalt mixture stiffness, 1 is the tensile strain at the This method does not consider the effect of the type
bottom of the asphalt layer, and a, b and c are constants of tire on the axles, i.e. single or dual tires, which has an
that are defined experimentally. important influence on the calculation of the EALF. The k
coefficient does not take into account the variety of flexible
N ¼ aE 2b 1 2c : ð7Þ pavements, both in terms of thickness and stiffness of the
pavement layers.
The substitution of Equation (7) into Equation (6) The report of the SPENS project (Kokot and Gaspar
leads to Equation (8), where 1x is the tensile strain for the 2009) summarises the models used to define the load
actual axle, and 180 is the tensile strain for the standard equivalency factors for various countries in Europe. The
axle, which in this case is the 80-kN axle load. only factor that considers the effect of the wheel type is the
 c Slovenian method that is defined in Equation (11).
1x
EALF ¼ : ð8Þ
180 EALF ¼ 1028 £ f 0 £ ðf k £ Lstat Þ4 : ð11Þ

Equation (8) is used to calculate the EALF for any where f0 is the factor of axle distribution (single ¼ 2.212,
pavement structure and for any type of axle and tire tandem ¼ 1.583), fk is the factor of wheel distribution
configuration by considering fatigue cracking as the (single ¼ 1.0, double ¼ 0.9) and Lstat is the static axle load
failure criterion. of individual vehicle (kN).
Considering the linear-elastic behaviour of the These equivalency factors depend not only on the type
pavement materials, the tensile strain at the bottom of and thickness of the pavement surface, as stated above, but
the asphalt layers is proportional to the applied load, and also upon the type of distress. Studies from FHWA (2001)
Equation (8) can therefore be rewritten as Equation (9), and ERES Consultants (2001) have shown the effects of
which is the usual expression to convert a given axle load the distress type, failure criteria and other parameters on
to the equivalent load, where Px is the actual axle load and the equivalency or damage factors used to calculate ESAL.

Table 1. Values of the k and a coefficients according to Equation (10) for the French method (LCPC, 1994).

k
a Single axle Tandem axle Tridem axle
Flexible pavement 5 1 0.75 1.1
Rigid and semi-rigid pavements 12 1 12 113
International Journal of Pavement Engineering 883

The concept of an equivalent applied load allows the different road pavements were modelled, with different
pavement designer to consider the damage caused by loads values for the thickness and stiffness of the pavement
of varying magnitudes and axle configurations. Although layers. The asphalt layer thickness was set to values of
pavement damage can be expressed per axle, it is more 0.10, 0.15, 0.20, 0.25 and 0.30 m, which are representative
convenient to express the damage in terms of the average of all of the flexible pavement structures from low to heavy
amount of damage caused by a particular vehicle, referred traffic. The asphalt layer stiffness was set to values of
to as a truck factor, which is the average number of ESAL 2500, 5000, 7500, 10,000 and 12,500 MPa, which are
applications per vehicle (Smith and Diefenderfer 2009). representative of the cracked asphalt layers up to high
Zaghloul and White (1994) studied the effect of heavy stiffness asphalt layers. The granular layer thickness was
loads on the Indiana highways and developed the EALF set to 0.10, 0.15, 0.20, 0.25 and 0.30 m, and its stiffness
based on an analytical approach by considering the was twice the subgrade stiffness. The subgrade was
permanent deformation of flexible pavements. The modelled for 20, 40, 60, 80, 100 and 120 MPa. The
approach was developed using a three-dimensional following Poisson’s ratios were considered: 0.35 for
dynamic finite element method for static and dynamic the asphalt layer, 0.40 for the granular layer and 0.45 for
analyses using multi-layer static analysis and actual field the subgrade layer. The complete factorial of the cases
measurements. The results showed that the EALF obtained presented leads to 750 different pavements that were
in that analysis agreed with the factors obtained by considered in this analysis. This factorial allows the
Downloaded by [Mr Pusjatan Bandung] at 02:35 27 August 2015

