Four Point Bending

Cracking is recognized as one of the main causes of pavement
Four Point Bending

deterioration, and is the primary cause of the need for maintenance
and rehabilitation. Researchers around the world are working on the
problem of cracking in asphalt pavements, with the goal of developing
better understanding of the mechanics of cracking, creating test
methods for assessing the risk of cracking for different materials and
designs, and implementing these results into improved design methods
and specications. This Third Conference on Four-point bending held
at the University of California, Davis, USA, follows two successful
previous conferences, held at the Delft University of Technology, The
Netherlands, in 2007, and at the University of Minho, Portugal in 2009.
The primary objective of these conferences is to provide an exchange
of ideas and experience and to disseminate that knowledge among
researchers, government and private agencies and consultants, about
the use of the four-point bending test to evaluate stiffness and fatigue
resistance of bituminous mixtures. These proceedings include 23 papers
from 15 countries that have been subjected to peer review by a scientic
committee composed of experts in asphalt materials, design and testing.
Themes of the papers cover a range of topics, including modelling of
the four-point beam test, applications to mechanistic design, asphaltic
materials evaluation, comparisons with other tests and non-asphaltic
materials evaluation. Editors: J. Pais and J. Harvey
Four Point Bending is of interest to academics and professionals
interested in pavement engineering.
Editors
Pais
Harvey
an informa business
4PB_def.indd 1 09-08-12 20:40

FOUR-POINT BENDING
This page intentionally left blank
PROCEEDINGS OF THE THIRD CONFERENCE ON FOUR-POINT BENDING,
DAVIS, CA, USA, 1718 SEPTEMBER 2012
Four-Point Bending
Editors
Jorge C. Pais
University of Minho, Guimares, Portugal
John T. Harvey
University of California, Davis, CA, USA
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
2012 by Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, an Informa business
No claim to original U.S. Government works

Version Date: 2012912
International Standard Book Number-13: 978-0-203-07306-3 (eBook - PDF)
This book contains information obtained from authentic and highly regarded sources. Reasonable efforts have been
made to publish reliable data and information, but the author and publisher cannot assume responsibility for the valid-
ity of all materials or the consequences of their use. The authors and publishers have attempted to trace the copyright
holders of all material reproduced in this publication and apologize to copyright holders if permission to publish in this
form has not been obtained. If any copyright material has not been acknowledged please write and let us know so we may
rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or uti-
lized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopy-
ing, microfilming, and recording, or in any information storage or retrieval system, without written permission from the
publishers.
For permission to photocopy or use material electronically from this work, please access www.copyright.com (http://
www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923,
978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For
organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for
identification and explanation without intent to infringe.
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
Four-Point Bending Pais & Harvey (Eds)
2012 Taylor & Francis Group, London, ISBN 978-0-415-64331-3
Table of contents
Preface VII
Conference organization IX
Modeling of the Four Point Beam test

A detailed FEM simulation of a 4-point bending test device 3
M. Huurman & A.C. Pronk
Description of a procedure for using the Modified Partial Healing model (MPH)
in 4PB test in order to determine material parameters 13
A.C. Pronk
Assessment of different flexure fatigue failure analysis methods to estimate
the number of cycles to failure of asphalt mixtures 27
M.I. Souliman, W.A. Zeiada, K.E. Kaloush & M.S. Mamlouk
Investigation into the size effect on four-point bending fatigue tests 35
N. Li, A.A.A. Molenaar, M.F.C. van de Ven, A.C. Pronk & S. Wu
Applications to Mechanistic Design

Application of four-point bending beam fatigue test for the design and construction
of a long-life asphalt concrete rehabilitation project in Northern California 51
V. Mandapaka, I. Basheer, K. Sahasi, P. Vacura, B.W. Tsai, C.L. Monismith,
J. Harvey & P. Ullidtz
Development of fatigue performance specification and its relation to mechanistic-empirical
pavement design using four-point bending beam test results 59
B.-W. Tsai, R. Wu, J.T. Harvey & C.L. Monismith
Development of a standard materials library for mechanistic-empirical fatigue and
stiffness evaluation 73
L. Popescu, J. Signore, J. Harvey, R. Wu, I. Basheer & J.T. Holland
Four point bending beam tests in Arizona and relationship to asphalt binder properties 85
G.B. Way, K.E. Kaloush & J.M.B. Sousa
First steps for the perpetual pavement design: Through the analysis of the fatigue life 99
N. Hernandez, J. Hernandez & R. Martinez
Asphaltic materials evaluation

Fatigue assessment of conventional and highly modified asphalt materials with
ASTM and AASHTO standard specifications 113
G.M. Rowe, P. Blankenship & T. Bennert
Mechanical characterization of dry asphalt rubber concrete for base layers
by means of the four bending points tests 123
C. Maggiore, G. Airey, G. Di Mino, C.M. Di Liberto & A. Collop
V
Impact of air void content on the viscoelastic behavior of hot mix asphalt 139
B. Hofko, R. Blab & M. Mader
Fatigue evaluation of reinforced asphalt concrete slabs 151
K.P. Biligiri, S. Said & H. Hakim
Influence of recycled asphalt shingles on fatigue properties of asphalt pavements 159
A.A. Cascione & R.C. Williams
Flexural fatigue tests and predictions models tools to investigate SMA mixes with
new innovative binder stabilizers 171
J.B. Sousa, I. Ishai & G. Svechinsky
Laboratory evaluation of fatigue and flexural stiffness of warm mix asphalt 189
A.N. Mbaraga, K.J. Jenkins & J. Van den Heever
Comparisons with other tests

Comparison of fatigue properties using 2-point and 4-point bending
tests Czech experience 205
P. Hyzl, M. Varaus, P. Mondschein, J. Valentin & V. Soucek
Hungarian experience with different bending devices 213
Z. Puchard & A. Gorgenyi
Development of quality control test procedure for characterizing fracture
properties of asphalt mixtures 223
S. Saadeh, O. Eljairi, B. Kramer & E. Hajj
Fatigue resistance: Is it possible having a unique response? 239
C. Maggiore, G. Airey, A. Collop, G. Di Mino, M. Di Liberto & P. Marsac
Non-asphaltic materials evaluation

A laboratory study of the influence of multiple axle loads on the fatigue
performance of a cemented material 253
M.A. Moffatt, G.W. Jameson & W. Young
Four point bending tests on cement treated materials 275
G. Gaarkeuken, B.W. Sluer & M.M.J. Jacobs
Repeated load flexural testing to characterize lightly stabilized granular materials 287
D.K. Paul & C.T. Gnanendran
Author index 295
VI
Preface
Cracking is recognized as one of the main causes of pavement deterioration, and is the primary cause
of the need for maintenance and rehabilitation. Researchers around the world are working on the
problem of cracking in asphalt pavements, with the goal of developing better understanding of the
mechanics of cracking, creating test methods for assessing the risk of cracking for different materials
and designs, and implementing these results into improved design methods and specifications.
Nowadays, there are several types of test equipment to measure the fatigue resistance of asphalt
mixtures in the laboratory, which are widely used by researchers and practitioners. Since the
implementation of the SHRP program in the USA, the four-point bending testing has been one
of the most frequently used testing apparatus to characterize the fatigue life and flexural stiffness
of asphalt mixtures. This test has been included in the European Standards and the number of
four-point bending beam devices in use continues to increase.
The First Conference on Four-point Bending held at the Delft University of Technology, The
Netherlands, 2007, organized by Ad Pronk, dealt with specific concerns regarding the test equip-
ment and methods such as boundary conditions, spatial evolution of damage in fatigue tests,
deformation due to shear force, fatigue and healing, support problems and neglect of shear at the
supports, calibration and finite element modelling, which were discussed around the table among
a limited number of persons from research institutions, consultants and device manufacturers.
The Second Conference on Four-point Bending held at the University of Minho, Portugal, was
open to all interested professionals and embraced the following themes: calibration, CEN standards,
damage and healing, application in finite element method analysis, statistical analysis of results, the
relationship between stiffness and fatigue, and different test procedures. The positive response to
the Second Conference, and the identified need to continue the international interaction regarding
the test apparatus and methods and particularly their use in the design of asphalt materials and
structures led to the organization of this Third Conference on Four-point Bending, held at the
University of California, Davis, USA.
The primary objective of this conference is to provide an exchange of ideas and experience and to
disseminate that knowledge among researchers, government and private agencies and consultants,
about the use of the four-point bending test to evaluate stiffness and fatigue resistance of bituminous
mixtures. Strong interest in the themes of the conference has been demonstrated by the 23 papers
from 15 countries finally accepted after a rigorous peer review by a scientific committee composed
of experts in asphalt materials, design and testing. Themes of the papers cover a range of topics,
including modelling of the four-point beam test, applications to mechanistic design, asphaltic
materials evaluation, comparisons with other tests and non-asphaltic materials evaluation.
We would personally like to thank each of the authors for sharing their four-point bending
expertise. We would also like to thank the Third Four-point Bending Conference Scientific and
Organizing Committees for reviewing the papers and helping to maintain the conference series
high standard. Finally, we want to thank the participants in the two-day conference for sharing their
time and experiences.
Jorge C. Pais and John T. Harvey,
Conference Chairs
VII
Conference organization
ORGANIZING COMMITTEE
John Harvey, University of California, Davis, USA (Chairman)

Jorge Pais, University of Minho, Portugal
Ad Pronk, ACP Consultants, Holland
Jorge Sousa, Consulpav, USA
SCIENTIFIC COMMITTEE
John Harvey, University of California, Davis, USA (Chairman)

Gordon Airey, University of Nottingham, UK
Hussain Al-Khalid, University of Liverpool, UK
Imad Al-Qadi, University of Illinois, USA
Maria Lurdes Antunes, LNEC, Portugal
Laurent Arnaud, ENTPE, France
Wojciech Bankowski, IBDiM, Poland
Didier Bodin, LCPC, France
Armelle Chabot, LCPC, France
Andrew Collop, University of Nottingham, UK
Ariaan de Bondt, Ooms Nederland Holding bv, Holland
Herv Di Benedetto, ENTPE, France
Abd El Halim, Carleton University, Canada
Mrcio Farias, University of Braslia, Brazil
Kamil Kaloush, Arizona State University, USA
Andreas Loizos, Nat. Techn Univ. of Athens, Greece
Massimo Losa, University of Pisa, Italy
Manuel Minhoto, Polytechnic Institute of Bragana, Portugal
Carl Monismith, University of California, Berkeley, USA
Jos Neves, Technical University of Lisbon, Portugal
Jorge Pais, University of Minho, Portugal
Manfred Partl, EMPA, Switzerland
Paulo Pereira, University of Minho, Portugal
Christophe Petit, University of Limoges, France
Lus Picado Santos, University of Coimbra, Portugal
Ad Pronk, ACP Consultants, Holland
Ezio Santagata, Politecnico di Torino, Italy
Jorge Sousa, Consulpav, USA
Dariusz Sybilski, Road & Bridge Research Institute, Poland
Akhtarhusein Tayebali, North Carolina State University, USA
Nick Thom, University of Nottingham, UK
Glicrio Trichs, University of Santa Catarina, Brazil
Bor-Wen Tsai, University of California, Berkeley, USA
Per Ullidtz, Dynatest, Denmark
Alex Visser, University of Pretoria, South Africa
George Way, Consulpav, USA
Rongzong Wu, University of California, Davis, USA
IX
SPONSORS
X
Modeling of the Four Point Beam test
A detailed FEM simulation of a 4-point bending test device
M. Huurman
BAM-Wegen, Utrecht, Delft, The Netherlands
A.C. Pronk
ACP Consultancy, Vlaardingen, The Netherlands
ABSTRACT: Asphalt stiffness and fatigue properties are important inputs for a mechanistic pave-
ment design. In the Netherlands these properties are traditionally determined by 4 Point Bending
tests (4PB). In this paper the accuracy of that test is discussed on the basis of Finite Element
Modelling (FEM) for a specific 4PB device. It is shown that application of the European-standard
results in an underestimation of the stiffness by 3.2% (at 7500 MPa) and an underestimation of
the phase lag by 0.4 (lag = 20 and 20 Hz.). It is shown that a thorough FEM analysis of the 4PB
may reduce these errors to a stiffness that is underestimated by 0.88% and a phase lag which is
overestimated by 0.1 . This project emphasizes the benefits of a FEM modelling of a 4PB device
in combination with calibration tests.
1 INTRODUCTION
Especially stiffness and fatigue properties are essential for the semi-mechanistic pavement design
models. In the Netherlands these properties are determined by use of the 4PB test. Today the
4PB test is standardised by European norms NEN-EN 12697-24 & 26 (2003 & 2007). The norms
describe that the beam is clamped in the test device at four locations. All clamps should allow
for free rotation and translation in the longitudinal direction. The outer clamps should be fixed to
prevent vertical movement; the inner clamps are excited by vertical loading (Figure 1).
Interpretation of test results is done on basis of the Euler-Bernoulli bending beam theory. In this
theory a 3Dimensional bending beam is represented by a beam without height and width, i.e. a line
or 1D beam. Resistance of the 1D beam to bending is dictated by the well known letter combination
EI. Here E stands for the stiffness modulus and I reflects the moment of inertia of the beam cross
section.
Where I = moment of inertia [m4 ], b = beam width [m], h = beam height [m].
The previous suggests that the 4PB is a simple test that can be analysed on basis of a simple
theory while results are of great value for mechanistic design purposes. However such a conclusion
cannot be justified because of the following reasons.
First of all the true beam has to be grabbed by clamps. This locally introduces stresses and strains
into the beam material. These are not considered during test interpretation on basis of a 1D beam
and therefore the effects are not considered in the NEN-EN formulas. These extra stresses and
strains may introduce extra (fatigue) damage or introduce local non-linear effects.
Secondly it is not possible to construct a 4PB test device that truly meets the conditions as
depicted in Figure 1. One has to consider things like friction and play which may both vary through
time due to wear and tear. Also deformation of set-up parts may affect test results.
3
Figure 1. Schematic representation of the 4PB as described in EN standards [2].
Figure 2. Overview of the 4PB device and its model representation.
Thirdly in the 4PB shear forces act on the beam in the area between the inner and the outer
clamps. These forces act to deform the beam whereas these deformations are not considered in the
prescribed test interpretation.
Fourthly the clamps limit the freedom of cross-section deformation leading to unwanted beam
deformations close to the clamps which are not dealt with in the bending beam theory.
The Dutch history in mechanistic design combined with the obligation to type testing led to a
strong increase in the availability of 4PB equipments in the Netherlands. At the moment at least
21 machines are available in the Netherlands of which 18 are produced by Zwick Roell.
From the above the following is concluded: 4PB is an important test for Dutch pavement design,
The test appears simple but is complex in reality, The test is highly appreciated in the Netherlands
and data is produced on daily basis, at least 18 machines are of the same type.
These conclusions triggered the authors to thoroughly investigate the accuracy of the 4PB device
as a first step towards the possibilities of utilisation of the 4PB in scientific research. Hereto the
most commonly used machine in the Netherlands is modelled in detail. Results of this work are
discussed hereafter. It also shows the benefits of FEM for testing devices.
2 FOUR POINT BENDING MACHINE
The Zwick Roell machine is the starting point of this study. Figures 2 and 3 give some pictures of
the device in combination with the FE model that was made.
4
Figure 3. Detail of model and set-up. By application of six elastic hinges the inner clamps are connected to
the hydraulic actuator of which the seating is formed by a depression in the central bridge.
Table 1. Properties of steel and aluminium.
E [MPa] [] Specific mass [kg/m3]
Aluminium 70000 0.25 2700

Steel 210000 0.2 7800
The machine measures the deflection in the middle of the beam with respect to the move less
rigid main frame. As shown in Figure 2 & 3 elastic hinges are applied to try and meet the boundary
conditions as depicted in Figure 1. These hinges are not susceptible to play, slip and stick or
wear and tear. The most important characteristics of the 4PB considered here are the following:
Effective beam length, L = 420 mm; Distance between inner clamps, l = 140 mm; Beam height,
h = 50 mm; Beam width, b = 50 mm; Location of deflection measurement, x = 210 mm; Width of
the clamps = 10 mm; Clamping force to follow viscous deformation = 130 N. For FE modelling
steel and aluminium the properties in table 1 are used.
3 ELASTIC RESPONSE
3.1 Application of the standard

Since AC is a visco-elastic material its stiffness is represented by a complex modulus, E*. Due to
its visco-elastic properties a phase lag, , exists between stress and strain. The stiffness and phase
lag of the tested material are determined by application of the following equations [1].
5
Where E = complex modulus, E1 = storage modulus [MPa], E2 = loss modulus [MPa], = phase
lag between force and displacement [rad], = geometrical factor [1/m], F = force applied to the
beam [N], Z = measured beam deflection [m], = mass factor [kg], = frequency of applied
force and displacement [rad/s].
The factors and are determined as follows [1]:
Where L = effective length of the beam [m], A = distance between an outer clamp and the next inner
clamp [m], x = location where deflection Z is measured [m], M = beam mass [kg], m1 = moving
masses [kg] at location A and m2 = moving masses [kg] at location y [m].
It should be noted that the equations above are based on a modified first order approximation of
the exact solution for the deflection due to pure bending. So, from the start on the deflection due to
shear forces is neglected in the European standards. The applied approximation is very accurate, the
difference with the exact solution (based on an infinite series development) is very small (Pronk,
2006).
To check whether the listed equations lead to an accurate test interpretation in the FEM sim-
ulation, first geometrical non-linear static elastic simulations are done. In these calculations the
phase lag, , is nil per definition. As a result E2 remains nil, also effects of inertia remain absent
(frequency is nil). As a result equations 2, 3 and 4 reduce to the following.
In the simulations the beam material is assigned a stiffness of 7500 MPa and the Poissons ratio
is set to 0.35. Application of the above equations on the FEM simulation results leads to an AC
stiffness that varies from 7260.75 MPa at very small beam excitement to 7260.72 MPa at 100 m/m
strain. From these results it is concluded that geometrical non-linear effects play a very limited role
only. Therefore these effects are neglected from here on. It is furthermore concluded that the back
calculated 4PB stiffness includes an error of 3.19% when the standard formulas are applied to
the data obtained with a perfectly well functioning machine without further correction.
In the sections hereafter effort is made to explain the indicated error in 4PB back calculated
stiffness. The most obvious reasons for the introduction of errors are addressed.
3.2 Shear effect

The beam in the 4PB is subjected to shear in the areas between the inner and the outer clamps. The
shear force to which the beam is subjected in this area equals F/2, i.e. half the applied total load.
Shear deformation is calculated by equation 9.
6
Table 2. Due to clamp restrained cross-section deformation the 4PB stiffness should be reduced with a factor
depending on the Poissons ratio.
Poissons ratio 0.00 0.15 0.25 0.35 0.45
Clamp correction, Cclamp , to work on shear corrected stiffness 1.29% 1.78% 2.33% 3.07% 4.02%
Figure 4. Left: Due to bending and the effect of the Poissons ratio the cross-section of the beam deforms,
at the inner clamps this deformation is restrained (deformation: 250). Right: Model overview (deformation:
10).
The shear deformation effect can be incorporated in the geometrical factor leading to *.
Where Zs = shear deflection of central part of the beam [m], G = shear modulus [MPa],
= Poissons ratio [] (mostly assumed 0.35 for AC mixtures) and = corrected geometrical
factor [1/m].
By correction for shear the 4PB back calculated stiffness increases to 7516.54 MPa at a Poissons
ratio of 0.35. This means that the error in the FEM calculation now becomes +0.22%.
3.3 Clamp effects

The 4PB specimen is clamped between 10 mm wide metal clamps. These clamps locally restrain
the beams freedom of cross-section deformation. This effect was earlier discussed by Huurman &
Pronk (2009), the magnitude of this effect is investigated using a model of an idealised 4PB. In this
idealised 4PB the boundary conditions as depicted in Figure 1 are fully met. The specimen is excited
by forces introduced via rigid 10 mm wide clamps. Figure 4 gives a visual impression of the model.
By comparison of the results of the idealised 4PB model with results of the bending beam theory it
was found that the restraints at the clamps lead to an increase of 4PB bending stiffness. This increase
is dependent on the Poissons ratio. Table I gives the overestimation of the stiffness by application
of the bending beam theory. Correction with the determined overestimation compensates for the
clamp effect.
As stated the correction for the clamp effect should work on the bending part of the beam stiffness
only. This leads to the following modification of *.
Where Cclamp = correction factor for clamp effect [].
7
Figure 5. Left: Clamps allowing for clamp translation (deformation factor: 1250). Right: Elastic hinges at
100 m/m imposed strain (deformation factor: 500).
Due to the clamp effect at = 0.35 the 4PB beam stiffness is reduced to 7285.79 MPa, implying
that the error in the 4PB stiffness is increased to 2.86%.
3.4 Clamp movements

During the (theoretical) analysis of 4PB results it is assumed that the outer clamps do not allow for
vertical movement. However, due to the non-infinite stiffness of the frame the FEM model indicates
that the centre line of the AC beam moves up and down at the outer clamps. In the case considered
here these movements are equal to 7.86581E09 m/N times the applied force. These deformations
effectively will reduce the central deflection, Z, if the deflection is measured with respect to the
main frame of the device (absolute deflection measure). For this reason the deflection Zc used for
the back calculation procedure needs to be corrected for this.
Where Zc = corrected beam deflection [m], Zm = measured beam deflection [m], and Ss = support
spring stiffness [N/m], i.e. 127.132 MN/m.
Taking the end support deformation into account the 4PB back calculated stiffness becomes
7454.43 MPa, so reducing the mistake in the back calculation to 0.61%.
3.5 Hinge stiffness

Figure 1 depicts the boundary conditions that need to be matched by 4PB machines. It should be
clear that it is very hard, if not impossible, to meet these boundary conditions. In the machine
considered here effort is made to meet the prescribed conditions by use of elastic hinges. These
hinges allow for limited clamp rotations and limited horizontal clamp translations. The nature of
the elastic hinges is such that their behaviour is constant, i.e. not affected by wear & tear, slip-stick
and maintenance such as regular cleaning or lubrication. Furthermore the nature of the hinges
guarantees a set-up which is absolutely free of play. A disadvantage of the hinges is that they
have a very small resistance to deformation. Figure 5 gives an impression of the elastic hinges
in action.
As the clamping force is applied material is squeezed away from the clamped area. This slightly
lengthens the beam resulting in clamp translation, see Figure 5 left. However, it should be clear
that the elastic hinges mainly need to allow for rotation. By an analysis of the elastic hinges that
is beyond the scope of this paper it was determined that the hinges have a rotational stiffness of
14.4 Nm.
8
From the bending beam theory it is known that rotations at the clamps are dependent on the
deflection due to bending in the centre of the beam.
Where Zb = bending deflection at the centre of the beam [m], Ri = rotation of inner clamps [],
Ro = rotation of outer clamps [], ci = constant for determination of inner clamp rotation = 3.727
[1/m], co = constant for determination of outer clamp rotation 25.455 [1/m].
Knowing the rotational stiffness of the elastic hinges equation (13) translates into the following
equation for moments applied at the clamps. Please note that two sets of two flexible elements in
series are applied to hold the outer clamps.
Where Mi and Mo are moments acting on inner and outer clamps respectively [Nm], Mt = moment
acting on the central part of the beam [Nm], si = rotational stiffness of inner clamps = 28.8 [Nm],
so = rotational stiffness of outer clamps = 14.4 [Nm]
The moment acting in the centre of the beam, (F/2) A, is reduced with Mt due to the rotational
stiffness of the clamps. This acts to reduce the 4PB back calculated stiffness. When applying the
equations from the standard this may be obtained by a correction on the force F.
Where Fb = force applied to bend the beam [N], Fm = force measured by the 4PB machine [N].
Taking the effects of clamp stiffness into account the 4PB back calculated stiffness becomes
7437.28 MPa, so reducing the mistake in the back calculation to 0.84%.
3.6 Conclusions
Non-linear geometrical effects in the 4PB considered here may be neglected.
When applying the formulas in the EN standard to FEM calculations a stiffness is back calculated
which is 3.19% too low. The main reasons for this are the following.
1. Neglecting shear deformation: underestimation of E by 3.52%
2. Neglecting clamp effects on cross section: overestimation of E by 3.07%
3. Neglecting clamp movements: underestimation of E by 2.31%
4. Neglecting hinge stiffness: overestimation of E by 0.23%
Taking the above into account the accuracy of the back calculated stiffness will increase
dramatically, theoretically leading to an unexplained error of 0.84%.
4 VISCO-ELASTIC RESPONSE
4.1 Simulation
To investigate the accuracy of the 4PB with respect to phase lag determination a visco-elastic
simulation was done on basis of an asphalt with a stiffness of 7500 MPa and a phase lag of 20 at
20 Hz subjected to 20 Hz loading. In the simulation the clamping force of 130 N per clamp was
applied in 10 seconds. After a 1 sec rest period the 20 Hz displacement controlled loading was
applied. Figure 6 gives an impression of the results obtained after the clamping procedure was
completed. The figure indicates that sinusoidal functions are fitted to the response signals from
the last load cycle applied in the simulated test. The input amplitude of actuator displacement was
0.03532 mm. This analysis leads to the following results.
9
Figure 6. Obtained force and beam deflection signals.
Table 3. Results of a simulated test at 20 Hz on asphalt (stiffness of 7500 MPa and phase lag of 20
Actuator displacement Central deflection Applied force
Amplitude 0.03531 mm 0.03883 mm 111.77 N

Phaselag 18.67 degree 19.61 degree 0.00 degree
When applying the formulas of the EN standard the listed results translate into a material with
a complex modulus of 7266.6 MPa and a phase lag of 19.6 . The error in obtained stiffness is thus
3.11% and the error in phase lag is 0.4 .
Hereafter effort is made to correct these results for the shear effect, clamp effect, clamp
movements and hinge stiffness.
4.2 Shear effect & Clamp effects

As discussed earlier neglecting the effects of shear results in an underestimation of stiffness by
3.52%. Opposite to this the clamps act to restrain cross section deformation resulting in an over-
estimation of stiffness by 3.07%. Following the procedures discussed earlier the back calculated
stiffness E becomes 7291.64 MPa resulting in an error of 2.78% when correcting for these
effects. The phase lag is not influenced by this correction and the error in the lag remains 0.4 .
4.3 Clamp movements

It was determined earlier that the vertical deformation in the outer clamps equals 7.86581E09 m
per N applied force. These deformations effectively reduce the central deflection, Z, and should thus
not be included in the 4PB back calculation procedure. For this reason Z needs to be corrected for
clamp deformation, see equation 12. Since the clamps are made of elastic materials this correction
should act on the elastic (i.e. non-delayed) deformation only. With this equation 12 translates into.
10
Where Zc = corrected beam deflection [m], Zce = corrected beam deflection [m] at maximum F,
Zme = measured beam deflection [m] at maximum F, Zmv = measured beam deflection [m] at
F = 0 N, c = corrected phase lag [ ], Ss = support spring stiffness [N/m], i.e. 127.132 MN/m.
With the above the 4PB back calculated stiffness becomes 7450.3 MPa, so reducing the mistake
in the stiffness back calculation to 0.66%. The correction also affects the phase lag which now
becomes 20.05 leading to an error of 0.05 .
4.4 Hinge stiffness

As discussed the hinges have minor resistance against rotation. As explained this implies that a
fraction of the applied force is used to rotate the clamps and not so much for actually bending the
beam. This effectively reduces the force for bending and thus reduces the back calculated beam
stiffness.
Equation 15 indicates the reduction of force applied to the beam at maximum deflection. To take
this effect into account the following equations are applied.
Where FZmax = force applied at maximum beam deflection corrected for clamp deformation [N],
Fm = amplitude of force applied by the 4PB machine [N], FZ=0 = force applied at zero beam
deflection corrected for clamp deformation [N], c = phase lag corrected for clamp deforma-
tion [ ], Zc = amplitude of beam deflection corrected for shear, and clamps deformation [mm],
Fc = amplitude of applied force corrected for clamp rotation stiffness [N].
When this reduction is taken into account the back calculated beam stiffness becomes 7434.2 MPa
resulting in an error of 0.88%. The phase lag is also affected by this correction and becomes 20.1
leading to an error of 0.1 .
4.5 Conclusions
When applying the EN standard the 4PB as discussed determines a stiffness that is 3.11% too low.
The main reasons for this error are the following.
1. Neglecting shear deformation: underestimation of E by 3.52%
2. Neglecting clamp effects on cross section: overestimation of E by 3.07%
3. Neglecting clamp movements: underestimation of E by 2.18%
4. Neglecting hinge stiffness: overestimation of E by 0.22%
Taking the above into account the accuracy of the back calculated stiffness will increase dramat-
ically, theoretically leading to an unexplained error of 0.88%. The phase lag that is determined is
0.39 too small. The main reasons for this error are the following.
1. Neglecting clamp stiffness: underestimation of lag by 0.44
2. Neglecting hinge stiffness: underestimation of lag by 0.05
Taking the above into account the accuracy of the back calculated lag will increase, theoretically
leading to an unexplained error of 0.1 .
11
5 CONCLUSIONS
It is shown that with Finite Element Modelling a detailed simulation of a 4PB device can be
made. Such a simulation can be used for an investigation of how much reality differs from the
theory. From the previous, taking into account the considered material (E = 7500 MPa, lag at
20 Hz = 20 ), it is concluded that the 4PB test executed at 20 Hz on an ideal machine leads to two
sources of error which both will affect the stiffness but not the phase lag.
Neglecting shear deformation between the outer and inner clamps, resulting in a reduction of
stiffness with 3.52%.
Neglecting the clamp restrained inner cross-sections, resulting in an increase of stiffness with
3.07%.
Other sources of error are found in the 4PB machine considered here. These sources of error are
related to the built quality of the machine and are the following.
Neglecting the hinge stiffness will lead to an increase of stiffness with 0.22% and a reduction
of lag by 0.05 (at 20 material lag)
Neglecting the vertical movements of the outer clams results in a reduction of stiffness with
2.18% and a reduction of lag with 0.44 .
Taking into account the above a source of unknown error remains. This source results in an
underestimation of stiffness with 0.84% to 0.88% (elastic or visco-elastic simulation) and an over-
estimation of lag by 0.1 . It is stated explicitly that the discussed 4PB machine compensates for
errors as discussed via calibration. It is recommended to use at least 3 reference beams with dif-
ferent EI values for the calibration of a device. However, it should be clear that the accuracy of any
machine increases with built quality reducing the need for correction of data. The built quality of
the machine discussed here is more than adequate since errors that follow from built quality remain
smaller than the intrinsic errors in the 4PB test itself.
REFERENCES
NEN-EN 12697-26, 2004, Bituminous mixtures Test methods for hot mix asphalt Part 26: Stiffness,
European Committee for Standardisation, Brussels, July 2004.
NEN-EN 12697-24+A1, 2007, Bituminous mixtures Test methods for hot mix asphalt Part 24: Resistance
to fatigue, European Committee for Standardisation, Brussels, July 2007.
M. Huurman & A.C. Pronk, 2009, Theoretical analysis of the 4 point bending test, Proceedings of the 7th Int.
RILEM Symposium Advanced Testing and Characterization of Bituminous Materials, May 2009, Rhodes,
Greece.
A.C. Pronk, 2006, Theory of the Four Point Dynamic Bending Test, Part I: General Theory, P-DWW.91-008,
2006, Delft, The Netherlands.
12
Description of a procedure for using the Modified Partial Healing

model (MPH) in 4PB test in order to determine material parameters
A.C. Pronk
ACP Consultants, Vlaardingen, The Netherlands
ABSTRACT: Fatigue testing of an asphalt beam in a 4PB device is fatiguing a specimen and not
the material itself. The amount of the local fatigue damage depends on the stresses and strains at
that location. Only the loading at the two inner supports and the deflection of the beam is measured.
A weighted stiffness modulus will be back calculated from the response.
This implicates that for each point in the beam the fatigue damage has to be calculated. After
integration over the beam the calculated stiffness modulus can be compared with the measured
response.
A procedure is described how the Modified Partial Healing model can be used for the calculation
of the weighted beam stiffness modulus. By varying the parameters of the MPH model the calculated
weighted beam stiffness modulus values can be fitted with the measured beam response.
The MPH model allows the introduction of an endurance limit which will eliminate time con-
suming fatigue tests at low strain levels. It is shown that the existence of a low endurance limit
is not in conflict with the Whler curve. Moreover the MPH model might explain the empirical
relationship between dissipated energy and the number of cycles to failure.
1 INTRODUCTION
Fatigue is a major issue in pavement design, but still only a few damage models are available
to describe the evolution of fatigue damage (decrease in the complex stiffness modulus). Many
damage models are based on fatigue tests using 4PB devices. [Shen, 2005; Castro, 2008].
However, the result or outcome of a 4PB fatigue test is the evolution of a weighted beam stiffness
modulus and not the evolution of the (local) stiffness modulus for the material itself.
Nevertheless this approach is worldwide used [AASHTO, 2008; Adhikari, 2010]. In order
to understand and use 4PB fatigue tests, a material damage model has to be used, which after
integration over the beam gives the decrease in the weighted stiffness for the beam.
Such a material model is the Modified Partial Healing model (MPH) which describes very well
the evolution of the complex stiffness modulus (modulus and phase lag) in an uni-axial push-pull
test (UPP or also known as a cyclic tension-compression test T/C), [Pronk, 2001; Pronk, 2005].
The UPP test can be considered as the best available test for which in theory the local stiffness
modulus will be constant throughout the specimen and equals the calculated overall stiffness for
the specimen [Di Benedetto, 2004].
This paper is meant as a guideline for using the MPH model in 4PB fatigue tests with the goal
to deduce material parameters from the measured overall response. But the outlined procedure can
also be used for other material fatigue models.
2 SYMBOLS
i,j = dummy variables (counter)

n = cycle number (equivalent of time)
13
x = Location along the beam (x = 0 to x = L) [m]
z = Location in height with respect to neutral line (H/2 < z < +H/2) [m]
y = Location in width with respect to neutral line (B/2 < y < +B/2) N[m]
A = Distance between an inner and an outer support. Normally A = L/3 [m]
B = Width of beam [m]
Cj,real & Cj,imag = Coefficients in equation for Vb,real and Vb,imag at cycle n []
Dj,real & Dj,imag = Coefficients (constant) in equation for Vb,real and Vb,imag at cycle n [m]
F = Force [N]
L = Distance between the two outer supports. [m]
H = Height of beam [m]
I = Bending moment of inertia [m4 ]
S{x,n} = Modulus of the local complex stiffness modulus [Pa]
Sj {n} = Modulus of the local complex stiffness modulus [Pa]
Sreal {x,z,n} = Local storage modulus [Pa]
Simag {x,z,n} = Local loss modulus [Pa]
Samp {x,z,n} = Modulus of the local complex stiffness modulus [Pa]
Sreal {x,n} = Weighted storage modulus over cross area at location x [Pa]
Simag {x,n} = Weighted loss modulus over cross area at location x [Pa]
Samp {x,n} = Modulus of the weighted complex modulus over a cross area [Pa]
Samp {n} = Modulus of the weighted complex modulus for a beam at cycle n [Pa]
Vb {x,n} = Deflectie due to pure bending of beam at location x for cycle n [m]
Vs {x,n} = Deflectie due to shear force at location x for cycle n [m]
Sreal {x,z,y} = Fatigue damage (decrease in local storage modulus) [Pa]
Simag = Fatigue damage (decrease in local loss modulus) [Pa]
j,real & j,imag = Decrease factor in modulus with respect to S0 cos (0 ) & S0 sin (0 ) []
real & imag = Parameter FL3 /(2S0 cos (0 )I) & FL3 /(2S0 cos (0 )I) [m]
= Fatigue damage parameter []
{x,z} = Local strain amplitude [m/m]
0 = Maximum strain amplitude at z = H/2 for A < x < L-A [m/m]
{x,n} = Phase lag between force and deflection at location x for z = H/2 [rad]
3 THEORY
The response in a 4PB fatigue test is the weighted overall complex stiffness modulus for the beam.
The formulas used for determining this modulus is based on the response of a homogenous beam
(local material stiffness modulus is constant).
The formulas used hereafter are based on the deflection due to bending for the pseudo-static
case but assuming a visco-elastic modulus. Furthermore it is assumed that a real controlled
strain mode is applied in the 4PB fatigue test. This can be achieved by keeping sthe difference
between the deflection in the centre Vb {L/2} and at one of the inner supports: Vb {A} constant
during the fatigue test. The local strain distribution in the beam is than given by equation 1. It
is assumed that the strain and material properties do not depend on the coordinate y (width B
of the beam).
14
The local decrease after one cycle in the complex stiffness modulus S{x,z,n} at cycle n is defined
by equation 2 and should be deduced from the fatigue damage model. It will be discussed later on.
4 WEIGHING OVER A CROSS AREA
The modulus of the weighted complex stiffness modulus per cross area S{x, n} (determined by
the storage and loss modulus (S real {x, n} & S imag {x, n}) at location x for cycle n is represented by
equations 3.
5 WEIGHING ALONG THE BEAM LENGTH L
The stiffness per cross area can now be written as Samp {x, n} ei{x,n} and the deflection due to
bending can be written as Vb,amp {x, n}.ei{x,n} . In view of the weighing along the beam the weighing
procedure for the loss modulus and storage modulus will be carried out separately.
Instead of Samp {x, n} ei{x,n} and Vb,amp {x, n} ei{x,n} the real part (Sreal {x, n} & Vb,real {x, n}) and
the imaginary part (Simag {x, n} &Vb,imag {x, n}) are used.
For the pseudo-static case or at low frequencies the influence of any mass can be ignored on the
phase lag. For high frequencies the original 4th order differential equation has to be the starting part
for the weighing procedure along the x axis introducing a small evolution in the phase lag between
the deflection at that location and the applied force.
For the proposed weighing procedure the real and imaginary parts of the deflections due to
bending along the beam are determined from the differential equations for a slender beam which
is loaded with a static load (equation 4).
15
Equation 4 with the boundary conditions and requirements at x = 0, x = A and x = L/2 has to be
solved for the determination the centre deflections Vb,real {L/2}, Vb,imag {L/2} and Vb,amp {L/2}. The
weighted modulus of the complex stiffness modulus for the beam is defined as the modulus of a
homogenous beam which gives the same deflection Vb,amp {L/2} (equation 5).
The procedure outlined above lead to a weighted modulus (amplitude) for the complex beam
stiffness modulus. In general it will be impossible to solve all equations analytically. Therefore a
discrete numerical approach is described in the next paragraph.
6 NUMERICAL APPROACH
6.1 General
Because of symmetry only a quart part of the beam has to be considered (0 < x < L/2; 0 < z < H/2).
The interval from x = 0 to A is divided in 10 sections (1 to 10) with a length A/10. The interval
from x = A to x = L/2 is divided in 5 sections but because the weighted stiffness is constant on
this interval only one section is needed. The beam height from z = 0 to z = H/2 is also divided
in 10 sections with a height length of H/20. Knowing the local stiffness modulus the weighted
stiffness modulus per cross area for a given value of x is calculated as given in equation 6.
The next step is to give each section j on the interval 0 < x < A the mean value of these weighted
stiffness for the real and imaginary parts at the borders of the section j. The ratio of these weighted
stiffnesss and the stiffnesss S0 sin (o ) and S0 cos (o ) at the start of the tests is denoted as j,imag
and j,real (equation 7).
The interval 0 < x < A is now divided in 10 sections with constant weighted loss and storage
stiffness modulus per section. For the part of the beam between x = A and x = L/2 the weighted
16
loss and storage stiffness modulus per cross area are constant and denoted as:
6.2 Deflection Vb on the interval 0 < x < A

On the sections from j = 1 to 10 ential equation and general solution is given by equation 9 for the
real part. A similar equation is valid for the imaginary part.
The boundary conditions are that both the deflection and its first derivative are continuous at the
crossings xj = (A j/10). For j = 1 the constants D1,real and D1,imag are equal to zero for all n.
The recursion formula for the constant Dj,real {n} is given by equation 10.
6.3 Deflection Vb on the interval A < x < L/2

On the interval A < x < L/2 the deflection has a parabolic shape as shown in equation 11 for the
real part of the deflection. Similar equations hold for the imaginary part.
At the crossing x = A both the deflection Vb,real and its first derivative have to be continuous
leading to a relation between the constants D10,real and D11,real (equation 12). The same is true for
17
D10,imag and D11,imag
The amplitude of the maximum deflection at x = L/2 is given by equation 13.
Because in the back calculation procedure a constant stiffness modulus is assumed for the beam
this results in the following expression for the weighted overall modulus (equation 14).
7 EXAMPLE
The numerical procedure for the weighing over the x axis is checked with the analytical solution
for a given distribution of the weighted (elastic) modulus over a cross area. Suppose that for cycle
n the weighted modulus over the cross area is given by equation 15.
The differential equations are:
The solution of the differential equation for Vb is than given by equation 17.
18
Figure 1. Evolutions of weighted moduli as a function of the parameter .
The constants and follow from the conditions at x = A and are given by equation 18.
For the constant A2 /(6L3 )

The maximum deflection at x = L/2 is given by equation 19.
The overall weighted modulus is calculated assuming a homogenous modulus distribution