AASHTO. inclusion of all typical pavement configurations existing in


More recently, in Hong et al. (2006) and Prozzi et al. roads.
(2007) as well as in various pavement design methods, the The load simulation was performed for single, tandem
EALF has been obtained by the mechanistic method and and tridem axles, with single and double wheels. The load
by using failure criteria based on fatigue laws for fatigue was modelled as a circular area of a 0.125 m radius, which
cracking and permanent deformation. The models from the is the typical value used in pavement design to simulate
Asphalt Institute, from the University of Nottingham and the wheel load area. For double wheels, the distance
from Shell, are the main criteria used in the development between the centres of the wheels was set to three times
of the EALF. their radius, i.e. 0.375 m. For multiple axles (tandem and
Based on a literature review, the methods and models tridem axles), the distance between axles was defined as
developed to calculate the EALF fail to consider the type 1.35 m, a typical value for the distance between axles in
of tire because the vehicle classification, by manual or multiple configurations.
automatic systems, identifies only the type of axle. As reported in the COST Action 344 (Addis, 2000),
However, the number of tires (single or dual) is extremely the tire – pavement contact area is more rectangular than
important for a correct evaluation of the EALF. circular, and the contact area is a function of the tire
In addition, the pavement constitution is not identified in configuration, load and inflation pressure. However, the
the models used to calculate the EALF due to the large contact areas indicated in the report for the usual tires have
variability in pavements. the same value as a circular area with a radius of 0.125 m.
This paper intends to develop a model for the Both these rectangular and circular areas produce identical
calculation of the EALF by considering the type of tensile strains at the bottom of the asphalt layers
wheel, i.e. single or dual, valid for flexible pavements considered in this work.
composed of three layers (an asphalt layer, a granular layer For the calculation of the k coefficients (Equation
and a subgrade layer), defined by the thickness and (10)), fatigue cracking was considered to be the main
stiffness of the pavement layers. Both the asphalt layer distress mode of the pavements. Thus, for each pavement,
and the granular layer are defined by their thickness and the tensile strain at the bottom of the asphalt layer was
stiffness, whereas the subgrade layer is defined by its considered for the calculation of the pavement life.
stiffness. The model was developed using the results of the For single axles, the tensile strain to calculate the k
simulation of 750 different pavement structures, assuming coefficient is the result of the application of the axle load,
different values for the thickness and the stiffness of the whereas for tandem and tridem axles, the tensile strain
pavement layers. In addition, an artificial neural network must be obtained by a damage analysis because the strain
was developed to calculate the EALFs. level installed at the passage of a second axle of multiple
axles has two components: the strain level due to the
second axle and the part of the strain level of the first axle
2. Cases studied that has not been recovered. The same phenomenon
This study developed a model for the calculation of the appears for the third axle of a tridem axle.
EALF by defining the k coefficient of Equation (10) For this study, the method proposed by Huang (2004)
through modelling a road pavement with an asphalt layer, to consider the damage analysis was used. The method of
a granular layer and a subgrade layer. A total of 750 Huang consists of the calculation of the strains installed in
884 S.I.R. Amorim et al.

Figure 1. Damage analysis for multiple loads (Huang, 2004).


Downloaded by [Mr Pusjatan Bandung] at 02:35 27 August 2015

the pavement according to the positions indicated in defines the location for the strain calculation based on the
Figure 1, where St is the distance between the axles of a type of axle and type of wheel.
tandem axle. For the first axle, the strain must be obtained For this study, the failure criterion used was based on
below the first axle (1a). For the second and third axles, the the fatigue cracking results of a set of 16 asphalt mixtures
strain must be the difference between the strain below the used in asphalt base layers tested by Pais et al. (2009)
first axle (1a) and the strain at the midpoint between two through four-point bending tests at a temperature of 208C
axles (1b), resulting in 1a 2 1b. The above calculations and at a frequency of 10 Hz.
must be done in the positions indicated in Figure 1, namely Each mixture was tested using the AASHTO T321-03
at the centre of a single wheel (position 1), at the edge of a standard, which establishes testing at least six specimens at
single wheel (position 2) and at the centre of the double two or three tensile strain levels. Some mixtures were
wheels (position 3). tested using 18 specimens at 3 strain levels, as
The strain level for each pavement and for each type of recommended by the European standard – EN 12697-26.
axle and wheel considered in this study was calculated The fatigue curves for these mixtures are represented in
using Burmister’s (1945) theory, which was implemented Figure 2, where one can conclude that they have identical
in the JPav software developed by the first author. This slopes despite the different vertical axis intercepts.
theory was adapted for single, tandem and tridem axles The results obtained for all of the asphalt mixtures are
and for single and dual wheels. The software automatically expressed by Equation (12), which relates fatigue life to

1000
Tensile strain (E-6)

Mix 31 Mix 04 Mix 07 Mix 08

Mix 10 Mix 11 Mix 12 Mix 18

Mix 19 Mix 20 Mix 24 Mix 25

Mix 26 Mix 27 Mix 14 Mix 29


100
1E+03 1E+04 1E+05 1E+06
Fatigue life (cycles)

Figure 2. Fatigue curves of the base course mixtures (Pais et al., 2009).
International Journal of Pavement Engineering 885

the tensile strain applied during the laboratory fatigue is 8.49 when considering single axles and 8.14 when
tests, where N is the number of repetitions to failure, 1t is considering tandem axles.
the tensile strain, and k1 and k2 are experimentally
determined coefficients.
3. Simulation results
 k 2
1 For the development of this study, the tensile strain at the
N ¼ k1 £ : ð12Þ
1t bottom of the asphalt layer of the 750 pavement structures
described earlier was calculated for six cases, namely three
The calculated coefficients k1 and k2 for all mixtures axle types and two wheel types, resulting in 4500 cases in
tested by Pais et al. (2009) are presented in Table 2. The this study.
average value for k2 is 4.147, which is inside the interval of For each case studied, the EALF is calculated by
the coefficient of the models defined by The Asphalt applying Equation (8), where 1x is the tensile strain for the
Institute (3.291) and Shell (5.671). The analytical analysis axle type and wheel type considered, 180 is the tensile
conducted in this paper used k2 value of 4, i.e. the strain for the standard axle (single axle with dual tire), and
coefficient a of Equation (10) was set to 4. c was set to 4 based on the results of the fatigue tests. If the
However, the coefficient for the EALF is a function of applied load is equal in both cases (for the calculation of 1x
the type of distress being analysed. For example, the value and for the calculation of 180), then the k coefficient of
Downloaded by [Mr Pusjatan Bandung] at 02:35 27 August 2015