(equation 5). This leads to the following expression for the overall weighted modulus.
In figure 1 a comparison is made between the analytical decrease in the overall weighted modulus
and the numerical approach (overlapping purple and red lines). The difference occurs in the third
digit. Also the decrease in the weighted modulus over a cross area for the midsection (A < x < L-A)
is given. The lesser fatigue damage (decrease in stiffness modulus) in the outer section has only a
minor influence of around 5%. The fourth line in the figure is a calculated overall modulus based
on a weighted average using the length of the section as a weight factor. It is clear that the overall
weighted modulus using that protocol is highly overestimated.
Not surprisingly the overall weighted modulus for the beam (which is the result of back calcu-
lation procedures) is only a few percent higher than the weighted modulus for the mid section. The
lesser fatigue damage in the outer sections (x < A & x > L-A) has only a minor increasing effect
19
on the overall weighted modulus. A simple average value for a overall modulus using the length of
a section as a weight factor overestimate the influence of the outer sections on the overall weigthed
modulus. It has to be marked that in an earlier publication (. . .) the simple average protocol was used.
8 DAMAGE MODEL
For the decrease in the loss and storage moduli a material damage model will be necessarily
[Bodin, 2004; Shen, 2005; Castro, 2008; Kim, 2008]. Another suitable model is the Modified
Partial Healing (MPH) model [Pronk, 2010]. This model can describe the evolutions of the loss and
storage moduli in an uni-axial push-pull (UPP) test very well. The UPP test can be considered as the
best available test for which the response can be seen as the response of the material [Di Benedetto,
2004]. In principle UPP tests should be carried out for the determination of the parameters of the
MPH model. However, the protocol describes in this paper will enable the determination of the
parameters using 4PB fatigue tests.
The MPH model describes two processes. The decreases in moduli due to the first process
are reversible. During the test an equilibrium will be reached. It can be seen as a mathematical
description of thixotropy. The second process causes real fatigue damage which during a continuous
fatigue test will not heal (irreversible). This process has also an endurance limit which mean that
below a certain strain level no (permanent) damage occurs.
The equations for the PH model are given by equations 21.
Where F{} = loss modulus [Pa], G{} = storage modulus [Pa], dQ{}/d = fatigue damage rate
[Pa/s], t = time [s], 1,2 , , and 1,2 are constants.
The fatigue damage rate is supposed to be related to the dissipated energy in one cycle (Wdis )
For a strain controlled test this leads to equation 22.
Where = constant (<<1), Wdis. = dissipated energy per cycle, T = duration of one cycle,
f = frequency [Hz]. Defining 1,2 20 = 1,2 f 20 and 1,2 2
0 = 1,2 f 20 and using the results of
the UPP tests [Di Benedetto, 2004] it was found that for the mix at issue the constant 1 was nil,
that the constant 2 was linear in the applied strain, that the constant was related to the square

of the applied strain and that the constants 1,2 have a trigger value lim. . Renaming the constants
2
in equation 21 (1,2 = 1.2 0 ; = 0 and 1,2 = 1,2 (0 lim . ) and assuming that the constant

1 (if not nil) will be also linear in the applied strain, equation 23 is deduced which is the base of
the MPH model. The constants 1,2 and 1,2 with a double asterix, when divided by the frequency
f, are assumed to be material constants.
20
Table 1. Strains in m/m and fatigue lives in kcycles.
Beam code
O O O A A A A A A B B B
501 503 504 501 503 504 601 602 604 601 602 603
0 137 218 177 137 178 217 138 219 179 218 137 177
N1 510 48 170 400 130 42 200 90 160 120 270 120
Nf 1200 110 340 1000 270 90 430 200 270 200 460 215
Table 2. Parameter values for the modified PH model in fitting the 4PB tests.
Beam 6
2 [10 ] 1 [106 ] 2 [106 ] [103 s1 ] lim [m/m]
0501 6.15 0.338 1.35 61.5 77

0503 6.15 0.426 1.32 60.9 75
0504 6.15 0.278 1.36 46.7 85
A501 6.15 0.344 1.35 60.8 65
A503 6.15 0.273 1.36 55.1 80
A504 6.15 0.342 1.35 71.0 36
A601 6.15 0.292 1.36 51.0 52
A602 6.16 0.397 1.63 69.9 137
A604 6.15 0.337 1.35 51.6 79
B601 6.15 0.340 1.35 66.5 111
B602 6.15 0.342 1.35 43.3 63
B603 6.15 0.249 1.37 64.4 78
Mean 6.15 0.33 1.37 58.6 78
St.dev. 0.00 0.05 0.08 9.0 26
UPP Mean 6.15 0.24 1.36 52 74
Equation 23 can be used directly for the results of UPP tests and for 4PB tests using the explained
numerical weighted procedure as described before.
9 APPLICATION
The procedure described above is applied to 4PB tests carried out in a RILEM project [DiBenedetto,
2004] These tests were carried out in controlled deflection mode with a frequency of 9.8 Hz and at
a temperature of 20 C. The applied strains and fatigue lives are given in table 1.
The fatigue life Nf is defined as the number of cycles at which the beam stiffness modulus Smix
is decreased to half of its initial value which is defined as the modulus obtained in the 100th cycle.
The fatigue life N1 is defined as the number of cycles at which a sudden change in the evolution
of the dissipated energy per cycle is observed. This moment can be visualized by plotting the ratio
of the dissipated energy in cycle n and the accumulated dissipated energy up to cycle n and was
already introduced in 1989 [Hopman, 1989; Pronk, 1990].
The test results of this RILEM project are already used in another publication [Pronk, 2010] but
instead of the weighing procedure as described in this paper, a simple linear weighing procedure
was used in that publication. That procedure leads to an overestimation of the overall beam stiffness
modulus (see figure 1). Also the strain values were not corrected for the deflection due to shear
forces.
21
Figure 2. Measured and calculated response for beam O-501 (0 = 132 m/m).
Using the weighing procedure as outlined above for each measured (captured) cycle the storage
modulus, loss modulus, complex modulus and phase lag for the weighted beam stiffness modulus
were calculated. These values were compared with the measured values. The summed squared
differences between measured and calculated values over the interval from cycle 1000 to N1 were
minimized, using the solver option in Excel by variation in the (material) constants, the endurance
limit lim. , the initial storage modulus and the loss modulus. The results are presented in table 2.
Some examples of the obtained regression fits are given in figures 2 to 4.
As shown in table 2 the values for the constants
2 and 2 do not vary at all. Also the variations
in the other constants are not too big given the fact that these 4PB tests were not strain but deflection
controlled. The results are also in good agreement with the values obtained from UPP tests (last
row in table 2) carried out on the same mix [Pronk, 2010].
As can be seen in the figures 2 to 4 the model starts to deviate from the measured response long
before the traditional fatigue life definition Nf . But the point of deviation is close to the fatigue
life definition N1 (O-501: 510 kcycles; A604: 160 kcycles; B603: 120 kcycles).
10 ENDURANCE LIMIT AND WHLER CURVE
Often the results of fatigue tests (fatigue lifes as a function of the applied strain) are plotted on
log-log scale and a linear fit is made leading to a relation as given by equation 24.
This presentation is called a Whler curve. Equation 24 does not contain an endurance limit as
in the MPH model. Suppose that only the last terms in equations 23 (with the endurance limit)
are responsible for fatigue damage. The first terms are related to the thixotropic behaviour of the
mix. The accumulated dissipated energy up to cycle N1 can be represented as a constant times N1
because the decrease in the dissipated energy per cycle is allmost linear as shown in figures 2 to
4 for the weighted loss modulus. Assume that failure at the top (and bottom) in the midsection of
22
Figure 3. Measured and calculated response for beam A-604 (0 = 173 m/m).
Figure 4. Measured and calculated response for beam B-601 (0 = 210 m/m).
the beam occurs when the accumulation of the fatigue damage reaches a certain level. In view of
equations 23 equation 25 can be a possible representation of measured results.
In figure 5 calculated (fictive) fatigue lives using equation 25 (constant = 5.00 1011 ; endurance
limit 50 micro strain) are plotted as a function of the strain. As clearly shown a power fit through
23
Figure 5. Fictive fatigue life N1 as a function of the applied strain according to equation 25 and a power fit
(equation 24) on the interval N1 = 100 to N1 = 1,000,000 cycles.
the points on the interval from N1 = 100 to 1,000,000 cycles gives an excellent fit. Only when the
strain approaches the endurance limit the deviations from a power fit increase.
Mark that in the equation of the power fit the exponent for the strain increases to 3.22. This is
close to the exponent of 3.9 of the Whler curve for the determined fatigue lives N1 and applied
strains in this project. In the authors view it will be worthwhile to investigate this concept more in
detail by using equations 23 for a theoretical determination of the accumulated fatigue damage up
to cycle N1(see Annex I). In the authors view it is questionable if the dissipated energy related to
thixotropy (first terms in the intehrals of equations 23) should be taken into account.
11 DISCUSSIONS & CONCLUSIONS
It is clear that the decrease of the local material complex modulus in a bending fatigue test depends
on the local stress and strain situation. As a consequence the back calculated modulus using the
deflection response is a weighted specimen stiffness modulus. To obtain the local material modulus
decrease a material damage model has to be used. After integration over the specimen a weighted
overall modulus for the specimen can be obtained and compared with the back calculated stiffness
modulus from the tests results.
A numerical procedure (protocol) is given for the calculation of the weighted overall modulus
for a beam with the aid of a suitable fatigue damage model.
It is assumed that the decrease in the local stiffness modulus depends only on the local strain
and stress situation.
The model can be used for bending fatigue tests in strain controlled mode. In that case the strain
distribution depends solely on the geometry of the specimen and the fatigue bending apparatus.
The weighted modulus per cross section has to be calculated from the local stiffness modulus
with the square of the distance to the neutral axis as the weighing factor.
For the determination of the weighted overall modulus for the specimen the response equation
for the deflection should be used for the weighing procedure
In case of a linear decrease with the length in the weighted modulus per cross section the
difference between the analytical calculated overall weighted modulus and the result of this
numerical procedure is nil.
24
The parameters for the Modified Partial Healing model (MPH) deduced from 4PB tests are in
good agreement with the values deduced from UPP tests on the same mix.
In the Modified Partial Healing model (MPH) an endurance limit is included which does not
conflict with the Whler equation.
REFERENCES
AASHTO T 321. 2008. Standard Method of Test for Determining the Fatigue Life of Compacted Hot-Mix
Asphalt (HMA) Subjected to Repeated Flexural Bending, AASHTO, Washington D.C.
Adhikari, S. & You, Z. 2010. Fatigue Evaluation of Asphalt Pavement using Beam Fatigue Apparatus. In The
Technology Interface Journal/Spring 2010, Volume 10 No. 3.
Di Benedetto, H., De La Roche, C., Baaj, A., Pronk, A.C. & Rundstrm, R. 2004. Fatigue of bituminous
mixtures. In Materials and Structures, 2004, Vol.37, no. 3, pp. 202216.
Bodin, D., Pijaudier-Cabot, G., De La Roche, C. & Piau, J-M. 2004. A continuum Damage Approach of
Asphalt Concrete Fatigue Tests, Journal of Engineering Mechanics/Volume 130/Issue 6/Technical papers.
Castro, M. & Snchez, J.A., 2008. , Estimation of Asphalt Concrete Fatigue Curves A Damage Theory
Approach. In Construction and Building Materials,2008, 22 (6), pp. 12321238
Hopman, P.C., Kunst, P.A.J.C. & Pronk, A.C. 1989. A renewed interpretation method for fatigue
measurements Verification of Miners rule. In Proc. of the 4th Eurobitume Symposium, 1989, Madrid,
Spain.
Kim, Y.R., Baek, C., Underwood, B.S., Subramanian, V. & Guddati, N. 2008. Application of viscoelastic con-
tinuum damage model based finite element analysis to predict the fatigue performance of asphalt pavements.
In Korean Soc. of Civ. Eng., 2008, issue 12(2), pp. 109120.
Pronk, A.C. & Hopman, P.C. 1990. Energy dissipation: The leading factor of fatigue. In Proceedings Strategic
Highway Research Program: Sharing the Benefits, 1990, London, UK.
Pronk, A.C., 2001. Partial healing in fatigue tests on asphalt specimens. In Int. Journ. Road Materials and
Pavement Design, 2001, Volume 4, Issue 4.
Pronk, A.C. & Cocurullo, A. 2009. Investigation of the PH model as a prediction tool in fatigue bending
tests with rest periods. In A. Loizos, M. Partl, T. Scarpas & I. Al-Qadi (eds.), Advanced Testing and
Characterization of Bituminous Materials, 2009, Taylor & Francis group, London, UK.
Pronk, A.C. & Molenaar, A.A.A., 2010. The Modified Partial Healing Model used as a Prediction Tool for
the Complex Stiffness Modulus Evolutions in Four Point Bending Fatigue Tests based on the Evolutions in
Uni-Axial Push-Pull Tests. In Proc. 11th Int. Conf. on Asphalt Pavements, 2010, Nagoya, Japan.
Shen, S. & Carpenter, S.H. 2005. Application of Dissipated Energy Concept in Fatigue Endurance LimitTesting.
In Transp. Res. Rec.: Journ. of the Transp. Res. Board, No. 1929, Washington, D.C., 2005, pp. 165173.
ANNEX I POSSIBLE ITEMS FOR FURTHER RESEARCH
The general form for the solution of equation 23 is given in Pronk, 2009. Suppose instead of a
linear dependency (0 lim .) a relation of the form (0 lim .) for the endurance limit. If the
coefficient 1 in the equation for the loss modulus turns out to be nil (as for this mix), the solution
is than given by equation 26 for the loss modulus at the top and bottom of the beam in the mid span
section..
It is assumed that the dissipated energy which is related to the thixotropic behavior of the material
will not contribute to the fatigue damage. Thus the accumulated dissipated energy Wfatigue up to
cycle N causing fatigue damage for the top and bottom of the mid span section (A < x < L-A) will
be given by equation 27.
25
Figure 6. The dissipated energy ratio for beam O-501 (test data and MPH model) as a function of the cycle
number n.
A first order approximation for equation 27 is given by equation 28.
The fatigue life N1 is based on a change in the evolution of the dissipated energy Wdis per
cycle. The change is visualized by plotting the ratio of the dissipated energy in cycle n and the
accumulated dissipated energy up to cycle n as a function of n. However it might be better to use
the model itself for a more realistic determination of the fatigue life based on a deviation between
model and measured data. For the regression fit of the damage model only the data up to N1 are
used. As can be seen in figures 2 to 4 the deviations between model and data occur later than N1.
Comparing the dissipated energy ratio with the value according to the model will lead always to a
bit higher number of cycles where the deviation between the test data and the model starts. That
number can be seen as the fatigue life NPH for the beam. An example is given in figure 6 for beam
O-501.
26
Assessment of different flexure fatigue failure analysis methods

to estimate the number of cycles to failure of asphalt mixtures
M.I. Souliman, W.A. Zeiada, K.E. Kaloush & M.S. Mamlouk

Arizona State University, Tempe, Arizona, US
ABSTRACT: Load associated fatigue cracking is one of the major distress types occurring in
flexible pavements. Beam fatigue testing has been used for several decades and is considered an
integral part of advanced characterization procedures. However, there have been several fatigue
failure analysis methods based on the failure stiffness of the mixture. These methods may not
produce the same results, and vary depending on the method of detecting the failure point. In
this paper, beam fatigue test results conducted at Arizona State University were used for the
analysis. The mixtures included conventional, Asphalt Rubber (AR), and Fiber-Reinforced (Aramid
and Polyolefin) asphalt mixtures. Deflection controlled fatigue tests were conducted according to
American Association of State Highway and Transportation Officials (AASHTO T321) procedure
at three temperatures. Paired-t statistical analysis approach was utilized to compare five different
fatigue failure methods. These methods are commonly known as the Pronk, Hopman, Carpenter,
ASU, and Rowe. An additional method using the Franken fitting model was also introduced. A
recommendation is presented to select a potential method that unifies the current fatigue analyses
using a rational energy-based approach. This new method provides a well-defined fatigue failure
point compared to other methods.
1 INTRODUCTION
The flexural fatigue test is used to characterize the fatigue life of Hot Mix Asphalt (HMA) at
intermediate pavement operating temperatures. This characterization is useful because it provides
estimates of HMA pavement layer fatigue life under repeated traffic loading. In a well-designed
pavement, strains in the pavement are low enough so that fatigue is not a problem. However, when
pavement is under-designed strains are sufficiently high to cause fatigue failures under repeated
loads. These failures ultimately result in fatigue cracking which will cause disintegration of the
pavement if not maintained on time.
The basic flexural fatigue test subjects a HMA beam to repeated flexural bending in a controlled
atmosphere. The standard beam fatigue procedure is found in: AASHTO T 321 (2003). The flexural
fatigue test has been used by various researchers to evaluate the fatigue performance of pavements
(SHRP A-404, 1994, Harvey and Monismith, 1993, Tayebali et al., 1995, Abojaradeh et al., 2007).
There have been several fatigue failure methods based on the failure stiffness of the mixture.
However, these methods may not produce the same results and vary depending on the method of
detecting the failure point. There is a need to select a reliable, accurate and simple fatigue failure
method.
2 STUDY OBJECTIVE
The primary objective of this study was to present different fatigue analysis methods including the
mechanistic dissipated energy and stiffness degradation methods. Subsequently, a comparison was
27
made between all the fatigue analysis results to asses which method is the most reliable, accurate,
and simplest to be implemented as the major method for fatigue failure analysis. To accomplish
this objective, a laboratory testing program was performed on three projects which included con-
ventional, Asphalt Rubber (AR)-, and fiber reinforced asphalt mixtures. A recommendation of the
potential fatigue failure method was presented.
3 BACKGROUND
There are two main fatigue failure definition concepts; dissipated energy based methods and
stiffness degradation based methods.
3.1 Dissipated energy based methods

When applying load to a material, the material will exhibit some strain induced by the acting stress.
The area under the stress-strain curve represents the energy being input into the material. When the
load is removed from the material, the stress is removed and the strain is recovered. If the loading
and unloading curves coincide, all the energy put into the material is recovered after the load is
removed. If the two curves do not coincide, there is energy lost in the material. This energy can
be altered through mechanical work, heat generation, or damage in the material in a manner that it
could not be used to return the material to its original shape. This energy difference is the dissipated
energy of the material caused by the load cycle. Therefore, dissipated energy can be defined as the
damping energy or the energy loss per load cycle in any repeated or dynamic test. Many researchers
have implemented the dissipated energy concept in defining fatigue failure (SHRP A-404, 1994,
Carpenter et al., 2003, Rowe, 1993, Tayebali et al., 1993, Van Dijk, 1975, Van Dijk and Vesser,
1992, Tayebali et al., 1992, and Baburamani and Porter, 1996).
3.2 Carpenter method

Carpenter defined failure as the change in dissipated energy between cycles a and a + 1 (DE)
divided by the dissipated energy of load cycle a (DE) (Carpenter et al., 2003). This change was
calculated approximately every 100 load cycles. The energy ratio (DE/DE) was plotted versus
load cycles gives a curve that decreases rapidly during the first portion of cycles then continues
constant for a long number of cycles and then increases rapidly at the end. The failure point (Nf )
was defined as the number of load cycles at which the change in the energy ratio begins to increase
rabidly as shown in Figure 1.
3.3 Pronk method

Pronk (1997) suggested a different expression of energy ratio to define failure as the ratio of the
cumulative dissipated energy up to cycle n (Wn ) to the dissipated energy for cycle n (wn ). Under
constant strain, plotting the energy ratio for different cycles versus the number of load cycles for
each specimen, the failure point (Nf ) was defined as the number of load cycles when the energy
ratio deviates from a straight line as shown in Figure 1.
3.4 Hopman method

Hopman (Pronk and Hopman, 1990) defined the energy ratio as: ER = n * wo /wn , where, ER = the
energy ratio; n = load cycle; wo = initial dissipated energy; wn = dissipated energy at n load cycle.
By plotting the energy ratio versus the number of load cycles under controlled deflection, the failure
point (Nf ), the transition between micro and macro crack formation, is defined as the number of
load cycles as the energy ratio deviates from a straight line as shown in Figure 1.
28
Figure 1. Fatigue failure methods.
3.5 Stiffness degradation based methods

This concept utilizes the stiffness reduction over time as the main indication of fatigue failure.
Three main methods use this concept to capture fatigue failure point; Rowe, ASU, and Franken
methods.
3.6 Rowe method

The definition of failure recommended by the Hopmans method was further improved by Rowe.
A peak value was produced rather than fitting an arbitrary straight line through the data set (Rowe,
1993). Stiffness ratio was simplified for both controlled force and controlled deflection as: SR =
n * S, where, SR = the stiffness ratio; n = load cycle; S = flexural stiffness. By plotting the energy
ratio versus the number of load cycles, the fatigue life was defined as the peak of the curve as
shown in Figure 1.
3.7 ASU method

ASU method was developed on the Rowe failure definition (Abojaradeh et al., 2007) Rowe ratio
(Ni Si ) was normalized by dividing it by the initial stiffness (So), a new stiffness ratio is developed
using this method (Ni *Si /So ), where Ni is the cycle number, Si is the stiffness at cycle i, and So
is the initial stiffness taken at cycle number 50. By plotting the stiffness degradation energy ratio
versus the number of load cycles, the fatigue life was defined as the peak of the curve as shown in
Figure 1.
3.8 Franken model fitting

This method utilizes Franken model that has been used before to fit the permanent deforma-
tion test data to determine the flow number for asphalt mixtures by Biligiri et al. (2007).
A similar approach can be used for fatigue test data. The Franken model is described as:
29
Table 1. Binder mix design data.
Design asphalt Target air

Mix type Binder type cement, AC (%) voids, Va (%)
Conventional PG 64-22 5.4 7

AR-Open* PG 58-22+R 8.8 18
Fiber-Reinforced PG 70-10 5.0 7
*Typical Arizona AR open graded mixtures have AC% between 8 and 10% and Va between 18 to 20%.
Y = 1 [A*XB + C*(expDX 1)], where, Y = stiffness; X = number of loading cycles; A, B,

C, and D = Beam specific regression constants. The Franken model is a composite of a power
function along with an exponential, which is able to fit all three stages (initial, secondary, tertiary
stages) of the fatigue test data as shown in Figure 1.
4 TEST RESULTS AND ANALYSIS
4.1 Mixtures characteristics

Three mixtures were utilized in this study, which were constructed in field projects in Arizona in the
United States. The first project evaluated in this study was a conventional mixture; named Kohls
Ranch and is a pavement preservation project, about 3 miles in length located within in the city
of Payson (Kaloush et al., 2008). The second project evaluated in this study was an AR mixture;
named Antelope Wash and is about 6.8 miles in length located on US 93 (Kaloush et al., 2008.
The third project was a Fiber-Reinforced mixture that was placed on a local road within the City
of Tempe (Kaloush et al., 2008, Kaloush et al., 2010). Table 1 shows the binder mix design, while
Table 2 shows the aggregate gradations for the each mixture.
4.2 Determination of Nf using all methods

Based on the earlier mentioned failure methods and procedures, Nf was calculated for all three
mixtures at three temperatures and three strain levels in order to investigate whether the change in
mixture type or testing temperature changes the Nf determination trend or not. A summary of Nf
values for the three mixtures at different temperatures are presented in Figures 2 through 4 which
illustrate the variability of determining Nf based on the fatigue failure methods utilized (Kaloush
et al., 2008 (1), Kaloush et al., 2008 (2), Kaloush et al., 2008 (3), Kaloush et al., 2010).
Figures 24 show that the ASU and Rowe methods produced the highest predicted Nf values
regardless of testing temperature or mixture type. The ASU method has the same Nf prediction as
the Rowe method since it only introduced a minor modification to the Rowe method by dividing
Rowe ratio (Ni Si ) by the initial stiffness (So ). These two methods are stiffness based methods; they
do not require any dissipated energy calculation. Furthermore, the definition of Nf used in these
methods is very simple (peak of the relation between N,Si /So and number of cycles).
The Franken method tends to predict the lowest Nf value among all the other methods. The main
reason behind this is based on the failure assumption used in this method, which defines Nf once
the SR-N curve enters the tertiary stage. Thus, the Franken model tends to be very sensitive to any
change in the slope between the secondary stage and the tertiary stage. Although this method holds
a reasonable failure assumption, further improvement needs to be addressed.
The Nf predictions based on Pronk, Hopman, and Carpenter methods require more date form the
acquisition system like energy dissipated for each loading cycle and the overall dissipated energy.
Moreover, the definition of Nf using these methods is hard to define and requires more time to
30
Figure 2. Nf estimation for conventional mixture at three temperatures.
Figure 3. Nf estimation for AR mixture at two temperatures.
Figure 4. Nf estimation for Fiber-Reinforced mixture at three temperatures
31
Table 2. Aggregate gradations.
Aggregate gradation Conventional AR Fiber-Reinforced
Gradation (% Passing 1 (26.5 mm) 100 100 100

by mass of each sieve) 3/4 (19 mm) 100 100 95
1/2 (12.5 mm) 85 100 85
3/8 (9.5 mm) 74 100 75
1/4 (6.3 mm) 65 72
No. 4 (4.75 mm) 62 40 58
No. 8 (2.36 mm) 46 8 44
No. 10 (2 mm) 42 8
No. 16 (1.18 mm) 32 6
No. 30 (600 m) 24 4 24
No. 40 (425 m) 20 4
No. 50 (300 m) 16 3
No. 100 (150 m) 8 2
No. 200 (75 m) 4.6 2.2 4
Table 3. Example of t-paired statistical outcomes.
Example of different methods Example of similar methods

Hopman vs ASU Pronk vs Rowe
Observations per method 3 Observations per method 3

Pearson Correlation 0.9866 Pearson Correlation 0.9881
Degree of freedom 2 Degree of freedom 2
t Stat 6.1943 t Stat 1.9247
P(T<=t) one-tail 0.0125 P(T<=t) one-tail 0.097
t Critical one-tail 2.9199 t Critical one-tail 2.9199
analyze the results. The Nf value is measured when the relation between energy ratio versus loading
cycle starts to deviate from a straight line.
4.3 Paired-t statistical analysis of the equality of Fatigue Analysis methods

Statistical analysis was performed for the six failure methods to investigate the statistical signif-
icance between them. The statistical analyses used a paired t-distribution for each two methods.
A statistical hypothesis, two-population test (Ho : 1 = 2 {Null Hypothesis} and H1 : 1 = 2
{Alternative Hypothesis}) were conducted. A significance level of 5% was assumed and the
acceptance method for a given hypothesis was obtained when tcritical tstat /2, , where is the
degree of freedom.
The statistical analysis confirmed the visual conclusions regarding the dissimilarity in predict-
ing Nf between different failure methods. Four methods are statistically the same; ASU, Rowe,
Carpenter, and Pronk. The paired-t test showed that Hopman and Franken methods produce statis-
tically different results than the other four. Table 3 shows an example of the t-paired comparison
test results that show a statistical difference between the Hopman and ASU methods, while the
Pronk and Rowe methods produce statistically comparable Nf predictions. Table 4 shows complete
paired-t statistical results for the conventional mixture at the three temperatures. Some methods
were statistically the same. Others were different. Similar statistical trends were obtained from
AR and Fiber- Reinforced mixtures.
32
Table 4. Example of t-paired statistical results for conventional mixture.
Temperature
Comparison pair 40 F (4.1 C) 70 F (21.1 C) 100 F (37.8 C)
Pronk vs Hopman Different Different Same

Pronk vs ASU Same Same Different
Pronk vs Rowe Same Same Different
Pronk vs Carpenter Same Same Different
Pronk vs Franken Different Different Same
Hopman vs ASU Different Different Same
Hopman vs Rowe Different Different Same
Hopman vs Carpenter Different Different Same
Hopman vs Franken Same Different Different
ASU vs Rowe Same Same Same
ASU vs Carpenter Same Different Different
ASU vs Franken Different Different Different
Rowe vs Carpenter Same Different Different
Rowe vs Franken Different Different Different
Carpenter vs Franken Different Different Different
5 CONCLUSIONS AND RECOMMENDATIONS
Constant-deflection flexural tests were performed according to the AASHTO T321-2003 procedure
to evaluate the prediction difference of Nf among six fatigue failure methods. Three mixtures were
used; conventional, AR and fiber reinforced dense graded mixtures.
Using the paired-t statistical analysis approach, it is recommended to use the Arizona State
University (ASU) and Rowe methods, which are independent of the testing temperature or mixture
type; these two methods were the most accurate, reliable, and simplest to analyze the beam fatigue
results. These methods have also the potential for unifying current fatigue analyses using a rational
stiffness-based approach and it has a well-defined fatigue failure point compared to other dissipated
energy approaches. In addition, the ASU method was recommended as part of NCHRP 9-19 project
(Abojaradeh et al, 2007).
REFERENCES
AASHTO Designation: T321-03. (2003), Determining the fatigue life of compacted Hot- Mix Asphalt (HMA)
subjected to repeated flexural bending.
Abojaradeh, M.A., Witczak, M.W., Mamlouk, M.S., and Kaloush, K., (2007) Validation of Initial and Failure
Stiffness Definitions in Flexure Fatigue Test for Hot Mix Asphalt. Vol. 35, No. 1, ASTM, Journal of Testing
and Evaluation, pp. 95102.
Baburamani, P. S., and Porter, D. W. (1996), Dissipated energy approach to fatigue characterization of asphalt
mixes. Proceeding combined 18th ARRB TR Conference Transit New Zealand Symposium, pp. 327347.
Biligiri, K., Kaloush, K., Mamlouk, M., Witczak, M., (2007), Rational Modeling of Tertiary Flow for Asphalt
Mixtures. Journal of the Transportation Research Board, No. 2001, pp 6372. Washington, D.C.
Carpenter, S. H., Ghuzlan, K, and Shen, S. (2003), Fatigue endurance limit for highway and airport pavements.
Transportation Research Record: Journal, No. 1832, pp. 131138.
Harvey, J., and Monismith, C.L., (1993), Effect of laboratory asphalt concrete specimen preparation variables
on fatigue and permanent deformation test results using strategic highway research program a-003a proposed
testing equipment. Transportation Research Board, Record 1417.
Kamil E. Kaloush and Krishna Prapoorna Biligiri, (2008), Performance Evaluation of Asphalt Concrete Mix-
tures In Arizona Kohls Ranch Project, Final Report Submitted to Arizona Department of Transportation,
Materials Group, 1221 North 21st Avenue, Phoenix, AZ 85009, USA.
33
Kamil E. Kaloush, Krishna P. Biligiri, Maria C. Rodezno, Waleed A. Zeiada and Luiz de Mello, (2008 ),
Performance Evaluation of Asphalt Rubber Mixtures in Arizona Burro Creek And Antelope Wash Projects,
Final Report Submitted to Arizona Department of Transportation, Materials Group, 1221 North21st Avenue,
Phoenix, AZ 85009, USA.
Kamil E. Kaloush, Krishna P. Biligiri, Carlos Cary, Smita Dwivedi, Jordan Reed, Carolina Rodezno, and
Waleed A. Zeiada, (2008), Evaluation of FORTA Fiber-Reinforced Asphalt Mixtures Using Advanced
Material Characterization Tests Evergreen Drive, Tempe, Arizona, Final Report Submitted to FORTA
Corporation, Grove City, PA.
Kamil Elias Kaloush, Krishna Prapoorna Biligiri, Waleed A. Zeiada, Maria Rodezno, and Jordan Reed, (2010),
Evaluation of Fiber-Reinforced Asphalt Mixtures Using Advanced Material Characterization Tests. Journal
of Testing and Evaluation, ASTM International, Vol. 38, No. 4, pp. 400411.
Pronk, A.C., Comparison of 2 and 4 point fatigue tests and healing in 4 point dynamic bending test based on
the dissipated energy concept. Proceedings of the eighth international conference on asphalt pavements,
Seattle, Washington, pp. 987994, 1997.
Pronk, A.C., and P.C. Hopman, (1990), Energy Dissipation: The Leading Factor of Fatigue. Highway Research:
Sharing the Benefits. Proceedings of the Conference the United States Strategic Highway Research Program.
London.
Rowe, G. M. (1993), Performance of asphalt mixtures in the trapezoidal fatigue test. Association of Asphalt
Paving Technologists, Vol. 62, pp. 344384.
SHRP-A-404. (1994), Fatigue response of asphalt-aggregate mixes. Asphalt research program. Institute Of
Transportation Studies, University Of California, Berkeley and Strategic Highway Research Program,
National Research Council.
Tayebali, A. A., Deacon, J. A., and Monismith, C. L. (1995), Development and evaluation of surrogate fatigue
models for SHRP. A-003a abridged mix design procedure. Journal of the Association of Asphalt Paving
Technologists, Vol. 64, pp. 340366.
Tayebali, A. A., Deacon, J. A., Coplantz, J. S., and Monismith, C.L. (1993), Modeling fatigue response of
asphalt aggregate mixtures. Association of Asphalt Paving Technologists, Vol. 62, pp. 385421.
Tayebali, A. A., Rowe, G.M., and Sousa, J.B. (1992), Fatigue response of asphalt aggregate mixtures.
Association of Asphalt Paving Technologists, Vol. 62, pp.385421.
Van Dijk, W. (1975), Practical fatigue characterization of bituminous mixes. Association of Asphalt Paving
Technologists, Vol. 44, p. 38.
Van Dijk, W., and Vesser, W. (1992), The energy approach to fatigue for pavement design. Association of
Asphalt Paving Technologists, Vol. 46, pp. 140, 1977.
34
Investigation into the size effect on four-point bending fatigue tests
N. Li, A.A.A. Molenaar & M.F.C. van de Ven

Delft University of Technology, Delft, The Netherlands
A.C. Pronk
ACP Consultancy, Vlaardingen, The Netherlands
S. Wu
Wuhan University of Technology, Wuhan, Hubei, P.R.China
ABSTRACT: Four point bending (4PB) tests have been widely used to determine the flexural
stiffness and fatigue properties of asphalt mixtures. However, the test results from different countries
or different test equipments are hardly comparable to each other. The reason is that besides the
test conditions, like temperature, frequency, waveform shape, strain level, etc., specimen size,
configuration of the set-up, frame stiffness and clamping force also play an important role. In
this paper, 4PB fatigue tests were carried out using beams with three different sizes. In the back
calculation of the stiffness modulus, the deflections due to shear force and frame deformation
were taken into account. The Modified Partial Healing (MPH) model was applied to simulate
the evolutions of the stiffness modulus and the phase angle. The endurance limit was determined
by the model parameters. Furthermore the fatigue lives, defined by means of classical approach,
dissipated energy ratio and the MPH model were compared. The results show that for the beam
with a high height over length ratio, shear deflection and frame deformation have a significant
influence. The specimen size dramatically impacts the measured flexural stiffness, fatigue life and
endurance limit.
1 INTRODUCTION
In asphalt pavement design, stiffness modulus and fatigue properties of asphalt mixtures are very
important parameters. Four point bending test using prismatic beams are widely used for measuring
the flexural stiffness and assessing the fatigue resistance of asphalt mixtures [EN 12697-24, 2004]
[AASHTO, 2003]. In theory, the loading is applied in the middle part of the beam through two
inner clamps during the test. A uniform strain distribution with respect to the length is created in
this area and the maximum stress occurs at the surface in this midsection of the bottom and/or top
of the beam.
As a standard laboratory test, the 4PB test is required to be repeatable and comparable. However,
so far there are no unified criteria for the 4PB test. The test results from different countries and
different test equipments are difficult to compare with each other even when the same test conditions
were used. Di Benedetto et al, [2004] reported on an interlaboratory investigation organized by the
RILEM 182-PEB Technical Committee. In this project, 4-point bending tests were conducted with
the same asphalt mixtures in four different countries. The results showed that the fatigue lives were
dependent on the test equipment and specimen size. To standardize the outcome of the 4PB test
devices, a collaborative study was started between 30 different institutes in Europe. Pronk [2009]
compared all the 4PB test equipments involved in the project. These equipments are different in
set-up configuration, clamping system, correction of the measured deflection, data collection and
processing. These differences are the main obstacles for harmonizing of the results from different
4PB test devices.
35
This paper focus on the investigation of specimen size effect on the 4PB fatigue test results. In the
back calculation of the stiffness modulus, the measured deflection was corrected by considering
shear and frame deformation. For simulating the evolution of the stiffness modulus and phase
angle during the test, the modified partial healing (MPH) model was applied to the test results. The
fatigue life NPH defined by the MPH model was compared with the classic fatigue life Nf ,50 and
the fatigue life NR defined by the dissipated energy ratio (DER).
2 THEORY OF THE MPH MODEL
When analyzing the results of a 4PB test, not the local stiffness of the material but the so-called
weighted stiffness of the specimen is used. Pronk developed the modified Partial Healing (MPH)
model based on the dissipated energy theory [Pronk 2001] [Pronk 2006] [Pronk 2010]. The MPH
model is a material model that is able to describe the evolution of the complex modulus and
phase angle for a volume unit during a fatigue test. After integration over the dimensions and the
stress-strain distribution of the specimen, a weighted complex modulus can be calculated.
During a cyclic loading test, energy is dissipated by the device and the specimen. It is assumed that
the total energy per cycle includes two parts, system losses Wsys and visco-elastic losses Wdis .
Visco-elastic losses Wdis is represented by the area enclosed by the stress-strain loop. Nearly all
energy is transformed into heat, leading to an increase in the temperature inside the specimen and
only a small part Wfat causes decrease of the stiffness due to fatigue damage [Pronk 1996]. In a
strain controlled fatigue test, this small part per cycle is expressed by Equation 1:
where = small parameter, i = stress amplitude at cycle i, [MPa]; 0 = applied strain amplitude,
a constant in the strain controlled mode, [mm/mm]; i = phase angle at cycle i, [ ]; Si = stiffness
modulus at cycle i, [MPa].
In the proposed model, a new parameter, stiffness damage Q, is introduced, which is directly
related to the Wdis . The damage Q decreases the stiffness modulus, including both the loss
modulus F and the storage modulus G. The loss modulus and storage modulus can be expressed
by Equation 2, respectively.
where F0 = initial loss modulus, [MPa]; G0 = initial storage modulus, [MPa]; t is the testing time,
[s]; 1 , 2 , 1 , 2 and = model parameters. In this formulation initial means the value at t = 0.
The terms with 1 , 2 and represent the reversible damage which will heal in time. It can also
be seen as a mathematical formulation for the thixotropic behavior of asphalt mixes. It is not really
healing of fatigue damage but partial healing of the stiffness modulus. The terms with 1 and 2
represent the irreversible damage which will accumulate during test. In the strain controlled mode,
the fatigue damage rate is given by Equation 3:
where t = time duration in one cycle, [s]; Fi is the loss modulus at the ith cycle, [MPa]; f is the
applied frequency [Hz].
36
To simplify the equation form, the following abbreviations are used:
For a volume unit, the solutions of the local loss and storage stiffness modulus are given by
Equation 6.
For the uniaxial tension and compression test, the stress-strain distribution is in principle uniform
throughout the cylindrical specimen. In theory, the evolution of the stiffness modulus for a small
volume unit should be equal to the evolution for the whole specimen measured from the test
equipment. So Equation 6 can be directly used to fit the measured values. However, the 4PB test
is not a homogenous test with respect to the stress-strain field. The stress and strain vary along the
length and height of the beam specimen during the test. Therefore, the decrease of the stiffness
modulus is different in each small volume unit. The back calculated stiffness during the test is not
the local stiffness but the weighted overall stiffness for the whole beam. Equation 6 can not be
applied directly to the test results and a weighing procedure is required to calculate the weighted
stiffness modulus per cross area (along the height) and the weighted overall stiffness modulus
(along the length) from the local stiffness. The details of the numerical analysis were demonstrated
by Pronk [Pronk, 2012] and will not be mentioned in this paper.
3 MATERIALS AND EXPERIMENTAL WORK
3.1 Materials
A dense asphalt concrete DAC 0/8 with a maximum aggregate size of 8 mm was designed in
accordance with the Dutch RAW specifications [CROW 2005]. The aggregate consists of Scottish
crushed granite, Bestone and crushed sand. A 40/60 penetration grade bitumen was used at a design
binder content by mass of the total mixture of 6.5%. A factory produced filler, Wigras 40K, was
used as the filler. Table 1 represents the mixture composition.
In order to research the effect of the specimen size, three different sizes (0.5, 1.0 and 1.5)
are used in this study, in which size 1.0 corresponds to the standard size with dimension of
Table 1. Composition of the mixture DAC 0/8.
Scottish crushed Norwegian Wigras

granite Bestones Crushed sand 40K
Sieve
(mm) 8-5.6 5.6-4 6-2 2-1 1-0.063 0.18-0.063 Binder Total
Wt. % 11.2 19.6 21.5 16.8 16.4 7.9 6.5 100
37
Table 2. Dimensions of the beam specimens.
Length Width Height

Specimen size (mm) (mm) (mm) h/L ratio
0.5 450 50 25 0.056

1.0 450 50 50 0.111
1.5 450 50 75 0.167
Figure 1. (a): Set-up of the 4PB fatigue test; (b): local view of inner clamps; (c): local view of outer clamp.
height width length of 50 50 450 mm. For all three specimen sizes, the length and width
are kept constant. Only the height is varied. The dimensions of the specimens with different sizes
are presented in Table 2. All beam specimens were cut horizontally and in parallel from slabs,
which were compacted with a PReSBOX compactor [IPC Global] [Qiu 2009]. The target air voids
content was 3%.
3.2 Test method

The 4PB test set-up is placed in a temperature chamber capable of maintaining a temperature
between 10 C to 60 C. Four clamps fix the beam specimen in the bending bed. The distance
between the clamps is 133.3 mm. The load is applied to the specimen through the inner clamps by
means of two loading jacks. The two outer clamps prevent vertical movement of the specimen. The
clamp consists of two thin steel sheets with a groove. A small steel roller was placed in the grooves
between clamp and the frame. All four clamps allow free rotation and horizontal movement of the
specimen, as shown in Figure 1.
The fatigue tests were conducted at a temperature of 20 0.5 C and a frequency of 10 Hz.
Specimens were preconditioned to 20 C for a minimum period of 2 hours. The tests were performed
in deflection controlled mode. During the test, the specimens were subjected to a continuously
sinusoidal loading controlled by the LVDT (0.5 mm) on the top surface between the inner clamps.
The force is measured using a load cell at the bottom. For each specimen size, the applied deflections
were chosen in such a way that the fatigue life could cover a wide rang of 5 103 to 1 107 cycles.
3.3 Deflection correction

The measured deflection includes three parts, pure bending deflection of the beam, shear deflection
of the beam at the locations of two inner clamps and the deflection of the frame. Only the pure
bending deflection is involved in the calculation of the stiffness modulus.
1. Correction for the shear deflection
38
The deflection of beam vbeam contains two parts, pure bending and shear deflection [Pronk
2009]. In the middle section between two inner clamps, the solution of bending deflection is
given by Equation 9.
where vbeam = deflection of beam, [m]; vs = shear deflection, [m]; vb = pure bending deflection,
[m]; Cs = correction coefficient; = Poissons ratio; h = height of the beam, [m]; L = distance
of two outer clamps, [m]; A = distance between inner and outer clamps, m; = shear factor,
0.859.
2. Correction for the frame deflection
The whole system including the beam and the frame could be considered as two serial springs.
So the following relations can be obtained:
where vframe = deflection of frame, [m]; vbeam = deflection of beam, [m]; vm = measured deflec-
tion, [m]; F = applied force, [N]; Sm = stiffness of the whole system, [N/m]; Sframe = frame
stiffness, [N/m]; Sbeam = beam stiffness, [N/m].
For the 4PB test, the beam stiffness in [MPa] has a following relation with the applied force and
the bending deflection.
where E = theoretical value of beam stiffness, [MPa]; Z = form factor; vb = bending deflection,
[m]; F = applied force, [N]; L = distance of two outer clamps, [m]; A = distance between inner
and outer clamps, [m]; h = height of the beam, [m]; b = width of the beam, [m].
The beam stiffness in [N/m] can be transformed from the beam stiffness in [MPa] by substituting
Equation 9 into Equation 10.
Therefore the frame stiffness is calculated by the following equation:
In theory, the frame stiffness keeps as a constant value. However, in reality the calculated frame
stiffness might be influenced by the applied loading and frequency. A frequency sweep with the steel
calibration beam (E = 200 GPa) was conducted to find out the relationship between the calculated
39
Figure 2. Calculated frame stiffness under different force levels.
frame stiffness and the loading conditions. Figure 2 gives the calculated frame stiffness at different
load levels and frequencies. It is clear that the frame stiffness is independent on the frequency
and slightly changes with the applied force. The relationship is approximately fitted by a linear
function.
where F = applied force, [N]; Sframe = frame stiffness, [N/m].