used in this work compares well with the value obtained by Equation (10) is equal to the calculated ESAL.
Archilla and Madanat (2000) for rutting in the case of a Using this methodology, the k coefficients can be
tandem axle, where they found a value of 3.89, but the value calculated for any pavement structure and load configur-
is different when considering single axles, where they ation. As an example, Table 3 presents the k coefficient for
found a value of 2.98. Archilla and Madanat (2001) found a pavement with an asphalt layer stiffness of 5000 MPa and
values of 2.44 (for single axles) and 2.86 (for tandem axles) a 0.20-m granular layer with a stiffness of twice the value of
using data from the AASHO Road Test and the WesTrack the subgrade layer. The analysis of these results reveals that
Road Test, and both values are significantly different from all of the variables analysed have a significant influence on
4. When using a recursive nonlinear model considering the the k coefficient and thus on the EALF calculation. For
pavement roughness, Prozzi and Madanat (2003) found example, the k coefficient for a 0.10-m asphalt layer
that the power was approximately 4.2, which is close to the thickness is twice the value for a 0.30-m asphalt layer
power of 4 considered in this study for pavement cracking. thickness. The inverse effect is observed for the subgrade
Prozzi and Madanat (2004) and Prozzi (2001) found a value stiffness influence. In addition, the axle configuration has
of 3.85 when considering the roughness and 4.15 when an important influence on the k coefficient.
considering the serviceability using data from the AASHO For the development of a model for the k coefficient of
Road Test jointly with data from the MnRoad Project, EALF (Equation (10)), Figures 3 – 5 present the typical
which are values that compare well with the value defined influence, respectively, of the asphalt layer thickness, the
in this work for fatigue cracking. More recently, Guler and asphalt layer stiffness and the subgrade stiffness on the k
Madanat (2011) used the AASHO Road Test data to find coefficient. For all variables, including the thickness of the
that the appropriate power for pavement cracking initiation granular layer, the k coefficient can be expressed as a
function of the power of the variable considered.
Table 2. Fatigue test results (Pais et al., 2009). The influence of the asphalt layer thickness on the k
Mixture k1 k2 R2 coefficient is presented in Figure 3 for a pavement with an
asphalt layer with a stiffness of 5000 MPa, a 0.20-m
1 7.977E þ 17 4.717 0.980 granular layer with a stiffness of 160 MPa and a subgrade
2 4.628E þ 16 4.316 0.960 layer with a stiffness of 80 MPa, where the k coefficient
3 1.419E þ 16 4.210 0.996
4 2.823E þ 17 4.728 0.979 decreases with the increase of the asphalt layer thickness.
5 1.074E þ 13 3.311 0.968 The influence of the asphalt layer stiffness on the k
6 2.353E þ 15 4.121 0.977 coefficient is illustrated in Figure 4 for a pavement with a
7 4.297E þ 16 4.510 0.993 0.10-m asphalt layer thickness, a 0.20-m granular layer
8 7.865E þ 18 5.140 0.995 with a stiffness of 120 MPa and a subgrade layer with a
9 2.060E þ 16 4.268 0.959
10 1.103E þ 15 3.948 0.954 stiffness of 60 MPa, where the k coefficient decreases with
11 5.833E þ 13 3.552 0.949 the increase of the asphalt layer stiffness.
12 3.785E þ 13 3.431 0.990 The influence of the subgrade stiffness on the k
13 1.078E þ 14 3.584 0.986 coefficient is illustrated in Figure 5 for a pavement with a
14 6.720E þ 17 4.714 0.968 0.10-m asphalt layer thickness with a stiffness of
15 2.228E þ 15 4.146 0.984
16 7.772E þ 13 3.652 0.992 5000 MPa and a 0.15-m granular layer with a stiffness
twice that corresponding to the subgrade layer, where it
886 S.I.R. Amorim et al.

Table 3. The k coefficient (Equation (10)) for the different pavements and axle configurations.

k
Subgrade Asphalt Single Single Tandem Tandem Tridem Tridem
stiffness layer axle –single axle –dual axle – single axle – dual axle – single axle – dual
(MPa) thickness (m) wheel wheel wheel wheel wheel wheel
40 0.10 5.394 1.000 0.502 0.128 0.068 0.016
40 0.15 3.886 1.000 0.334 0.076 0.046 0.009
40 0.20 3.077 1.000 0.274 0.059 0.038 0.007
40 0.25 2.531 1.000 0.247 0.053 0.037 0.006
40 0.30 2.144 1.000 0.233 0.051 0.039 0.008
60 0.10 6.268 1.000 0.614 0.163 0.078 0.020
60 0.15 4.345 1.000 0.379 0.090 0.052 0.011
60 0.20 3.346 1.000 0.290 0.064 0.041 0.008
60 0.25 2.705 1.000 0.249 0.053 0.037 0.006
60 0.30 2.252 1.000 0.227 0.049 0.037 0.006
80 0.10 6.924 1.000 0.703 0.190 0.085 0.022
80 0.15 4.703 1.000 0.422 0.103 0.057 0.013
80 0.20 3.576 1.000 0.307 0.069 0.044 0.008
Downloaded by [Mr Pusjatan Bandung] at 02:35 27 August 2015

80 0.25 2.846 1.000 0.253 0.054 0.038 0.007


80 0.30 2.343 1.000 0.225 0.048 0.037 0.006
100 0.10 7.390 1.000 0.771 0.211 0.091 0.024
100 0.15 5.105 1.000 0.464 0.116 0.062 0.014
100 0.20 3.788 1.000 0.326 0.074 0.047 0.009
100 0.25 2.976 1.000 0.260 0.056 0.040 0.007
100 0.30 2.425 1.000 0.224 0.048 0.037 0.006
120 0.10 7.792 1.000 0.830 0.230 0.095 0.026
120 0.15 5.444 1.000 0.504 0.128 0.066 0.016
120 0.20 3.972 1.000 0.343 0.080 0.049 0.010
120 0.25 3.083 1.000 0.267 0.058 0.041 0.008
120 0.30 2.496 1.000 0.226 0.048 0.038 0.006