Then the deflection of beam can be calculated by substituting Equation 14 into Equation 10.
3.4 Fatigue life definitions

The determination of the fatigue failure has been a controversial topic for a long time. Various
fatigue life definitions were proposed based on the phenomenology, dissipated energy and fracture
mechanics approaches [van Dijk 1975] [Pronk 1991] [Rowe 1993] [Paris 1963] [Jacob 1995]. In
this study three different fatigue criteria are used:
1) Nf ,50 : the point of failure is defined as the moment at which the calculated stiffness of the
specimen is reduced to 50% of its initial value (stiffness at the 100th cycle).
2) NR : the fatigue life is defined as the point at which the slope of the dissipated energy ratio (DER)
versus number of load cycles deviates from a straight line. DER is the ratio of the accumulated
dissipated energy up to cycle N and the dissipated energy in cycle N, given as Equation 16.
3) NPH : the fatigue life is defined as the point where the stiffness modulus deviates from the fitted
evolution based on the MPH model.
40
Figure 3. Comparison of measured and bending deflection.
Figure 4. Initial stiffness of three specimen sizes.
4 RESULTS AND DISCUSSION
4.1 Comparison of measured and corrected deflection and initial stiffness

The bending deflection was calculated from the measured deflection by Equation 9 and 15. The
results for each specimen size are represents in Figure 3. For size 0.5, the bending deflections are
close to the measured values, which indicates that during the test the frame and shear deflection
are very small. This is because the specimen size 0.5 has a low h/L ratio (h/L = 0.056) and the low
applied forces hardly deform the frame. With increasing specimen size, the frame and shear deflec-
tion became larger. So the difference between total and bending strain also becomes significant.
The bending deflection of the size 1.5 is only around 84% of the measured deflection.
The initial stiffness under different strain levels for each size is plotted in Figure 4. Their average
value, the number of samples, standard deviation and coefficient of variation (CV) are given in
Table 3. It is can be seen that for each size, the initial stiffness does not depend on strain levels,
but drops dramatically with increasing of specimen size. Especial for the size 1.5, the reduction of
the average initial stiffness is around 30004000 MPa compared to size 0.5 and size 1.0. The CV
values for all specimen sizes are from 4% to 12%, which is within the acceptable error tolerance
due to HMAC mixture inhomogeneity and test variability.
41
Table 3. Initial stiffness of three specimen sizes.
Size 0.5 Size 1.0 Size 1.5
Mean value of Sinitial [MPa] 11629 10537 7598

Number of specimens 6 7 7
Stdev [MPa] 512 445 813
CV [%] 4.40 4.22 11.73
Figure 5. Measured and simulated evolutions of stiffness and phase angle of beam B-1-3 (size1.0) at the
strain level of 165 m/m.
4.2 Size effect on the MPH model parameters

Figure 5 gives the evolutions of the stiffness modulus and phase angle of beam B-1-3 (size1.0) at
the strain level of 165 m/m. According to the changing rate of the stiffness and the phase angle,
the fatigue process is divided into the three phases. The model parameters 1 , 2 , 1 , 2 and
are determined by minimizing the differences between the measured and fitted values for S, , F
(S sin) and G (S cos) in the interval from N = 1000 to N = Nr . Figure 5 shows that in the
first and second phase, the fitted lines of stiffness and phase angle correspond very well with the
measured values.
The relationships between all the model parameters and bending strain levels are illustrated
in Figure 6. The parameter 1 is always nil at different strain levels for all specimen sizes. The
parameters 2 , 1 , and 2 could be expressed as the linear relationship with the bending strain. The
parameter , the time decay, is dependent on the squared value of the bending strain. It seems that
size effect does not influence the functions of 2 and . For the parameters 1 and 2 , the slopes
of the curves are influenced by the specimen size. The smaller specimen has a steeper slope. The
parameters 1 and 2 represent the irreversible damage of the specimen. Its intersection point with
the x-axis means that the irreversible damage will not happen when the bending strain is below a
certain value. This indicates the existence of the endurance limit.
Table 4 presents the regression equations of all the model parameters and the ranges of the
calculated endurance limit. The results show that the value of limit decreases with the increase of
the specimen height. With the functions gathered in Table 4, the evolutions of stiffness and phase
angle can be predicted for other strain levels.
42
Figure 6. Model parameters vs. strain level for each specimen size.
4.3 Size effect on the fatigue life

As mentioned in section 3.4, three different fatigue life definitions are used in this paper. An
explanation is given in Figure 7.
Table 5 presents all three different fatigue live definitions. The fatigue life NPH and NR are found
near the transition point between the second and the third phase. In nearly all cases, NPH and NR at
43
Table 4. Regression equations of the model parameters for each specimen size.
Size 0.5 Size1.0 Size1.5
1 = 0,
2 = 4.15 0.68,
= (3.92E 08) 2
1 = 0.372 ( 68) 1 = 0.146 ( 44) 1 = 0.091 ( 46)
2 = 1.198 ( 58) 2 = 1.05 ( 44) 2 = 0.7 ( 37)
limit = 5868 m/m limit = 44 m/m limit = 3746 m/m
Figure 7. Determinations of fatigue lives of Beam B-13-2 (size 1.5) at the strain level of 130 m/m.
Table 5. Comparison of the different fatigue lives.
Void
Sample content 0 NPH NR Nf ,50 (NR - NPH )/ Nf ,50 - NPH )/
Size code [%] [ 106 ] ( 106 ) ( 106 ) ( 106 ) NPH NPH
Size 0.5 B-2-6 2.3 80 7.0 6.6 8.7 5.6% 24.2%

B-2-1 2.4 100 2.3 1.8 2.5 18.0% 12.5%
B-2-3 2.3 121 1.4 1.1 1.7 22.2% 16.4%
B-2-2 2.4 140 0.55 0.35 0.63 36.2% 15.1%
B-2-5 2.4 161 0.17 0.16 0.21 8.8% 25.0%
B-2-4 2.3 180 0.18 0.13 0.27 27.1% 48.2%
Size 1.0 B-6-2 3.0 73 4.6 3.4 6.8 26.0% 49.6%
B-1-2 2.6 89 1.2 1.2 2.5 0.7% 98.6%
B-10-4 3.8 92 0.86 0.95 1.7 9.8% 102%
B-10-1 2.8 112 0.78 0.63 1.1 19.8% 38.0%
B-10-2 3.0 126 0.4 0.33 0.56 15.9% 41.6%
B-10-3 3.8 151 0.31 0.22 0.42 29.7% 33.4%
B-1-3 2.5 165 0.14 0.15 0.24 4.5% 68.0%
Size 1.5 B-9-1 2.7 80 1.2 1.5 2.3 23.3% 94.5%
B-4-1 3.3 87 1.1 1.2 1.4 9.0% 28.2%
B-12-1 3.2 110 0.63 0.61 1.0 2.9% 65.6%
B-13-2 3.4 130 0.39 0.38 0.52 0.9% 34.9%
B-12-2 3.5 145 0.32 0.35 0.58 8.4% 80.3%
B-9-2 3.5 161 0.28 0.35 0.5 24.2% 76.6%
44
Figure 8. Fatigue lives NPH of the specimens with different sizes.
Table 6. Regression equation and material coefficients of the fatigue lines.
Material coefficients
Specimen size k b R2
Size 0.5 9.31E + 15 4.78 0.97

Size 1.0 1.29E + 13 3.56 0.92
Size 1.5 1.86E + 10 2.20 0.99
the same strain level are close to each other, because both of these two fatigue definitions are based
on the dissipated energy theory. At the end of the third phase, the stiffness reduced to 50 % of the
initial value. The classical fatigue life Nf ,50 is much higher than the other fatigue lives, especially
for the bigger specimen. For some samples, the classical fatigue life Nf ,50 is about two times larger
than the fatigue life determined by the MPH model. This paper used the NPH to characterize the
fatigue property of the sample.
Fatigue relationships are usually represents by means of a straight line in a double logarithmic
coordinate system.
where N is the fatigue life; 0 is the strain amplitude, m/m; k and b are the material coefficients.
The fatigue lines obtained for size 0.5, 1.0 and 1.5 specimens are plotted in Figure 8 and all
the regression equations and material coefficients are listed in Table 6. The slope and the intercept
of the curve are strongly influenced by specimen size, which both decrease with the increasing
of the specimen size. In general, the larger specimen has a shorter fatigue life at the same strain
level. Because the specimen with larger volume has a higher chance to create weak spots during
the testing. But at high strain level, the fatigue life of size 1.5 is longer than the other two sizes.
The reason for this is unknown at this moment. More tests are required to find out the answer.
In this paper, the results were presented and discussed of the four-point bending fatigue tests con-
ducted with three different specimen sizes. The MPH model was applied to simulate the evolutions
45
of the stiffness modulus and the phase angle. The size effect on the fatigue results and the model
parameters was investigated. Based on the analysis of the test results, the following conclusions
have been drawn:
1. The shear and frame deflection depend on the H/L ratio of the sample and the frame stiffness.
When the H/L ratio of specimen is high (>0.1) or the frame stiffness is not stiff enough, the
deflection correction have to be taken into account at relative high load levels.
2. The measured initial stiffness is strongly influenced by the specimen size. The difference of the
initial stiffness between size 0.5 and size 1.5 is around 4000 MPa.
3. The MPH model is able to simulate the evolutions of the stiffness and the phase angle. Based
on the test results, the model parameters can be determined. All parameters are the functions of
the bending strain level.
4. The relation between parameters 1 and 2 the bending strain indicates the existence of an
endurance limit. The calculated endurance limit decreases with the specimen size.
5. From the comparison of three different fatigue lives, the traditional fatigue life Nf ,50 is higher
than the fatigue lives represented by NR and NPH . This is especially true for the big specimen
size.
6. The larger specimen has a short fatigue life at the same strain level. The slope and the intercept
of the fatigue lines decrease with the increasing of the specimen size.
These conclusions indicate that besides the test conditions, the fatigue test results strongly
depend on the specimen size, clamping force, deflection correction and fatigue life definition.
Therefore before doing a 4PB fatigue test or analyzing the 4PB test results, these important factors
should be carefully considered.
REFERENCES
AASHTO, AASHTO standard specifications for transportation materials and methods of sampling and testing.
23rd Edition, Part 2B. T321-03: Determining the fatigue life of compacted hot-mix asphalt (HMA) subjected
to repeated flexural bending, AASHTO, Washington, D.C., 2003.
CROW 2005. In: Standaard RAW Bepalingen 2005, C.R.O.W., Ede (In Dutch).
Di Benedetto H., de la Roche C., Baaj H., Pronk A. C. and Lundstrom R. 2004. Fatigue of bituminous mixtures.
Materials and Structures, 2004, 37, No. 3, 202216.
EN 12697-24, Bituminous mixtures Test methods for hot mix asphalt Part 24: Resistance to fatigue, 2004.
IPC Global, PReSBBOX Compactor Technical Reference, Version 1d0, September 18, 2007.
Jacobs, M. M. J. 1995 Crack Growth in Asphaltic Mixes, Ph.D. dissertation, Delft University of Technology,
Delft, The Netherlands.
Qiu, J., D. Xuan, et al. (2009). Evaluation of the shear box compactor as an alternative compactor for asphalt
mixture beam specimens. AES ATEMA2009, 3rd Int. Conf. on Advances and Trends in Engineering
Materials and their Applications. Montreal, Canada.
Paris, P.C., and Erdogan, K. 1963 A Critical Analysis of Crack Propagation Laws, from: Transactions of the
ASME, Journal of basic Engineering, Series D, 85, No. 3.
Pronk, A. C. and Hopman, P. C. 1991. Energy dissipation: the leading factor of fatigue. Proceedings of the
Conference on the United States Strategic Highway Research Program, pp. 255267.
Pronk, A.C. 1996. Analytical description of the heat transfer in an asphalt beam, tested in the 4 point dynamic
bending apparatus, DWW report W-DWW-96-006.
Pronk, A.C. and Huurman, M. 2009. Collaborative study with 4PB devices in Europe. Round Robin test with
three reference beams. Preliminary results. 2nd Workshop on Four Point Bending, University of Minho,
Guimaraes, Portugal.
Pronk, A.C. and Huurman, M. 2009. Shear deflection in 4PB tests. 2nd Workshop on Four Point Bending,
University of Minho, Guimaraes, Portugal.
Pronk, A.C. 2001. Partial healing in fatigue tests on asphalt specimen, Road Materials and Pavement Design,
vol. 4, n. 4.
Pronk, A.C. 2006. Partial Healing, A new approach for the damage process during fatigue testing of asphalt
specimen, Proceedings of the Symposium on Mechanics of Flexible Pavements, ASCE, Baton Rouge.
46
Pronk, A.C. and Molenaar, A.A.A. 2010. The Modified Partial Healing Model used as a Prediction Tool for
the Complex Stiffness Modulus Evolutions in Four Point Bending Fatigue Tests based on the Evolutions in
Uni-Axial Push-Pull Tests, Proceedings of the 11th Int. Conf. on Asphalt Pavements, Nagoya, Japan.
Pronk, A.C. 2012. Description of a procedure for using the Modified Partial Healing model (MPH) in 4PB
test in order to determine material parameters. To be published at the 3rd 4PB conference, Davis, USA.
Rowe, G. M. 1993. Performance of asphalt mixtures in the trapezoidal fatigue test. Proceedings of Associations
of Asphalt Paving Technologists, 62, 344384.
van Dijk, W. 1975. Practical Fatigue Characterization of Bituminous Mixes, Proceedings The Association of
Asphalt Paving Technologists, 38.
47
Applications to mechanistic design
Application of four-point bending beam fatigue test for the design and
construction of a long-life asphalt concrete rehabilitation project in
Northern California
V. Mandapaka, I. Basheer, K. Sahasi & P. Vacura

California Department of Transportation, Sacramento, US
B.W. Tsai, C.L. Monismith, J. Harvey & P. Ullidtz

University of California Pavement Research Center, Davis, US
ABSTRACT: This paper presents how the agent, Caltrans, applies the four-point bending beam
fatigue test results to develop statistically-based fatigue performance specification and determine
the fatigue model parameters for each material used, and how these results are used to assist in the
rehabilitation design and construction of a long-life pavement structure in Northern California. The
asphalt mixes considered for use in rehabilitating the existing pavement consisted of three layers
of hot mix asphalt comprised of PG64-28PM (polymer-modified) in the surface course, PG64-10
(25%) RAP (recycled asphalt pavement) in the middle layer, and PG64-10 RB (rich bottom) in the
bottom layer; all produced with lime-treated aggregates. The fatigue test results were subsequently
integrated into the database, together with the other material properties and used to determine the
required layer thicknesses using the California Mechanistic-Empirical (CalME) pavement design
program. Test results along with the specification developed for this project are presented.
1 INTRODUCTION
The fatigue resistance of an asphalt concrete mix is its ability to withstand repeated loading with-
out fracture. In California, fatigue cracking is considered to be the main distress affecting the
performance of asphalt concrete on major Interstate and State highways (Harvey et al., 1995).
Caltrans has recently adopted the Mechanistic-Empirical (M-E) pavement design method to ana-
lyze, design, and rehabilitate flexible pavement systems. The M-E method combines mechanistic
models that calculate the pavement primary response (stresses, strains, and deflections) induced
due to interaction of climate and traffic loading, and empirical models which relate the calcu-
lated response to pavement performances. The California M-E design software (CalME) has been
recently developed to perform the M-E analysis and used for the design of a number of long-life
pavement projects. Several models were developed for various distress mechanisms and incor-
porated in CalME including reflective cracking, rutting, HMA aging, etc. (Ullidtz et al., 2010;
Mandapaka et al., 2012). Caltrans has adopted the four-point bending beam (LLP-AC1, 2011) test
based on the AASHTO T-321 test [Standard test method for determining the fatigue life of hot
mix asphalt (HMA) subjected to repeated flexural bending], among other tests for the design of
the HMA materials to be used in constructing long-life flexible pavements designed with the M-E
procedure. Performance-related specifications are associated with the laboratory performance tests
and fundamentally based predictive models (Shatnawi, 1997).
The pavement section used in this paper for illustration of the use of fatigue testing data to develop
fatigue performance specification lies on interstate I-5 in Tehama county in northern California,
USA. The existing pavement section consists of two lanes in each direction and the pavement
structure comprised of HMA layer, cement treated base (CTB) layer, and aggregate subbase (AS)
51
layer over a subgrade classified as SW based on the Unified Soil Classification System (USCS).
The pavement section suffered from extensive fatigue cracking (especially reflection cracking)
and occasional rutting. Cores retrieved from various locations within the project limits revealed
some delaminated and stripped areas in the existing HMA layer at different depths. Due to the
presence of delaminated areas in the existing pavement, it was recommended to entirely remove
the existing HMA down to the CTB layer. A three-HMA layered system concept (Monismith et. al.,
2009) was proposed for use for this long-life rehabilitation design using the materials that met the
performance specifications. The desired design life was 40 years with a traffic index (TI) of 16.0
and corresponding 125 million ESALs (HDM 2008). The climate in the project site is considered
Inland Valley according to HDM (2008).
The objective of this paper is to present the methodology for utilizing the four-point bending
(4PB) beam fatigue test (AASHTO 321) to (1) obtain the stiffness master curves required for
determining the fatigue damage model parameters necessary for CalME, and (2) determine the
fatigue performance specifications for the three HMA materials proposed for use on the project.
2 DETERMINING HMA STIFFNESS AND FATIGUE DAMAGE PARAMETERS
Four-point bending flexural fatigue testing is used to measure the fatigue damage or stiffness dete-
rioration characteristics of specific HMA materials over a range of traffic (in terms of testing stress
or strain levels in the laboratory) and environmental conditions (in terms of testing temperatures) so
that the fatigue considerations can be incorporated into the process of designing asphalt concrete
pavements. One of the advantages of using 4PB beam fatigue test is that the middle one-third
portion of the beam is theoretically subjected to pure bending without any shear deformation. This
state of loading is of particular interest because the fatigue resistance of an HMA is essentially its
ability to withstand repeated bending without fracture. Therefore, the four-point bending beam test
can efficiently characterize the fatigue properties of asphalt mixes (Wu, 2009).
2.1 HMA stiffness master curves

The HMA stiffness master curve is developed using the AASHTO T 321 test; which is also called
the four-point bending beam frequency sweep test. Several tests were performed at 11 frequencies:
15, 10, 5, 2, 1, 0.5, 0.2, 0.1, 0.05, 0.02, and 0.01 Hz; three temperatures:10 C, 20 C and 30 C; and
two strain levels: 200 and 400 microstrain for each HMA material used in this project. The obtained
data is used to construct a stiffness master curve for each of the HMA materials by determining the
values of the various model parameters in the stiffness equation expressed as (Ullidtz et al., 2010;
ARA Inc, 2004):
where Ei = the intact modulus in MPa; tr = reduced time in sec; 2 , 1 , 1 , and 1 = experimentally
determined constants, log = logarithm with base 10, and tr is the reduced time calculated from
(ARA Inc., 2004):
where lt = the loading time in sec; viscref = the binder viscosity at the reference temperature;
visc = the binder viscosity (in cpoise) at the present temperature, and aT = constant. This model
accounts for the temperature and loading time dependency of the HMA modulus. The binder
viscosity is calculated using the following equation (ARA Inc., 2004):
52
Table 1. Fatigue damage model parameters of AC material for use in CalME.
Material type A 0 1 ref Eref RMS
PG6428PM 4887.232 1.4535 0 200 7.6899 3000 3.845 0 3.1343

PG6410RAP 151.2162 0.35766 0 200 5.9161 3000 2.9581 0 7.3679
PG6410RB 2491.208 0.74099 0 200 6.4907 3000 3.2454 0 7.6002
where T is binder temperature in Rankine, and A and VTS are model parameters. The detailed
procedure for determining the stiffness model parameters is presented in (Wu et al., 2009).
2.2 CalME HMA fatigue damage model

In CalME, it is assumed that fatigue damage causes the HMA modulus to decrease according to
the following relationship (Ullidtz et al., 2010):
where = the damage, which can be calculated from the following series of equations:
where MN = the number of load applications in millions; MNp = the permissible number of load
applications in millions, = the tensile strain at the bottom of the asphalt layer; E = the (current)
damaged modulus; Ei = the intact modulus; T = HMA average temperature; A, 0 , 1 , , , =
model parameters; and ref and Eref = reference strain and modulus, respectively, used for nor-
malizing the ratios. In order to comply with the concept that fatigue damage is driven by strain
energy, it is typically assumed that = 2 (ARA Inc., 2004)
As mentioned previously, three HMA materials were selected for use on the project. The materials
will be successively placed on top of the CTB exposed by milling of the existing asphalt materials.
The first HMA material to be placed consist of PG64-10 RB (rich bottom, i.e., 0.5% higher
binder content), the middle layer consists of PG64-10 and RAP material at 25% by weight, and
the surface HMA material consists of polymer modified HMA with PG64-16PM. The PG grades
were selected based on climatic region where the project is located per the Caltrans HDM (2008).
Table 1 summarizes the fatigue model parameters for all the three HMA materials used in the
design of the pavement structural section, along with the corresponding root mean squared (RMS)
error parameter. These model parameters were obtained by performing 4PB beam fatigue tests on
beam specimens prepared using the standard procedure developed for Caltrans (LLP-AC1, 2011)
at a temperature of 20 C. The standard data obtained from the tests is processed using non-linear
optimization against the fatigue model proposed in Equations 47 above. A comparison between the
calculated stiffness (based on Eq. 4) to the measured stiffness for each HMA material is observed by
53
Figure 1. Comparison between measured E/Ei and calculated E/Ei (PG64-28PM).
Figure 2. Comparison between measured E/Ei and calculated E/Ei (PG64-10RAP).
plotting the stiffness ratios (E/Ei ) as can be seen in Figures 13. It is to be noted that in Figures 13
the notation 20_205_5.5_5 refers to Temperature C _strain level(s)_airvoids%_Binder content%.
It can be observed from Figures 13 that the calculated stiffnesses agreed reasonably well
with corresponding measured values for stiffness ratios higher than 0.50.6. For stiffness ratios
lower than 0.50.6, the calculated stiffnesses are higher than measured values indicating a more
conservative prediction and less fatigue damage. The deviation between calculated and measured
stiffness grows as the stiffness ratio continues to decrease. This is expected since data with residual
stiffness ratio below 0.3 is ignored. The reason for disregarding the low stiffness ratios is that once
the beam starts to develop macro-cracks it will behave quite differently from a pavement layer that
54
Figure 3. Comparison between measured E/Ei and calculated E/Ei (PG64-10RB).
develops macro-cracking. The beam risks rapid failure, whereas the pavement will probably fail
slower due to different boundary conditions. The model parameters shown in Table 1 were uploaded
into the CalME software with other material parameters and the pavement structure was analyzed
using the Incremental-Recursive (I-R) method described in Ullidtz et al. (2010) and Mandapaka
et al. (2011).
3 DEVELOPMENT OF PERFORMANCE SPECIFICATIONS
Performance specifications have been developed in order to provide the contractor a quantitative
measure of the quality of materials that can be used in each HMA layer. The development of
the fatigue performance specification where quality assurance is required for HMA construction
makes use of a confidence band concept to statistically determine the lower bound of fatigue life
at specified strain levels that must be achieved by the Contractor through testing by 4PB tests. In
order to develop these bands, the procedure based on Selvin (1995) has been adopted and used in
this project. In this procedure, a (1 )-confidence band of fatigue regression line; a band that
has a probability of 1 of containing the true regression line of fatigue life as a function of
tensile strain level, is not constructed from a series of confidence intervals based on y0 [calculated
Ln(Nf )], which is an estimate of the mean of the population of values associated with x0 [calculated
Ln(strain)]. Instead, the width of the geometric region at the point x0 that has a probability of 1
of containing the true regression line is given by:
where F1 is the (1 )-percentile of an F-distribution with 2 and n 2 degrees of freedom. Notice

that the number 2 of n 2 degrees of freedom comes from the fact that the regression equation
has two parameters and has the formulation with y = a + bx, i.e., Ln(Nf ) = a + bLn(strain). For
estimating a mean value from a regression line with the formulation of Ln(Nf) = a + b Ln(strain),
an estimate of variance of y0 .
55
Table 2. Lower bounds for 95% confidence band for PG64-10RB and PG64-10RAP mixes tested at 20 C.
PG64-10RB PG64-10RAP
w/Lime w/Lime PG64-28PM w/Lime
Strain Ln(Strain) (LB) Ln(Nf) (LB) Ln(Nf) Lower Bound(LB) Ln(Nf)
0.0001 9.21034 15.81985 15.63268 20.20609

0.000164 8.7139 15.21146 14.34777 19.86629
0.000229 8.38366 14.52459 13.30796 19.45809
0.000293 8.13583 13.45893 12.18087 18.78389
0.000357 7.93738 12.04937 10.98033 17.74516
0.000421 7.77186 10.61595 9.8585 16.56064
0.000486 7.62989 9.29322 8.85378 15.41701
0.00055 7.50559 8.09695 7.95648 14.36458
0.000614 7.39505 7.0147 7.14983 13.40497
0.000679 7.29552 6.03023 6.41876 12.5285
0.000743 7.20501 5.12893 5.751 11.72418
0.000807 7.12201 4.29864 5.13683 10.98212
0.000871 7.04538 3.52939 4.56846 10.29393
0.000936 6.9742 2.81308 4.03966 9.65264
0.001 6.90776 2.14304 3.54533 9.05246
Figure 4. The 95% confidence band for PG64-28PM HMA material (with 1.2% lime added, AC = 5.2%,
AV = 6.0%) tested at 20 C.
(SY2 ) is calculated from:

0

where SY2 |x = (yi yi )2 /n 2 is the square of residual standard error of the regression equation.
A summary of the lower bound fatigue life values for each of the three HMA materials used
in this project corresponding to a particular strain level is as shown in Table 2. Additionally, the
56
Figure 5. The 95% confidence band for PG64-10RAP HMA material (with 1.2% lime added, AC = 5.38%
by weight of virgin aggregate), AV = 6.0%) tested at 20 C.
Figure 6. The 95% confidence band for PG64-10RB HMA material (with 1.2% lime added, AC = 5.5%,
AV = 3%) tested at 20 C.
95% confidence bands of PG 64-10RAP, PG64-10RB, and PG64-28PM are plotted as shown in
Figures 4, 5, and 6, respectively. It is apparent from Figures 46 and Equation 10 that the estimate
of variance of y0 is large at extreme values of x0 and smallest when x0 = x. For fatigue specification,
it is suggested that the mean of natural logarithm of fatigue life, Ln(Nf ), of three fatigue tests at the
specified strain level should be above the specified lower bound and the regression line obtained
from specified strain levels should be in the range above the lower bound.
57
4 CONCLUSIONS
Four point bending (4PB) beam fatigue test (AASHTO T321) has been adopted by the California
Department of Transportation (Caltrans) to assist in the design and construction of a long-life
flexible pavement on I-5 in Northern California. The 4PB tests were used to determine the fatigue
model parameters and master curves as necessary inputs to the California M-E design software,
CalME. The tests were also necessary to develop the fatigue performance specifications for the
project. The minimum fatigue life at a given strain level for each material for a 95% confidence level
was specified as the performance criteria for each material. The 4PB fatigue test has enabled the
integration of construction quality requirements (performance specifications) with the ME design
of flexible pavement. Although the main focus of this paper is to present the application of the 4PB
beam fatigue test in assisting the design and construction of a long-life flexible pavement, rutting
performance has also been incorporated in the design of this project, but not presented in this paper
due to being outside the scope of the paper.
REFERENCES
ARA Inc. (2004). Guide for Mechanistic-Empirical Design of New and Rehabilitated Pavement Structures,
ERES Consultants Division, ARA Inc., National Cooperative Highway Research Program, Transportation
Research Board, National Research Council.
HDM (2008). Highway Design Manual, California Department of Transportation. Chapter 600, July 1.
http://www.dot.ca.gov/hq/oppd/hdm/hdmtoc.htm. Last accessed on November 21, 2011.
Harvey J., Deacon J.A., Tsai B.W., Monismith C.L. (1995). Fatigue performance of asphalt concrete mixes
and its relationship to asphalt concrete pavement performance in California. Report for the California
department of transportation, Asphalt Research program, CAL/APT program, Institute of Transportation
Studies, University of California, Berkeley, October 2011.
LLP-AC1 (2011). Sample preparation and testing for long life asphalt concrete pavements. Lab procedure,
California Department of Transportation, May 20, 2011. http://www.dot.ca.gov/hq/esc/Translab/ofpm/pdf/
AC1-LLP_Dist-02.pdf. Last accessed on November 21, 2011.
Mandapaka V., Basheer, I., Sahasi K., Ullidtz P., Harvey J., and Sivaneswaran N. (2012). Mechanistic-empirical
and life-cycle cost analysis for optimizing flexible pavement maintenance and rehabilitation. To appear in
May 2012 Edition of the ASCE Journal of Transportation Engineering.
Mandapaka V., Basheer I., Sahasi K., Harvey J., Wu R., Monismith C.L. (2011). Mechanistic-Empirical
design of a long life flexible pavement in Northern California, paper accepted for the 1st Conference of
transportation research group, held in Bangalore, India, December 710, 2011.
Monismith C.L., Harvey J., Tsai B.W., Long F., Signore J. Summary report (2009). The phase one I-710
freeway rehabilitation design project: Initial design (1999) to performance after five-years of traffic
(2009), prepared for California department of Transportation, 2009.
Shatnawi S.R. (1997). Fatigue performance of asphalt concrete mixes using a new repetitive direct tension test.
Prepared for California Department of Transportation, October 1997.
Selvin, S. (1995). Practical Biostatistical Methods. Duxbury Press, 1995.
Ullidtz, P., Harvey J., Basheer I., Jones D., Wu R., Lea J., and Lu Q. (2010). CalME: A new mechanistic-
empirical design program for flexible pavement rehabilitation, transportation research record of the
transportation research board, no. 2153, pp. 143152.
Wu.R., Tsai B.W., Harvey J., Ullidtz P., Basheer I., Holland J. (2009). Using four-point bending tests in calibra-
tion of the California mechanistic-empirical pavement design system. 2nd Workshop on 4PB conference,
held in Minho, Portugal. http://www.civil.uminho.pt/4pb/workshop2.htm. Last accessed on November 21,
2011.
58
Development of fatigue performance specification and its

relation to mechanistic-empirical pavement design using
four-point bending beam test results
B.-W. Tsai, R. Wu, J.T. Harvey & C.L. Monismith

University of California Pavement Research Center, Davis, Berkeley, US
ABSTRACT: This paper describes a general procedure of how to apply four-point bending beam
test results to develop fatigue performance specification and to relate the developed specification to
mechanistic-empirical pavement design. Flexural fatigue and frequency sweep tests are two major
four-point bending beam test types that respectively evaluate the asphalt mixs fatigue damage
and time-temperature-stiffness relationships which are two key elements in mechanistic-empirical
pavement design. A three-stage Weibull function was used to describe the mix fatigue damage and
a Gamma function was utilized to express the complex modulus master curve and the temperature
shifting relationship. The development of the fatigue performance specification where Quality
Assurance (QA) is required for Hot-Mix Asphalt (HMA) construction makes use of a confidence
band concept to statistically determine the lower bound of fatigue life at specified strain levels.
A pavement structure evaluated in I-710 project, Long Beach, California, was used to demonstrate
how performance specification relates to mechanistic-empirical pavement design via the SHRP-
developed fatigue analysis system.
1 INTRODUCTION
This paper presents a general procedure of how to apply four-point bending beam test results to
develop fatigue performance specification and to relate the developed specification to mechanistic-
empirical pavement design.
Flexural fatigue and frequency sweep tests are two major four-point bending beam test types
and are two key elements in mechanistic-empirical fatigue pavement design. A flexural fatigue
test can evaluate the fatigue damage process (or stiffness deterioration process) of asphalt mix,
subject to a specified stress/strain condition that produces pure bending with no shear deformation
in the middle third of the beam. A flexural frequency sweep test can provide useful information
of time-temperature-stiffness relationship of asphalt concrete in mechanistic-empirical pavement
design while subjected to climate alteration and change of vehicle speed.
Tsai et al. (Tsai et al., 2005) demonstrated the applicability of the three-stage Weibull equation
to characterize the fatigue deterioration properties of asphalt concrete subjected to various material
variables and testing conditions. The three-stage Weibull equation not only describes the fatigue
damage process but also provides a rational methodology to extrapolate the fatigue life, especially
for rubberized and polymer-modified asphalt mixes that can not reach 50% reduction in stiffness
after testing. The complex modulus master curve obtained from flexural frequency sweep tests is a
useful tool to characterize the effects of loading frequency and temperature on the initial stiffness
of an asphalt mix.
59
The nonconformity of construction quality with design requirements for public works, espe-
cially for civil engineering infrastructure systems, can result in increased expenditures over time.
Thus, development of a rational performance specification to ensure that the construction quality
complies with the design requirements should have a high priority. The development of the fatigue
performance specification where quality assurance (QA) is required for hot-mix asphalt (HMA)
construction makes use of a confidence band concept to statistically determine the lower bound of
fatigue life at specified strain levels.
Figure 1 illustrates a general procedure of pavement design, which includes the following
elements: input of design requirements, selection of pavement structure, selection of trial mix,
conduction of performance testing, determination of performance specification, and performance
simulation to validate the pavement design. The performance specification is statistically deter-
mined using laboratory test results. Performance simulation can be achieved by the recently
developed California Mechanistic-Empirical pavement design program also known as CalME
(Ulidtz et al., 2005), the NCHRP 1-37A: Guide for Mechanistic-Empirical Design of New and
Rehabilitated Pavement Structures (MEPDG) (NCHRP, 2004), and the SHRP-developed analysis
system (Tayebali et al., 1994). In an attempt to rationalize the performance specification, the per-
formance models embedded in performance simulation must be cautiously calibrated based on not
only the laboratory tests but also the in-situ data. The SHRP-developed fatigue analysis system is
used in this paper to judge whether the trial mix would perform satisfactorily in service and meet
the design requirements.
How performance specification relates to performance simulation was demonstrated using a
pavement structure evaluated in the I-710 Phase II project. The I-710 Freeway Rehabilitation
Project in Long Beach, California was one of the first major freeway rehabilitation projects in the
US to incorporate 55-hour weekend closures in the construction of long-life asphalt pavements
(Monismith et al., 2009).
Figure 1. General procedure of pavement design.
60
2 FATIGUE LIFE AND INITIAL STIFFNESS
2.1 Determination of fatigue life using the three-stage Weibull equation

The main purpose to represent the stiffness deterioration process of a four-point bending beam
result using the three-stage Weibull equation is to provide a rational approach to extrapolate the
fatigue life, especially for rubberized and polymer-modified asphalt mixes that can not reach 50%
reduction in stiffness after testing.
The stiffness ratio (SR) at repetition n, which is defined as the ratio of stiffness at repetition n
relative to the initial stiffness (taken at approximately 50 repetitions), will be utilized as an index
in characterizing the phenomenon of the fatigue damage process or stiffness deterioration process.
The stiffness deterioration curve obtained from the flexural controlled-deformation fatigue test can
be adequately expressed as a three-stage Weibull equation with the following form (Tsai et al.,
2005):
where SR = the stiffness ratio, n = loading repetitions, and , = the experimentally determined
coefficients. The subscript numbers indicate the corresponding stages of a stiffness deterioration
curve.
The equation consists of ten parameters 1 , 2 , 3 , 1 , 2 , 3 , 1 , 2 , n1 , and n2 (Figure 2), and
needs to comply with the following continuity conditions.
(1) SR1 = SR2 , when n = n1
SR1 SR2
(2) = , when n = n1
n n
(3) SR2 = SR3 , when n = n2
SR2 SR3
(4) = , when n = n2
n n
Figure 2. Three-stage Weibull curve defined in stages.
61
Figure 3. Example fatigue stiffness deterioration curves with various shapes.
In the fitting of the three-stage Weibull curves, the shapes of stiffness deterioration curves can be
chiefly categorized into four types as shown in Figure 3. Shapes S1 and S4 indicate the appearance
of the cracking propagation stage which requires the interpolation to determine the fatigue life.
Shapes S2 and S3 are usually found in the test situations of high temperature, low strain level,
rubberized or polymer-modified asphalt mixes. In such cases, the three-stage Weibull equation
was found to extrapolate the fatigue life reasonably.
2.2 E* Master curve

The complex modulus master curve obtained from the four-point bending beam tests also known
as flexural frequency sweep tests is a useful tool to characterize the effects of loading frequency
and temperature on initial stiffness of an asphalt mix. A set of Gamma functions (Stone, 1996),
shown in the following, that describes flexural stiffness as a function of temperature and time of
loading can be used for pavement design (Tsai et al., 2004).
The modified gamma fitting of E master curve has the following formulation:
where x = reduced loading frequency (Hz) in natural logarithm and A, B, C, and D are
mathematically determined coefficients.
The modified gamma fitting of the temperature-shifting relationship has the same formulation
as used in the E master curve but with n = 1, D = 0 and can be expressed as follows:
62
Figure 4. Construction and application of E* master curve and time-temperature relationship using genetic
algorithm and Gamma functions (Tsai et al. 2004).
where Ln(aT ) is the temperature shift factor, A and B are mathematically determined coefficients,
and Tref is the reference temperature (20 C).
As an example, Figure 4a presents the flexural frequency sweep test results for temperatures at
5 C, 20 C, 30 C, and 40 C. The loading frequencies used were 15, 10, 5, 2, 1, 0.5, 0.2, 0.1, 0.05,
0.02, and 0.01 Hz. To construct a complex modulus master curve, the test results at temperatures
other than the reference temperature have to horizontally shift toward the curve at the reference
temperature. Figure 4b indicates a satisfactory shifting result obtained using the genetic algorithm
(Tsai et al., 2004). Figures 4c and 4d illustrate a way to correct the initial stiffness with various
loading frequencies and temperatures. The general steps to calculate the initial stiffness at loading
frequency and temperature T are as follows:
(1) Find the temperature difference with respect to the reference temperature.
(2) Map the temperature difference through the temperature-shifting relationship function to obtain
the temperature shift factor (Figure 4c).
(3) Locate the loading frequency in Figure 4d, add up the signed temperature shift factor, and
then map through the matter curve function to obtain the corrected stiffness.
Another way to acquire the time-temperature-stiffness relationship is to collect the initial stiffness
information when conducting flexural fatigue tests at various temperatures and loading frequencies
and then to analyze the collected data with regression analysis. In the following analysis, the
regression equations of initial stiffness will be used.
63
3 FATIGUE LIFE PERFORMANCE SPECIFICATION
The development of the fatigue performance specification where quality assurance (QA) is required
for hot-mix asphalt (HMA) construction makes use of a confidence band concept to statistically
determine the lower bound of fatigue life at specified strain levels.
A (1-)-confidence band of fatigue regression line, a band that has a probability of 1- of con-
taining the true regression line of fatigue life as a function of tensile strain level, is not constructed
from a series of confidence intervals based on y0 , which is an estimate of the mean of the population
of values associated with x0 . Instead, the width of the geometric region at the point x0 that has a
probability of 1 of containing the true regression line is given by (Selvin, 1995):
where F1 is the (1 )-percentile of an F-distribution with 2 and n 2 degrees of freedom.