can be concluded that the k coefficient increases with the because the tensile strain for multiple axles is the addition
increase of the subgrade stiffness. An identical behaviour and subtraction of the tensile strain for the individual
is expressed by the k coefficient due to the variation of the axles, which for several cases, changes the typical
granular base thickness. behaviour of the k coefficient.
For some cases studied in this work, the influence of The separation between the typical and the atypical
the variables that were considered on the k coefficient conditions presented earlier is a function of the pavement
presents a different trend, as observed in Figure 6. For properties, namely the thickness and stiffness of the
example, for the tridem axle with dual wheels, the k pavement layers. A value of 1.7 for the Stiffness Ratio
coefficient increases for asphalt layer thicknesses . 0.15 m defined as in the Equation (13) was found to correctly

10.00 Single axle;


single wheel
Single axle;
1.00 dual wheel
Tandem axle;
single wheel
0.10
k

Tandem axle;
dual wheel
0.01 Tridem axle;
single wheel
Tridem axle;
0.00 dual wheel
0.1 0.15 0.2 0.25 0.3
Asphalt layer thickness (m)

Figure 3. Typical influence of the asphalt layer thickness on the k coefficient.


International Journal of Pavement Engineering 887

10.00 Single axle;


single wheel
Single axle;
dual wheel
1.00 Tandem axle;
single wheel
k Tandem axle;
0.10 dual wheel
Tridem axle;
single wheel
Tridem axle;
0.01 dual wheel
2500 5000 7500 10,000 12,500
Asphalt layer stiffness (MPa)

Figure 4. Typical influence of the asphalt layer stiffness on the k coefficient.


Downloaded by [Mr Pusjatan Bandung] at 02:35 27 August 2015

Single axle;
10.00
single wheel
Single axle;
dual wheel
1.00 Tandem axle;
single wheel
k

Tandem axle;
0.10 dual wheel
Tridem axle;
single wheel
Tridem axle;
0.01
dual wheel
40 60 80 100 120
Subgrade stiffness (MPa)

Figure 5. Typical influence of the subgrade stiffness on the k coefficient.

10.00 Single axle;


single wheel
Single axle;
1.00 dual wheel
Tandem axle;
single wheel
0.10
k

Tandem axle;
dual wheel
0.01 Tridem axle;
single wheel
Tridem axle;
0.00 dual wheel
0.1 0.15 0.2 0.25 0.3
Asphalt layer thickness (m)

Figure 6. Atypical influence of the asphalt layer thickness on the k coefficient.


888 S.I.R. Amorim et al.

Table 4. Axle parameter (AP).

Single axle – single Single axle – dual Tandem axle – single Tandem axle – dual Tridem axle –single Tridem axle – dual
wheel wheel wheel wheel wheel wheel
1.0 2.2 2.7 4.1 3.6 5.1

10.00
Esubg=20 MPa Esubg=80 MPa
Esubg=40 MPa Esubg=100 MPa
Esubg=60 MPa Esubg=120 MPa

1.00
k

0.10
Downloaded by [Mr Pusjatan Bandung] at 02:35 27 August 2015

0.01
0 1 2 3 4 5 6
Axle Parameter - AP

Figure 7. Influence of the AP value on the k coefficient for some subgrade stiffness conditions.