Notice that the number 2 of n 2 degrees of freedom comes from the fact that the regression equa-
tion has two parameters and has the formulation with y = a + bx, i.e., Ln(Nf ) = a + bLn(strain). For
estimating a mean value from a regression line with the formulation of Ln(Nf ) = a + bLn(strain),
an estimate of variance of y0 (SY2 ) is.
0
2
i yi )
where SY2 |x = (yn2 is the square of residual standard error of the regression equation.
As an example, the 95% confidence band of AR8000 mix, which was the mix used in the
following demonstration, is estimated and illustrated in Figure 5. It is apparent from Equation
6 and Figure 5 that the estimate of variance of y0 is large at extreme values of x0 and smallest
Figure 5. 95% confidence band for AR8000 mixes.
64
when x0 = x. Based on the lower bound of the 95% confidence band, the fatigue performance
specification is then established as: Nf = 2,000,000 at 200 microstrain and Nf = 50,000 at 400
microstrain. The line passing these two points (the circles shown in Figure 5) will be utilized to
verify whether the mix of the pavement structure meets the design requirements.
In engineering practice, to determine the fatigue specification, it is suggested that the mean of
natural logarithm of fatigue life, Ln(Nf ), of three fatigue tests at the specified strain level should
be above the specified lower bound and the regression line obtained from specified strain levels
should be in the range above the lower bound.
4 FATIGUE LIFE PERFORMANCE PREDICTION
The primary purpose of conducting fatigue performance prediction is to verify whether or not the
performance specification can meet the design requirements of fatigue cracking.
A reliability-based mix design system for mitigating fatigue distress (Harvey 1997; Tayebali
1994) illustrated in Figure 6 was utilized to verify the rationality of performance specification. This
fatigue analysis system integrates the laboratory fatigue test results with three factors to evaluate the
allowable ESALs. The factors include (1) shift factor (SF), which was calibrated against Caltrans
design procedure, accounts for traffic wander and crack propagation; (2) temperature conversion
factor (TCF) presented as a function of thickness of asphalt concrete and had been calculated for
California desert, mountain, and coastal environments; and (3) reliability multiplier (M ) takes into
account of variances of traffic demand estimate, constructed thickness, air void content, asphalt
content, and laboratory fatigue test results. In this paper, the TCF was locally determined based on
the temperatures at Long Beach, California. The TCF calculation also involves the strain calculation
for evaluating fatigue life. A multi-layer linear elastic program, ELSYM5, was used to complete
the task.
Figure 6. SHRP-developed fatigue analysis system to determine the allowable ESAL (Harvey 1977).
65
Table 1. Pavement structure and material properties for I-710 project.
Thickness Stiffness Poisson AC AV

Layer [mm (in.)] [MPa (psi)] ratio (%) (%)
PBA6A 76 (3 ) Varied 0.35 5.0 6.0

AR8000 178 (7 ) Varied 0.35 4.7 6.0
AR8000 (Rich bottom) 76 (3 ) Varied 0.35 5.2 3.0
Granular base 152 (6 ) 130 (18854 psi) 0.35
Subgrade 65 (9427 psi) 0.45
Where,
AV is the percent air void content, and
AC is the percent asphalt content.
4.1 Pavement structure, traffic, environment, and materials

A pavement structure evaluated in I-710 Phase II project (Monismith 2009) was used to validate the
rationality of the fatigue specification with the help of the fatigue analysis system. Figure 7 presents
a five-layer system of the I-710 pavement structure and the associated loading configurations for
ELSYM5 input. The layer properties were also shown in Table 1.
Two asphalt binders were utilized: (1) AR-8000 paving asphalt (AASHTO MP1 designation
PG64-16) and (2) polymer-modified asphalt, PBA-6a* (AASHTO MP1 designation PG64-40).
(The * refers to the fact that this material contains additional elastomeric components and exceeds
the normal PBA-6a specification requirements). The aggregate was obtained from the Vulcan
Materials plant in the San Gabriel River Valley at Azusa, California. The aggregate grading used
for mix design is a 19 mm (0.75 in.) nominal maximum size (Monismith 2009).
The standard traffic loading consisted of an 80 kN (18 kips) standard axle load with dual tires
spaced at 330 mm (13 in.) and having an inflation pressure of 710 kPa (104 psi).
The stiffness of asphalt concrete mix varies subjected to the change of temperature. Given
a temperature condition, the layer stiffness of each asphalt concrete sublayer is represented as
that of the middle position of the sublayer (i.e., at the depths of 38 mm, 165 mm and 292 mm)
while conduct the ELSYM5 calculation with the consideration of temperature gradient. The layer
stiffnesses were evaluated using the regression equations of initial stiffness, which is normally a
function of temperature.
Maximum principal tensile strains at the bottom of the asphalt layer (330 mm [13 in.]) were
computed at 10 output locations along the x-axis as shown in Figure 7. The largest (most critical)
maximum principal tensile strain was then used to evaluate the temperature equivalency factor
(TEF) and temperature conversion factor (TCF).
Traffic analysis (Monismith 2009) indicates that the bulk of the truck traffic occurred in the
period 5:00 A.M. to 7:00 P.M. Subsequent analysis was done under the assumption that the daily
truck traffic applied uniformly during the 14-hour period. Also, it is assumed that no wander traffic
was applied onto the specified full-depth pavement structure.
The temperature data used in this calculation is based on the temperature data of Long Beach,
California, which is embedded in the EICM program (Lytton 1990). The pavement temperature pro-
files, temperature gradients, and temperatures at bottom of asphalt concrete were calculated using
EICM program in an hourly base from 09/01/1996 to 01/31/2001 which resulted in a total of 24,195
ELSYM5 runs. Figures 8a and 8b display the temperature histograms for surface temperatures and
temperatures at bottom of asphalt concrete layer respectively. Figure 8c plots the corresponding
temperature distribution curves. Figure 8c illustrates large variation of surface temperature when
compared with the relatively steady temperatures at the bottom of the asphalt concrete layer.
Table 2 lists the regression equations of initial stiffness and fatigue life for PBA-6a* and AR8000
respectively. Notice that the E (Lnstif ) term is the expected value of Lnstif and the number enclosed
66
Figure 7. Pavement structure and ELSYM5 input.
Table 2. Regression equations of initial stiffness and fatigue life for PBA-6a* and AR8000 mixes.
Mix type Response variables Regression equation R2
PBA-6A* Lnstif E(Lnstif) = 9.1116 0.1134 Temp 0.93

AR8000 Lnstif E(Lnstif) = 14.6459 0.1708 AV 0.8032 AC 0.0549 Temp 0.82
LnNf E(LnNf) = 36.5184 0.6470 AV 6.5315 Lnstn 0.83
LnNf LnNf = 30.8192 5.3219 Lnstn NA
Where,
Lnstif is the natural logarithm of initial stiffness (MPa),
LnNf is the natural logarithm of fatigue life,
Lnstn is the natural logarithm of tesnsile strain level,
AV is the percent air void content,
AC is the percent asphalt content, and
Temp is the temperature in C.
by the parentheses is the standard error of the estimated coefficients. The determined fatigue
life equation of AR8000 [without expected symbol, E()] for validation of fatigue performance
specification was obtained according to the fatigue performance specification: Nf = 2,000,000 at
200 microstrain and Nf = 50,000 at 400 microstrain. The other fatigue life equation of AR8000 was
used for the development of reliability multiplier. The regression analysis shows that the stiffness
of PBA-6a* depends only on temperature. Notice that the stiffness of AR8000 depends not only
on temperature but also on air void content and asphalt content.
67
Figure 8. Temperature distributions from 5:00AM 7:00PM: (a) histogram of surface temperatures,
(b) histogram of temperatures at bottom of asphalt concrete layer, and (c) cumulative distribution functions
for surface temperature and temperature at bottom of asphalt concrete layer.
4.2 Pavement structure, traffic, environment, and materials

The temperature conversion factor accommodates temperature variations on fatigue life of asphalt
mix and is calculated as follows:
1. obtain the laboratory fatigue life and initial stiffness equations (Table 2),
2. run ELSYM5 to obtain the maximum principal tensile strain and then calculate the fatigue life,
Nf @Tref = 20C
3. calculate TEFi = , and
Nf @Ti
n
4. TCF = fi TEFi
i
Table 3 lists the TCF calculation for temperatures at the bottom of the asphalt concrete layer
from 5:00AM to 7:00PM. It should be noted that the value of TEFi in each temperature category
is the mean value of that temperature category. Figure 9 exhibits an exponentially increasing trend
of the values of TEFi as temperature increases. High temperature usually softens the mix stiffness
and results in high maximum principal tensile strain occurred at the bottom of asphalt layer. From
the regression equations of fatigue life in Table 2, it shows that higher tensile strain will reduce the
fatigue life. Hence, the higher the temperature the larger the value of TEFi .
4.3 Calculation of Shift Factor (SF)

The shift factor was empirically determined by the formulation of SF = 2.7639 105 1.3586 ,
where is the maximum principal tensile strain. The maximum principal tensile strain,
68
Table 3. TCF calculation (5:00 AM to 7:00 PM).
Mean temperature
Temperature (C), Temperature equivalency factor,
Tb frequency, fi TEFi fi TEFi
12.5 0.007109 0.139388 0.000991

15.0 0.135317 0.259942 0.035175
17.5 0.250589 0.445474 0.111631
20.0 0.175532 1.000000 0.175532
22.5 0.204629 1.976370 0.404423
25.0 0.173631 3.706408 0.643547
27.5 0.051953 6.323180 0.328507
30.0 0.001240 8.399118 0.010414
TCF 1.71022
Figure 9. Temperature equivalency factor with frequency distribution.
55 microstrain, applied to this equation was calculated at bottom of asphalt concrete layer with
a constant temperature, 20 C, across the asphalt concrete layers and resulted in a shift factor of
16.93034.
4.4 Calculation of reliability multiplier (M)

The formulation of reliability multiplier M for a reliability level of 90 percent is shown in the
following (4, 9),
The regression equation of fatigue life of AR8000 mix shown in Table 2 was used to evaluate
Var( ln Nf ) of the reliability multiplier. Notice that the residual standard error of this regression
equation is 1.414. Let x0 = (x00 , x01 , . . . , x0k ) represent a set of values of the covariates in the
regression equation, where x00 = 1 if an intercept is present. In this calculation, for 3.0 per-
cent air void and tensile strain of 55 microstrain, x0 = (1, AV , ln stn) = (1, 3.0, ln(0.000055)).
69
Then, the predicted value of y, i.e., ln Nf , at the point x0 is y0 = x0 b. Hence, the variance
Var(y0 ) = Var( ln Nf ) = 2 [x0 (X X )1 x0 ] (Sen & Srivastava 1990), where X is the design matrix
and 2 the variance that is square of the residual standard error of the regression equation, in this
case, 2 = (1.414)2 = 1.999396. Finally, we have Var( ln Nf ) = 1.207278. Then, by substituting the
value into the formulation, we have, M = exp(1.28 0.22 + Var( ln Nf )) = 4.614483.
4.5 Fatigue performance calculation

The allowable ESAL for this pavement design can be calculated according the formulation,
Lab SF
ESALallowable = Nf
TCFM
, where the NfLab is the laboratory fatigue life of AR8000 (Rich Bottom)
at 20 C with the tensile strain 0.000055. Under the assumption that the trafficking is applied uni-
formly from 5:00 A.M. to 7:00 P.M., the resulting estimated fatigue life was 4.13 109 ESALs,
which is approximately 20 times the design life of 200 106 ESALs. This is more than adequate
to accommodate the assumed design traffic.
5 FINDINGS AND DISCUSSION
An attempt has been made to illustrate an approach using a performance test four-point bending
beam test to ensure that reasonable fatigue performance specification were determined in pave-
ment design. From four-point bending beam test results, the brief discussion of how to determine
fatigue life and initial stiffness, conceptual development of fatigue performance specification, and
the fatigue analysis system, the following observations and suggestions are offered:
1. The methodology introduced in this paper successfully provides a straight answer to the question
of how to relate fatigue performance specification to an acceptable pavement design that meets
the design requirements. However, it should be recognized that the rationalization of perfor-
mance specification resides in the fact that it relates to the accuracy of the performance models
embedded in the performance simulation of analysis system.
2. The SHRP-developed fatigue analysis system is conceptually similar to those conventional
fatigue analyses using the linear-summation-of-cycle-ratios hypothesis, for example, Miners
Law. Regardless of the similarity, it makes several significant improvements by means of the
introduction of shift factor, temperature conversion factor, and reliability multiplier.
3. Even though the fatigue analysis system adopted herein includes shift factor, temperature con-
version factor and reliability multiplier, it still does not take into consideration the fact that
the stiffness of the asphalt concrete layer deteriorates as the fatigue damage gradually accumu-
lates, similar to many conventional fatigue analyses using the linear-summation-of-cycle-ratios
hypothesis to assess the fatigue damage. The incremental-recursive mechanistic-empirical
approaches, such as CalME and MEPDG, by contrast, have circumvented this problem. It
is then suggested that the SHRP-developed fatigue analysis system may be used appropriately
in preliminary pavement design and the incremental-recursive approaches may be introduced
when refinery pavement design is required.
4. In the SHRP-developed fatigue analysis system, the calculation of the temperature conversion
factor requires intensive computing time in finding the Maximum principal tensile strains.
The intensive computing was improved by modifying the ELSYM5 as a subroutine and being
integrated into the program. Once the temperature conversion factors at various climate regions
have been determined by the agency, the contractor only needs to conduct fatigue tests at 20 C
and specified strain levels.
REFERENCES
Harvey, J.T., Deacon, J.A., Tayebali, A.A., Leahy, R.B., and Monismith, C.L. 1997. A Reliability-Based Mix
Design and Analysis System for Mitigating Fatigue Distress, Proceedings, 8th International Conference on
Asphalt Pavements, Seattle, Washington, August, Volume 1, pp. 301323.
70
Lytton, R.L., Pufahl, D., Michalak, C., Liang, H., and Dempsey B. 1990. An Integrated Model of the Climate
Effects in Pavements. FHWA-RD-90-33, Federal Highway Administration, Washington, D.C.
Monismith, C.L., Harvey, J.T., Tsai, B.-W., Long, F., and Signore, J. 2009. Summary Reprot: The Phase
One I-710 Freeway Rehabilitation Project: Initial Design (1999) to Performance after Five-Plus Years of
Traffic (2009). Report prepared for the California Department of Transportation (Caltrans) Division of
Research and Innovation by the University of California Pavement Research Center, Davis and Berkeley.
UCPRC-SR-2008-04.
NCHRP 2004. NCHRP Report 1-37A: Guide for Mechanistic-Empirical Design of New and Rehabilitated
Pavement Structures. Transportation Research Board of the National Academies, Washington, D.C., March
2004.
Sen, A. & Srivastava, M. 1990. Regression Analysis Theory, Methods, and Applications. Springer-Verlag,
Selvin, S. 1995. Practical Biostatistical Methods. Duxbury Press.
Stone, C.J. 1996.A Course in Probability and Statistics, Duxbury Press.
Tayybali, A.A., Deacon, J.A., Coplantz, J.S., Harvey J.T., and Monismith, C.L. 1994. Response of Asphalt-
Aggregate Mixes, SHRP-A-404, Strategic Highway Research Program, National Research Council,
Washington, D.C.
Tsai, B.-W., Kannekanti, V.N., and Harvey, J.T. 2004. The Application of Genetic Algorithm in Asphalt Pavement
Design. In Transportation Research Record: Journal of the Transportation Research Record Board, No.
1891, TRB, National Research Council, Washington, D.C., 2004, pp. 112120.
Tsai, B.-W., Harvey J.T., and Monismith, C.L. 2005. Using the Three-Stage Weibull Equation and Tree-Based
Model to Characterize the Mix Fatigue Damage Process. In Transportation Research Record: Journal of
the Transportation Research Record Board,No. 1929, TRB, National Research Council, Washington, D.C.,
2005, pp. 227237.
Ulidtz, P., Harvey J.T., Tsai, B.-W., and Monismith, C.L. 2005. Calibration of Incremental-Recursive Flex-
ible Damage Models in CalME using HVS Experiments. Report prepared for the California Department
of Transportation (Caltrans) Division of Research and Innovation by the University of California Pave-
ment Research Center, Davis and Berkeley. UCPRC-RR-2005-06. This and several related reports my be
downloaded from www.its.berkeley.edu/pavementresearch/. Accessed July 2, 2007.
71
Development of a standard materials library for mechanistic-empirical

fatigue and stiffness evaluation
L. Popescu, J. Signore, J. Harvey & R. Wu

University of California, Pavement Research Center, Davis & Berkeley,
California, US
I. Basheer & J.T. Holland

California Department of Transportation Caltrans, California, US
ABSTRACT: The Standard Materials Library is a vital component of CalME, the new ME flexible
pavement design program for the California Department of Transportation. The key feature of this
library are fatigue data taken from 4PB tests on laboratory produced HMA mixes representative of
those used throughout the state and from field specimens. This library consists primarily of data on
mixes from multiple aggregate and binder sources and suppliers found in California. The library
affords the engineer the ability to evaluate the influence of binder type and source and, aggregate
type and source/type, and supplier on the performance of a pavement design. The engineer can also
develop rehabilitation designs utilizing properties of the in-place in conjunction with properties of
new HMA materials, which is particularly effective in a low-bid project delivery system, where
the designer does not know the material that will be used ahead of time. Regional databases will
be developed to broaden designers options for alternative materials.
1 INTRODUCTION
The University of California Pavement Research Center (UCPRC) has been supporting the
California Department of Transportations effort in developing and implementing the departments
own MechanisticEmpirical design software, CalME. CalME contains an Incremental-Recursive
Mechanistic-Empirical (IRME) method in which the material properties for the pavement are
updated in terms of damage as the pavement life simulation progresses. The Incremental-Recursive
(IR) approach is based on a model describing the stiffness degradation of each layer caused by
fatigue damage under traffic loading. Two of the key components of CalME supporting the IR
approach are the stiffness master curve and fatigue damage model for the hot mix asphalt material
(Wu et al., 2009). Four-point bending (4PB) tests are crucial for determining the materials param-
eters for these models. Master curve and fatigue damage model parameters for various hot mix
asphalt materials are stored in CalME Material Library database.
The objective of this paper is to introduce the CalME Materials Library database, present its
advantages, and demonstrate its use through a case study. The case study optimizes the use of
the 4PB test data on which the fatigue damage model is based. This approach provides a more
accurate evaluation of various rehabilitation or new construction scenarios allowing performance
comparisons of various materials from different sources. Having the possibility to compare per-
formance of several materials presents the advantage of a more efficient decision in terms of
resource use.
73
2 FATIGUE DAMAGE MODULE IN CALME
The fatigue damage module in CalME comprises the master curves for asphalt materials, the Hot
Mix Asphalt (HMA) fatigue damage model, and the moduli of the unbound layers. The fatigue
damage analysis of flexible pavement structures in CalME is based on the assumptions of multilayer
elastic system, dependence of the HMA layer stiffness on loading time and loading temperature,
and dependence of the unbound layers stiffness on confinement provided by the upper layers as
well as the load amplitude applied by the truck wheel loads (Wu et al., 2009).
2.1 Master curves for asphalt materials

The master curve parameters may be determined from various sources such as: frequency sweep
tests on beams, initial values during fatigue tests, FWD tests, etc.
Equation (1) which follows the model in MEPDG describes the intact HMA modulus versus
reduced time:
Where Ei = the intact modulus in MPa,; tr = reduced time in sec; , , , = experimentally

determined constants, and log = logarithm with base 10.
Reduced time is found from:
Where lt = the loading time in sec; viscref = the binder viscosity at the reference temperature;
visc = the binder viscosity at the present temperature, and aT = constant. This model accounts for
the temperature and loading time dependency of HMA modulus. The binder viscosity is calculated
from Equation 3:
Where T is binder temperature in Rankine; A and VTS are model parameters.
2.2 HMA fatigue damage model

The fatigue damage parameters may be determined from four point, constant strain fatigue tests.
In CalME it is assumed that fatigue damage causes HMA modulus to decrease as in Equation 4:
Where = damage, which can be calculated from the following equations:
74
Where MN = the number of load applications in millions; = the tensile strain at the bottom of
the asphalt layer; E = the (current) damaged modulus; Ei = the intact modulus; t = HMA average
temperature; = t the shift factor that will need to be calibrated; A, 0 , 1 , , , = model
parameters; ref and Eref = reference strain and modulus, respectively, used for normalizing the
ratios. In general = 2 in order to comply with the concept that fatigue damage is driven by strain
energy.
Note: the constants and exponents in equations 47 are not related to those in Equation 1.
3 DETERMINATION OF HMA MATERIAL PARAMETERS
The 4PB beam tests necessary to generate the material parameters for the CalME IR method
are performed according to the AASHTO T-321 (Determining the fatigue life of compacted hot
mix asphalt (HMA) subjected to repeated flexural bending).In order to characterize the HMA
properties, two applications of the 4PB beam test are required: 1) flexural fatigue testing for
characterizing the fatigue damage model and 2) flexural frequency sweep testing for characterizing
the stiffness master curve.
3.1 Determination of HMA fatigue damage model parameters from flexural fatigue tests
Four-point bending flexural fatigue tests are used to measure the fatigue damage of specific mixes
over a range of traffic, which in terms of laboratory testing, translates into stress or strain levels.
Environmental conditions in terms of testing temperatures are also accounted for in these tests.
The concept of fatigue life is based on the universal idea that most materials undergo a gradual
deterioration under repeated loads that are much smaller than the ultimate strength of the material.
Based on the strain failure concept, a fatigue crack initiates at the bottom of a HMA layer and
grows towards the surface. Its development is proportional to the strain level at the bottom of the
layer and it depends on the HMA thickness (thicker pavements give lower strain values), stiffness,
and other properties.
In order to mimic the conditions found in an in-situ structure, flexural fatigue tests can be
performed in a constant stress (stress is maintained constant through the test, while strain is
allowed to vary) or constant strain (strain is maintained constant while stress is allowed to vary)
mode. Experience has shown that in general, thick HMA pavements (>125 mm) perform closer to
a constant stress mode in the field, while thin HMA pavements (<125 mm) perform closer to a
constant strain mode in the field. Based on these observations, it can be stated that the constant
stress mode favors stiffer materials whereas the constant strain mode favors more flexible mixes.
Beam fatigue testing is performed at intermediate temperatures, usually 20 C and sometimes
at 10 C and 30 C, because fatigue cracking is thought to be a primary HMA distress at these
intermediate temperatures. At higher in-service temperatures (above about 38 C) rutting is usually
the HMA distress of greatest concern, while at lower temperatures (below about 4 C) thermal
cracking is usually the HMA distress of greatest concern.
In the four-point bending beam test, the middle one third part of the beam is subjected to pure
bending (constant moment) without any shear deformation. This makes the four-point bending
beam test a good candidate to characterize the fatigue properties of asphalt mixes.
The power fatigue damage model defined by Equations 47 is used to predict laboratory fatigue
test results at 20 C and various temperatures and test strain levels. Model parameters are determined
by matching calculated and measured stiffness reduction curves through a nonlinear optimization
process. During this process the shift fatigue factor is set to 1.0.
75
Table 1. PG Asphalt grades for California.
Asphalt grade California climate regions
PG 64-10 Central Coast, Inland Valley, South Coast

PG 64-16 North Coast, Low Mountain, South Mountain
PG 64-28 High Desert, High Mountain
PG 70-10 Desert
3.2 Determination of HMA stiffness master curve model parameters from

flexural frequency sweep tests
The four-point beam flexural frequency sweep tests are used to characterize the effects of load-
ing frequency on the initial flexural complex modulus of an HMA mix. This permits CalME to
account for vehicle speed and temperature variations when conducting an incremental-recursive
mechanistic-empirical prediction of fatigue performance. In general, flexural frequency sweep tests
are performed at three temperatures, 10 C, 20 C, and 30 C, and a range of frequencies, normally
from quick to slow at 15, 10, 5, 2, 1, 0.5, 0.2, 0.1, 0.05, 0.02, and 0.01 Hz. The flexural complex
modulus (E*) master curve is then constructed based on the MEPDG stiffness model (Equations
13) by fitting the laboratory results.
4 CalME MATERIALS LIBRARY
CalMEs complex analysis and performance prediction capabilities imposed the need to finding a
solution that will allow an easy access to the material characterization information. The solution
materialized in a Materials Library database which is part of CalMEs software installation package.
Presently, the Materials Library database contains a data set of materials characteristics including
mixes, granular materials, recycled materials ranging in source from Heavy Vehicle Simulator
(HVS) test sites, laboratory mixes, and in-situ materials.
In 2006, California Department of Transportation transitioned from the viscosity grading binder
classification system to the performance-graded (PG) binder classification system. Based on
Californias main climate regions, four PG binder grades are currently identified (Table 1).
A comprehensive set of four-point flexural bending beam tests was performed by the UCPRC
laboratory on laboratory mixed, laboratory compacted test specimens in order to cover the most
common mixes used in projects throughout California. This test program comprises three phases
of which two have been already completed. The first stage included fabrication and testing of
different mixes designed with each of the four PG binder categories produced by two refineries,
and aggregates coming from two of most common aggregate sources in California representing a
granite and a river crushed material totaling 48 tests (4 binders times 2 binder sources 2 aggregate
sources 3 replicates). The second phase of testing focused on mixes produced with polymer
modified from one binder source, and same two sources of aggregates used in phase 1 for a total of
6 tests. Rich bottom mix designs with same binder source and same two aggregate sources were also
part of phase 2 (total of 6 tests). Third phase will include mixes made with selected binder grades
from two other refineries and same two aggregate sources. The purpose of this phase is to allow
a sensitivity analysis among same grade binders but produced by different sources. Experience
showed that in many situations a mix made from same grade binder from one source has different
performance when compared with same binder produced from another source. This can have a
significant impact during the mix design, construction and post-construction performance.
Not only material characteristics of mixes produced with PG binders are available in the CalME
Materials Library database, but also mixes produced based on the old viscosity grading system.
These mixes were placed between mid 1990s to mid 2000 on the (HVS) test sections. Specimens
76
prepared in the laboratory from the mixes placed as well as cores extracted at the end of the HVS
tests were tested in flexural bending and their material characteristics are augmenting the Materials
Library database. Mixes include overlays of rubberized asphalt gap-graded, rubberized asphalt with
modified binder, HMA type A 19 mm with AR-4000 binder. These were all standard Caltrans
mixes used throughout the state at the time. The importance of having this information available
is that it allows engineers to be able to represent the old pavement layers properties when using
CalME for new overlay design analysis as it will be illustrated next.
More recently, HMA materials found in-situ at project locations in northern California scheduled
for long life pavement rehabilitation design approach were tested in flexural bending and the
material characteristics were added to the Materials Library database. Another category of mixes
whose properties are characterized in the Materials Library database are mixes from WesTrack
project.
In terms of granular materials, most common materials described according to Unified Soil
Classification System are also characterized in the Materials Library database. Along with the
standard classes of unbound materials, new recycled materials produced from cold-in-place
recycling (CIPR), or full-depth recycling with foam (FDR-Foam) or through pulverization (FDR-
Pulverization), asphalt treated permeable base (ATBP), lean concrete base (LCB), etc. are included
in the Materials Library database.
The structure of CalME Materials Library database is very flexible, permitting the addition of
new materials characteristics from flexural bending tests according to the project needs.
5 OVERLAY DESIGN ANALYSIS CASE STUDY
Following is an example of how the information in CalME Materials Library database could be used
to compare the performance of different overlay scenarios. The example is based on a case study
on one route in Northern California (Popescu et al., 2009). This is a two-lane road (one lane in each
direction) and the transverse profile is characterized by often changes between cut and fill sections.
The field information gathered consisted of as-built plans, time history of traffic data, condition
survey observations, Falling Weight Deflectometer (FWD) measurements, coring, and Dynamic
Cone Penetrometer (DCP) measurements. Several small representative sections were investigated
during the field study. This was due to frequent profile changes (vertical and horizontal) resulting in
safety issues and traffic control difficulties. Each section was considered representative of a larger
stretch of the pavement in terms of condition and structure (from as-built data). This example
focuses on one of the sampled sections. The overall design strategy is to take FWD data from field
measurements and perform backcalculation with Caltrans new backcalculation program CalBack.
CalME then imports CalBack moduli data as the basis for subsequent design work.
Figure 1 shows the variation of the peak deflection (sensor D1) for the selected section. The
average peak deflection is approximately 0.420 mm at an average surface temperature of approxi-
mately 12 C. The large scatter of deflection values within each section could be explained by the
combination of thickness variation across the section and pavement damage. This section exhibited
a moderate to high degree of fatigue and thermal cracking throughout.
Table 2 shows the variation in thickness from cores, the average layer thickness used in
backcalculation, and the average values of the backcalculated moduli for the selected section
The variation of backcalculated layer stiffness is presented in Figure 2.
Backcalculated layer stiffnesses and their corresponding variability were imported from the
backcalculation application (CalBack) into CalME. Structure, material types, traffic, and climate
region were also required CalME inputs. Table 3 lists these inputs for the referred case study.
Appropriate material selection is a key feature enhanced in CalME through, a fairly significant
list of entries into the Materials Library. For the case study presented here, the existing surface
material was a HMA Type A 19 mm for which the master curve and fatigue damage model material
characteristics were determined from beams produced from extracted field ingots and submitted
to four-point flexural frequency sweep and four-point bending flexural fatigue tests respectively.
77
Figure 1. Peak deflection (D1).
Table 2. Section structure and backcalculated moduli.

Existing section Avg. backcalculated stiffness
HMA
Thick. Avg. Avg.
Section range HMA AB Thick.* 80th % Air
length (mm) Thick. (mm) SG soil Defl. Temp HMA AB SG
(m) (Cores) (mm) (Type)** type** (microns) (C) (Mpa) (MPa) (MPa)
760 114 to 147 130 300 (GW-GC) SC-SM 545 12 4,958 299 143
*Thickness of the aggregate base layer was estimated from DCP measurements.
**Classification of the granular materials was based on the Unified Soil Classification System.
For overlay design the performance of the following materials was compared: HMA with PG 64-16
(the recommended grade to be used in mix designs for the climate region (North Coast) in which
the project is located) and a rubberized hot mix asphalt gap graded (RHMA-G).
The master curve characteristics of these materials are listed in Table 4, and the fatigue damage
model characteristics for the same materials are listed in Table 5. The in-situ material characteris-
tics were added later on whereas the other materials characteristics were already available in the
Materials Library, as described in section 4 of this paper.
Figure 3 shows the mix stiffness versus loading frequency in fatigue. From the plot it is apparent
that PG 64-16 HMA is slightly softer than the in-situ HMA at loading frequencies below 10Hz,
whereas the RHMA-G mix has a lower stiffness than both HMA mixes across the overall range of
frequencies. Figure 4 shows a fatigue damage curve for these same mixes. In this figure the same
78
Figure 2. Backcalculated layer stiffness.
Table 3. Analysis section and CalME inputs.
Number of axles in the Traffic growth HMA thickness Base thickness

Climate region first year of analysis (%) (mm) (mm)
North Coast 478456 1.4 130 300
Table 4. Master curve material parameters.
Name aT A VTS Eref (Mpa)
In-situ HMA 2.301 0.9964 0.5993 1.2968 9.6307 3.5047 9065.1

PG 64-16 HMA 19 mm 2.301 0.4806 0.5797 1.2469 9.6307 3.5047 8243.2
RHMA-G 12.5 mm 2.301 0.4642 0.6799 1.3434 9.6307 3.5047 6073.9
Table 5. Fatigue damage model parameters.
Name A 0 ref Eref (MPa)
In-situ HMA 966.684 0.7028 200 4.629 3000 2.3143

PG 64-16 HMA 19 mm 170.136 0.2438 200 7.6749 3000 3.8375
RHMA-G 12.5 mm 36.875 0.0316 200 5.3896 3000 2.6948
79
Figure 3. Master curve of modulus versus frequency.
Figure 4. Modulus versus temperature.
pattern as in Figure 3 is noticed: the RHMA-G mix is noticeably softer when compared to any of
the two HMA mixes (material library and in-situ HMA). The information in Figure 3 and Figure 4
suggests that RHMA-G in general exhibits a better fatigue cracking resistance for structures with
higher levels of traffic.
The thickness of the HMA overlay was determined following the Caltrans Highway Design
Manual (HDM), Topic 635 guidelines (Highway Design Manual, 2008). Based on the deflection
study, the HMA overlay required to maintain the structural capacity is 45 mm. In order to prevent
reflective cracking, the minimum HMA overlay thickness for 20 years design should be 125%
80
Table 6. Results of Incremental-Recursive analysis for overlay design options.
1 2 3 4 5
Rut depth (mm) Cracking (m/m2 )

Rut depth (mm) at the end of Cracking (m/m2 ) at the end of
Design Overlay design at the end of design life with at the end design life with
ID option design life* 90% reliability of design life 90% reliability
1 80 mm PG64-16 2.4 4.2 0 0

HMA Overlay
2 30 mm RHMA-G 3.15 6.5 0 0
overlay
3 Mill 30 mm and fill 4.24 7.8 4.8 10
with 45 mm
PG64-16 HMA
4 Mill 30 mm 4.7 9.0 0 0
and fill with
30 mm RHMA-G
*Design life = 20 years.
of half of the existing thickness, which in this case is 80 mm. The higher of the two (80 mm) is
recommended to be selected as the new overlay. The thickness of an overlay with RHMA-G it was
also determined from HDM Topic 635 guidelines which recommended a minimum of 30 mm to
a maximum of 60 mm provided that appropriate compaction is ensured. In this case an overlay of
30 mm RHMA-G was selected.
Both overlays were verified with CalME Incremental-Recursive method.
The HMA and RHMA-G mill and fill were picked from CalME list of design options to emphasize
other aspects of the use of material library information in the mechanistic-empirical analysis. The
performance of these two designs was also checked using CalME incremental-recursive analysis
method.
Results of all four designs are summarized in Table 6.
New overlay performance was analyzed in terms of fatigue cracking and rutting. Failure criteria
are 0.5 m/m2 for fatigue cracking and 10 mm for rutting. The amount of cracking and rutting at the
end of the design period was also reported for 90% reliability (e.g. there is a 10% chance that by
the end of the design period the amount of fatigue cracking or rutting will exceed the failure limit
for the considered distress).
In CalME Incremental-Recursive simulations it was assumed that the applied traffic is chan-
nelized; the reflective cracking from underneath layers was also accounted for. Design life was
20 years. Results in Table 6 show that the fatigue damage limit of 0.5 m/m2 will not be reached
before the end of design life for Design #1. Similarly, Design #230 mm RHMA-G overlay will
not attain its fatigue cracking limit before the end of design life. Rutting will be below the failure
limit for the two designs and for the 90% reliability criteria.
Figure 5 shows that for Design #3 (mill 30 mm and fill with 45 mm PG64-16 HMA) the fatigue
cracking damage limit is exceeded after approximately 13 years (blue line). Accounting for a higher
risk (the 90% reliability value in Column 5 of Table 6) the fatigue cracking failure criteria has a 10%
chance to be exceeded after the first year (magenta line). Based on the two aspects this design option
is not viable. Premature fatigue cracking failure for this design requires increasing the structural
adequacy by placing a thicker overlay or by selecting a different mix.
Different scenarios can be analyzed for this case using same incremental-recursive method.
81
Figure 5. Mill and fill PG64-16 HMA overlay.
Design #4 (mill 30 mm and fill 30 mm with RHMA-G) shows good fatigue cracking performance
(both fatigue cracking limit at the end of the design life and at the end of the design life with 90%
reliability show no cracking at all). Rutting gets very close to failure limit (10 mm). In this case
the 9 mm total rut accumulated at the end of the 20 year design life it is most likely due to rutting
of the unbound layers since rutting of the surface layer is expected to take place in the first 5 years
of service of the new overlay.
Materials in Designs 1, 2, and 4 who showed good performance could then be compared in terms
of cost, availability of resources and thus the most efficient alternative could be selected.
6 CONCLUSIONS
The Materials Library database is a continually expanding and complex powerful tool to aid the
designer of HMA pavements in consideration of fatigue cracking performance. The example pre-
sented briefly illustrated how material characteristics extracted from four-point flexural bending
tests in conjunction with a fatigue performance modeling tool could aid engineers in making better
decisions in terms of designs and best materials available.
This approach is particularly effective in a design-bid-build (low-bid) project delivery system,
where the designer does not know the material that will be used ahead of time. Having access
to a material library database and a performance model, designers could run several scenarios
including most commonly used materials and compare performance and costs with new materials
or materials available from nearby sources.
Phase 3 of laboratory testing will augment the Material Library with characteristics of mixes
with similar binder grades evaluated in Phase 1 but from different sources. This will allow for
sensitivity analysis of mixes with similar designs and PG grades but with binders from different
sources.
82
ACKNOWLEDGMENTS
This work was undertaken with funding from the California Partnered Pavement Research Program
of the California Department of Transportation, Division of Research and Innovation, which is
greatly appreciated. The opinions and conclusions expressed in this paper are those of the authors
and do not necessarily represent those of Caltrans or the Federal Highway Administration.
REFERENCES
Highway Design Manual, Topic 635 Engineering Procedures for Flexible Pavement and Roadway
Rehabilitation, July 2008
Popescu, L., Signore, J., Harvey, J., Wu, R., Guada, I., Steven, B. Rehabilitation Design for 01-LAK-53, PM
3.1/6.9 Using Caltrans ME Design Tools: Findings and Recommendations, Sept. 2009.
Wu, R., Tsai, B.W., Harvey, J., Ullidtz, P., Basheer, I., and Holland, J. Using four-point bending tests in
calibration of the California mechanistic-empirical pavement design system, 2nd Workshop on Four Point
Bending, 2009.
83
Four point bending beam tests in Arizona and relationship to asphalt

binder properties
G.B. Way
Consulpav, Scottsdale, Arizona, US
K.E. Kaloush
Arizona State University, College of Engineering, Tempe, Arizona, US
J.M.B. Sousa
Consulpav, Walnut Creek, California, US
ABSTRACT: Four Point Bending beam tests (4PB) were conducted on asphalt (bituminous)
hot mixes placed on Arizona Department of Transportation (ADOT) test projects. The evaluated
mixes were constructed over a wide range of climate zones and traffic loadings. The 4PB tests were
primarily conducted by Arizona State University for the ADOT. The tested hot mixes included dense
graded, gap graded and open graded. In addition various asphalt binders were evaluated including
several different performance graded asphalts and asphalt rubber. This paper evaluates the results
of the 4PB tests in terms of the asphalt content, air voids and the asphalt binder properties, as well
as the viscosity temperature intercept and slope values, A and VTS, in terms of the long term field
cracking performance.
1 PURPOSE OF PAPER
The purpose of this paper is to further review the history of the experimental use of the four point
bending beam (4PB) test in the design of several paving projects constructed in Arizona from 1993
to the present. Also, to report on the results of the 4PBB testing associated with these projects
designed and constructed from 1993 to the present. From this historical body of information
and data, discuss the positive qualities and areas of concern of the 4PBB testing method as well as
suggestions for possible improvements. Results of the testing may also be used to estimate cracking
performance and to improve the predictive qualities of the analysis of test results. Considerable
background information related to this study can be found in the first paper presented at the Four
Point Bending proceedings second workshop (Way, 2009).
2 BACKGROUND OF USE BY ADOT OF FOUR POINT BENDING BEAM TEST
As noted in the previous paper the Arizona Department of Transportation (ADOT) has long been
interested in improving the design, construction and performance of its pavements. ADOT has used
several different volumetric asphalt hot mix design methods including the Hveem, Marshall and
Superpave method.
In 1993 the Strategic Highway Research Program (SHRP) recommended the four point bending
beam (4PB) test to estimate the fatigue cracking (Monismith, 1994). The ADOT became very
interested in implementing the 4PB test just as it supported the implementation of the asphalt binder
PG grading system, aggregate consensus properties and the gyratory compactor (Sousa, 1993),
(Way, 1997). However, both the shear test and 4PB tests required very specialized and expensive
85
Figure 1. Original Cox SHRP 4PBB and RSST-CH, shear testing equipment at University of California
at Berkeley.
equipment. Nevertheless, ADOT contracted with a private consulting testing service to conduct
shear test and 4PB testing on several experimental projects designed and built during 1993 to 1995.
Later in about 1999 Arizona State University (ASU) acquired the 4PB equipment and another set of
ADOT projects were designated for 4PB tests in order to further advance the future implementation
of the Mechanistic-Empirical Pavement Design Guide (MEPDG),(AASHTO, 2008).
2.1 4PBB testing in the 1990s

In the 1990s the 4PB test equipment at the University of California at Berkeley (UCB) was used for
4PB tests requested by the ADOT in a manner as described as part of the SHRP Research Project
A-003A (Harvey, 1993), (SHRP, 1994). In the early 1990s 4PB test was conducted at UCB with
a very unique and expensive electro-mechanical-hydraulic loading device within an environment
chamber. The device was designed and built by Cox in cooperation with input from UCB, Figure 1.
2.2 ADOT support of 4PBB testing 19931995

The ADOT contracted with a consulting company to design and test HMA mixes in Arizona using
the new 4PB test equipment. The consulting company recognized the cumbersomeness and cost
of the UCB testing equipment would make the use of this equipment impossible to be used in a
field construction situation. Therefore the consulting company invented a means of placing the
4PB equipment, along with temperature control chamber, in a trailer to be used at a construction
field site. The ADOT contracted with the firm to do such testing on three major overlay projects
86
Figure 2. Trailer with Cox equipment to test RSST-CH, shear test and 4PBB specimens in the field
(Cox, 2009).
from 1993 to 1995. One of the authors of this paper Dr. Sousa co-developed the equipment in the
trailer with assistance from the Cox equipment company and was involved in the design and testing
with this equipment in Arizona, Figure 2. Several project reports that summarized all of the design
and field 4PB testing were developed and submitted to the ADOT (Sousa,1993-1995, 1997). Beam
specimens for design were compacted with a rolling wheel compactor. Beam specimens were
also taken during construction from the roadway pavement and trimmed to the proper size and
orientation for the testing.
2.3 ADOT support of Arizona State University 4PBB testing 1999 to the present
Following the earlier 19931995 4PB efforts and subsequent national research theADOT contracted
with ASU to sample various project HMA mixes and conduct testing consistent with the new
Mechanistic Empirical Pavement Design Guide (MEPDG) pavement design procedure. This testing
included 4PB testing. The selected projects included SHRP gyratory mixes as well as Marshall
design asphalt rubber gap graded (ARAC) and specially designed asphalt rubber open graded
(ARFC) mixes. The mixes were sampled as loose hot mix on the project and then brought back
to the ASU laboratory. The loose mix was compacted using vibratory loading applied by a servo-
hydraulic loading machine. A beam mold was manufactured at ASU with structural steel that is not
hardened at the appropriate dimensions to develop a specimen that can be trimmed to the proper
dimensions for testing. Since the earlier testing conducted from 19931995 there had been much
improvement in the testing protocol and documentation. The 4PB testing conducted by ASU is
essentially in compliance with the AASHTO protocols, (AASHTO, 2003). Additional details about
the ASU specimen preparation, compaction and testing procedure and 4PB fatigue calculations see
the previous 4PB report (Way, 2009).
87
2.4 Summary background of the flexural beam fatigue test
Load associated fatigue cracking is one of the major distress types occurring in flexible pavement
systems. The action of repeated loading caused by traffic induced tensile and shear stresses in
the bound layers, which will eventually lead to a loss in the structural integrity of a stabilized
layer material. Fatigue initiates cracks at points where critical tensile strains and stresses occur.
Additionally, the critical strain is also a function of the stiffness of the mix. Since the stiffness of an
asphalt mix in a pavement layered system varies with depth; these changes will eventually effect
the location of the critical strain that varies with depth; these changes will eventually effect the
location of the critical strain that causes fatigue damage. Once the damage initiates at the critical
location, the action of traffic eventually causes these cracks to propagate through the entire bound
layer.
Over the last three to four decades of pavement technology, it has been common to assume that
fatigue cracking normally initiates at the bottom of the asphalt layer and propagates to the surface
(bottom-up cracking).
During the 1990s testing the 4PB fatigue model shown in equation 1 was used to estimate the
loading repetitions to failure as a function of the strain applied to the beam.
t = tensile strain at the critical location. Strain is in actual test units, example 0.000100 inch/inch.
Nf is the number of load cycles which represent a surrogate for 18 kip Equivalent Single Axle
Loads which ultimately lead to fatigue cracking in the pavement.
Later and currently the most common model form used to predict the number of load repetitions
to fatigue cracking is a function of the tensile strain and mix stiffness (modulus), equation 2.
where:
Nf = number of repetitions to fatigue cracking
t = tensile strain at the critical location
E = stiffness of the material
K1 , K2 , K3 = laboratory calibration parameters
In the laboratory, two types of controlled loading are generally applied for fatigue characteriza-
tion: constant stress and constant strain. All of the 4PB testing conducted and reported in this study
is a constant strain controlled loading.
2.5 Testing equipment

Flexural fatigue tests were performed according to the AASHTO T321-03 (AASHTO, 2003), and
SHRP M-009 (SHRP, 1998). The flexural fatigue test has been used by various researchers to
evaluate the fatigue performance of pavements (Harvey, 1993), (Tayebali, 1995). Figure 3a shows
the flexural fatigue apparatus. The device is typically placed inside an environmental chamber to
control the temperature during the test.
The cradle mechanism allows for free translation and rotation of the clamps and provides loading
at the two center points as shown in 3b. Pneumatic actuators at the ends of the beam center it later-
ally and clamp it. Servomotor driven clamps secure the beam at four points with a pre-determined
clamping force. Haversine or sinusoidal loading is applied to the beam via the built-in digital
servo-controlled pneumatic actuator. The innovative floating on-specimen transducer measures
and controls the true beam deflection irrespective of loading frame compliance. The test is run
under either a controlled strain or a controlled stress loading. Note the AASHTO T321-03 states
the following in relation to the application of the testing device loading and resultant deflection:
88
Figure 3a. ASU 4PBB testing apparatus.
Figure 3b. Loading characteristics of the 4PB apparatus.
. . . The loading device shall be capable of (1) providing repeated sinusoidal loading at a frequency
of 5 to 10 Hz, (2) subjecting specimen to 4-point bending with free rotation and horizontal transla-
tion at all load and reaction points, (3) forcing the specimen back to its original position (i.e. zero
deflection) at the end of each load pulse.
3 RESULTS OF FOUR POINT BENDING BEAM TESTING (IPC, 2009)
3.1 Projects and materials

Since 1993 several projects have been sampled and tested. Figure 4 shows their approximate
location. The projects along with their traffic loading in 18 kip single axle loads per year are shown
in Table 1. The projects are mostly overlay projects of varying thickness from 50 mm to 100 mm in
thickness. All the projects have a nominal 19 mm thick ARFC surfacing except for the Perryville
project (project number 2) which did not have an ARFC surface. The projects represent various
89
Figure 4. Project locations.
asphalt mixes including dense graded HMA, gap graded asphalt rubber (ARAC) and open graded
asphalt rubber (ARFC). An average gradation of each type of mix is shown in Table 2, along with
their respective typical average binder content, air voids and voids in the mineral aggregate (VMA).
3.2 4PB test results in terms of Equation 1

For each project a full report was prepared with considerable detailed information (Sousa, 1993
1995), (Kaloush, 20022008). In the first 4PB paper (Way, 2009) the 4PB test results in the 1990s
and 2000s were shown in extensive tables in terms of equation 1. This was done so that the early
4PB test results in the 1990s could be compared to the ASU test results in the 2000s. The 4PB tests
were conducted at the appropriate estimated fatigue temperature or range of temperatures. During
design or using construction field samples to prepare the 4PB test at least six test specimens were
prepared and a straight line fitted to the test results as shown in Figure 5. The correlation R2 values
90
Table 1. Project location and traffic loading.
Project ID Name Route # Milepost Begin & End Year built ESALs/yr (000s)
1 Deer Valley I-17 224.5225.8 NB 1993 1000

2 Perryville I-10 112.3122.3 WB 1995 3000
3 Sedona I-17 299.0311.6 NB 1995 600
A Antelope Wash US 93 95.1101.9 2006 200
B Burro Creek US 93 138.0142.5 2006 200
C Buffalo Range I-40 224.7229.9 2001 2200
D Salt River Int. Various 2001
E Silver Springs I-40 79.586.2 2006 1500
F Badger Springs I-17 256.0263.0 2005 600
G Kohls Ranch SR 260 266.0269.0 2007 100
H Two Guns I-40 230.0240.0 2003 1500
I Jackrabbit I-40 268.0277.4 2003 2000
Table 2. Average percent passing and various mix properties.
Sieve mm Dense HMA ARAC ARFC
25.40 100 100 100

19.00 98 100 100
12.70 81 82 100
9.50 69 73 100
6.40 57 49 68
4.75 50 37 36
2.36 36 20 7
2.00 32 18 6
1.18 25 12 4
0.60 17 8 3
0.42 14 6 2
0.30 11 5 2
0.15 6 3 2
0.075 4 1.7 1.2
% Binder 4.9 7.0 9.2
% Air Voids 6.6 8.6 18
% VMA 16.0 20.0 32.0
were typically greater than 0.90. Table 3 shows a summary of the average 1990s test results for the
three projects. Table 4 shows the ASU 4PB test results.
The Table 3 results are for typical dense grade mixes of various asphalt binder content, con-
struction air voids. From the derived k1 and k2 coefficients it was possible to calculate the number
of load repetitions to failure (Nf) at a constant strain. For this paper a 100 micro-strain (.000100)
level was selected for the calculation of the load repetitions to failure. As can be seen for the 1990s
data the overall average Nf value for the dense mixes was calculated to be 1.26E+08 at an average
temperature of 18.8 C.
Table 4 shows the test results for the mixes tested by ASU in the 2000s (note air voids measured
using the Corelok device (Corelok, 2009)). Reviewing the ASU test results conducted on the dense
graded mixes at an average temperature of 21.1 C gave an average Nf value of 1.29E+08 which
indicates that both sets of tests are comparable in estimating the Nf which is a good measure of
91
Figure 5. Example number of load cycles versus strain at 50 percent of initial stiffness. Each data point is a
separately compacted beam.
Table 3. 4PB equation 1 dense graded average test results for 19931995 projects.
PG/AC % Test
Mix binder % Air temp Calc. Nf at
Site type grade Bind voids C k1 k2 = 0.0001
1 SHRP H 70-10 4.2 7.2 20 2.37E10 5.36 2.44E+08

2 SHRP 76-10 4.4 6.3 30 4.31E04 3.23 7.30E+07
3 SHRP 70-22 4.8 5.8 15 1.60E08 4.19 1.24E+08
All Average All 4.6 6.0 18.8 9.39E05 4.10 [1.26E+08]
Table 4. 4PB equation 1 ASU average test results for 20022008 projects.
PG/AC % Test
Mix binder % Air temp Calc. Nf at
Site type grade Bind voids C k1 k2 = 0.0001
All AC 70-10 4.8 7.1 21.1 8.79E12 5.07 [1.29E+08]

All ARAC 58-22 7.2 10.0 21.1 7.50E09 5.53 7.75E+09
All ARFC 58-22 9.2 18.4 21.1 3.14E07 6.09 3.84E+13
the usefulness and accuracy of the 4PB test. It should be noted that it is probably not appropriate
to average the k1 and k2 coefficients albeit the end result Nf values can be averaged.
3.3 4PB test results in terms of Equation 2

It is recognized that currently 4PB test results are expressed in a fatigue equation as shown in
equation 2 and expressed in Figure 6 with further testing at standard temperatures of 4.4, 21.1 and
37.8 C. with selected project results shown in Table 5 (note this is not all of the test results).
Table 5 test results were further reviewed with regard to the calculation which are shown in
Table 6. The Nf values were derived by averaging all of the generalized equations at a constant
strain and representative stiffness. The representative stiffness are an average from the 4PB testing.
92
Figure 6. Example number of load cycles versus strain at various temperatures.
Table 5. 4PB equation 2 ASU generalized fatigue equations.
Project 50% of Initial stiffness, So @ N = 50 Cycles
Material type Mix type K1 K2 K3 R2
G Kohls Ranch PG 64-22 Conventional HMA 1.61E04 4.48 1.06 0.92

C Buffalo Ranch PG 64-22 Conventional HMA 1.32E03 4.95 1.53 0.97
E Silver Springs PG 70-22 Conventional HMA 3.28E03 4.20 1.28 0.74
H Two Guns PG 64-22 Conventional HMA 1.94E04 5.14 1.55 0.72
I Jack Rabbit PG 64-22 Conventional HMA 6.26E04 3.53 0.62 0.97
C Buffalo Ranch PG 70-40 ARAC 2.50E02 4.32 1.27 0.75
F Badger Springs PG 70-28 ARAC 2.97E05 6.68 2.12 0.66
H Two Guns PG 76-34 ARAC 1.18E08 8.18 2.60 0.79
C Buffalo Ranch PG 70-40 ARFC 7.81E+03 3.00 1.53 0.99
B Burro Creek PG 70-22 ARFC 1.10E04 4.86 3.07 0.71
A Antelope Wash PG 70-22 ARFC 9.00E09 5.30 1.87 0.96
E Silver Springs PG 70-28 ARFC 1.045 2.36 0.45 0.33
F Badger Springs PG 70-28 ARFC 0.749 4.18 1.46 0.95
H Two Guns PG 76-34 ARFC 4.58E+03 3.70 2.01 0.65
I Jack Rabbit PG 76-40 ARFC 5.99E+02 5.12 2.74 0.65
As can be seen the mixes with asphalt rubber and a higher binder content have the higher level of
load repetitions to failure.
4 CRACKING PERFORMANCE 1990S TO THE PRESENT
Table 7 is a summary of the ADOT percent cracking for each of the projects under review from their
time of construction to 2011. The percent cracking shown in Table 7 is representative of the degree
of cracking shown in Figure 7, (Way, 1979). The percent cracking is representative of the percent of
a 1000 square foot area of the travel lane that has cracking (Way, 1976). The percent cracking can be
converted to the lineal feet of cracking in a 1000 square foot area of the travel area by multiplying
93
Table 6. 4PB equation 2 ASU calculated Nf load repetitions.
Strain Material type E Stiffness (ksi) Nf load repetitions
.000100 AC 400 9.15E+07

.000100 ARAC 160 8.48E+10
.000100 ARFC 50 3.81E+11
.000500 AC 400 5.42E+04
.000500 ARAC 160 1.17E+06
.000500 ARFC 50 2.16E+07
.001000 AC 400 2.75E+03
.001000 ARAC 160 2.20E+04
.001000 ARFC 50 2.32E+05
Figure 7. Example load cycles versus strain.
the percent cracking by 13.8 (Way, 1979). As can be seen in Table 5 the Perryville project, number
2, pavements with low binder content and no ARFC surface have the greatest degree of surface
cracking. All of the other projects have an ARFC surfacing and their percent cracking is 4 percent
or lower. Rut depth and skid resistance are also satisfactory for all of the reported projects.
5 ASPHALT BINDER A AND VTS PROPERTIES
The neat unmodified binders used in this study were characterized by using the PG grading system.
However, the AR binders at present cannot be characterized with the PG grading system mainly
because the AR binders contain rubber particles that are about 1 mm in size. The PG grading system
uses a dynamic shear rheometer (DSR) that uses parallel plates to test the asphalt binder and the
plates are separated by a 1 mm gap. Thus the AR particles are too large to test in the DSR rheometer.
ASU devised a method of estimating the PG grade by plotting routine binder tests on a log log plot
and determining A-VTS parameters, Figure 8. In the MEPDG manual there is a table where the
A-VTS values can be associated with a particular PG grade. For example Table 8 is derived from
94
Table 7. Cracking performance of experimental projects.
Year YR YR YR YR YR YR YR YR YR YR YR YR YR YR
built 3 4 5 6 7 8 9 10 11 12 13 14 15 16
1-Deer Valley 1993 0 0 0 0 0 0 0 0 0 3 3 3 RP

2-4.3% 1995 0 0 0 0 0 0 1 4 7 8 RP
2-4.3% 1995 0 0 0 0 0 1 2 7 8 8 RP
2-3.8% 1995 0 0 1 15 15 15 20 25 30 30 RP
2-4.0% 1995 0 0 0 0 0 2 4 6 10 10 RP
2-ARAC 6.3% 1995 0 0 0 0 0 0 0 0 0 0 0 0 0 0
3-SHRP 1995 0 0 0 1 1 1 3 3 3 3 3 4 4 4
3-SHRP 1995 0 0 0 0 0 0 0 0 0 1 1 1 2 2
3-SHRP 1995 0 0 0 0 0 0 0 0 4 4 4 4 4 4
3-ARAC 1995 0 0 0 0 0 0 0 0 0 0 0 1 1 1
3-ARAC 1995 0 0 0 0 0 0 0 0 0 0 0 0 1 1
A-Antelope Wash 2006 0 0 0 0 0 0
B-Burro Creek 2006 0 0 0 1 1 1
C-Buffalo Range 2001 0 0 1 1 2 2 3 3 3 3
E-Silver Springs 2006 0 0 0 0 0 0
F-Badger Springs 2005 0 0 0 0 0 0 0
G-Kohls Ranch 2007 0 0 0 0 0 0
H-Two Guns 2003 0 0 0 0 0 0 0 0
I-Jackrabbit 2003 0 0 0 0 0 0 0
RP means section
has been re-paved
Figure 8. MEPDG Binder A-VTS properties for Asphalt and Asphalt-Rubber.
95
Table 8. Equivalent AR PG grades by using the A-VTS values.
Orig. Original RTFO RTFO

Binder type A VTS A VTS
PG 58-22 base asphalt 11.164 3.764 11.076 3.722

PG 58-22 after rubber added (AR) 8.3595 2.726 8.0475 2.598
AR binder equivalent A-VTS like a PG 70-40 8.129 2.648 8.129 2.648
PG 64-16 base asphalt 11.163 3.755 11.116 3.728
PG 64-16 after rubber added (AR) 8.39 2.738 8.543 2.781
AR binder equivalent A-VTS like a PG 76-34 8.532 2.785 8.532 2.785
a paper presented in South Africa (Kaloush, 2011). As the paper showed how the A-VTS values
for unmodified asphalt PG grades PG 58-22 and PG 64-16 changed after blending with scrap tire
rubber to make AR. The A-VTS values become equivalent to PG 70-40 and PG 76-34 as indicated
in the MEPDG manual PG, A-VTS crossover table. This method allows a means to use AR binder
in the MEPDG until such time as a test is developed to determine the DSR PG grade according to
SHRP.
6 DISCUSSION OF FOUR POINT BENDING BEAM TEST RESULTS
The use of the 4PBB test as a pavement analysis tool and/or design tool has been under research
investigation since at least the late 1980s. Over the years the manner in which this tool is used to
estimate or predict fatigue cracking should be used has evolved.
Three asphalt mix and binder types were evaluated including the typical dense grade mix, asphalt
rubber gap and open graded mixes. The 4PB test discriminated between the three mixes indicating
that the asphalt rubber mixes would have longer fatigue lives which are consistent with field
observations over a 16 year period.
Various neat unmodified asphalt binder types were evaluated including PG 70-10, PG 76-10, PG
70-22, PG 64-22 as well asphalt rubber which is a highly modified asphalt. The PG graded asphalt
binders used in the dense grade mix indicated similar predicted fatigue lives which are probably
due to the asphalt binder contents being of a similar amount of 4.6 to 4.8 percent by weight of
the mix. The 4PB test results for the asphalt rubber gap graded mixes and open grade mixes had
higher amounts of binder content of 7.2 and 9.2 respectively, also by weight of the mix. This higher
amount of binder along with the gap graded and open graded gradations provided for a greater film
thickness which along with the binder content contributes to the higher fatigue life.
Surprisingly even though the asphalt rubber mixes have higher binder content they have higher
air voids as well. Dense graded mixes average about 6.0 to 7.1 percent air voids whereas the asphalt
rubber gap and open graded mixes average 10.0 and 18.4 respectively. Normally such high air
voids would be suspected to lower fatigue lives. However, even with high air voids the mixes gave
very much higher fatigue lives. This is probably due to the nature of the asphalt rubber elastic
properties which contribute to higher binder amount, higher film thickness as well polymer like
elastic recovery properties.
Asphalt rubber presently cannot be PG graded with the DSR due to the rubber particle size,
however ASU used the A and VTS properties of the asphalt rubber binder to estimate the PG grades
which varied from PG 70-22 to PG 76-40. Such PG grades coupled with the higher binder content
and elastic properties of the asphalt rubber all contributed to the longer fatigue life.
As this paper and the previous paper has shown even with changes in the 4PB fatigue equations
comparable values of predicted fatigue failure can be obtained. In addition the 4PB appears to
predict the relative order of field cracking performance as shown in Figure 9 even though there is
96
Figure 9. General cracking trend of ADOT dense graded with no asphalt rubber and asphalt rubber mixes.
beam variability. In addition unusual mix types like the gap graded and open graded asphalt rubber
mixes can be successfully tested.
Although this is only a small example of what the 4PB test can predict, it is a start and will
greatly help to reinforce confidence in using this test method in the future. Much work still remains
in how to interpret and integrate the 4PB test into a practical pavement design and mix control
procedure.
ACKNOWLEDGMENTS
The authors thank Mark Bouldin for all his expert contribution to the testing and reporting on the
19931995 projects. Thanks, also to several ASU students for their contribution to sampling, testing
and conducting numerous 4PBB research studies and preparing reports. In particular we thank
Aleksander Zborowski, Mohammad Abojaradah, Andres Sotil, Krishna Biligiri, Maria Rodezno,
Walled Zeiada, Luiz de Mello and Luiz Guilherme and of course Kenny Witczak for keeping track
of all the field sampling and laboratory testing.
DISCLAIMER
The contents of this report reflect only the views of the authors. The authors do not endorse
specific standards, proprietary products or manufacturers. Associations, trade or manufacturers
names appear herein solely because they are considered essential to the object of this report.
REFERENCES
AASHTO, 2003, Designation: T321-03. Determining the Fatigue Life of Compacted Hot-Mix Asphalt (HMA)
Subjected to Repeated Flexural Bending. 2003.
AASHTO, 2008, Mechanistic-Empirical Pavement Design Guide, Interim Edition: A Manual of Practice.2008.
97
Corelok, 2009. www.instrotek.com/corelok.htm
Cox, 2009. www.jamescoxandsons.com
Harvey, J. 1993, Effect of Laboratory Asphalt Concrete specimen Preparation Variables on Fatigue and Per-
manent Deformation Test Results Using Strategic Highway Research Program A-003A Proposed Testing
Equipment, Transportation Research Record 1417, Transportation Research Board, Washington, D.C.,
1993.
IPC, 2009, www.ipcglobal.com.au
Kaloush, ASU, 2007, Performance Evaluation of Asphalt Mixtures in Arizona Silver Springs and Badger
Springs Projects Final Report, Arizona State University, Report prepared for the Arizona DOT. March,
2007.
Kaloush, ASU, 2008, Performance Evaluation of Asphalt Mixtures in Arizona Burro Creek and Antelope
Wash Projects Final Report, Arizona State University, Report prepared for the Arizona DOT. July, 2008.
Kaloush, ASU, 2006, Performance Evaluation of Asphalt Mixtures in Arizona Two Guns and Jackrabbit
Projects Final Report, Arizona State University. Report prepared for the Arizona DOT. July, 2006.
Kaloush, ASU, 2002, Performance Evaluation of Arizona Asphalt Rubber Mixtures Using Advanced Dynamic
Material Characterization Tests Buffalo Range Project Final Report, Arizona State University, Report
prepared for the Arizona DOT, July, 2002.
Kaloush, ASU, 2008. Performance Evaluation of Asphalt Mixtures in Arizona Kohls Ranch Project Final
Report, Arizona State University, Report prepared for the Arizona DOT, July, 2008.
Kaloush, 2011, Kaloush, Kamil, Maria Carolina Rodenzo, Krishna P. Biligiri, George B. Way, Mark Belshe,
Mechanistic-Empirical Pavement Design Guide Implementation and Pavement Preservation Strategies
with Asphalt Rubber", The 30th Annual Southern Africa Transport Conference, CSIR International
Convention Center, Pretoria, South Africa, July 1114, 2011.
Monismith, C. L, 1994, Permanent Deformation Response of Asphalt-Aggregate Mixes, Strategic Highway
Research Program, National Research Council, SHRP-A-415, 1994.
Monismith, C. L, 1994, Fatigue Response of Asphalt-Aggregate Mixes, Strategic Highway Research Program,
National Research Council, SHRP-A-404, 1994.
SHRP, 1998. Designation: M-009. Standard Method of Test for Determining the Fatigue Life of Compacted
Bituminous Mixtures Subjected to Repeated Flexural Bending. 1998.
SHRP, 1994, Designation: SHRP-A-404. Fatigue Response of Asphalt-Aggregate Mixes, Strategic Highway
Research Program, National Research Council. Washington D.C. 1994.
Sousa, J., APT, 1993. Mix Design for Test Section I-17, Deer Valley, Applied Paving Technology. Report
prepared for the Arizona DOT. December, 1993.
Sousa, J., APT, 1995. Mix Design for Test Section I-17, Sedona, Applied Paving Technology. Report prepared
for the Arizona DOT. December, 1995.
Sousa, J., APT, 1995. Mix Design for Test Section I-10, Perryville, Final Report, Applied Paving Technology,
Report prepared for the Arizona DOT. November, 1995.
Sousa, J., 1997, Comparison of Mix Design Concepts, Transportation Research Record No. 1492,
Transportation Research Board, 1995.
Tayebali, 1995, Development and Evaluation of Surrogate Fatigue Models for SHRP A-003A Abridged Mix
Design Procedure, Journal of the Association of Asphalt Paving Technologists Vol. 64, 1995, pp. 340-366.
Way, G., 1997, Arizona SHRP Experience, Progress of Superpave, Evaluation and Implementation. In ASTM
STP 1322; ASTM Publication Number 04-013220-08. ISBN 0-8031-2418-X. September, 1997.
Way, G., 1976, Prevention of Reflective Cracking in Arizona Minnetonka-East (A Case Study), Report
Number FHWA-AZ-HPR-224, Arizona Department of Transportation, May 1976.
Way, G. B., 1979, Prevention of Reflective Cracking Minnetonka-East, Report Number 1979GWI, Arizona
Department of Transportation, August 1979.
Way, G., 2009, Arizonas 15 Year Experience Using the Four Point Bending Beam Test. Four Point Bending
proceedings of the second workshop, University of Minho, 2425th September 2009, Guimaraes, Portugal,
IBSN 978-972-8692-42-1, September, 2009.
98
First steps for the perpetual pavement design:

Through the analysis of the fatigue life
N. Hernandez
SemMaterials Mxico, Puebla, Mxico
J. Hernandez & R. Martinez

SemMaterials Mexico, Puebla, Mexico
ABSTRACT: The main structural distress issues in Mexico for asphalt pavements is bottom-up
fatigue cracking, mainly due to thin pavement structures unable to withstand the damage caused
by heavy loads and increased traffic volume. This article analyzes the fatigue life of two asphalt
mixtures designed in our laboratory in order to be part of a Perpetual Pavement, for which fatigue
tests were performed using four-point bending beams. The fatigue limit approach is used to esti-
mate the Fatigue Endurance Limit, and later compared with the value of 120 proposed in the
pavement design. The test results demonstrate how the two mixtures exceed the design parameters,
guaranteeing the performance of the Perpetual Pavement.
1 INTRODUCTION
In Mexico, asphalt pavements often exhibit premature failures in a short period of time after their
construction. Some of the possible causes of this behavior are the following: poorly designed
pavements, use of inappropriate materials, excessive vehicle loads, poor construction procedures
and little or no quality control, etc.
Due to the importance of the highway network in Mexico, it is necessary to eliminate the prema-
ture failure of asphalt pavements by correcting the causes listed above. One step is to implement
new pavement structures such as the structure type known as Perpetual Pavement, defined as fol-
lows: an asphalt pavement designed and built to last longer than 50 years without requiring major
structural rehabilitation or reconstruction, and needing only periodic surface renewal in response
to distresses confined to the top of the pavement (APA, 2002).
In a classical pavement design, as design load applications increase, pavement thickness must
increase also. There is a growing belief that bottom-up fatigue cracking does not occur for thick
pavements, limiting it only to top-down cracking on the surface resulting in easier and less expen-
sive repairs. The concept of the HMA fatigue endurance limita level of strain below there is no
cumulative damage over an indefinite number of load cyclesis proposed to explain this occur-
rence. Therefore, pavement thickness greater than the one required to keep strain levels at the
bottom of the HMA layer below the endurance limit would not provide additional life. This is
the main parameter used in the design of perpetual pavements and has a significant economical
implication.
The objective of this paper is to describe the procedure and the tools used to design a perpetual
pavement, focusing on the evaluation of the fatigue strength of a rich asphalt bottom asphalt layer.
Two types of asphalt mixtures are designed and analyzed in the study; each has to meet different
parameters according to the function that has to meet in the pavement. The two different types of
asphalt binders considered in the study are classified by the PG system as PG 76-22 and PG 82-22
(according to AASHTO T315-06).
99
Figure 1. Perpetual pavement design concept (Newcomb et al., 2000).
Fatigue resistance is analyzed in both design mixtures, estimating the fatigue endurance limit of
each one which is compare with the proposed in the pavement design.
2 STRUCTURAL DESIGN OF THE PAVEMENT
2.1 Design concepts

The perpetual pavement concept derives from the mechanistic principles, where the design of the
structure consists of asphalt layers, each with different characteristics, using appropriate materials
and carrying a suitable construction procedure. By following this, the structure will stand the
traditional design life while simultaneously sustain high traffic volumes/loads.
According to Walubita and Scullion (2010), the philosophy of perpetual pavement design is such
that the structure must meet the following principles:
Have enough strength to resist structural distresses such as bottom-up fatigue cracking,
permanent deformation, and/or rutting; and
Be durable enough to resist damage due to traffic forces (abrasion) and environmental effects
(e.g., moisture damage).
The perpetual pavement mechanistic design principle consists in providing enough stiffness in
the upper pavement layers to prevent rutting and enough total pavement thickness and flexibility
in the lowest HMA layer to avoid bottom-up fatigue cracking (Figure 1). Like any other pavement
structure, the extended performance relies on a solid/stable foundation to provide long-term support
to the pavement structure under traffic loading as well as to reduce seasonal support variation due
to environmental effects (e.g., freeze-thaw and moisture changes).
2.2 Pavement design

The study was conducted to design a pavement in the state of Nuevo Leon, Mexico. The road is
identified as MEX-054, the design speed is 110 km/hr, with a driving lanes width of 7.0 m and
shoulder width of 5.0 m and the total length of the project is 11 km. The proposed structure is
described below.
The bottom layer is called Fatigue Resistant Layer (FRL), designed to mitigate fatigue cracking.
The intermediate layer or High Modulus (HM), designed to combine the features of stability and
durability.
The surface or wearing layer, designed to protects the pavement from top down cracking and the
environmental effects, mainly from the introduction of surface water and aging, as well as good
100
Table 1. Monthly average temperature and rainfall.
Average temperature Average rainfall

Month ( C) (mm)
January 13.0 17.8

February 13.7 14.0
March 19.3 17.5
April 22.8 129.0
June 26.1 25.9
May 28.4 62.0
July 26.0 283.0
August 27.9 11.2
September 25.6 257.0
October 3.7 1.5
November 19.3 0.3
December 12.8 0.3
Table 2. Proposed modulus for the layers of the pavement.
Layer Mr (MPa) |E*| (MPa) Reference standard
High Modulus 10,000a AASHTO TP 62-07

Fatigue Resistant 3,000a AASHTO TP 62-07
Subgrade 120 N CMT 1 03/02
a Notes: Dynamic modulus specified at 20 C and 10 Hz.
skid resistance and ride quality. Due to its thickness (30 mm) this is not considered within the
structural design.
The main factors associated with pavement design were traffic data, the environmental conditions
of the area and the characteristics of the materials. The traffic information was obtained for the
project location from data published annually by the General Director of Technical Services (DGST,
for its acronym in Spanish). The Annual Average Daily Traffic for this section of road is 7022 from
a 2009 survey with a planned growth rate of 3.8%.
The Information about weather condition in the area was obtained according to the monthly
average temperature and rainfall (Table 1). The data used in the study was obtained from the
weather station UDEM Prepa Founders, located in Escobedo, Nuevo Leon, with geographical
coordinates Lat: N 25 46 23 Lon: W 100 17 42 and elevation of 496.82 m.
The main materials parameter used for the design is the modulus of each layer, used as an input
parameter in the design procedure. The moduli were proposed according to the requirements of the
regulations in Mexico for the non-stabilized layers case. Moduli were proposed to asphalt layers
according to the role that each one has to fulfill. Table 2 shows the proposed values.
Currently Mechanistic Empirical methodologies (M-E) are the most suitable for the design of
perpetual pavements. For Von Quintius (2006) Mechanistic-Empirical (M-E) based procedures are
used to design the flexible pavement to limit load related cracking. All M-E based design procedures
can be grouped into three types related to load-induced cracking, which are listed below.
1. Design procedures that use the equivalent axle load and equivalent temperature concepts. The
equivalent temperature is determined based on an annual or monthly basis. These procedures
usually use the cumulative damage concept to determine the amount of fracture damage over
the design period for each structure. The Asphalt Institutes DAMA program falls within this
category.
101
Table 3. Thickness obtained for the design structure.
Thickness (mm)
Layer DAMA PerRoad 3.5 DISPAV-5
High modulus 180 200 210

Fatigue resistant 100 100 100
Subgrade
Figure 2. Pavement structure for the beltway.
2. Design procedures that use the equivalent temperature concept and axle load distribution for
each axle type. These procedures also use the cumulative damage concept to determine the
amount of fracture damage for each pavement structure. The PerRoad program falls within this
category.
3. Design procedures that use the axle load distribution and pavement temperature distributions at
specific depths over some time interval, generally less than a month. These procedures usually
use the incremental damage concept to determine the amount of fracture damage within specific
time and axle load intervals at specific depths within the pavement structure. The Mechanistic-
Empirical Pavement Design Guide (M-E PDG) falls within this category.
The first two methods described above were used for the pavement design: Asphalt Institute
(DAMA) and PerRoad 3.5. A program called DISPAV-5 was also used, which is based on M-E
principles and is commonly used in Mexico for pavement design. The load configuration used in
the design calculations depends on the method. DAMA uses the tire contact pressure, wheel load,
the space between dual wheal as well as the number of repetitions per month to characterize the
load configuration. PerRoad uses the philosophy of characterizing traffic by load spectra; in this
case, traffic is separated by axle type (percent single, percent tandem, percent tridem, and percent
steer); after determining the percentage of each axle from the total traffic, traffic is then subdivided
into weight classes on 2 kip intervals. DISPAV-5 uses the concept of equivalent single axle (ESALs)
of 80 kN load configuration design, 67 106 ESALs were used to perform the bottom-up fatigue
damage analysis. The results of the procedures used are shown in Table 3.
The thicknesses obtained with the three design procedures are very similar, however, the pave-
ment structure selected was the one obtained with DAMA. In Figure 2 it is shown the final structure
for the pavement.
102
Table 4. Coarse aggregate properties.
Test Standard Aggregate Results
Los Angeles ASTM C 131 Gravel #2 29%

Seal 24%
Abrasion in the Micro-Deval Apparatus ASTM D 6928 Gravel #2 8.1%
Seal 11.5%
Fractured Particles ASTM D 5821 Gravel #2 79%
Seal 78%
Flat and Elongated Particles ASTM D4791 Gravel #2 23.4%
Seal 27.4%
Table 5. Fine aggregate properties.
Test Standard Results
Sand equivalent ASTM D 2419 67.1%

Uncompacted void content AASHTO T-304 43.0%
Methylene blue RA-05/2008 3.0%
Bulk specific gravity ASTM C-128 2.6 g/cm3
3 ASPHALT MIX DESIGN
Currently, Mexico has implemented a methodology for the design of asphalt mixtures where dif-
ferent levels are set depending on the characteristics of the project determined by the level of
traffic expected in the design lane. This methodology is called the AMAAC Protocol (Mexican
Association of Asphalt). Within the protocol are included requirements for dynamic modulus tests
and four-point bending fatigue tests. This methodology was used in the design of mixtures of this
study.
To ensure that the asphalt mixes comply with the parameters proposed in the pavement design
by meeting the following specifications in the mix design.
1. For FRL comply with a fatigue life 9.0E+07 at 20 C, 10 Hz, and strain level of 120 and
2. For HM comply with dynamic modulus |E*| 10000 MPa at 20 C, 10 Hz, and strain level of
75 to 125 .
3.1 Aggregates
Aggregates used for the mix designs was 100% crushed limestone (very common aggregate in the
region), made using three fractions consisting of two coarse fractions (gravel # 2, seal ) and one
fine fraction (sand # 4). The properties of the aggregates are summarized in Table 4 and 5 for the
coarse and fine aggregate, respectively.
Once the properties of the aggregates were obtained, different gradations were evaluated
according to recommendations in the AMAAC protocol. The final design gradation is shown
in Figure 3. It was decided to use the same gradation for the design of both mixes (FRL and HM),
changing only the type of binder to achieve the desired properties of each mixture.
3.2 Asphalt binders

A PG 76-22 asphalt modified with polymer SBS with an anti-stripping additive were use to mitigate
fatigue cracking and to improve the adherence with the limestone aggregate. Their characteristics
are shown in Table 6.
103
Figure 3. Aggregate gradation for mixes.
Table 6. Fatigue Resistant binder properties.
Viscosity Rotational @ 135 C AASHTO T 316-10 1423 c.P

Penetration @ 25 C ASTM D5-18572 54 dmm
Softening Point AASHTO T53-06 62.6 C
Complex modulus (G*/sin @ 76 C) AASHTO T315-06 1.25 kPa
Phase angle (, @ 76 C) AASHTO T315-06 73.8
Table 7. High modulus binder properties.
Viscosity Rotational @ 135 C AASHTO T 316-10 2055 c.P

Penetration @ 25 C ASTM D5-18572 18 dmm
Softening Point AASHTO T53-06 66 C
Complex modulus (G*/sin @ 82 C) AASHTO T315-06 1.7 kPa
Phase angle (, @ 82 C) AASHTO T315-06 82.3
To comply with the special characteristics needed in the high modulus layer a special binder
had to be developed by SemMaterials Mexico Liquid Research and Development Laboratory.
The characteristics of the developed binder are shown in Table 7.
3.3 Characteristics of asphalt mixtures

The desired properties in the two mixtures appear to be similar (high stiffness), but while in FRL
is achieved by designing a mixture with low content of air voids (3%), in the HM mixture using a
binder more rigid to a content air voids of 4%. The function each one is, for FRL mitigate fatigue
104
Table 8. Properties of asphalt designed mixtures.
Mixtures Asphalt type Asphalt content (%) Void content (%)
High Modulus PG 82-22 4.0 4.0

Fatigue resistant PG 76-22 4.3 3.0
Figure 4. CS4300 Universal Testing System for Soils and Asphalt.
cracking and the HM has to be rigid enough to distribute properly the efforts in the pavement
structure. Table 8 shows a summary of the properties of asphalt designed mixtures.
4 TESTS RESULTS
The modulus and fatigue tests were carried in the Universal Test System (UTM) developed by
the James Cox and Sons, Inc. The machine is a versatile, fully automated in a single axis, with
a closed-loop hydraulic testing system designed specifically to perform tests on soils and asphalt
concrete mixtures over a wide range of stresses and frequencies.
The equipment is controlled by a ATS software which provides a closed-loop control feedback
in the servo-hydraulic system, in order to control temperature and acquire data. The tests devices
were placed in a temperature chamber that maintained the temperature test by air flowing.
4.1 Dynamic modulus

Dynamic moduli of the asphalt mixtures were evaluated according to AASHTO TP 62-07 at 10
Hz of frequency and 20 C, placing a dummy specimen in the chamber temperature to identify
the time required for the test specimen to reach 20 C (due average annual temperature for the
project location 21 C). A 150 mm diameter and 180 mm of height specimens were fabricated by
the gyratory compactor to be sawn in to a 100 mm diameter and 150 mm height specimens.
Table 9 shows the results obtained in testing dynamic modulus. It can be seen that these values
exceed the design criterion for the HM proposed in the pavement (10,000 MPa).
105
Table 9. Dynamic modulus and phase angle.
|E*| Phase
|E*| average Phase angle Angle
Mixture Strain () (MPa) (MPa) (degrees) average
HM_1 78 13210 13690 9.5 10.3

HM_2 94 14673 11.7
HM_3 81 13188 9.6
FRL_1 125 4320 4450 21.9 22.1
FRL_2 99 4202 22.3
FRL_3 115 4822 22.1
Figure 5. Cox fatigue testing apparatus and load and freedom characteristics.
4.2 Fatigue
The four-point bending test was used to evaluate the fatigue resistance in a controlled strain
mode. The tests were performed according to ASTM D 7460-10 (Determining Fatigue Failure
of Compacted Asphalt Concrete Subjected to Repeated Flexural Bending).
ASTM D 7460-10 sets that the loading device must be capable of (1) providing cyclic haversine
(=SIN2 (degrees/2)) loading for a 5 to 10 Hz frequency range, (2) subjecting the specimen to 4-point
bending with free rotation and horizontal translation at all load and reaction points, and (3) forcing
the specimen back to its original position (that is, zero deflection) at the end of each loading cycle.
The test device used in the study complies with the characteristics specified in ASTM D7460-10;
Figure 5 shows the loading points and free rotation device for the CS7600 manufactured by James
Cox & Sons.
Fatigue tests were conducted using compacted slabs with the linear kneading compactor and
sawn in to the following dimensions: 380 mm (length) 63 mm (width) 50 mm (height). Stiffness
reduction was used to determine the failure of the specimens with failure defined as 50% reduction
of initial stiffness.
The tests were run at a controlled temperature of 20 C and 10 Hz for each mixture, three strain
levels were selected (700 , 500 , 300 ) according to the NCHRP 646 Report (Validation of
Endurance Limit of Fatigue for Hot Mix Asphalt).
106
Table 10. Bending beam fatigue results.
Mixture Va (%) Initial stiffness (MPa) Strain () Cycles to failure
FRL_01B 4.1 4151.6 714 6000

FRL_01A 4.3 4176.8 738 4000
FRL_02A 4.3 5485.7 742 7000
FRL_05A 4.2 5232.2 508 120000
FRL_04B 4.5 6456.4 479 540000
FRL_05B 4.3 7971.5 395 3900000
FRL_03A 3.9 5189.5 296 17894429a
FRL_03B 4.1 5604.0 309 5600000
HM_01A 5.6 14600.6 640 300
HM_02B 5.3 10676.3 666 300
HM_03A 5.3 6315.1 494 965
HM_05B 5.1 4923.4 472 4000
HM_04B 5.1 7710.1 530 2000
HM_05A 4.9 6301.9 299 100000
HM_06A 5.2 12514.6 262 60000
HM_06B 5.1 7004.2 299 120000
a Notes: predicted fatigue life with the Weibull function (ASTM D7460-10).
Figure 6. Fatigue laws obtained in the study.
Table 10 shows the characteristics of the beams tested, including the results of fatigue tests
for the two mixtures designed in the laboratory. Figure 6 shows the fatigue laws obtained for the
mixtures analyzed.
The Fatigue Endurance Limit (FEL) was determined by using the Equation 1 from the procedure
documented in the NCHRP Report 646 (Prowell et al., 2010). This equation is based on a 95
percent lower prediction limit of a linear relationship between the log-log transformation of the
strain levels of the beams tested and cycles to failure. All calculations were conducted using a
spreadsheet shown in NCHRP Report 646.
107
Table 11. Predicted endurance limits.
Va average
Mixer (%) FEL ()
High Modulus 5.2 120