identify that separation. The axle – wheel configuration is considered in the


model by a parameter named the axle parameter (AP). For
the definition of the AP, the k coefficients presented in
ðH asp Þ3 £ Easp þ ðH gra Þ3 £ Egra
Stiffness Ratio ¼ ; ð13Þ Table 3 were plotted against the AP. The defined AP values
Esubg presented in Table 4 are the ones that lead to a regular
variation of the k parameter with the AP values, as
where Hasp is the thickness of the asphalt layer (m), Easp is observed in Figure 7. The AP increases from single to
the stiffness of the asphalt layer (MPa), Hgra is the tridem axles and from single to dual wheels. The analysis
thickness of the granular base layer (m), Egra is of this figure reveals that the fit by an exponential law
the stiffness of the granular base layer (MPa) and Esubg simulates variation of the k coefficient with the AP
is the stiffness of the subgrade (MPa). parameter.
Given the influence of the properties of the pavement
layers (thickness and stiffness) as well as the AP, Equation
4. Model development (14) is proposed to model the k coefficient.
The main objective of this work was the development of a k ¼ a1 £ ðH asp Þa2 £ ðH gra Þa3 £ ðEasp Þa4 £ ðEsubg Þa5
model for the calculation of the EALFs considering the
type of axle, the type of wheel and the constitution of the £ exp ða6 £ APÞ; ð14Þ
pavement. This was achieved by developing a model to
quantify the k coefficient (Equation (10)) function of the where Hasp is the thickness of the asphalt layer (m), Hgra is
pavement properties (thickness and stiffness of the the thickness of the granular layer (m), Easp is the stiffness
pavement layers) and the axle –wheel configuration. of the asphalt layer (MPa), Esubg is the stiffness of the

Table 5. Constants for Equation (14).

Stiffness Ratio a1 a2 a3 a4 a5 a6
# 1.7 1.275Eþ01 2 9.307E-01 8.280E-02 2 2.939E-01 2.963E-01 21.421Eþ00
. 1.7 6.416Eþ00 1.027E-01 2 2.597E-03 1.176E-01 21.335E-01 21.380Eþ00
International Journal of Pavement Engineering 889

10

y = 0.986x0.988

k predicted by the model


1 R ² = 0.994

0.1

0.01

0.001
0.001 0.01 0.1 1 10
k calculated by the M-E approach

Figure 8. Predicted versus observed values of the k model (Equation (14)).


Downloaded by [Mr Pusjatan Bandung] at 02:35 27 August 2015

subgrade (MPa), AP is the axle parameter as defined in the single wheel leads to a k value that can be 10 times
Table 4, and a1, a2, a3, a4, a5 and a6 are the constants greater than the value obtained for dual wheels. This
presented in Table 5, which were developed using the least conclusion highlights the importance of the type of wheel
squares method. For a single axle with a single wheel, in the ESAL evaluation and thus in the analysis and design
k ¼ 1 and thus the model does not need to be applied. of road pavements.
The p-value of each variable considered in the model This conclusion can also be observed in Figure 12,
equals 0 for a level of significance of 0.05, which means where the average values of the k coefficient are
that all of the coefficients are indispensable to the model. represented for the entire axle and wheel configuration.
The quality of this model can be observed in Figure 8 This figure can also be used to define simplified values for
in which the k values calculated using the pavement the k coefficient (the average value for each type of axle
software are plotted against those predicted by the model and wheel). However, the high standard deviation observed
developed in this work. for each type of axle and wheel indicates that the use of
The consideration of the wheel type of the trucks in these values must be accompanied by some precaution, i.e.
the ESAL analysis is extremely important because this the average values must be used only when it is impossible
consideration results in different k values for single wheels to use the model presented in Equation (14).
and dual wheels, as shown in Figures 9– 11 for single Despite the quality of the statistical model that was
axles, tandem axles and tridem axles, respectively. Based developed, the use of artificial neural networks has a huge
on the analysis performed in this study, for the same load, potential in the modelling of complex problems, such as

12.00
Single wheel
10.00
Dual wheel

8.00

6.00
k

4.00

2.00

0.00
0 200 400 600 800
Simulation number

Figure 9. Difference between single and dual wheels on single axles on the k value.
890 S.I.R. Amorim et al.

10.00
Single wheel
Dual wheel

1.00

k
0.10

0.01
0 200 400 600 800
Simulation number

Figure 10. Difference between single and dual wheels on tandem axles on the k value.
Downloaded by [Mr Pusjatan Bandung] at 02:35 27 August 2015

1.00
Single wheel
Dual wheel

0.10
k

0.01

0.00
0 200 400 600 800
Simulation number

Figure 11. Difference between single and dual wheels on tridem axles on the k value.