Fatigue Resistant 4.2 269
Figure 7. Plot Ln(Ln(SR)) versus Ln(N).
where: y0 = log of the predicted strain level (microstrain); t = value of t distribution for n 2
degrees of freedom = 1.9432 for n = 8 with = 0.05; s = standard error from the regression
analysis; n = number of samples = 6; Sxx = ni=1 (xi x)2 (Note: log fatigue lives); x0 =
log(50,000,000) = 7.69897; x = log average of the fatigue life results.
Table 11 shows the 95 percent one-sided lower prediction values for the endurance limit for
each of the two mixes tested in this study based on the number of cycles to the failure determined
at the 50% decrease of the initial stiffness.
Finally to verify that there is a fatigue endurance limit for the mixture, a beam was tested to
10,000,000 loading cycles. NCHRP Report 646 recommends that for the strain level calculated in
Table 10 (269 ), however, in this study it was performed at the level of deformation proposed in
the pavement design (120 ). Figure 7 shows the results of the fatigue tests in a double natural
logarithm scale (ln of the (ln stiffness ratio, SR)) versus the number of repetitions of the tensile
strain. Stiffness ratio is defined as the stiffness at a specific number of repetitions divided by the
initial stiffness.
According to Tsai et al. (2006), when plotting a double log of SR versus log of repetitions, in
the beam fatigue test three stages are presented in the fatigue damage curve:
1. Stage I, initial or warm-up stage during where the temperature of the beam increases with energy
dissipation until it reaches a fairly steady temperature;
2. Stage II, crack initiation, where there is a steady rate of stiffness reduction versus repetitions;
and
3. Stage III, crack propagation, where the rate of stiffness reduction versus repetitions is greater
than in Stage II.
Figure 7 shows each of the scenarios described above, Stage I the temperature equilibrium under
the initial repetitions is reached; Stage II, the cracks are initiated and Stage III is designated as
108
the crack propagation stage, although mixes with modified binders often do not exhibit a lot of
crack propagation due to a very slow rate of fatigue damage and in this case for the strain low level
(120 ). As can be seen, the flattening of slope near the calculated endurance limit is indicative
of the endurance limit for the mix.
5 CONCLUSIONS
Mechanistic-Empirical methodologies were used to design a perpetual pavement, based on

specified parameters (asphalt mix stiffness) that had to be met for the mix in the laboratory.
The results of the tests indicate that dynamic modulus of the mixes met and even exceeded the
proposed modulus in the design for the two mixes studied.
The fatigue resistant layer is 100% higher than high modulus layer in the fatigue test for all stain
levels (700 , 500 and 300 ). This was expected based on the type of binder used in each
mixture and the difference of 1% less of air voids in the fatigue resistant mix.
The endurance limit of fatigue was estimated according to the methodology proposed by Report
NCHRP 646, and was found to be greater for the fatigue resistant mix (269 ) than for the high
modulus (120 ).
To verify that the proposed strain level in the pavement design to was met by the endurance
limit of fatigue in the fatigue resistance mix, a beam test was performed to ensure that there was
not fatigue failure at 10,000,000 load cycles. The test results confirm that for the strain level of
deformation used, the mix does not show fatigue failure using the criterion of 50% reduction of
the initial stiffness. The presence of the endurance limit was also seen in the tendency of the plot in
double log of SR versus log of repetitions (Weibull curve) to show a flattening of the slope when
tested at the endurance limit strain, indicating an infinite fatigue life.
REFERENCES
Asphalt Pavement Alliance. 2002. Perpetual Pavements: A Synthesis, APA 101, Asphalt Pavement Alliance.
L. Fontes, G. Trichs, J.C. Pais & P.A.A. Pereira. 2009. Fatigue Laws for Brazilians Asphalt Rubber Mixtures
Obtained in 4 Point Bending Tests. Four Point Bending proceedings of the second workshop. University of
Minho. 2425 September 2009. Portugal.
Newcomb, D.E., M. Buncher, & I.J. Huddleston. 2000. Concepts of Perpetual Pavements. Transportation
Research Circular No. 503. Perpetual Bituminous Pavements. Transportation Research Board. Washington,
DC. pp. 411.
Prowell, B.D., E.R. Brown, R.M. Anderson, J. Sias-Daniel, H. Von Quintus, S. Shen, S.H. Carpenter,
S. Bhattacharjee & S. Maghsoodloo. 2010. Validating the Fatigue Endurance Limit for Hot Mix Asphalt.
NCHRP Report 646, Transportation Research Board, Washington, D.C.
Timm, D. H., M. M. Robbins, J. R. Willis, N. H. Tran & A. J. Taylor. 2011. Evaluation of Mixture Performance
and Structural Capacity of Pavements Utilizing Shell Thiopave Phase II: Construction, Laboratory Eval-
uation and Full-Scale Testing of Thiopave Test Sections One Year Report. Draft Final Report, National
Center for Asphalt Technology, Auburn University, AL.
Von Quintus, H. 2006. Application of the Endurance Limit Premise in Mechanistic-Empirical Based Pavement
Design Procedures. International Conference on Perpetual Pavements. Columbus, Ohio.
Walubita, L. & Scullion, T. 2010. Texas perpetual pavement New Design Guidelines. Transportation
Researcher. Texas Transportation Institute. The Texas A&M University System College Station, Texas.
Tsai, B.W., Jones, D., Harvey, J. & Monismith, C. 2006. Reflective Cracking Study: First level Report on
Laboratory Fatigue Testing. Davis and Berkeley, CA: University of California Pavement Research Center.
(UCPRC-RR-2006-08).
109
Asphaltic materials evaluation
Fatigue assessment of conventional and highly modified asphalt

materials with ASTM and AASHTO standard specifications
G.M. Rowe
Abatech Inc., Blooming Glen, Pennsylvania, US
P. Blankenship
Asphalt Institute, Lexington, Kentucky, US
T. Bennert
Rutgers University, Piscataway, NJ, US
ABSTRACT: During a fatigue test, damage is accumulating and the characteristic properties
for the material are changing. Typically, this is expressed as a change in modulus, and when this
has reached some critical value, the test is terminated. This has resulted in different definitions of
failure for the two types of test, controlled stress and strain, and results in different definitions of
failure and consequently different fatigue relationships. The use of dissipated energy concepts has
been used as a method to bring the results of controlled stress and strain testing closer together.
However, some of the analysis conducted with dissipated energy has not considered the way a
specimen damages during a test. The flexural fatigue life of asphalt mixes is currently assessed by
the AASHTO T321 procedure in the USA. Meanwhile, the ASTM D 7460 test method offers an
alternate analysis procedure that does not rely upon a curve fit to determine the point of failure,
but rather considers a maximum value in the nS curve as the definition of cracking. The nS
curve is based upon an understanding of how energy is dissipated during the test. Recently, some
mixtures with asphalt binders that contain approximately 10 to 15% polymer (defined as heavily
modified in this paper) have been evaluated in these test devices. The data demonstrates that crack
initiation occurs beyond the 50% stiffness reduction, which in AASHTO T321 is used as the value
to terminate the test, with the assumption that all necessary data has been collected to determine
the specimens fatigue life. The origins of both methods are discussed and data is presented that
demonstrates the fatigue process that occurs in different modes of testing. In addition, analysis
is presented of heavily modified mixes and these are contrasted to those made with conventional
binders. Recommendations are made with regard to the interpretation of data from the various two
methods.
1 INTRODUCTION
The use of fatigue tests for analysis of asphalt mixes has been in existence since the late 1950s
(Pell, 1961). Several different loading arrangements exist and those commonly used today include
two, three and four point bending tests and those conducted in direct tension-compression mode.
The current methods for conducting bending beam fatigue tests in the USA were refined and
adopted in specifications in the early 1990s following work conducted as part of the Strategic
Highway Research Program (Tayebali et al., 1994; Deacon et al., 1994). In the work conducted at
this time the use of computers for controlling sophisticated test was relatively new. For example in
work conducted at Nottingham (Rowe, 1996), use was made of Hewlett Packard HP85A Personal
Computer for controlling push-pull testing while bending beam fatigue tests were conducted using
personal computers with Intel 286 chips. Some testing and analysis was conducted using the 386
113
chips at this time. The reliance on these older computers and associated software resulted in a need
for efficient programming and limitation of data file size. With the rapid advance over the past
two decades the limitations that existed have simply vanished. Previously, programmers kept file
sizes limited and notes in the AASHTO 321 (2007) specification indicate that this may be an issue.
The data presented in this specification is collected on a semi-log basis in order to control/limit
the amount of data being collected. When using regression analysis in the determination of test
parameters the amount and frequency of data can affect the analysis parameters and consequently
the results of a test significantly. TheASTM D7460 (2008) version of the fatigue specification uses a
different determination of the failure point which was based upon dissipated energy considerations.
The parameter n.S (number of load cycles multiplied by stiffness) peaks at the location in a test
where the dissipated energy per cycle starts to change significantly. This signifies the point where
damage merges to form micro-cracks which then propagate more rapidly through the beam. This
peak occurs in both controlled stress and strain tests. However, in the controlled strain test the peak
at this point can be followed by a secondary increase in this parameter when the test is continued
to very low stiffness values. Since the ASTM method relies less on curve fitting the variability
that results from inclusion or exclusion of data is lower but the data collected scheme will result in
some deviations. The parameters that can affect the result in this method are also discussed.
2 BENDING BEAM FATIGUE TESTING
In the USA the two standards applied to bending use a beam subjected to bending action with a
continuous 10 Hz haversine or sinusoidal loading, ASTM D7460 and AASHTO T321. Typically
the test is conducted at a temperature between 15 and 25 C. While this paper presents data from
this type of testing it should be noted that the same type of analysis can be applied to the trapezoidal
fatigue test conducted in Europe or other cyclic loading fatigue test configurations.
During the bending beam fatigue test a cyclic load is applied to a specimen and the evolu-
tion of damage is monitored. For the work presented herein analysis had been conducted in IPC
fatigue machines (Figure 1) at either the Asphalt Institute (Lexington, Kentucky) or in the Rutgers
Asphalt Pavement Laboratory (RAPL) located within the Center for Advanced Infrastructure Test-
ing (CAIT) at Rutgers University (New Jersey). The data from these tests is stored in as ASCII
Figure 1. IPC fatigue device.
114
format comma separated variable (CSV) file. This makes subsequent analysis of the data by other
software programs and or spreadsheet routines relatively easy.
In this paper we have presented data from several studies. Conventional asphalt mixes include
those used in New Jersey and Kentucky made in accordance with state specifications. These
include a PG64-22 and PG76-22 binders (the latter is typically a modified binder). We have also
included SMA mixtures made with modified binders and specialized bridge deck wearing course
(BDWC) mixes that are intended for water-proofing applications. We have provided aggregate size
information in the various figures. All the analysis information of fatigue data presented in this
paper has been collected by the two laboratories during the last 4-years.
Fatigue test data is often plotted on log stiffness vs. log loading cycles as reported by many early
researchers in this field (Pell et al., 1961; Pell and Cooper, 1974, Cooper, 1976). However, it should
be noted that the stiffness reduction that occurs can also be plotted on linear graphs. Often linear
plots enable a more clear assessment of the quality of curve fits to the data sets. Rowe and Bouldin
(2000) have previously demonstrated that the change in stiffness during a bending beam fatigue
test is best represented by graphs with linear scales.
During the research conducted during the SHRP program this practice was adopted by Tayebali
et al. (1992, 1994). The analysis of data was then used to estimate the location where the stiffness
(S) had dropped by 50% using an exponential equation fitted to the data as follows:
where:
e = natural logarithm to the base e;
A = constant;
b = constant; and
N = number of loading cycles.
Whereas the wording in the AASHTO T321 standard excludes the data beyond the point where
the estimate of the initial stiffness has dropped below 50%, this data is often included in analysis
by various researchers, extending the analysis. We have referenced the analysis performed strictly
in conformance with the AASHTO method by a (S) whereas the analysis including additional data
is indicated by a (R). An example of a data set collected and analyzed to both schemes is shown
in Figure 2 using both linear and log axis. The regression parameter (R2 ) is high in both versions
of the analysis (0.95 and 0.98). However, this statistic is often misleading and although the high
number is obtained the relationship may be non-descriptive. By simple inspection it can be seen
that the exponential fit appears to fit the data when log scales are used but clearly is non descriptive
when the data is inspected on a linear basis.
The ASTM standard contains definition of failure which is based upon determining the peak
location from a curve of normalized stiffness versus cycles to failure, as shown in Figure 3.
During SHRP data was typically stored for analysis from data in the 4-point bending beam test
using a semi-log basis. This data format was used in the development of the AASHTO T321 standard
and is typical of that presented in various reports and papers published from that research, Tayebali
et al. (1992, 1994). However, one key aspect of any parameter analysis such as determination of the
constants A and b in the AASHTO equation using regression methods is the number and spacing
of data points that are used in the analysis.
In the orginal specification the number of data points was collected on a semi-log basis whereas
in the newer test devices data is collected on an equal spacing within each decade of testing. In order
to assess the difference introduced by the new data collection scheme versus that used in the original
test development, data was analyzed from several test results using a semi-log consideration and
all the collected data. We have referenced the two data collection schemes as IPC (the equipment
used in the testing) and SHRP (to reflect the orginal limitations of data collection that was used to
develop the AASHTO test method) for convenience.
The result of using more data in the fatigue analysis per the current IPC data collection scheme
increases the average fatigue life by approximately 24 to 26% depending upon the amount of data
115
Figure 2. Representations of fatigue curves on linear and log scales with AASHTO.
Figure 3. Stiffness cycles plotted against cycles.
included in the analysis as demonstrated by the data shown in Figure 4. The inclusion of data beyond
the 50% stiffness reduction has the potential to change the data analysis more significantly. In the
data set analyzed below the biggest variation when more data is included in the analysis occurs
with the soft highly modified materials. The analysis of conventional materials using the limited
log based data (SHRP) often produced the same result.
The variation obtained with an AASHTO analysis scheme can be further studied by investigating
the change in life with varying percentages of terminal stiffness included in the analysis. The data
in Figure 5 demonstrates that significant differences can exist with modified materials whereas the
changes that occur with conventional materials are significantly reduced. However, it should be
116
Figure 4. Fatigue life from analysis method AASHTO 321 using different data inclusion/exclusion schemes.
Figure 5. Fatigue life from analysis method AASHTO 321 using different data inclusion/exclusion schemes.
noted that the variations in excess of 40% obtained with conventional mixes is not a desirable basis
to form a specification.
The adoption of functional form must be carefully assessed to determine if it is consistent with
the observed data. Additional analysis of data collected during the SHRP project was presented by
Rowe (1993) and Rowe and Bouldin (2000). This analysis led to the conclusion that the functional
form adopted in the AASHTO T321 test method did not adequately describe the stiffness reduction
that occurs in a fatigue test. Further work conducted by Ghuzlan, and Carpenter (2000) considered
the use of a dissipated energy ratio and its rate of change during a test. Other methods used to
117
Figure 6. Data from fatigue test analyzed using ASTM method with Cubic spline fit through data in region
of test termination.
define failure in fatigue have included the peak in phase lag between the stress and strain response.
However, these latter methods have problems in consistently defining the failure point and currently
have not been adopted in any nationally recognized procedure. It is interesting to note that these
additional procedures all tend to define the same location which occurs as a crack is initiated in
a specimen but the peak in the dissipated energy ratio appears to be best method to define where
micro-cracks have propagated and coalesced to form a macro-cracks (Ajideh et al., 2010). The
principle advantage of the ASTM method which is based upon the peak in the dissipated strain
energy concept is that failure is defined by a simple and reproducible parameter. The definition of
the peak with data collection using the IPC method is very reproducible and historically two main
methods have been used to determine this. Rowe (1993, 1996) used a 6-order polynomial fit to the
data to a location just beyond the peak. The differential of the polynomial curve fit allows for some
smoothing of the data and possibly a better definition of the maximum value. This is considered
as an alternate method of defining the peak. In the ASTM method the peak is simply taken as the
data point which gives the maximum value with no smoothing applied. The normalized stiffness
modulus is effectively the stiffness ratio of the modulus at any cycle to that measured initially and
this ratio could be notated as n.S (S = stiffness modulus ratio). It should be noted that while the
ASTM method uses the normalized stiffness modulus the definition of the peak will be identical
regardless of if the normalized modulus or the actual modulus is used. Consequently, the method
is not sensitive to the definition of the initial modulus. In some of the following graphs the vertical
axis shown as n.S rather than n.S.
In a controlled displacement (strain) fatigue test that a secondary increase in the n.S or n.S
parameter can occur. This data should be removed from the analysis since is represents data collected
while the specimen is in a failed condition. A typical set of data showing this effect is shown in
Figure 6 where the test data continued to 14% of the stiffness obtained at 50-cycles. For this
specimen the peak in n.S or n.S occurred at 49% stiffness reduction. Figure 7 shows the effect of
including or excluding the tail from the polynomial analysis. While this method is not currently
within the ASTM method the fitting of a curve function does help to remove any spikes in the
data analysis and produce more consistent analysis. The differences obtained between different
analysis methods used (spline, polynomial or simple maximum) are very small. In the example
118
Figure 7. Effect of data trimming on 6-order polynomial smoothing to ASTM analysis.
Figure 8. Analysis of fatigue data using the ASTM method compared against method AASHTO (R).
shown in Figure 6 and Figure 7 a difference of 72 cycles exists from the minimum to maximum
value, which represents 2% of the mean value. The variation in this method is further illustrated
in the analysis of results presented earlier in Figure 4. We have shown the polynomial smoothing
method compared to the strict ASTM definition (max) in Figure 8 which compares the results from
the ASTM analysis to that obtained from the AASHTO method (IPC data collection scheme using
extended data sets) which is most commonly implemented in the USA. The data shows reasonable
agreement with the range of deviation of the ASTM analysis being between 0.3 and 4.4%. The
difference between the AASHTO (R) and the ASTM method gives a life difference that is between
17 to +43% of the life as determined by AASHTO (R).
3 DISCUSSION
The results from fatigue tests are used in several contexts in specifications and assessment of
performance. In some cases a specifying agency requires a given level of performance at a specified
strain level. This has recently become a common practice with the use of some highly modified
materials such as BDWCs which have performance requirements in several states (for example
New York, New Jersey, etc.). In addition, with the recent introduction of the Mechanistic-Empirical
Pavement Design Guide (MEPDG) several states have been assessing material performance to
119
produce relationships between strain level and fatigue life. The AASHTO standard does not contain
any information on precision and bias whereas the ASTM quotes the within-lab repeatability on
two specimens as two results should not differ by 0.787 when the logs of the number of cycles to
failure is considered. The use of log averaging is important for the expressing fatigue data since
the results are typically considered to be log normally distributed. When power law relationships
are established for use in the MEPDG the results are effectively treated in the same manner.
The analysis of the data did not demonstrate that any significant difference exists regarding the
within-lab repeatability on two specimens as this was further tested by evaluating an additional
ten mixes. The principle improvement by using the peak in n.S or n.S is that the analysis is not
sensitive to the amount of data included in the regression analysis including; 1) the initial value of
stiffness or cycles, 2) the definition of an arbitrary failure stiffness, and 3) the type of smoothing
used to define the peak that occurs. These aspects should greatly improve the repeatability between
laboratories and remove the practice of customized analysis by laboratories.
The choice of the n.S or n.S is a simple parameter that is consistent with other damage/failure
definitions. The variation in stiffness, phase angle, dissipated energy per cycle and n.S is shown in
Figure 9 for the same result as shown in Figures 6 and 7. It can be clearly observed that the peak in
the ratio of n.S (or n.S) is consistent with the deviation of stiffness reduction and energy dissipated
per cycle deviate from a linear fit (as shown by the red dashed lines in the figure). It is interesting to
note that the deviation from this line is the also the location identified by the change in dissipated
energy ratio (DER) as proposed by Ghuzlan, and Carpenter (2000). However, the definition of
Figure 9. Variation in stiffness modulus, phase angle, dissipated energy per cycle and n*.
120
this point from the method proposed by them is difficult due to inherent scatter/noise in the data.
The change in the phase angle has a minor peak before this occurs but the most significant change
in phase angle occurs after the specimen has significant cracking. In the later stages of the test
when the specimen has a significant fatigue crack the shape of the curve from the load cell and
displacement transducer becomes very irregular and can no longer be considered sinusoidal (Rowe,
1996). In this data set that location occurs at approximately 4,700 cycles. If the peak in phase angle
is considered before this location the result obtained would be approximately 2,600 cycles. The
most significant problem in using the phase angle peak or shape to define the failure is the clear
definition of failure. The peak is not always defined and if curve fitting to sine curve is used then
some criterion is needed to define the failure point. Conceptually, the error calculation procedures
as used for dynamic modulus testing as specified in AASHTO T62 could provide a basis for this
type of analysis. However, this type of method is really unnecessary since the calculation of n.S or
n.S is very easy and different procedures for data smoothing result in very little error in the result
calculated.
The collection of fatigue test data during the test can be curtailed by tracking the n.S or n.S
parameter and terminating the test at the maximum value if this parameter has decreased by 10%.
This is an easy parameter to track in data acquisition and would ensure that a crack exists in the
specimen at point of termination. The results that have been analyzed in this paper suggest that the
stiffness at which the test would be terminated for a fatigue test would be lower for a less stiff, highly
modified material. In addition, since the fatigue damage process is linear within a test, the data
collection could easily be implemented on a stiffness reduction basis rather than as a function of the
cycles number. This would provide more equally spaced data throughout the test. Previously, data
has been collected on a 5% stiffness reduction basis (Rowe, 1996) and this has proved adequate for
defining the peak of the n.S or n.S parameter. This data could be collected on any arbitrary stiffness
reduction, for example a 1% stiffness trigger to data collection would provide approximately 50 to
80 data points for a specimen.
4 SUMMARY
This paper presents an analysis of fatigue data using two methods originally developed as part of the
SHRP work. One of these methods forms the analysis procedure contained within the AASHTO
method while the other is used in the ASTM method. The AASHTO method currently requires
a regression analysis to a functional form that is non-descriptive of the data within the method.
Different research laboratories and different analysis methods can obtain significantly different
results depending upon the data collection scheme used for the testing and the subsequent data
selection for analysis. In order to avoid these ambiguities it is recommended that the AASHTO
method is changed to define failure as the peak in n.S or n.S. Additional concluding comments can
be made as follows:
The newer data collection schemes increase the calculated fatigue life by approximately 25%
over those used to originally develop the specification.
Differences in fatigue life calculated by the AASHTO method are more sensitive to the data
included for highly modified products.
Different methods of smoothing the n.S or n.S data in the ASTM format have minimal effect on
the calculated fatigue life.
Trimming of data which is representative of cracked beams is recommended to improve analysis.
Consideration of the n.S or n.S parameter could be used for the test termination.
The data collection scheme could be improved by storing data on a stiffness reduction basis.
Some additional work could be conducted to look at the errors produced in fitting sinusoidal
curve to the data sets. This could provide some useful guidance on data trimming and/or test
termination. Analysis of sine curves could follow the same scheme as currently being used for
dynamic modulus testing in AASHTO TP62.
121
REFERENCES
AASHTO, Standard Method of Test for Determining the Fatigue Life of Compacted Hot-Mix Asphalt (HMA)
Subjected to Repeated Flexural Bending, AASHTO Designation: T 321-07, American Association of State
Highway and Transportation Officials, 2007.
AASHTO, Determining Dynamic Modulus of Hot Mix Asphalt (HMA), AASHTO Designation: TP 62,
American Association of State Highway and Transportation Officials.
ASTM, Standard Test Method for Determining Fatigue Failure of Compacted Asphalt Concrete Subjected to
Repeated Flexural Bending, ASTM Designation: D 7460 08, ASTM International, 100 Barr Harbor Drive,
PO Box C700, West Conshohocken, PA, 2008.
Ajideh, H., Bahia, H. and Earthman, J., Damage Resistance and Performance Evaluations of Engineered
Asphalt Materials and Asphalt-Aggregate Mixtures with High Percentage of RAP, Research Report, City of
Los Angels, the University of California, Irvine and the University of Wisconsin, Madison, 2010.
Cooper, K.E., The Effect of Mix and Testing Variables on the Fatigue Strength of Bituminous Mixes, M.Phill
Thesis, University of Nottingham, 1976.
Deacon. J.A., Tayebali. A.A., Coplantz. J.S., Finn F.N. and Monismith C.L., Fatigue Response of Asphalt
Aggregate Mixes: Part III Mix Design and Analysis, Report SHRP-A-404, Asphalt Research Program,
Strategic Highway Research Program, National Research Council, Washington, DC 1994.
Pell, P.S. and Cooper, K.E., The Effect of Testing and Mix Variables on the Fatigue Performance of Bituminous
Materials, Proceedings of the Association of Asphalt Paving Technologists, 1975, Vol. 44, pp. 137.
Pell, P.S., McCarthy, P.F. and Gardner, R.R., Fatigue of Bitumen and Bituminous Mixes, International Journal
of Mechanical Science, Pergamon Press Ltd, 1961, pp. 247267.
Rowe, G.M., Performance of Asphalt Mixtures in the Trapezoidal Fatigue Test, Journal, Association of Asphalt
Paving Technologists, Volume 62, 1993. pp. 344384.
Rowe, G.M., Application of the Dissipated Energy Concept to Fatigue Cracking in Asphalt Pavements, PhD
Thesis, University of Nottingham, January 1996.
Rowe, G.M. and Bouldin, M.G., Improved Techniques to Evaluate the Fatigue Performance of Asphaltic
Mixtures, Eurobitume Book 1, September 2000, pp. 754763
Ghuzlan, K.A. and Carpenter, S.H., Energy-derived, Damage-based Failure Criterion for Fatigue Testing,
Transportation Research Record: Journal of the Transportation Research Board, No. 1723, pp. 141149,
2000.
Tayebali, A.A., Deacon, J.A., Coplantz, J.S., Harvey, J.T. and Monismith, C.L., Fatigue Response of Asphalt-
Aggregate Mixes Part I Test Method Selection, Report SHRP-A-404, Asphalt Research Program, Strategic
Highway Research Program, National Research Council, Washington, DC 1994.
Tayebali, A.A., Rowe, G.M. and Sousa, J.B., Fatigue Response of Asphalt-Aggregate Mixtures, Paper Presented
at theAnnual Meeting of theAssociation ofAsphalt Paving Technologists, Charleston, 2426 February 1992,
pp. 333360.
122
Mechanical characterization of dry asphalt rubber concrete for base

layers by means of the four bending points tests
C. Maggiore & G. Airey

University of Nottingham, Nottingham, UK
G. Di Mino & C.M. Di Liberto

University of Palermo, Palermo, Italy
A. Collop
De Montfort University, Leicester, UK
ABSTRACT: The mechanical characterisation of the asphalt concrete in terms of both the fatigue
resistance and the stiffness modulus is necessary to use any design method of the flexible road
pavements. Even more the determination of these mechanical properties is necessary if the asphalt
concrete investigated is an innovative material as a Dry Asphalt Rubber Concrete DARC (i.e. a
bituminous mixture with rubber via dry process). Such material is less known and investigated than
the asphalt rubber concrete via wet process, even if its application implies peculiar economical and
environmental advantages such as no specialized equipment or significant plant modifications and
large quantity of recycled waste tires compared to the wet process. This paper focused on the
mechanical behaviour of some DARCs with different rubber content that varies within 0, 3.0% by
weight of the aggregates; these materials are designed to the base layer of the flexible pavements. In
particular an experimental survey was carried out in order to determine the stiffness modulus of the
DARCs by means of the four point bending test on prismatic specimens (UNI EN 12697-26:2004).
Keywords: fatigue, stiffness modulus, dry asphalt rubber concrete, 4 Point Bending test.
1 INTRODUCTION
In Italy, in light of the environmental issues, a law D.M. 203/2003 established that the use of the
recycled rubber from the waste tires has to be increased both in the road infrastructures and in the
railways.
In the field of road pavement nevertheless the applications of the rubber asphalt via dry process
are very few mostly because both its complex stiffness modulus and fatigue behaviour are not
investigated enough.
In this paper, the authors have focused on the mechanical properties of the Dry Asphalt Rubber
Concrete (DARC) as material for base layer within the road flexible pavement, by means of an
experimental survey carried out in the Laboratory of the Road Materials of Department of Civil,
Environmental and Aerospace Engineering at the University of Palermo.
Four materials were analyzed and compared: the first was a bituminous mixture without rubber;
the others were: Dry Asphalt Rubber Concrete (DARC) with respectively a rubber content equal
to 1.0, 2.0 and 3.0 by weight of the aggregates.
The test procedures were conducted by means of the four bending points (4PB-PR) apparatus
according to the standard UNI EN 1269726: 2004.
123
2 BACKGROUND
The complex stiffness modulus is a complex number that relates stress to strain for linear viscoelastic
material subjected to continuously sinusoidal loading in the frequency domain.
HMA mixture can be considered as linear viscoelastic material under small strain levels
(100 ); thus the HMA stress-strain relationship under continuous sinusoidal loading in the
linear viscoelastic region can be defined by the complex dynamic modulus.
The complex modulus E (at any given temperature) is defined as the ratio of the complex ampli-
tude of the sinusoidal stress (t) (at any given time and load frequency) and complex amplitude
of the sinusoidal strain (t) (at the same time and frequency) that results in a steady state response
(Di Benedetto et al., 2001):
where:
(t) = 0 eit , with (t) = Im[ (t)]
(t) = 0 ei(t) , with (t) = Im[ (t)]
0 = peak (maximum) stress;
0 = peak (maximum) strain;
= 2f = pulsation;
f = load frequency;
= phase angle that represents the lag time between stress and strain;
t = time;
i = imaginary unit.
The complex modulus can also be defined as (UNI EN 12697-26:2004):
The real component E1 represents the elastic property of the material while the imaginary
component E2 describes the viscous behaviour. They are defined as follows:
The absolute value of the complex modulus is called stiffness modulus and it can be calculated as:
The stiffness modulus is used as overall measure of the asphalt stiffness for quasi-static flexible
pavement analyses. This means analyses where time is not considered explicitly, but as nominal
loading rates to determine the appropriate asphalt stiffness.
The phase angle can be defined as:
The Annex B of 12697-26 CEN European Standard Determination of the stiffness of bituminous
mixture describes the procedure for evaluating the real component E1 and the imaginary component
E2 of the complex modulus, the stiffness modulus and the phase angle by means of four point
bending test on prismatic specimens.
124
Figure 1. Boundary conditions, loading scheme and cross section.
Evaluation of the phase angle is the key point in this work because such parameter plays a
decisive role for the attenuation of the borne vibrations due to the traffic.
Generally the interpretation of a bending test is based on the solution of the motion equation of
an elastic beam subjected to a force F, in which the Youngs modulus is replaced by the stiffness
modulus. Boundary conditions and loading scheme is shown in Figure 1.
The solution, in terms of vertical displacement v of a point at x position and any given time t, is
(Giuliana and Nicolosi, 2003):
where:

n = n = n L2 EmI is the frequency of the n-harmonic wave;
2 2
I is the moment of inertia of the beams cross section (see Figure 1) ;

qn = L2 P sin n
3
+ sin 2n
3
;
P is the applied load equal to an half force (see Figure 1);
is the angular velocity for unit of length;
m is the mass for unit of length;
L is the length of the beam.
The complex mechanical impedance is defined as:
where max is the maximum vertical displacement that occurs for x = L/2
125
The expression of Z extended only to the first term of the series (n = 1) and the complex modulus
can be written as:
where = tan is the damping factor.
It is possible get E1 imposing the equality of the real parts,
where M = Lm.
Knowing E1 , E2 is determined through the following equation:
According to UNI EN 12697-26, 2004, the two components of the complex modulus are:
where
F = 2P is the load amplitude (peak-peak);
D is the vertical displacement measured;
is the phase;
is the angular frequency;
is a form factor that depends by the type geometry of specimen.
For a prismatic specimen, is given by:
It is easy to verify that the two equations (12) and (14) related to the real component E1 are
identical but the equation (13) and (15) related to the imaginary component E2 are different; this
because the UNI EN 12697-26, 2004 does not take in account the following inertial term.
126
The inertial term is missing; this could influence the value of the phase angle (due to the
presence of the tangent function) and its value might be characterized by a high experimental
dispersion due to directly measure of force and displacement. In this case it would be better to
carry out a pre-processing data.
The AASHTO T321-03 standard does not consider the weight of specimen and the inertial
contributions of the moving mass (clamps and sensors) to determine the bending stiffness.
The bending stiffness, at n cycle, is defined as the ratio between the maximum tensile stress and
the maximum tensile strain. According to the standard AASHTO T321-03, the maximum tensile
stress (Pa) and the maximum tensile strain (m/m), are given by the relations:
where:
F is the load applied by the actuator (N);
b e h are, respectively, the average width and the average height of the prismatic specimen (m);
is the center-beam deflection (m);
a is the space between inner clamps (0.119 m);
L is the space between outer clamps (m).
Therefore the bending stiffness S (Pa) is given by the following equation:
Among the outputs there is not the maximum deflection at the center of the specimen. The
registered value is the Beam Fatigue Displacement (BFD), recorded by the transducer at the bottom
of the beam between the supports placed below at a distance from outer clamps equal to L/6.
The Beam Fatigue Displacement is not the total displacement of the beam because the deflection
that occurs in the outer parts (between the external clamps and support devices) is not computed.
The displacement was calculated and added to the BFD; thus the total deflection of the beam is
obtained using the deflection equation.
Finally, the displacement at L/6 is exactly half BFD; so the total displacement is twice value of
the BFD. (see Figure 2).
Figure 2. Scheme of the beam with force and constraints.
127
Table 1. Physical properties of the aggregates.
Retained at UNI n 5 Passing at UNI n 5
Test Value Standard Test Value Standard
Shape coefficient Cf 1.25 EN 933-3:1997 Sand Equivalent 89% EN 933- 8

Flakiness index Ca 1.54 EN 933-4:1997 Plasticity index NP C.N.R.
UNI 0014
Shape index (%) 1.2 Rigden air voids (filler) 31.30% EN 1097-4
L.A. Abrasion Loss (%) 20.7 EN 1097-2:2003
Micro-Deval Wet (%) 6.9 EN 1097-1:2003
Abrasion Loss
Coefficient of 113 C.N.R. n 4/35
fragmentation
Polished Stone Value 0.39 EN 1097-8 Delta ring and ball with 10 C PA EN 13179
Resistance to Freezing (%) 1.30 EN 1367-6 filler/bitumen = 1.5
Stripping (%) 0 C.N.R. B. U.
n 138/92
Voids (%) 0.72 C.N.R. B. U.
n 65/78
Bulk density (g/cm3 ) 2.86
The values of the stiffness modulus determined by the AASHTO are slightly lower than those
calculated using 12697-26 CEN European Standard and this difference is more evident at higher
frequencies. This depends on the lack of the inertial and viscous parameters in the formulas (21)
and (22).
3 EXPERIMENTAL SURVEY
The main aim of the experimental work was to determine the influence of the rubber content,
of the load frequency and of the temperature on the mechanical and dissipative properties of the
materials for the base layer; in particular the asphalt mixtures were HMA_RFI, DARC1, DARC2
and DARC3, respectively with 0, 1%, 2% and 3% of rubber content.
The recycled crumb rubber comes from the waste tires of truck and so it can be considered
as natural rubber which is more suitable to the mixing with the asphalt concrete (Scofield, 1989;
DAndrea & Mariotti, 2001; Zeng et al., 2001; Zhong et al. 2002; DAndrea et al., 2004).
The stiffness and the damping ratio of each material were determined using the four bending
points test on prismatic specimens (4BP-PR) according to the UNI EN 12697-26 2004, at the
temperatures of 10 C, 17 C and 25 C.
The 4 Point Bending tests were performed for the experimental work with a servo-hydraulic
testing system at University of Palermo. Prismatic beams are manufactured with dimensions of
380 mm in length, 50 mm in height and 63 mm in width.
In this study, the determination of the stiffness modulus tests were carried out applying a contin-
uous sinusoidal cyclic waveform, in controlled displacement (strain) mode at constant temperature.
Strain level equal to 30 was chosen for the laboratory activity.
3.1 Materials
The limestone aggregates here used for the mixtures are commonly adopted in Western Sicily.
Table 1 reports the properties of the aggregates. The properties of the binder are reported in table 2.
128
Table 2. Physical and rheological properties of the binder.
Binder Brookfield viscometer
Required Observed
Characteristics Standards values value
Pen @ 25 C (mm) EN 1426:2007 5070 56
Softening point ( C) EN 1426:2007 70 49
Equi-viscosity 0.28 Pa s AASHTO 145
temperatures T316-04
(compaction and 0.17 Pa s 150
mixing) by
Brookfield ( C)
Bulk gravity 1.03
(g/cm3 )
Figure 3. RFI master band gradation.
It was considered the gradation master band of the RFI standards and the average gradation was
selected as showed in figure3.
The ambient crumb rubbers which dimensions vary within [0.425 4.76] mm were used.
Figure 4 shows the gradation of the rubber which has a bulk gravity value equal to 1.15 gr/cm3 .
The Dry Asphalt Rubber Concrete was obtained by means of the substitution of the limestone
aggregates with the crumb rubber in volumetric proportion (DAndrea & Mariotti, 2001).
3.2 Method
According to the UNI EN 12697-26 2004, stiffness is calculated at the 50th cycle, but during the
earliest tests a clear dispersion of data occurred within this range.
After analsing the experimental data, stiffness was calculated considering a different range of
initial cycles depending on the frequency. (see Table 3).
129
Figure 4. Rubber gradation.
Table 3. Load cycles number for each frequency.
Frequency Cycle
0.1 200
0.5 200
1 300
5 300
10 300
20 400
30 400
Figure 5. Isothermal curves of stiffness modulus at T = 10 C e 30 .
3.3 Results
Figure 5, 6 and 7 shows the isothermal curves, plotting stiffness modulus versus frequency at
different temperatures.
The complex modulus results of experimental work can be summarized in master curve which
is the representation of the variation of |E*| as a function of the reduced frequency, obtained as the
product of the frequency of the test and the shift factor (T ). The shift factor is obtained by means
130
of the following:
where:
DH/R is the relationship between the universal gas constant and the enthalpy, or activation energy
of the system (DH = 265 kJ/mole; R = 0.008314 kJ/ Kmole);
T is test temperature (K);
TS is reference temperature (17 C = 290.15 K).
The reference temperature shows the variation of modulus as a function of reduced frequency.
This transformation is legitimate because the principle of equivalence (or time-temperature super-
position) is valid for asphalt material. This principle essentially describes the mechanical behavior
131
Figure 8. Master curves of HMA_RFI.
Figure 9. Master curves of DARC1.
of the material: the same effect due to an increase/decrease of the frequency can be generated by the
equivalent decrease/increase of the temperature. In this way, it is possible to estimate the complex
modulus for the frequency values that fall outside the domain of the experimental frequencies.
Master curves of each material at the reference temperature 15 C (290.15 K) are shown in figure
8, 9, 10 and 11.
The analysis of the results allows a more immediate understanding of the relationship between the
complex modulus and frequency and temperature, at which the material is subjected. Comparing the
values at the extreme test conditions, the stiffness modulus shows values ranging from 21600 MPa
for HMA_RFI (f = 20 Hz, T = 10 C) and 1600 MPa for the DAR3 (f = 0.1 Hz, T = 25 C).
The experimental results show that the same material (same gradation curve) can have very
different mechanical response varying content of rubber, binder and loading conditions. Thus, a
more detailed analysis is needed.
132
Another interesting aspect is that complex modulus values at higher frequency (30 Hz) are lower
compared to the values at the previous lower frequency (20 Hz). A reasonable explanation to this
phenomenon is the achievement of the resonance frequency of some elements compared to the
higher stiffness of the sample; in fact, this phenomenon is registered at low temperatures and
higher test frequency. This behaviour is slight at lower stiffness (T = 25 C).
Regarding the phase angle, experimental results show that phase angle decreases when the
frequency increases; it increases when the temperature increases. This is due to the viscoelas-
tic behaviour of the mixtures at high temperatures, that means an increasing of the imaginary
component of the complex modulus, E2 . (see Figure 12, 13, 14 and 15).
As seen in Figure 12 and 13, the phases of DARC1 are characterised by greater values than del
reference mixture (HMA_RFI); an opposite trend is shown in the case of the DARC2 and DARC3
in which the phase angle values are smaller compared to the HMA_RFI. (see Figure 16 and17)
133
Figure 12. Phase angles versus frequency for HMA_RFI.
Figure 13. Phase angles versus frequency for DARC1.
134
Figure 16. Comparison between the phase angles of the four mixtures at 17 C.
To better clarify this phenomenon, was plotted against the percentage of rubber in the mixture,
according to the isochrones at 0.5, 1, 5, 10 Hz. It was found that, for all temperatures and frequencies
tested, a peak phase angle value was recorded at 1% of rubber content; after that the phase angle
decreases when the rubber content increases or decreases. (see Figure 18, 19 and 20)
In black curves, phase angle is plotted against stiffness modulus at the given temperature and
frequency. Also this diagram describes the phenomenon mentioned before.
The red curve represents the mixture DARC1; it can be seen that the phase angle values are
bigger compared to the other mixtures for the same modulus values. This increasing is underlined
at higher temperatures and lower frequency (when the stiffness modulus is lower). Thus, the mixture
DARC1 is characterised by a higher damping ratio; thus an increasing in the viscoelastic component
is recorded resulting in an increasing in the dissipative properties of the material.
135
Figure 17. Comparison between the phase angles of the four mixtures at 25 C.
Figure 18. Comparison between the phase angles of the four mixtures for increasing frequency at 10 C.
136
Figure 21. Black curves for the different mixtures studied at frequencies of 0.5, 1, 5, 10 Hz.
4 CONCLUSIONS
The mechanical characterisation of the asphalt concrete in terms of both the fatigue resistance
and the stiffness modulus is necessary to use any design method of the flexible road pavements,
especially for innovative material such as dry asphalt rubber concrete (DARC).
This paper analysed the mechanical behaviour of some DARCs with different rubber content
that varies within 0 and 3.0% by weight of the aggregates.
The following are some of the conclusions that can be made based on the experimental work
carried out by means of the four point bending test on prismatic specimen (UNI EN 12697-26,
2004):
After testing all the mixtures at different loading conditions (different frequency and tempera-
ture), it was found that the values of the stiffness modulus determined by means of the American
Standards (AASHTO T321-03) are slightly lower than those calculated using the European stan-
dards (UNI EN 12697-26). This difference is more evident at higher frequencies and depends
on the lack of the inertial and viscous parameters in the calculation of maximum tensile stress
and strain.
137
Four different mixture were been analysed: a traditional one and other three charcterised by
different rubber content. Master curves show that the same material (same grading curve) has
different mechanical results depending on rubber content.
Complex modulus decreases when rubber content increases. Also complex modulus is bigger at
higher frequency.
Phase angle decreases when frequency increases, it increases when temperature increases; this
is due to viscoelastic behavior of the mixtures.
DARC1 (1% rubber content) is characterised by higher values of phase angle compared to the
reference mixture (0% rubber content), DARC2 (2% rubber content) and DARC3 (3% rubber
content). It has a higher dumping ratio, thus has a major dissipative properties of the material.
In conclusion, the four point bending test was suitable to characterize the mechanical behavior
of several innovative materials. Also it was able to point out small differences in the valuation of
stiffness but mostly of phase angle.
REFERENCES
AASHTO T321-03. Standard Method of Test for Determining the Fatigue Life of Compacted Hot-Mix Asphalt
(HMA) Subjected to Repeated Flexural Bending. AmericanAssociation of State and Highway Transportation
Officials
DAndrea, A. & Mariotti, T. (2001). Asphalt mixes containing crumb rubber (1st Part). Rassegna del bitume,
39/01, pp. 2133.
DAndrea A., Urbani L., Bonin G. (2004). Traffic Vibration Camping Whit Innovative Materials: Development
And Calibration Of a Simulation Model. 2th International S.I.I.V. Congress. 2729 October. Florence, Italy.
Di Benedetto H., Partl, M.N., Francken, L. and De La Roche C. (2001). Stiffness testing for bituminous mixture.
Materials and structures/Materiaux et Constructions, Vol. 34, March 2001, pp. 6670.
EN-12697-24-2007. Bituminous mixtures Test methods for hot mix asphalt Part 24: Resistance to fatigue
EN-12697-26-2004. Bituminous mixtures Test methods for hot mix asphalt Part 26: Stiffness
Giuliana, G. and Nicolosi, V. (2003). Caratterizzazione del comportamento viscoso reversibile dei conglomerati
bituminosi attraverso le prove di flessione. XIII Convegno Nazionale S.I.I.V. Padova. 3031 October 2003.
Scofield L.A. (1989). The History, Development and Performance of Asphalt Rubber at ADOT. Report AZ-
SP-8902. Arizona Transportation Research Center, Arizona Department of Transportation, 1989.
Zhong X. G., Zeng X. and Rose J. G. (2002). Shear modulus and damping ratio of rubber-modified asphalt
mixes and unsaturated subgrade soils. Journal of Materials in Civil Engineering 14(6): 496502.
Zeng X., Rose J. G., Rice J. S. (2001). Stiffness and Damping Ratio of Rubber-Modified Asphalt Mixes:
Potential Vibration Attenuation for High-Speed Railway Trackbeds. Journal of Vibration and Control, 7:
527538.
138
Impact of air void content on the viscoelastic behavior of

hot mix asphalt
B. Hofko, R. Blab & M. Mader

Vienna University of Technology, Institute of Transportation,
Research Center of Road Engineering, Vienna, Austria
ABSTRACT: Volumetric mix design properties of Hot Mix Asphalt (HMA) have a significant
impact on the viscoelastic behavior of the material. This study presents results from an extensive
lab testing program on two different HMAs: an Asphalt Concrete (AC) with a maximum aggregate
size of 22 mm and an unmodified bitumen 50/70, as well as an AC 22 with an SBS-modified
binder PmB 45/80-65. For both materials the dynamic modulus |E*| and the phase lag were
obtained by testing specimens in the 4-Point-Bending Beam test (4PBB) and the Direct Tension and
Compression test (DTC). The stiffness tests were run with a temperature and frequency sweep. Since
the air void content of the specimens was varied in a wide range, the impact of this volumetric mix
design property on the viscoelastic behavior can be analyzed in detail. It was found that the Time-
Temperature Superposition Principle (TTSP) can be extended to an Air Void-Time-Temperature
Superposition (AVTTSP) where the viscoelastic behavior of HMA can be derived for an arbitrary
air void content at a chosen reference temperature and reference frequency. In addition results from
the two test setups (4PBB and DTC) used within the study are compared and discussed.
1 INTRODUCTION
Air voids have a significant influence on the performance of HMA. Several research projects in
the past studied the effect of air voids on different performance parameters (Bell et al., 1984;
Linden et al., 1988; Harvey et al., 1994 and Harrigan 2002). More recently (Youngguk et al., 2007)
presented an air void model for performance prediction of HMA, and (You et al., 2010) studied the
effects of air voids on 2D and 3D discrete element models. (Kassem et al., 2011) worked on the
influence of air voids on the resistance to fatigue cracking, permanent deformation and moisture
damage. From the results presented in these studies, it can be concluded that the dynamic modulus
of HMA tends to decrease with increasing void content and the phase lag characterizing the
viscosity of the material tends to increase with increasing void content. Thus, HMAs react less stiff
and more viscous to loading when the content of air voids increases.
Since the behavior of viscoelastic materials is influenced by temperature as well as time (i.e.
frequency of loading), its parameters like the dynamic modulus and phase lag are a function of time
and temperature, e.g. |E*| = |E*|(t,T). One way to describe this function is to carry out a number of
dynamic tests at large number of temperatures and frequencies. For so-called thermo-rheological
simple materials like HMA, there is a more efficient way. For these materials, time and temperature
can be combined to one parameter. When dynamic tests are carried out on a thermo-rheological
simple material at different temperatures, the results show the same function in terms of time if the
time domain is scaled. A lower temperature leads to a shift of material parameters to shorter times
or higher frequencies. A master curve can be derived from a limited number of tests at different
temperatures and frequencies by taking advantage of the time-temperature superposition principle
(TTSP). This means that the viscoelastic parameters can be obtained for an arbitrary temperature
and a wide range of frequencies, without having to test the material at this specific temperature
139
or frequency (Findley et al. 1989). (Rowe et al. 2009) showed that the shape of master curves is
significantly affected by the air void content of an HMA.
2 OBJECTIVES AND TASKS
The main objective of the presented study was to combine the effect of time, temperature and air
void content on the viscoelastic behavior of HMA under dynamic loading. The TTSP includes
the effect of time and temperature. From previous studies mentioned above it seemed that there
should be a further link to also include the air void content into the TTSP and extend it to an air
void-time-temperature superposition principle (AVTTSP). To achieve this objective, the following
tasks were carried out:
Produce HMA-slabs by roller compaction in the lab with different target contents of air voids
and obtain specimens from the slabs by coring and cutting for stiffness tests.
Conduct stiffness tests on the specimens with different contents of air voids with temperature
and frequency sweep.
Analyze results in terms of dynamic modulus |E*| and phase lag at different temperatures,
frequencies and air void contents.
Develop master curves combining the effect of time, temperature and air void content by
introducing the AVTTSP.
Discuss results gathered from different HMAs and different test setups.
3 MATERIALS AND SPECIMEN PREPARATION
Two different HMAs were tested in the project: An asphalt concrete (AC) with a maximum aggregate
size of 22 mm with unmodified bitumen 50/70 (penetration between 501 /10 mm and 701 /10 mm),
referred to as AC 22 50/70 and an AC 22 with an SBS-modified bitumen PmB 45/80-65 (penetration
between 451 /10 mm and 801 /10 mm, softening point >65 C). The mineral type used for both mixes
is crushed limestone.
For the test program a total of 69 slabs were compacted from the two mixes. The mean value
(MV) of the grading curves of all 69 slabs is given in the table in Figure 1. The table contains the
minimum and maximum grading curve as well. The right diagram in Figure 1 shows the maximum
and minimum grading curve. The given data show that the difference in gradation between the slabs
Figure 1. Mean value, minimum and maximum of the gradation (left table) and graphical representation of
minimum and maximum grading curve (right figure).
140
is small with a maximum difference between minimal and maximal grading curve of 3.5% (m/m).
Thus, it can be assumed that the grading curve does not influence the results presented in the further
course.
The binder content for both mixes was set to 4.5% (m/m), which is the optimal binder content
according to Marshall.
HMA slabs were produced in the lab by roller compaction in a path-controlled way to a specified
target density. The dimensions of the slabs are 500 260 mm with a variable height. For specimens
tested in both setups (4PBB and DTC) the height of the slabs was set to 80 mm. For 4PBB, 3
prismatic specimens (500 60 60 mm) were cut from each slab according to the pattern shown
in the left picture in Figure 2. The x-direction of the co-ordinate system shows the path of the
roller compactor, the z-axis marks the direction of the compaction force. For DTC, 6 cylindrical
specimens ( = 63.5 mm, H = 200 mm) were cored and cut from each slab. The pattern is depicted
in the right picture in Figure 2.
Table 1 shows the number of slabs produced for each HMA and each test setup. It also contains
information on how many specimens were tested.
The bulk density and the volume of air voids were determined for each specimen. The left
diagram in Figure 3 shows the distribution of the air void content for specimens tested in the 4PBB
for both mixes. Each bar represents a percentage of specimens which lies within a range of air voids.
Obviously, specimens from the AC 22 50/70 exhibit a wide range in terms of air void content ranging
from 1.0% (v/v) up to 7.0% (v/v). The distribution for specimens from AC 22 PmB 45/80-65 shows
less scattering. Still, specimens with an air void content ranging from 2.0% (v/v) to 5.0% (v/v)
can be found. The right diagram in Figure 3 presents analogous data for specimens tested in the
DTC. The air void distribution is narrower than for the specimen tested in 4PBB. Specimens from
Figure 2. 4PBB specimens cut from HMA-slabs (left) and DTC-specimens cored and cut from HMA-slabs
(right).
Table 1. Number of slabs produced for testing and number of

tested specimens.
Number of slabs produced

(number of tested specimens)
4PBB DTC
AC 22 50/70 34 (88) 10 (52)

AC 22 PmB 45/80-65 23 (67) 9 (51)
141
Figure 3. Distribution of air void content for specimens tested in 4PBB (left) and DTC (right).
Figure 4. Principle of the 4PBB (left) and of the DTC (right).
AC 22 50/70 range from 2.0% (v/v) to 5.0% (v/v) and specimens from AC 22 PmB 45/80-65 range
from 1.5% (v/v) to 4.0% (v/v).
4 TEST METHODS AND TEST CONDITIONS
To investigate the viscoelastic behavior of HMA two different test setups were employed. In the
4PBB, the specimen is subjected to dynamic (sinusoidal) flexural loading perpendicular to its
longitudinal axis. Figure 4 shows the principle of the setup. While the two outer supports of
the specimen are fixed and allow rotation and translation in horizontal direction, the bending is
realized by the two inner supports in vertical direction perpendicular to the longitudinal axis of the
specimen. The test device is located within a temperature chamber to control the test temperature
throughout the test. The sinusoidal bending is carried out deformation-controlled symmetrically
around the zero position with a constant amplitude. In order not to damage the specimen during
stiffness testing, the amplitude of the horizontal strain on the lower side of the specimen is set to
50 m/m. The deformation of the specimen in vertical direction as well as the loading is recorded
as a function of time.
The DTC represents another test setup to assess the viscoelastic behavior of HMA. The principle
is shown in Figure 4. The specimen is fixed to two load plates on both of its end planes. The
specimen is subjected to cyclic (sinusoidal) axial loading around the zero position. The test is
carried out deformation controlled with a strain amplitude of 10 m/m in order not to damage the
specimen during the stiffness test. Again, the device is located within a temperature chamber to
control the test temperature.
142
Figure 5. Stress and strain for a viscoelastic material as a result of cyclic tests.
For both test setups, the dynamic modulus |E*|, as well as the elastic (E1 ) and viscous (E2 ) part
of the dynamic modulus and the phase lag () can be obtained according to the following equation
(Di Benedetto et al. 2001) and the definition shown in Figure 5.
Tests with both setups were run at temperatures of 10 C, +5 C and +20 C and frequencies
ranging from 0.1 Hz to 40 Hz. At the end of each test a second frequency packet of 0.1 Hz was
run and test results in terms of dynamic modulus were compared to the results of the first 0.1 Hz
frequency package. If the results of the last and first 0.1 Hz packet exhibited a difference larger than
5%, the test results were not considered for further analysis since it is assumed that the specimen
was subject to fatigue in the course of testing.
5 AIR VOID-TIME-TEMPERATURE SUPERPOSITION
The TTSP takes advantage of the fact that time (i.e. testing frequency) and temperature are
interchangeable variables for thermo-rheological simple materials like HMA. This means that
an increase in temperature is equal to a decrease in frequency in terms of viscoelastic material
behavior, vice versa. The TTSP is a powerful concept to derive a single master curve from material
tests carried out at different temperatures and frequencies for a user-defined reference temperature
and a wide range of frequencies. Thus, the master curve enables the user to obtain viscoelastic
material parameters even for not tested temperatures and frequencies (Findley et al. 1989).
The following section aims towards an extension of the TTSP to include another variable into
the concept, in this case the air void content of the HMA. By using the air void-time-temperature
superposition principle (AVTTSP), viscoelastic material parameters can be derived for a reference
temperature and reference frequency for a wide range of air void contents. The derivation of the
AVTTSP is presented exemplarily for the AC 22 50/70 tested in the 4PBB in the following.
143
Figure 6. Dynamic modulus for different frequencies and air void contents at 20 C for the AC 22 50/70 tested
in the 4PBB (left) and air void-time superposition (right).
When the impact of time and air void content on the dynamic modulus are taken into consideration
at a constant temperature of 20 C (see left diagram in Figure 6) it is obvious that an increase in
frequency leads to an increase in the dynamic modulus. If the focus is laid upon the air void content,
the dynamic modulus decreases with increasing void content. The decrease of the dynamic modulus
can be approximated by a linear for each test frequency, although the linearity tends to decrease with
decreasing frequency. At 40 Hz the linear regression of the modulus vs. air void content exhibits a
coefficient of determination R2 of 0.8528. R2 drops to 0.59 for a frequency of 0.1 Hz. The sensitivity
of the material stiffness in terms of dynamic modulus to a change in the air void content increases
with increasing frequency. At 0.1 Hz the slope of the linear is 273.87 and increases by a factor of
2.6 to 709.45 at 1.0 Hz and by a factor of 1.75 to 1241 at 10 Hz. The slope of the linear reaches a
maximum at 40 Hz with 1400.2. The more time the material is given to react to loading, the less
important becomes the air void content. Since the elastic part of the material behavior is more
dominant at high frequencies, it can be stated that the elastic material behavior is more sensitive to
a change in the air void content.
Analogous to the TTSP, it is possible to scale the air void content of test results at different
frequencies to derive one single master curve for a reference frequency for one test temperature.
The equation for this shifting procedure, which can be called air void-time superposition, is
where VCf = scaled void content in the air void-time superposition [% (v/v)]; VC = void content
[% (v/v)]; af = shift factor depending on test frequency; f0 = reference frequency [Hz]; f = test
frequency [Hz].
The procedure explained above was carried out for test data shown in the left diagram in Figure 6.
The right diagram in Figure 6 shows the master curve for a reference frequency of 10 Hz. The shift
factor af was determined to be 1.3. The shape of the master curve resembles one part of an S-shaped
sigmoidal function. This is analogous to the master curve according to the TTSP. At higher void
contents the dynamic modulus converges towards zero asymptotically. By employing the air void-
time superposition the material behavior can be determined for a wide range of air void contents
at the test temperature for a reference frequency.
When test data is available for more than one test temperature, the air void-time superposition can
be extended even more. The left diagram in Figure 7 shows master curves according to the air void-
time superposition for stiffness tests carried out on the AC 22 50/70 in the 4PBB at temperatures of
10 C, +5 C and +20 C. The reference frequency was set to 10 Hz. Again, these master curves
can be shifted in the air void domain to create a more general master curve including test data for
all test temperatures and frequencies. This air void-time-temperature superposition (AVTTSP) is
144
Figure 7. Air void-time superposition at different temperatures for AC 22 50/70 tested in the 4PBB (left) and
Air Void-Time-Temperature Superposition (AVTTSP) (right).
Figure 8. Master curve for |E | (left) and (right) for AC 22 50/70 tested in the 4PBB.
presented in the right diagram in Figure 7 and can be derived according to the following equation:
where VCf ,T = scaled void content in the AVTTSP [% (v/v)]; aT = shift factor depending on test
temperature; T0 = reference temperature [K]; T = test temperature [K].
By employing the AVTTSP, the viscoelastic material behavior can be determined for an even
wider air void range at a reference temperature and a reference frequency. Two shift factors have
to be determined. The best way to derive these two shift factors is to start with the shift factor af ,
which is dependent on the test frequency and continue with the shift factor aT , which depends on
the test temperature. It was found that one set of shift factors describes the behavior of a material
tested at different frequencies and temperatures. For the presented example of an AC 22 50/70
tested in the 4PBB, the shift factor af was found to be 1.3 and the shift factor aT to be 140.
The right diagram in Figure 7 shows the master curve according to the AVTTSP for a reference
frequency of 10 Hz and a reference temperature of 10 C. The master curve presented in the
right diagram in Figure 7 states that an increase in the void content is equal to an increase in test
temperature or a decrease in test frequency. The AVTTSP gives valuable information on how a
material reacts to dynamic loading when the air void content of the material is changed. Thus,
it can be used for a sensitivity analysis in terms how an increase in the void content affects the
viscoelastic material behavior. It can also be used for quality assurance in the field when the void
content of samples taken from a pavement is determined in accompanying quality control.
The left diagram in Figure 8 shows the same master curve as the right diagram in Figure 7 but
the dynamic modulus is shown in log scale. The right diagram in Figure 7 contains the master curve
of the phase lag according to the AVTTSP for the same test data as presented above. The shift
145
Table 2. Shift factor af for the two mixes and test setups.
af 4PBB DTC
AC 22 50/70 1.27 1.10

AC 22 PmB 45/80-65 1.40 1.10
Table 3. Shift factor aT for the two mixes and test setups.
aT 4PBB DTC
AC 22 50/70 140 135

AC 22 PmB 45/80-65 133 140
Figure 9. Master curves of |E*| (left) and (right) comparing AC 22 50/70 and AC 22 PmB 45/8065 tested
in the 4PBB.
factors are the same as for the dynamic modulus. The presented data prove that the AVTTSP is
also valid for the phase lag . Again the master curve was determined for a reference frequency of
10 Hz and a reference temperature of 10 C. An increase in the void content goes along with an
increase in the viscosity of the material. The master curve also shows that an increase in the void
content at a specific reference frequency and temperature has the same effect as an increase of the
test temperature or decrease in test frequency at a constant void content.
6 RESULTS AND DISCUSSION
Since two different mixes were tested with two different test setups, an analysis of the difference
between mixes and test setups is presented in the following. Table 2 and Table 3 show the shift
factors af and aT for both mixes and test setups. The master curves presented in the following are
always for a reference frequency of 10 Hz and a reference temperature of 10 C.
Figure 9 shows the master curves of the dynamic modulus (right diagram) and the phase lag
(left diagram) for the two mixes, AC 22 50/70 and AC 22 PmB 45/80-65, tested in the 4PBB. The
material stiffness in terms of the dynamic modulus is comparable for both materials in the low and
high air void range. Between air void contents of 5.0% (v/v) and 20.0% (v/v), the mix with the
modified binder exhibits lower stiffness. When it comes to the phase lag, both materials behave
similar in the low air void range. In the intermediate air void range from 5.0% (v/v) to around
20% (v/v) the mix with the unmodified binder shows lower phase lags and is thus more elastic
than the mix with the PmB. This situation is turned around for air void contents above 20.0% (v/v).
146
Figure 10. Master curves of |E*| (left) and (right) comparing AC 22 50/70 and AC 22 PmB 45/8065 tested
in the DTC.
Figure 11. Master curves of |E*| (left) and (right) comparing AC 22 50/70 tested in the 4PBB and DTC.
The PmB mix seems to reach a threshold value of around 30 to 35 , whereas the unmodified mix
ranges between 35 and 45 . This goes along with the fact that mixes with PmB exhibit more elastic
behavior in the high temperature or low frequency range which is equal to a high air void range.
When both mixes are tested in the DTC, no clear difference between the materials can be seen.
Figure 10 shows the results. Both mixes exhibit similar dynamic moduli throughout the air void
contents. In terms of the phase lag (right diagram in Figure 10), the situation is similar to the results
from 4PBB testing. The unmodified mix shows lower phase lags in the low air void range up to
15.0% (v/v) and higher phase lags in the high air void range from around 20.0% (v/v) on.
The two diagrams in Figure 11 compare results for the AC 22 50/70 tested in the 4PBB and the
DTC. The left diagram shows the master curves for the dynamic modulus. For the low air void
range (equal to low temperatures or high frequencies) the material reacts stiffer when tested in
axial direction in the DTC. From around 10.0% (v/v) on both test setups result in similar dynamic
moduli. The opposite tendency can be seen in terms of the phase lag (right diagram in Figure 11). In
the low air void range up to around 10.0% (v/v) both setups produce similar phase lags. When the
air void content is increased further, specimens tested in the DTC react more viscous, i.e. produce
higher phase lags than specimens tested in the 4PBB. Thus, it can be stated that both setups produce
comparable results in terms of material stiffness in the high air void range and in terms of viscosity
in the low air void range.
Figure 12 shows an analogous comparison of test setups for the AC 22 PmB 45/80-65. The
left diagram contains data for the dynamic modulus. Interestingly enough there is no difference
between both setups if the intermediate range of the air void content is left out of the consideration
for the DTC test data. Due to problems in the test control this data shows large scattering and cannot
be seen as valid. When the focus is put upon the phase lag (right diagram in Figure 12) it seems
147
Figure 12. Master curves of |E*| (left) and (right) comparing AC 22 PmB 45/80-65 tested in the 4PBB
and DTC.
that the modified mix tested in the DTC results in slightly lower phase lags in the intermediate air
void range (from 10.0% (v/v) to 20.0% (v/v)).
7 SUMMARY AND CONCLUSIONS
The presented research study aimed towards an extension of the TTSP to include the effect of
the air void content of HMA into the superposition principle. Therefore, two different HMAs, an
AC 22 50/70 and an AC 22 PmB 45/80-65, were tested in dynamic stiffness tests with two different
setups, the 4PBB and the DTC. The specimens were obtained by cutting and coring from HMA
slabs compacted with a roller compactor. A total of 69 HMA slabs were produced with a wide
range of air void contents to investigate the impact of this volumetric mix design property on the
viscoelastic material behavior in terms of dynamic modulus and phase lag. The dynamic stiffness
tests were carried out with temperature and frequency sweep.
From the test results it can be concluded that there is a linear interrelation between the dynamic
modulus and the air void content for all test frequencies. A higher void content leads to lower
material stiffness (i.e. dynamic modulus). The linearity is more significant for high frequencies
than for low frequencies. In addition the HMA reacts more sensitive to a change in the void content
the higher the test frequency is.
In the further course of the research it was found that the TTSP can indeed by extended by
including the effect of the air void content. As a consequence, an air void-time-temperature super-
position principle (AVTTSP) was introduced. Two shift factors take into account the test time and
test frequency. By employing the AVTTSP the viscoelastic material reaction (dynamic modulus
and phase lag) can be obtained for a wide range of air void contents at a user-defined reference
temperature and reference frequency. Thus, the impact of a change in the air void content can
be investigated for the tested material. The shape of the derived master curves is similar to those
master curves derived from the TTSP. The shape resembles an S-shaped sigmoidal function. Thus,
there are two threshold values in the low and high air void range, where the viscoelastic material
parameters approximate an asymptotical value.
A comparison of both materials at a reference frequency of 10 Hz and a reference temperature
of 10 C shows that clear differences can be found in terms of dynamic modulus only when the
materials are tested in the 4PBB, whereas differences in terms of the phase lag become obvious in
both test setups. The modified mix reacts stiffer in the intermediate air void range, more viscous
in the intermediate and less viscous in the high air void range compared to the unmodified mix.
When the two test setups are compared it was found that they produce comparable dynamic
moduli in the high air void range from 10.0% (v/v) on and comparable phase lags in the low air
void range up to 10.0% (v/v). However, differences between both setups are not substantial outside
of these ranges.
148
REFERENCES
Bell C.A., Hicks R.G., an Wilson J.E., Effect of Percent Compaction on Asphalt Mixture Life. ASTM
STP829, American Society of Testing and Materials, Philadelphia, USA, 1984.
Di Benedetto H., Partl, M. N., Francken, L., and De La Roche, C., Stiffness Testing for Bituminous Mixtures.
Journal of Materials and Structures, Vol. 34, 2001.
Findley W. N., Kasif O., and Lai J., Creep and Relaxation of Nonlinear Viscoelastic Materials, Dover
Publications Inc., 1989.
Harrington E.T., Significance of As-Constructed HMA Air Voids to Pavement Performance from Analysis
of LTPP Data. NCHRP Research Results Digest, Issue No. 269, Transport Research Board, Washington,
USA, 2002.
Harvey J.S, Monismith C.L., and Eriksen K.T., Effect of Laboratory Compaction Method on Test Results
Using SHRP A-0003 Equipment. Report Number: VTI 1A, Part 3, Swedish National Road and Transport
Research Institute, VTI, Linkping, Sweden, 1994.
Kassem E., Masad E., Lytton R., and Chowdhury A., Influence of Air Voids on Mechanical Properties of
Asphalt Mixtures. Journal of Road Materials and Pavement Design, Vol. 12, No. 3, 2011.
Lytton R.L., Uzan J., Fernando E.G., Rogue R., Hiltmen D., and Stoffels S., Development and Validation
of Performance Prediction Models and Specifications for Asphalt Binders and Paving Mixtures. SHRP
A-357, Strategic Highway Research Program, Washington, USA, 1993.
Rowe G. M., Khoee S. H., Blankenship P., and Mahboub K. C., Evaluation of Aspects of E* Test by Using
Hot-Mix Asphalt Specimens with Varying Void Contents. Journal of the Transport Research Boards No.
2127, Transport Research Board, Washington, USA, 2009.
Youngguk S., Omar E.-H., Mark K., Joon L, and Kim Y. R., Air Void Models for Dynamic Modulus, Fatigue
Cracking and Rutting of Asphalt Concrete. Journal of Materials in Civil Engineering, Vol. 19, No. 10,
2007.
149
Fatigue evaluation of reinforced asphalt concrete slabs
K.P. Biligiri
Arizona State University, Tempe, Arizona, United States of America
Swedish National Road & Transport Research Institute, Linkping, Sweden
S. Said & H. Hakim

Swedish National Road & Transport Research Institute, Linkping, Sweden
ABSTRACT: Reinforcement of asphalt layers is an alternative method of rehabilitating and

strengthening of the existing roads. The objective of this study was to understand the effect of
reinforcement in asphalt concrete slabs and assess crack growth. The scope of the work included
evaluation of crack propagation mechanism of the laboratory prepared asphalt slabs with and with-
out steel reinforcement using laboratory accelerated load testing with a Wheel Tracking Test device.
Comparisons of the test results between reinforced and unreinforced asphalt slabs revealed that the
steel net increased the fatigue life of the asphalt slab by delaying the propagation of cracks through
the asphalt slab. Overall, the resistance of asphalt concrete against reflective cracking using rein-
forcement with steel fabric increased by a factor of about 35% in comparison with the unreinforced
asphalt mix. These results were also comparable to the seven-year old serviced field test sections
placed with and without steel reinforcement.
1 INTRODUCTION
Tensile stresses along with the other stresses are produced at the bottom of the asphalt concrete
(AC) layers due to traffic loadings on the pavement surfaces. However, continuous applications of
repeated traffic loads initiate fatigue cracks, which then propagate through the AC layer, generally
in a vertical direction to reach the top surface. The cracks in the old pavement propagate usually
quickly through the bituminous overlay. This type of cracking mechanism is called reflective
cracking.
Reflective cracking is the most common and serious type of distress in asphalt overlays. There
is a great technical difficulty in characterizing the reflective cracking distress and the different
forms of crack movement that can produce a reflective crack. Reflective cracking is dependent
on a myriad of factors including the extent and severity of existing asphalt or concrete pavement
cracking, concrete joint movements, traffic loads, climatic conditions, structural characteristics,
interface conditions, overlay mix characteristics and general overall pavement damage. All of
these factors contribute to the complexity of predicting the reflective cracking behavior of the
AC overlays.
Crack propagation assumes an important portion of the cracking mechanism since that is the area
which can be investigated over a period of time interval unlike the crack initiation which is instan-
taneous. Furthermore, regular monitoring and assessment of crack propagation (and resistance) of
the actual pavement sections in the field is deemed important for national and local agencies. In
addition, AC reflective cracking resistance is one of the most significant distresses imminent at the
rehabilitation stage of the fatigued roads. Overlaying of pavement surfaces with weak materials
to counter reflective cracking would potentially instigate heavy and large scale cracking in the
pavement system within a very short duration. Therefore, practical methods to distinguish between
mixtures reflective cracking potential are needed.
151
This paper presents a research study that was undertaken to understand crack propagation of
laboratory prepared asphalt mixtures with steel reinforcement. The effort aimed at assessing if the
steel reinforcement in asphalt layers will delay the propagation of the cracks through the bituminous
overlay, and if it could be an effective method in rehabilitation and strengthening of pavement. This
paper presents excerpts from a larger part of the study performed at VTI-Swedish National Road &
Transport Research Institute, Sweden (Salmenkaita et al, 2002). The major aim of the study was to
study the effects of steel reinforcement against reflective cracking using laboratory Wheel Tracking
Test (WTT) device (SS-EN 12697-22, 2004).
Following the laboratory studies, overlays to an old cracked asphalt highway pavement was
designed using an asphalt mix comprising of steel fabric reinforcement. Three full-scale 100-meter
test sections were built. Two test sections were reinforced with steel fabrics and one section was
left unreinforced (considered a reference section). These sections were instrumented with strain
gauges. Strain measurements were taken at the bottom surface of the overlays and on the steel
bars, and manual distress surveys were conducted. Strain measurements at the bottom surface of
the asphalt layers showed lower strains in the reinforced test sections than in the reference section
without reinforcement after seven years of service (Said et al, 2009).
It is worth noting that during the late 1980s, Brown et al investigated reflective cracking mecha-
nism on larger samples than permitted by the beam testing method, and under more representative
loading conditions using large WTT equipment (Brown et al, 1989). A slab or beam representing
the test structure was submitted to the action of a moving wheel very similar to the one used in this
research study. Furthermore, WTT methodology was used in several other projects to also evaluate
permanent deformation characteristics of the different asphalt pavement materials (Said, 2004).
The Extra-Large WTT was found to be reliable and practical for laboratory studies of flow rutting
assessment.
Although normally cracked roads are rehabilitated by a thin AC overlay, reinforcement of asphalt
layers could be an alternative for rehabilitation and strengthening of existing roads. This research
study was carried out to quantify the effect of reinforcement in asphalt concrete mixes and assess
crack growth in the laboratory that closely simulates field conditions.
2 OBJECTIVE OF THE STUDY
The aim of this research study was to quantify the effect of steel reinforcement in asphalt con-
crete slabs, and assess crack growth and propagation in the asphalt layers. The scope of the work
included laboratory testing using an accelerated load testing equipment, namely, WTT appara-
tus at low temperatures to understand the crack propagation phenomenon in reinforced asphalt
concrete mixes.
3 BACKGROUND TO WHEEL TRACKING TEST
The WTT equipment used is described as extra-large size device in the testing standard EN 12697-
22. Asphalt slabs required for testing normally have dimensions of 50 cm 70 cm 6 cm (with
a 30 kg mass). The compaction is performed using a vibratory footpath roller, 1700 kg Dynapac
CG11 as illustrated in Figure 1. The prepared asphalt slab rests on an intermediate layer of plywood
with a 10-mm transversal gap at the center representing a crack under the overlay. The system is
supported with an elastic foundation made of rubber with shore 40. The slab is subjected to repeated
loading by a rolling wheel.
The test box made of marine plywood is illustrated in Figure 2 (a). The asphalt slab is placed
over two pieces of plywood with 10-mm gap. The wood pieces are supported by a rubber mat
with shore 40 at the bottom of the box. The vacuums between the slab and the box frame are
grouted by gypsum to eliminate any undesirable movements of the slab during testing. The test
152
Figure 1. Compaction of the test slab.
Figure 2. (a) Sample plate configuration; (b) Wheel Tracking Test (WTT) equipment with rubber supported
asphalt slab used in reflective cracking tests at 5 C.
Figure 3. (a) Schematic picture of the gluing technique of strain gauge on asphalt slab; (b) Strain gauges
glued at the bottom surface of the asphalt test slab in an actual setup.
box is positioned in the wheel-tracking machine and brought to the test temperature overnight. The
wheel-tracking machine is as illustrated in Figure 2 (b).
Two strain gauges are mounted in the lower and upper edges of the surface so that deformations
could be registered. The strain gauges are glued at the bottom surface of the slab under the wheel
path. Two slots are milled up, 3 mm in depth, 20 mm wide and 150 mm in length. The strain gauges
are glued by a thin layer of bitumen in the slot and then covered by elastic silicon glue. Cables from
the strain gauges are also glued in small traces by quick glue as shown in Figures 3 (a) and (b).
The two strain gauges at the top surface of the slab are placed just outside the wheel path above the
bottom strain gauges. Testing is carried out separately on the different mix types with three sample
replicates. The test temperature used is 5 C and the load is adjusted for a desired deformation until
153
Figure 4. Strain gauges: (a) on steel net (b) instrumentation on the slab.
cracks propagated to the surface for a finite number of passes. This methodology simulates the
influence of traffic loading on crack growth in an AC overlay.
4 REINFORCEMENT IN ASPHALT SLABS
As described previously, the main aim was to understand the effect of reinforcement on crack prop-
agation rate of asphalt slabs with respect to their reflective cracking mechanisms. To this end, the
efforts of research included manufacture of asphalt slabs with and without steel reinforcement; and
then utilize WTT to compare crack propagation mechanisms in both the asphalt concrete mixtures.
The slabs were manufactured in the laboratory facilities with dimensions of the slabs being
700 500 60 mm. The slabs with and without steel net were compacted in a mold with a steel
frame and a marine (high quality) plywood ground as described earlier. In reinforced slabs, the
steel net was placed about 2-cm from the bottom surface of the slab. The steel nets had the notation
NPS 500 SP. The pitch size was 50 50 mm with bars of 5 mm in diameter. The small pitch size
was used in this work because the tire of the WTT machine was smaller than the truck tires; this
guaranteed that the steel net would be loaded during the test.
Two strain gauges were used to measure longitudinal and transverse strains on the steel net.
The strain gauges were glued on the steel bar. The gauge was wrapped up with a thin layer of a
vulcanization tape, covered with a plastic layer, and wrapped up again by the tape. This procedure
protected the gauge during the compaction. Figure 4 (a) shows strain gauges glued in one pitch.
The strain gauges were located at the bottom surface of the steel net in order to be protected during
compaction (Figure 4 b). The instrumented pitch would be located just under the wheel path.
The tests were conducted at +5 C with a tire pressure of 0.6 MPa. Different loads were used to
introduce various strain levels in the slab. The width of the contact area between the tire and the slab
was about 168 mm compared to the truck tire in the field, which is about 300 mm. A rolling tire at a
speed of 2.5 km/h loaded the asphalt slab. Strains at the bottom and upper surfaces of the slab as well
as on the steel net were measured continuously with the number of wheel passages. The fracture
life was determined when the strain increased dramatically with the number of wheel passages in
linear-logarithmic scale diagram. This was also captured when the strain gauge ruptured, which
was actually visualized on the computer screen. The crack elongation was recorded by double strain
gauges in the bottom and top of each sample plate. For each test, the initial elongation (strain after
100 passages reflecting the maximum force which was applied to the sample) and the difference in
the number of passages from the time crack initiated on the lower side until it reached the upper edge
of the sample plate. This difference, which was a measure of how quickly the cracks developed,
was also recorded.
154
Table 1. Crack propagation test results using the WTT apparatus, unreinforced asphalt slabs.
Initial Number of passes

Slab Air voids Load deformation Nf
number (%) (kN) () Lower side Upper side Difference
4 7.0 10 260.2 7000 40000 33000

14 5.4 5 205.4 15500 91000 75500
15 7.0 7.8 391.9 4800 11000 6200
19 5.5 2.3 143.6 90000 175000 85000
20 6.6 8.1 472.7 4600 11500 6900
21 6.9 4.3 235.3 7500 79000 71500
Table 2. Crack propagation test results using the WTT apparatus, reinforced asphalt slabs.
Initial Number of passes

Slab Air voids Load deformation Nf
number (%) (kN) () Lower side Upper side Difference
12 6.7 10 297.0 5000 28000 23000

13 7.1 12 483.9 850 9750 8900
16 6.6 5.3 216.3 20000 51000 31000
17 7.2 4.3 183.5 39000 120000 81000
18 6.0 3.9 194.3 30000 135000 105000
5 TEST RESULTS & ANALYSES
5.1 Test results