10.000 4.247
Average
1.000 Standard deviation
1.000 0.400

0.098
0.100 0.054
k

0.012
0.010

0.001
Single axle; Single axle; Tandem axle; Tandem axle; Tridem axle; Tridem axle;
single wheel dual wheel single wheel dual wheel single wheel dual wheel
Axle type and wheel type

Figure 12. Average values for the k coefficient for all of the types of axles and wheels.
International Journal of Pavement Engineering 891

the calculation of the EALF function of the variables


considered in this work. The advantage of this model is its
possible implementation in an Excel spreadsheet. Thus,
for this work, an artificial neural network model was
developed with the k coefficient predicted according to the
characteristics of the pavement and the type of wheel
and axle.
The model has four levels, namely the entry level
(level 1) with five nodes, two intermediate levels (levels 2
and 3, respectively, with four and two nodes) and the
output level (level 4) with one node, corresponding to a 5–
4– 2 –1-type architecture, as shown in Figure 13.
The synaptic weight and bias of each connection were
calculated using the QNET2000 software and are
presented in Table 6. The quality of the model, expressed
Figure 13. Architecture used for the artificial neural network. by the comparison between the original k coefficients and
Downloaded by [Mr Pusjatan Bandung] at 02:35 27 August 2015

Table 6. Synaptic weights and biases for the artificial neural network that was developed.

Layer 2
Output layer (layer 4)
Node 1 Node 2 Node 3 Node 4 Bias k
Input layer (layer 1) Hasp 2.29933 25.42569 2 2.68083 20.82218
Hgran 2 0.34464 0.43581 0.34726 20.02513
Easp 0.86772 21.04906 2 6.38637 20.15277
Esubg 5.51466 1.07156 0.65546 0.16987
AP 2 1.29847 25.90283 2 5.52637 27.70048
Bias 0.00925 20.59196 2 0.69706 0.62223
Layer 3 Node 1 2.33351 20.81311 2 2.14685 21.97248 2 3.50215 2 3.8601
Node 2 1.88271 23.29252 2 4.48686 22.72487 3.37904 2 5.59374
Bias 3.74243

Figure 14. Quality of the artificial neural network that was developed.
892 S.I.R. Amorim et al.

the values predicted by the artificial neural network allowed us to obtain a value for r 2 of 1.000, which is
developed, can be observed in Figure 14, where an r 2 of between the original and predicted values.
1.000 was obtained. The work performed in this paper is based on a linear
elastic analysis that considered fatigue cracking through a
set of results obtained in the laboratory. However, a field
5. Conclusions validation is important for a suitable application of the
In this paper, a study to define the EALFs for flexible developed model.
pavements by considering the type of axle (single, tandem The developed model can be used with weigh-in-
and tridem) and the type of wheels (single and dual) was motion data, despite requiring the knowledge of the wheel
conducted. A model was developed based on the type, because in several countries, it is easy to identify the
calculation of 750 flexible pavements by considering type of wheels that are used for each truck.
fatigue cracking as the main distress mode that appears in
pavements represented by the fourth power in the equation
used to calculate the EALF. The fourth power used in this References
work was validated for fatigue cracking based on laboratory Association of State Highway and Transportation Officials
fatigue test results that were carried out for a set of asphalt (AASHTO), 1993. AASHTO guide for design of pavement
mixtures typically applied as asphalt base layers. structures. Washington, DC: AASHTO.
Downloaded by [Mr Pusjatan Bandung] at 02:35 27 August 2015

Association of State Highway and Transportation Officials


The calculation of the EALFs can be conducted as a (AASHTO), 1962. The AASHO road test. Washington, DC:
function of the quotient between the actual load and the Highway Research Board of the NAS-NRC Division of
load of the standard axle powered to the exponent of Engineering and Industrial Research, Report 5, Special
the failure criterion. In this function, a coefficient (k) Report 61E.
exists that expresses the type of pavement and the type Addis, R.R., 2000. Effect of wide single tyres and dual tyres.
Bruxelles: Cost – European Cooperation in Science and
of axle. Technology. COST action 334 Executive Report.
According to the results of this work, the following Archilla, R. and Madanat, S., 2000. Development of a pavement
conclusions can be drawn: rutting model from experimental data. ASCE Journal of
Transportation Engineering, 126 (4), 291– 299.
. The k coefficient decreases with the increase in the Archilla, R. and Madanat, S., 2001. Estimation of rutting models
asphalt layer thickness and stiffness and increases by combining data from different sources. ASCE Journal of
with the increase of the granular layer thickness and Transportation Engineering, 127 (5), 379– 389.
stiffness. For pavements with a Stiffness Ratio . 1.7 Burmister, D.M., 1945. The theory of stresses and displacements
in layered systems and applications to the design of airport
and for multiple axles, the k coefficient presents an runways. Highway Research Board, 23, 126– 144.
inverse behaviour. ERES Consultants, 2001. DataPave software (version 3.0).
. The axle type and wheel type were considered in Washington, DC: Federal Highway Administration.
the model for the k coefficient through the AP, FHWA, 2001. Guide to LTPP traffic data collection and
which is approximately the number of wheels in processing. Washington, DC: Federal Highway Adminstration.
Guler, I. and Madanat, S. 2011. Axle load power for pavement
the axle. fatigue cracking: empirical estimation and policy impli-
. The model developed for the k coefficient predicts cations. Transportation Research Record, 2225, 21 – 24.
the calculated models with good quality. Huang, Y.H., 2004. Pavement analysis and design. 2nd ed.
. The type of wheel in the axles has a significant Upper Saddle River, NJ: Pearson Prentice Hall.
influence on the k coefficient, as demonstrated by Hong, F., Pereira, F.M., and Prozzi, J.A., 2006. Comparison of
equivalent single axle loads from empirical and mechanistic-
the value of k for a single wheel, which is empirical approaches. In: Transportation Research Board
approximately 10 times greater than the k Annual Meeting, Washington, DC, USA.
coefficient for a dual wheel. Kokot, D. and Gaspar, L., 2009. Recommendation for traffic
. The following average k coefficients were equivalency factors. Project sustainable pavements for
obtained: European new member states. European Commission DG
Research. Sixth Framework Programme. Bruxelles: Euro-
+ Single axle/single wheel ¼ 4.247. pean Cooperation in Science and Technology.
+ Single axle/dual wheel ¼ 1.000. LCPC, 1994. French design method for flexible pavements. Paris:
+ Tandem axle/single wheel ¼ 0.400. Laboratoire central des Ponts et Chausse´es (LCPC).
+ Tandem axle/dual wheel ¼ 0.054. Pais, J.C., et al., 2009. The prediction of fatigue life using the k1-
+ Tridem axle/single wheel ¼ 0.098. k2 relationship. In: Second Workshop on Four-point Bending,
Guimarães, Portugal: University of Minho, 39 – 46.
+ Tridem axle/dual wheel ¼ 0.012. Prozzi, J., 2001. Modelling pavement performance by combining
An artificial neural network was developed with great field and experimental data. Thesis (PhD). University of
California, Berkeley, CA, USA.
success to model the k coefficient, demonstrating the Prozzi, J. and Madanat, S., 2003. Incremental nonlinear model
ability of this technique to solve complex prediction for predicting pavement serviceability. ASCE Journal of
problems. The model based on the 5 –4 – 2– 1 architecture Transportation Engineering, 129 (6), 635– 641.
International Journal of Pavement Engineering 893

Prozzi, J. and Madanat, S., 2004. Development of pavement Smith, B.C. and Diefenderfer, B.K., 2009. Development of truck
performance models by combining experimental and field equivalent single-axle load (ESAL) factors based on weigh-in-
data. ASCE Journal of Infrastructure Systems, 10 (1), 9 –22. motion data for pavement design in Virginia. Charlottesville,
Prozzi, J.A., Hong, F., and Grebenschikov, S., 2007. Equivalent VA: Virginia Transportation Research Council. Final Report
VTRC 09-R18.
damage factors based on mechanistic-empirical pavement Zaghloul, S. and White, T.D., 1994. Guidelines for permitting
design. In: Transportation Research Board Annual Meeting, overloads – Part I. Effect of overloaded vehicles on the Indiana
Washington, DC, USA. highway network. West Lafayette, IN: Pardue University.
Downloaded by [Mr Pusjatan Bandung] at 02:35 27 August 2015

You might also like