Table 1 and Table 2 show the WTT results for the unreinforced and reinforced asphalt mixtures,
respectively. The tables include the air voids measured for each asphalt test slab, along with the load
applied during the test, initial deformation (incurred at the bottom of the slab), the number of passes
required for the crack to propagate from the bottom of the slab to the surface. Strain measurements
were recorded at a predefined schedule with the number of wheel passages. Figure 5 shows an
example of signals from the asphalt slab and the steel net during a test.
As known, steel is very good in tension, and the figure shows the same characteristic trend.
Furthermore, tensile strains are predominately observed on the bottom of the asphalt slab surface,
which is also evident from the figure. Also, the top of the asphalt surface produces compressive
strain as also shown in the figure. These trends are comparable with the field reinforced sections
that show the effectiveness of steel reinforcement resulting in decreasing tensile stresses at the
bottom of the asphalt surficial layers (Said et al, 2009).
5.2 Analyses
Figure 6 illustrates measurements on the asphalt slab with reinforcement of the steel net. Defor-
mation signals from the strain gauges allocated on the test slabs show increasing strain levels with
the number of load applications. As expected, the fracture commenced at the bottom of the asphalt
slab. Then the strain level increased on the steel net as illustrated in the figure. The strain on the
top surface is in compression and also increased with the number of applications until the cracks
propagated to the top surface of the slab. The difference in number of applications to fracture
between bottom and top strain gauges indicates the propagation time of a crack through the asphalt
slab. It is obvious that the time of crack propagation through the asphalt slab was much longer in
the reinforced slab than in the slab without reinforcement (Figure 7).
155
Figure 5. Tensile strain signals on asphalt slab surfaces and on steel bar under rolling wheel.
Figure 6. Strain levels & number of applications of a reinforced asphalt concrete slab.
The initial strain of the steel net is approximately about half of the initial strain at the bottom of
the asphalt slab (Figure 6). Initial steel net strains varied between 75 and 250 micro-strains. The
maximum strains of the steel net measured in these experiments occurred at the end of the test
when the cracks had gone through the asphalt slab. Steel net strains when the slabs were thoroughly
cracked varied between 270 and 580 micro-strains.
Comparisons between reinforced and unreinforced asphalt slabs with respect to resistance against
reflective cracking are presented in Figure 8. The figure presents fatigue relationships based on
the strain versus number of load applications. The number of load applications is based on the
difference between the number of wheel passes when a crack starts at the bottom surface of the slab
(disconnection of the bottom surface gauges) and reaches the top surface of the slab (disconnection
of the top surface gauges). The crack growth time is used here since the steel net effect is primarily
to delay the crack propagation to the surface of the pavement. Although the experimental program
156
Figure 7. Strain levels & number of applications of an unreinforced asphalt concrete slab.
Figure 8. Fatigue relationship of reinforced and unreinforced asphalt slabs.
consisted of only a very few number of samples, there is a clear discernment between the reinforced
and unreinforced asphalt slabs in that one can observe the merit of using steel reinforcement in
delaying crack propagation. Figures 9 (a) and (b) show examples of cracks produced at the bottom
and top surfaces of the asphalt slabs. The cracks as expected were at the top of the gap between the
wood pieces. Furthermore, the nature of the total number of cracks at the top surface were usually
many small cracks for the reinforced slabs while there was only one sharp and relatively a wide
crack on the top surface for the unreinforced slab.
6 CONCLUSIONS
The aim of this research study was to quantify the effect of steel reinforcement in asphalt concrete
slabs, and assess crack growth and propagation in the asphalt layers. The scope of the work included
laboratory testing using an accelerated load testing equipment, namely, Wheel Tracking Test (WTT)
apparatus at low temperatures to understand the crack propagation phenomenon in reinforced
asphalt concrete mixes.
157
Figure 9. Example of a crack at the: (a) bottom surface of an asphalt slab; (b) top surface of an asphalt slab.
The scope of the work included manufacture of asphalt slabs with and without steel reinforce-
ment; and then utilize WTT device to compare crack propagation mechanisms in both the mixtures.
The slabs were manufactured in the laboratory with dimensions being 700 500 60 mm. The
tests on both the asphalt mixture types were conducted at +5 C with a tire pressure of 0.6 MPa.
Strains at the bottom and upper surfaces of the slab as well as on the steel net were measured
continuously with the number of wheel passages. The fracture life was determined when the strain
increased dramatically with the number of wheel passages in linear-logarithmic scale diagram. This
was also captured when the strain gauge ruptured, which was actually visualized on the computer
screen.
From the laboratory test results, it is concluded that the steel net increased the fatigue life of
the asphalt slab by delaying the propagation of cracks through the asphalt slab. It was also noted
that the nature of the total number of cracks at the top surface were usually many small cracks for
the reinforced slabs while there was only one sharp and relatively wide crack on the top surface
for the unreinforced slab. Overall, the resistance of asphalt concrete against reflective cracking
using reinforcement with steel fabric increased by a factor of about 35% in comparison with the
unreinforced asphalt mix. The results should be taken as a trend rather than significant differences
between reinforced and unreinforced asphalt slabs, mainly due to the limited number of samples
used in the experimentation program.
Following the laboratory studies, measured strains on an old Swedish cracked asphalt highway
pavement designed with steel fabric reinforcement within the mix showed lower fatigue strain
values than the unreinforced sections after seven years of service.
REFERENCES
Brown S. F. Brunton J. M. and Armitage R. J. 1989. Grid reinforced overlay, 1st RILEM conference on
reflective cracking in pavement, Liege, 1989, pp. 6370.
Said, S 2004. Prediction of flow rutting, International Society for Asphalt Pavements-sponsored International
Symposium on Design and Construction of Long lasting Asphalt Pavements, Auburn, June 2004.
Said S., Carlsson H., and Hakim, H. 2009. Performance of flexible pavements reinforced with steel fabric.
Bearing Capacity of Roads, Railways and Airfields Tutumluer & Al-Qadi (eds), Taylor & Francis Group,
London, ISBN 978-0-415-87199-09.
Salmenkaita S., Kolisoja P., Lechner B., Simon C., Said S., Belt J. and Lms V.P. 2002. Laboratory studies
on asphalt-steel net compound system and unbound granular material steel net compound system. EU-
project REFLEX Report T3:2, 2002.
SS-EN 12697-22:2004 Bituminous mixtures Test methods for hot mix asphalt Part 22: Wheel tracking.
158
Influence of recycled asphalt shingles on fatigue properties

of asphalt pavements
A.A. Cascione & R.C. Williams

Iowa State University, Ames, Iowa, US
ABSTRACT: As the use of Recycled Asphalt Shingles (RAS) in asphalt pavements gains wider
acceptance, questions remain about the effects of RAS on the fatigue performance of asphalt pave-
ments. To compare the fatigue performance of RAS pavements to non-RAS pavements, pavement
mixtures may be evaluated with accelerated laboratory fatigue tests. For this paper, repeated flexu-
ral beam performance testing was conducted to study the load-associated fatigue characteristics of
several types of RAS mixtures. The RAS mixtures were field produced samples of pavements from
several states designed with different RAS sources, percentages, specifications, and mix types.
Repeated flexural loading on the four-point bending beam apparatus was conducted on samples
to compare how the different RAS variables affect fatigue cracking. The test results indicate that
low percentages of RAS can enhance the fatigue characteristics of asphalt pavements with some
construction specifications being more beneficial than others.
1 INTRODUCTION
Significant, widespread interest in modifying asphalt pavements with recycled asphalt shingles
(RAS) has been shown among state highway agencies (agencies) in the United States. Many
agencies share common questions about the effect of RAS on pavement fatigue performance.
Previous research has allowed for only limited laboratory testing and field surveys. The complexity
of RAS materials and lack of past experiences generates a need for additional research to study the
performance of pavements with RAS, in particular, fatigue cracking performance.
The test data evaluated in this paper is laboratory beam fatigue data collected at Iowa State
University as part of the Transportation Pooled Fund (TPF) Program TPF-5(213). TPF-5(123) is a
partnership of several state agencies in the United States with the goal of researching the effects
of RAS on the performance of asphalt applications. As part of the Pooled Fund research program,
multiple state demonstration projects were conducted to provide adequate laboratory and field test
results to comprehensively answer design, performance, and environmental questions about asphalt
pavements with RAS. These questions include their fatigue performance.
2 RAS IN ASPHALT PAVEMENTS
In the United States, an estimated 8 to 11 million tons of RAS is buried in landfills each year (EPA
1998). Recycling and reusing this construction waste material in asphalt pavements could yield
significant cost savings while reducing the impact on the environment. The components of RAS
allow it to be a good candidate as a secondary material in asphalt mixtures. New roofing shingles
contain between 19 to 30 percent asphalt and include fine angular granules which can improve the
resistance to permanent deformation. Shingles also contain fiberglass or cellulose backing, that
159
Table 1. Components of asphalt shingles, % (Brock 2007).
Post-Manufacturer Post-Manufacturer
(Organic) (Fiberglass) Post-Consumer
Asphalt cement 30 19 31
Mineral filler 26 41 25
Granules 34 38 32
Fiberglass mat 0 2 0
Cellulose felt 10 0 12
when crushed during the recycling process, break up into fiber-like particles that may improve the
fatigue cracking resistance of asphalt.
With these benefits in mind, more state highway agencies are beginning to see the potential
impact RAS could have in lowering the cost of pavements. However, little information about the
long-term fatigue performance of pavements with RAS is known because it is a new material
that agencies are beginning to use. The challenge agencies have when implanting the use of RAS
materials, is developing a construction specification for RAS mixtures that ensures a product with
similar qualities and performance to non-RAS mixtures. Several nuances about the source and
processing of RAS make it important for agencies to understand which factors about RAS affect
the material properties essential for good fatigue performance.
The TPF-5(213) research team identified five primary factors about RAS that interest state
highway agencies when developing specifications. They want to know how these factors influ-
ence the performance of pavements with RAS. These factors are the percentage of RAS in the
pavement, the source of RAS (either post-consumer or post-manufacturer), the consistency of
RAS after being processed (a fine grind versus a coarse grind), the performance difference
between RAS and recycled asphalt pavement (RAP), and the compatibility of RAS with warm mix
asphalt (WMA).
2.1 Percentage of RAS in asphalt pavements

RAS can be added to asphalt pavements at different percentages. Due to the stiffness of the RAS
binder, the performance grade (PG) of a virgin binder will increase when RAS is added. To
ensure the virgin materials are not detrimentally altered, agencies have developed preliminary RAS
specifications with different limits on the amount of RAS that can be added to mixtures. These
limits range from as low as three percent to as high as seven percent. Understanding the different
effects percentages of RAS have on pavement fatigue performance will greatly aid in specification
development.
2.2 Source of RAS

RAS comes from two different shingle sources, waste produced from the manufacturing pro-
cess (referred to as post-manufacturer RAS) and construction demolition debris generated
from reroofing projects (referred to as post-consumer RAS). The difference between post-
manufacturer and post-consumer RAS is presented in Table 1. Post-consumer RAS will contain
a higher percentage of asphalt in the shingle by mass due to the loss of granules on the shin-
gle over time. Post-consumer shingles are also older and are more likely to be made with a
cellulose felt backing and be more absorptive than fiberglass. A cellulose felt backing will
absorb more asphalt than a fiberglass mat backing and therefore contain a higher percentage of
asphalt.
The binder used in RAS is stiffer than performance graded binder used for pavements which may
impact fatigue cracking potential. The binder in roofing shingles go through an oxidation process
160
during manufacturing so it has a higher viscosity. Additionally, post-consumer shingles have been
in service and aged for a number of years. While binder with a higher modulus is desirable for
highway pavement surface courses at high service temperatures to avoid rutting, stiffer binders are
susceptible to premature fatigue cracking in pavements performing in strain-controlled conditions,
i.e. thin pavement surfaces (Roberts et al. 1996). However, RAS contains organic or synthetic
fibers which are desirable for enhancing the fatigue life of pavements.
The different backings on the shingle will ultimately provide different fiber particles in the RAS.
Various fiber modifiers, such as cellulose, polyester, and mineral fibers, have been widely used in
asphalt mixtures (Wu et al. 2007). Fiber modification has shown promising fatigue results. Fibers
appear to increase the stiffness of the asphalt binder resulting in stiffer mixtures. This improves
the resistance to permanent deformation. The tensile strength and toughness of the mixtures with
fibers is also improved (Cleven 2000). An investigation by Lee et al. (2005) showed that fiber
reinforcement of extended the fatigue life of asphalt mixtures by 20 to 25 percent due to the
increase in fracture energy. However, fibers with greater tensile strength (polyester, carbon fiber)
have improved these properties to a greater extent than cellulose fibers (Putman 2004). This raises
the question of how the non-bituminous particles of RAS from different sources benefit the fatigue
properties of an asphalt mixture.
2.3 RAS grind size

When RAS is crushed and screened during processing, recyclers can change its end-product gra-
dation (grind size). Past research has shown that a finer grind produces a more consistent mixture
(Johnson et al. 2010) and can increase the tensile strength more than a coarser grind (Button et al.
1996). Obtaining a consistent RAS gradation that blends well with the other mix components can be
the most challenging aspect of processing RAS. Test results that demonstrate differences between
grind size underscore the necessity for proper field quality control specifications.
2.4 Replacing RAP with RAS

Most agencies have a RAP specification that allows contractors to replace a portion of the virgin
materials with RAP. Extensive research on RAP has verified its benefits to pavements (Al-Qadi
et al. 2007). The stiffer binder can increase a pavement modulus thereby increasing its ability to
resist permanent deformation. However, the increase in modulus can have a negative effect on low
temperature cracking and fatigue performance. Since RAS shares similar material components as
RAP, agencies are highly interested in knowing if RAS can replace all or a portion of RAP materials
used in asphalt mixtures.
With the addition of different recycled products in one mixture, agencies are moving toward a
maximum percent binder replacement specification so the blended grade of the virgin and recy-
cled binder still meet the desired grade. Agencies are then ultimately interested in the performance
difference between RAS and RAP mixtures that have similar percent binder replacements.
2.5 RAS with warm mix technology

RAS and warm mix asphalt (WMA) are two new technologies in the pavement industry. As a
result, little to no research has been conducted on pavements that utilize both technologies. The
concern is that the temperature of the plant when producing WMA is not hot enough to fully
melt and blend the RAS with the virgin materials. RAS mixes that are not fully blended could
result in a poor performing pavement with not all of the economic opportunity of RAS being real-
ized. Field trials that combine these technologies will be instrumental in evaluating a RAS/WMA
pavement.
161
Table 2. Multi-State experimental plan.
State agency Factors invested in field demonstration projects
Iowa Effect of different RAS percentages

Minnesota Difference between post-manufacturer and post-consumer RAS Comparison of RAS mixture
to traditional RAP mixture
Missouri Difference between RAS with a fine grind and a coarse grind Effect of replacing a percentage
of RAP with RAS
Indiana Compatibility of RAS with WMA foaming technology Difference between asphalt mixtures
containing RAP and RAS
3 RAS MIX DESIGN EXPERIMENTAL PLAN
To evaluate how the five factors of RAS identified by the research team effect pavement perfor-
mance, an experimental plan was developed that included a laboratory fatigue testing component.
Four state highway agencies in the pooled fund study proposed a unique field demonstration project
that investigated different aspects of RAS mixes. Field demonstration projects were sponsored by
the Indiana Department of Transportation (INDOT), the Missouri Department of Transportation
(MODOT), the Minnesota Department of Transportation (MNDOT), and the Iowa Department of
Transportation (IADOT).
INDOTs field demonstration project was conducted on US Route 6 in 2009; MODOTs
field demonstration project was conducted on US Route 65 in 2010; MNDOTs field demon-
stration project was conducted on Interstate 94 at the MnRoads facility in 2008; and IADOTs
field demonstration project was conducted on State Highway 10 in 2010. The factors that each field
demonstration project investigated are presented in Table 2.
During each field demonstration project, representative samples of each asphalt mixture were
obtained and shipped to the laboratory testing facility at Iowa State University for binder character-
ization and mixture laboratory performance testing. To evaluate the mixtures fatigue performance
properties, flexural beam testing was conducted on the four-point bending beam apparatus.
4 MATERIALS
The properties of the RAS used in the mix designs are presented in Table 3. The mix designs
developed for the field demonstration projects and their properties are presented in Tables 4 and
5. Mix design gradations within each demonstration project remained constant so that would not
be a variable for the experiments. However, the mixes from the three Minnesota DOT were each
different. The 5% Post-Consumer RAS mix design was coarse-graded, and the 5% Post-Consumer
RAS and 30% RAP mix designs were fine-graded.
5 TEST PROCEDURES
Four-point beam fatigue testing was conducted according to AASHTO T321, Determining the
Fatigue Life of Compacted Hot-Mix Asphalt (HMA) Subjected to Repeated Flexural Bending.
Samples of field produced asphalt were compacted to 7 0.5 air voids in a linear kneading com-
pactor to obtain a compacted slab with dimensions 380 mm in length, 210 mm in width, and 50 mm
in height. Each slab was saw-cut into three beams with dimensions 380 mm in length, 63 mm in
width, and 50 mm in height. Two slabs were compacted for each asphalt mixture to produce six
beams for testing.
The equipment used to conduct the four-point bending beam test included a digitally con-
trolled, servo-pneumatic closed loop beam fatigue apparatus from IPC Global. A control data
162
Table 3. RAS properties.
% Passing
sieve size Iowa Missouri Missouri Minnesota Minnesota Indiana
(mm) RAS fine RAS coarse RAS Post-Manuf. Post-Cons. RAS
19.0 100 100 100 100 100 100

12.5 99 99 97 100 100 100
9.5 98 99 96 100 100 99
4.75 95 94 90 99 100 90
2.36 90 91 85 97 99 87
1.18 72 73 67 80 85 69
0.60 51 53 46 58 65 47
0.30 40 46 39 40 49 40
0.15 30 37 31 28 35 34
0.075 21.3 26.1 21.9 22.0 24.1 26.5
% Asphalt 21.1 25.0 21.7 14.6 20.5 26.8
content in RAS
(by total wt.)
Table 4. Asphalt mix design experimental plan.
State RAS RAP RAS grind Mix

Mix ID agency (%) (%) RAS source size (mm) type
0% RAS Iowa 0 0 HMA

4% RAS Iowa 4 0 Post-Consumer <12.5 HMA
30% RAP Minnesota 0 30 HMA
5% Manuf. RAS Minnesota 5 0 Post-Manufacturer <4.75 HMA
5% Post-Cons RAS Minnesota 5 0 Post-Consumer <4.75 HMA
0% RAS 15% RAP Missouri 0 15 HMA
5% Fine RAS 10% RAP Missouri 5 10 Post-Consumer <4.75 HMA
5% Crse RAS 10% RAP Missouri 5 10 Post-Consumer <12.5 HMA
15% RAP HMA Indiana 0 15 HMA
3% RAS HMA Indiana 3 0 Post-Consumer <12.5 HMA
3% RAS WMA Indiana 3 0 Post-Consumer <12.5 WMA
and acquisition system (CDAS) was connected to the beam fatigue apparatus which connected to
a computer that controlled the load during the test. The beam fatigue apparatus was housed in an
environmental chamber maintained at the testing temperature of 20 0.5 C. Beams were placed
in the environmental chamber at least two hours prior to testing to allow them to equilibrate to the
testing temperature. The mode of loading used for the test was strain controlled. Haversine wave
pulses were applied to the specimen during the test at 10 Hz.
Testing was conducted at varying strain levels to generate a fatigue curve for each asphalt mixture.
For each of the six beam specimens prepared for each asphalt mixture, strain levels of 375, 450,
525, 650, 800, and 1000 micro-strains were applied. Testing at these strain levels were repeated for
all the mixtures tested in this study except for the two Indiana mixtures containing 3% RAS. Due
to a limited amount of material, only 3 three beams of these mixtures were tested at 400, 700, and
1000 micro-strain levels.
163
Figure 1. Fatigue curve for Iowa mixture with 0% RAS.
During testing of a beam specimen, properties of flexural stiffness, modulus of elasticity, dissi-
pated energy, and phase angle were recorded by the software every 10 cycles. On the 50th cycle,
the stiffness of the beam specimen was recorded as the initial stiffness. The beam specimens were
tested until failure, which was defined as the cycle corresponding to a 50 percent reduction of the
initial beam flexural stiffness.
6 FATIGUE ANALYSIS
A phenomenological approach for fatigue analysis was selected as the chosen methodology to
evaluate the fatigue life properties of the mixtures. The phenomenological approach relates the
tensile strain at the bottom of an asphalt pavement layer to the number of load repetitions to failure
(Ghuzlan et al. 2006). In this approach, fatigue life is plotted versus stress or strain on a log-log
scale.
Since strain-controlled was used as the mode of loading, a log-log regression was performed
between strain and the number of cycles to failure (Nf ) as presented in Figure 1. The relationship
between strain and Nf can be modeled using the power law relationship as presented in Equation 1.
where Nf = cycles to failure; o = flexural strain; K1 = regression constant; and K2 = regression

constant.
The fatigue model can be calibrated to relate laboratory to field conditions by applying a shift
factor, the hypothesis being that laboratory fatigue tests can simulate field conditions. Because of
the challenging nature of duplicating field conditions in a laboratory, no universal shift factor has
been measured. Rather, shift factors have ranged between 4 and 100 (NCHRP 2010).
Pavements that have a higher resistance to tensile strains that develop at the bottom of an
asphalt layer due to repeated traffic will have a greater resistance to fatigue cracking. Therefore,
fatigue curves of several asphalt mixtures can be used to rank the mixtures resistance to fatigue
cracking. However, the results must take into consideration the mode of loading. Research from
the Strategic Highway Research Program (SHRP) A003-A project (Tangella et al. 1990) showed
that materials that are more flexible (lower stiffness) perform better in constant strain. The constant
164
strain mode of loading best represents the performance of thin pavements (less than 4 inches) while
the constant stress mode of loading best represents the performance of thick pavements (greater than
6 inches). Materials that are stiffer may not perform as well under constant strain in the laboratory,
but when used in thick pavements, lower tensile strains will develop under field loading. Therefore,
when fatigue testing will is done in a constant strain mode of loading, fatigue evaluations should
be made in the context of the pavement structure.
If tensile strains are low enough in a pavement structure, the pavement has the ability to heal
and therefore no damage cumulates over an indefinite number of load cycles. The level of this
strain is referred to as the fatigue endurance limit (FEL). Identifying the fatigue endurance limit in
a laboratory is somewhat elusive because it is impossible to test a sample to an infinite number of
cycles. The researchers under NCHRP Report 646 (2010) developed a practical definition of FEL
as the strain level at which a sample could withstand 50 million load cycles. If a shift factor of 10
was applied to the test results, it would be estimated that the pavement could withstand 500 million
loading cycles which represents 40 years of traffic.
Because it can take up to 50 days of testing to see if a sample reaches 50 million cycles, the
NCHRP Report 646 researchers developed a procedure to estimate the FEL of asphalt mixture from
a fatigue curve. They found that the lower 95% prediction limit at 50 million load cycles from a
regression analysis of fatigue data corresponded reasonably close to the FEL. This technique uses
Equation 2 to estimate the fatigue life.
where yo = the one-sided lower 95% prediction interval at the micro-strain level corresponding to
50,000,000 cycles;
t = value of t distribution for n 2 degrees of freedom for a significance level of 0.05;
s = standard error of the regression analysis;
n = number of samples;
Sxx = sum of squares of the x values;
xo = log 50,000,000; and
x = average of the fatigue life results.
7 FATIGUE TEST RESULTS
7.1 Evaluation of different percentages of RAS

The beam fatigue test results for the Iowa mixtures, as shown by the strain versus loading cycles
to failure curves, are presented in Figure 2. The four mixtures contain vary similar gradations and
volumetric properties. They all have approximately the same asphalt content as shown in Table 5.
The only difference between the mixtures is percentage of RAS. Because RAS contains stiffer binder
than virgin binder, it is expected that an increase in RAS percentage would increase the stiffness
of the mixture. Yet, the average initial beam stiffness of the 0 percent RAS mixture was 3497 MPa
while the average initial beam stiffness of the 4 percent, 5 percent, and 6 percent RAS mixtures was
3090 MPa, 3106 MPa, and 3156 MPa respectively. Past beam fatigue studies in controlled strain
mode of loading showed that when stiffness decreases from a change in binder type or grade, beam
fatigue life is typically increased (SHRP-A-404). This study appears to follow the same trend as
well, since the mixes with lower initial stiffness demonstrated longer fatigue lives. However, as the
percentage of RAS increases from 0 to 4 to 5 percent in the mixture, which stiffens the binder grade,
the fatigue life uncharacteristically increases. A possible explanation of this phenomenon, could be
from the complex RAS-aggregate-binder interactions and the contribution of fibers from the RAS.
As the percent RAS content increases from 5 to 6 percent, the fatigue life no longer increases
but decreases. While still significantly higher than the fatigue life of the 0 percent RAS mixtures,
165
Figure 2. Fatigue curves for Iowa mixtures.
Table 5. Asphalt mixture properties.
State NMAS % % Binder Design

Mix ID agency (mm) PG AC replacement gradations
0% RAS Iowa 12.5 58-28 5.5 0 76

4% RAS Iowa 12.5 58-28 5.4 15.1 76
5% RAS Iowa 12.5 58-28 5.5 17.5 76
6% RAS Iowa 12.5 58-28 5.4 19.8 76
30% RAP Minnesota 12.5 58-28 5.3 33.4 90
5% Manuf. RAS Minnesota 12.5 58-28 4.9 14.9 90
5% Post-Cons RAS Minnesota 12.5 58-28 5.0 20.5 90
0% RAS 15% RAP Missouri 12.5 70-22 4.7 19.1 80
5% Fine RAS 10% RAP Missouri 12.5 70-22 5.3 30.2 80
5% Crse RAS 10% RAP Missouri 12.5 70-22 5.3 30.2 80
15% RAP HMA Indiana 9.5 70-22 5.7 18.0 100
3% RAS HMA Indiana 9.5 70-22 6.2 12.6 100
3% RAS WMA Indiana 9.5 70-22 6.2 12.6 100
the decrease could result from the effect of the stiffer binder (now at 19.8 percent replacement)
having a more influential effect on the fatigue properties.
The fatigue curve model coefficients, R2 values, and estimated endurance limits are presented
in Table 6. Because the fatigue life increases as with the addition of RAS in a controlled strain
mode of loading, the results indicate that RAS will improve the fatigue life of a thin lift pavement.
7.2 Evaluation of different RAS sources

For the Minnesota mixtures, the fatigue curves are presented in Figure 3 with the fatigue model
coefficients and fatigue endurance limit estimates in Table 7. In this experiment, two different
mixtures with RAS are compared to a traditional Minnesota DOT mixture containing 30 percent
RAP. The two RAS mixtures were designed with different sources of RAS, one containing post-
consumer RAS, the other containing post-manufacturer RAS. In the controlled-strain mode of
166
Table 6. Iowa fatigue curve results.
% Binder Endurance limit

Mix ID replacement K1 K2 R2 (Micro-strain)
0% RAS 0 1.43E13 5.45 0.987 144

4% RAS 15.1 6.75E14 5.68 0.987 182
5% RAS 17.5 1.97E12 5.27 0.982 175
6% RAS 19.8 7.07E14 5.65 0.967 162
Figure 3. Fatigue curve for Minnesota mixtures.
Table 7. Minnesota fatigue curve results.

30% RAP 33.4 9.19E12 4.90 0.994 131

5% Post-Manufacturer RAS 14.9 2.22E09 4.19 0.996 123
5% Post-Consumer RAS 20.5 6.66E11 4.51 0.982 89
loading, the post-manufacturer RAS mixture has a longer fatigue life at higher strain levels and a
similar fatigue endurance limit to the 30 percent RAP mixtures. The combined effect of a decrease
in percent binder replacement and possible contribution of fiberglass particles in this mixture may
have contributed to a comparable fatigue life.
For the post-consumer RAS mixture, it exhibited reduced fatigue performance than the 30
percent RAP mixture. In contrast to the post-manufacturer RAS mixture, the post-consumer RAS
had a stiffening effect on the total mixture. While in a controlled-strain test, the stiffening negatively
impacted the fatigue life. However, this increase in stiffness would benefit a thick pavement section
as it would reduce the tensile strains at the bottom of the pavement layer.
With opposite fatigue performance properties between the post-consumer and post-manufacturer
RAS mixture, the information may serve useful in optimizing mixture characteristics. The qualities
of post-manufacturer RAS may serve better in a pavement predicted to have high tensile strains;
whereas the qualities of post-consumer RAS may serve better in a pavement where a stiffer mix
167
Figure 4. Fatigue curve for Missouri mixures.
Table 8. Missouri fatigue curve results.

0% RAS 15% RAP 19.1 5.15E17 6.40 0.968 139

5% Fine RAS 10% RAP 30.2 7.25E19 6.91 0.992 145
5% Coarse RAS 10% RAP 30.2 2.07E20 7.37 0.968 159
with good fracture properties and needed. In addition, this highlights the importance of verifying
mix properties when the source of RAS changes.
7.3 Evaluation of different RAS grind size

The Missouri mixture fatigue curves are presented in Figure 4 with the fatigue model coefficients
and fatigue endurance limit estimates in Table 8. The three mixtures contain RAP. Two of the
mixtures replace 5 percent RAP with RAS. Of those two mixtures, one contains a fine grind RAS
and one contains a coarse grind RAS. The no-RAS mixture out performed the two RAS mixtures
at higher strain levels, but demonstrated a lower fatigue endurance limit.
The results are significant on two levels. First, the higher percent binder replacement mixes have
similar (even slightly higher) endurance limits. Second, the use of a coarse grind RAS does not
appear to negatively impact the fatigue properties of the mixture. The results attest to similar (and
possibly even improved) laboratory fatigue performance of the Missouri mixtures when 5 percent
RAP is replaced with 5 percent RAS, regardless of grind size.
7.4 Evaluation of RAS with WMA

For the Indiana mixtures, the fatigue curves are presented in Figure 5 with the fatigue model
coefficients and fatigue endurance limit estimates in Table 9. Here, hot mix asphalt with RAS
is compared to warm mix asphalt with RAS. Both mixtures are also compared to a traditional
15% RAP mixture in Indiana. The fatigue curves and endurance limits of all three mixtures are
essentially the same. The HMA with RAS has similar fatigue properties as the HMA with no RAS.
Additionally, the WMA mixture with RAS performed the same as the other mixes.
168
Figure 5. Fatigue curve for Indiana mixtures.
Table 9. Indiana fatigue curve results.

15% RAP HMA 18.0 7.04E12 4.87 0.993 114

3% RAS HMA 12.6 1.41E11 4.77 0.970 118
3% RAS WMA 12.6 1.17E11 4.81 0.985 110
8 CONCLUSIONS
This paper presented the laboratory beam fatigue tests conducted on asphalt mixes obtained by
Transportation Pooled Fund (TPF) Program TPF-5(213). The mixes tested were field produced
samples of asphalt pavements from several states designed with different RAS sources, percentages,
specifications, and mix types. By using the phenomenological approach to analyze the results, the
following conclusions can be made:
Adding RAS to an Iowa DOT asphalt mixture can increase its ability to resist the amount of
fatigue damage accumulated in a strain-controlled mode of loading. For this study, mixes with
4, 5, and 6 percent RAS demonstrated greater fatigue lives than the same mix with 0 percent
RAS. The longer fatigue lives may be the result of the RAS-aggregate-binder interactions along
with the fibers contributed by the RAS.
A Minnesota DOT asphalt mixture with 5 percent post-manufacturer RAS demonstrated a similar
predicted fatigue endurance limit as an asphalt mix with 30 percent RAP. The mix with 5 percent
post-consumer RAS demonstrated a much lower fatigue life. While not performing well in a
strain-controlled environment, this mix would perform better in a thick pavement structure where
higher stiffness is desirable. The difference highlights the importance of verifying mix properties
when the source of RAS changes.
Replacing 5 percent RAP with 5 percent RAS in a Missouri DOT asphalt mixture lowered the
fatigue live at higher strain levels but did not impact the predicted fatigue endurance limit.
Rather, it slightly improved it. The results did not significantly change when either a coarse
grind or find grind RAS was used.
169
Based on the Indiana asphalt mixture test results, combining RAS with WMA foaming technol-
ogy does not change the laboratory fatigue performance of HMA. An HMA sample with RAS
demonstrated similar fatigue properties as WMA with RAS. Additionally, the WMA-RAS and
HMA-RAS mixtures performed similar as the HMA-RAP sample.
ACKNOWLEDGEMENTS
The researchers acknowledge the support of Debra Haugen, Mathew Kirby at Iowa State University,
and Joe Schroer at the Missouri DOT. The research work was sponsored by Transportation Pooled
Fund (TPF) Program TPF-5(213). Partners providing funding to TPF-5(213) include the Missouri
DOT (the lead state), the Iowa DOT, the Minnesota DOT, the Indiana DOT, the Colorado DOT, the
Wisconsin DOT, the California DOT, the Illinois DOT, and the Federal Highway Administration.
Any opinions, findings, and conclusions or recommendations expressed in this material are those
of the writers and do not necessarily reflect the views of any agencies.
REFERENCES
AASHTO T321. 2007. Standard Method of Test for Determining the Fatigue Life of Compacted Hot-Mix
Asphalt (HMA) Subjected to Repeated Flexural Bending, Standard Specifications for Transportation
Materials and Method of Sampling and Testing, Washington, D.C.
Al-Qadi, I.L., Elseifi, M., Carpenter, S.H., 2007. Reclaimed Asphalt Pavement A Literature Review,
Research Report FHWA-ICT_07_001, Illinois Center for Transportation , University of Illinois at Urbana-
Champaign, Urbana, IL, 2007
Brock, B. November 12, 2007. Economics of RAS in HMA, Presentation at the 3rd Asphalt Shingle
Recycling Forum, Chicago, Illinois
Button, J., Williams., D., and Scherocman, J. 1996. Roofing Shingles and Toner in Asphalt Pavements, Texas
Transportation Institute, Research Report 1344-2F
Cleven, M.A., 2000. Investigation of the Properties of Carbon Fiber Modified Asphalt Mixtures, MS Thesis,
Michigan Technological University, Houghton, Michigan
EPA. 1998, From Roofs to Roads, NAHB Research Center
Ghuzlan, K.A., Carpenter, S.H. 2006. Fatigue damage analysis in asphalt concrete mixtures using the
dissipated energy approach, Canadian Journal of Civil Engineering, Volume 33, pp. 890901
Institute of Transportation Studies. 1994. University of California, Berkeley, Fatigue Response of Asphalt-
Aggregate Mixes. Report SHRP-A-404. Strategic Highway Research Program, National Research Council,
Washington D.C.
Johnson, E., Johnson, G., Dai, S., Linell, D., McGraw, J., Watson, M. 2010. Incorporation of Recycled
Asphalt Shingles in Hot Mixed Asphalt Pavement Mixtures, Minnesota Department of Transportation, St.
Paul, MN
Lee, J., Rust, J.P, Hamouda, H., Kim, Y.R., Borden, R.H. 2005. Fatigue Cracking Resistance of Fiber-
Reinforced Asphalt Concrete, Textile Research Journal, 75: 123
NCHRP Report 646. 2010. Validating the Fatigue Endurance Limit for Hot Mix Asphalt, Transportation
Research Board, National Highway Research Council, Washington D.C.
Putman, B.J., Amirkhanian, S.N., 2004, Utilization of waste fibers in stone matrix asphalt mixtures,
Resources, Conservation and Recycling , Vol. 42, 265274
Roberts, F.L., Kandhal, P.S., Brown, E.R., Lee, D. 1996. Hot Mix Asphalt Materials, Mixture Design, and
Construction, 2nd Ed. Nation Asphalt Pavement Association Research and Education Foundation, Lanham,
Maryland
Tangella, S.C.S.R., Craus, J., Deacon, J.A., Monismith, C.L., 1990. Summary Report on Fatigue Response
of Asphalt Mixtures, Prepared for Strategic Highway Research Program Project A-003-A
Wu, S., Ye, Q., Li, N. 2008. Investigation of rheological and fatigue properties of mixtures containing polyester
fibers, Construction and Building Materials, 22, 21112115
Ye, Q. & Wu, S. & Li, N. 2009. Investigation of the dynamic and fatigue properties of fiber-modified asphalt
mixtures, International Journal of Fatigue, doi: 10.1016/j.ijfatigue.2009.04.008
170
Flexural fatigue tests and predictions models tools to investigate

SMA mixes with new innovative binder stabilizers
J.B. Sousa
CONSULPAV Lda., Milharado Mafra, Portugal
I. Ishai
Department of Civil & Environmental Engineering, Technion, Haifa, Israel
G. Svechinsky
Consulting Engineer, Haifa, Israel
ABSTRACT: A new binder stabilizer was developed mainly to prevent excessive drainage of the
bitumen during hauling and placement of SMA mixes. This stabilizer is an activated fine-ground
raw silica mineral, which is the waste by-product of the mining for the Phosphate Industries. The
activation was aimed for obtaining unique thixotropic rheological mechanism for draindown pre-
vention. In comprehensive studies it was found that SMA mixes, combined with the new Activated
Mineral Binder Stabilizer AMBS (or iBind), exhibit low acceptable bitumen draindown values
which are comparable to those with the fibers. By systematic mix designs, European, American
and Israeli SMA mixes also show comparable and better mechanical properties. These results were
obtained with 0.5% less binder content and between 1020 C lower mixing temperatures, as com-
pared to the fibers. Thereafter, the main effort was to investigate if SMA mixes stabilized with
the new AMBS have the comparable fatigue life as Fiber SMA Mixes, despite the lower bitumen
content. This paper describes the summary of fatigue testing, including the characterization of
materials, production of fatigue beams, the 4PBB equipment used for fatigue testing and, finally,
the analysis and interpretation of the results.
1 INTRODUCTION
Due to problems associated with the addition of cellulose fibers to SMA mixes, a new binder
stabilizer was developed mainly to prevent excessive drainage of the bitumen during hauling and
placement of the mix. This stabilizer is an activated fine-ground raw silica mineral, which is
the waste by-product of the mining for the Phosphate Industries. The activation was aimed for
obtaining Thixotropic and Shear-Thinning properties for the bitumen, by which the mastic in the
mix possesses high viscosity at rest (hauling and after placing) for reducing draindown, and low
viscosity at motion (mixing and placing) for maintaining the proper workability.
In comprehensive laboratory studies it was found that SMA mixes, combined with the new
Activated Mineral Binder Stabilizer AMBS (or iBind), exhibit low acceptable bitumen draindown
values which are comparable to those with the fibers. By systematic mix designs, European,
American and Israeli SMA mixes also show comparable and better mechanical properties related
to: resistance to water damage, wear resistance, indirect tensile strength, and rutting resistance.
These results were obtained with 0.5% less binder content and between 1020 C lower mixing
temperatures, as compared to the fibers (CONSULPAV 2010a, Ishai et al 2011, Watson and Moore
2011).
171
Table 1. Granulometric composition of the SMA mixture.
Figure 1. South Carolina aggregate gradation limits for the SMA mixtures tested (in yellow).
Thereafter, the main effort was to investigate if SMA mixes stabilized with the new AMBS have
the comparable fatigue life as Fiber SMA Mixes, despite the lower bitumen content. It is argued
that the extra 0.5% binder in Fiber SMA mixes are absorbed and lost in the cellulose and provide
no additional benefit to the mix than preventing draindown during the construction phase.
This paper describes the summary of a research and fatigue testing performed at CONSULPAV
Laboratories (CONSULPAV 2010b), and it includes the characterization of materials, production
process and the compaction of asphalt mixes, the laboratory equipment used for testing of resistance
to fatigue and, finally, the interpretation of the results of the tests performed.
2 PRODUCTION OF SMA MIXTURES AND SAMPLES
2.1 Materials and gradation

All tests were performed on South Carolina SMA 12.5 Mixture composed of Benafessim
Montemor-o-novo aggregates and filler and Raio Maior hydrated lime as stripping additive. The
mix gradation is described in Table 1, and illustrated in Figure 1. The bitumen was supplied by
CEPSA on in March 2010. It graded, as per supplies certification at a Pen 35-50 and a PG 70-22
bitumen.
172
Figure 2. View of the compacting machine.
The bitumen stabilizers cellulose fibers and AMBS were added at a quantity of 0.3% of
the mixture. Fibers were added to the dry hot aggregates, while the AMBS was added after initial
mixing with the bitumen.
2.2 Mixture compaction

The methodology adopted in this study for the preparation of bituminous mixing and compaction
is based on the standard AASHTO PP3-94D (Standard Practice for Hot Mix Asphalt Preparing
Specimens by Means of the Rolling Wheel Compactor) and is summarized below:
1. According to the percentage of each sieve fraction, and taking into account the percentage of
voids to be achieved, the weight of each particle size fraction is calculated in order to reconstitute
gradation required.
2. The aggregate, as well as the bitumen is heated to a required mixing temperature.
3. Adequate mixing time is used for the bitumen to fully coat the entire surface of the aggregates.
For the type of aggregates and bitumen used in the mixing time it was around 4 minutes. This
task was performed in a 25 liters mixer.
4. After mixing the mixture was kept in oven at 135 C 3 for 4 hours 5 min. This operation,
called a cure, is a phase of short term aging performed according to standard AASHTO PP2-94
(Standard Practice for Short and Long Term Aging of Hot Mix Asphalt), which tries to simulation
the aging that takes place during plant mixing.
5. After four hours of oven aging the mix is again set to the compaction temperature. When
the mixing reaches this temperature is placed in a mold, the compression is performed using a
segmented steel plate (see Figure 2) to represent the rolling wheel compaction (a metal drum
173
Table 2. Dimensions of the beams extracted from the slabs for testing fatigue life.
rolling in static mode i.e. no vibration). The 750 kg on the machine to transmit an asphalt mix
similar to the pressure applied on site.
2.3 Cutting and preparation of samples for testing

After releasing the slabs from the mold prismatic specimens are cut with a diamond saw. The
prismatic specimens are approximately 5.1 cm in thickness (height) by 6.3 cm base (width) and
38.0 cm in length.
The four sawn beams are measured in height and width at three different points along their
length. The average of three values for each dimension are used and summarized in Table 2.
These geometric characteristics are introduced into the ATS testing program that controls the 4
point flexural bending testing fixture. These are fundamental for determination of the strain levels
applied during testing and for stiffness determination.
3 VOLUMETRIC PROPERTIES OF BEAM SAMPLES
The beam samples were tested for determination of specific gravity, according to the ASTM D 2726-
96a. The Theoretical Maximum Specific Gravity (max ) was determined on the mixture prepared in
the laboratory. The value of Max. Theoretical Specific Gravity was used to calculate the porosity
(% air voids) of the beams according to the following equation:
where:
Vv (%) percentage of porosity or air voids in the mix
mx theoretical maximum specific gravity in the mix (g/cm3 )
specific gravity of the beam (g/cm3 )
The values of specific gravity and porosities of the SC SMA mixtures with AMBS and Fibers
that were determined in the laboratory are summarized in Table 3. The same table presents also the
values of percentage of bitumen volume Vb , the percentage of voids in the mixture of aggregates
VMA, and the percentage of voids filled with asphalt mix of bitumen VFB (degree of saturation
in bitumen), respectively, with calculated values by the following equations:
174
Table 3. Specific gravity and porosity of the beams.
Figure 3. Global view of the servo-hydraulic equipment used in the tests.
4 EQUIPMENT FOR TESTING FATIGUE LIFE
There are several tests used to characterize the fatigue resistance of asphalt mixtures, which involve
a variety of testing techniques, types of equipment, configuration, type and mode of loading, test
conditions and analysis procedures. To study the mechanical behavior of bituminous mixtures, the
research team has adopted the latest and the most advanced technology in the area of fatigue tests,
as recommended by SHRP A-003A. This allows in particular conducting the test based on pure
bending prismatic beams where loads are applied at four points.
The fatigue machine is called a CS7800 and was manufactured in the USA. Figure 3 presents
an overview of the servo-hydraulic equipment, comprising a loading structure, a hydraulic power
175
Figure 4. Detailed view of the equipment of the 4 point bending inside the climatic chamber.
Figure 5. Schematic representation of the device 4 point bending.
system and a climate chamber installed in the laboratory, Figure 4 shows a detailed view of the
loading device in flexion. This device has a special free rotation and translation in the four load
points (Figure 5). This appears to be of extreme importance, since the flexing causes the deviation
of the axis of the specimen and consequent rotation along each load cycle.
The load structure comprises a vertical actuator connected to a servo valve that allows the
application of loads dynamic with frequencies of up to 20 Hz. At the end of the servo valve
is a load cell with capacity 25 kN. The application of this load is carried by a sinusoidal load, in
accordance with the Equation (5).
176
Figure 6. Simply supported beam.
The response of the material applied to the request results in a deformation that follows a law
characterized by the same frequency but with a lag to the load, according to Equation (6). These
delays are assigned the name of the phase angle ().
The environmental chamber allows the temperature control from 20 up to +70 C with an
accuracy of 0.5 C. This is essential to keep the temperature constant throughout the test. In this
work the tests were conducted at a frequency of 10 Hz and the default extension with controlled
deformation or magnitude of the extension held constant during the test. A 20 C temperature
adopted for the testing.
It should be mentioned that the main advantages of simple bending tests performed on four
points are:
1. Firstly it is considered that the bending test simulates simple four points in a fairly correct the
appearance of fatigue cracking due to the efforts to develop traction on the bottom of asphalt
layers;
2. These tests are well known and in widespread use and easy to understand;
3. In 4 point bending rupture of the specimen occurs in a zone of uniform bending, which helps
the reduction of the coefficient of variation of test results, requiring less testing than in other
types of testing to determine mix fatigue life;
4. The results can be used directly (with an appropriate correlation factor between laboratory
findings and behavior in service) in the structural design of pavements.
5 DEFINITION OF BASIC PARAMETERS
5.1 Basic configuration

The resolution of the mechanical system of simply supported beam (Figure 6), using the theory of
elasticity, allows the calculation of the different parameters needed for the interpretation of fatigue
tests, namely:
the maximum stress (t ) suffered by the beam in response to extension (t) imposed;
the modulus of rigidity (Stiffness Modulus) (E);
the phase angle ().
177
5.2 Tensile stress
The tension experienced by the beam, for each cycle is calculated from the following relationship:
where
t maximum tensile stress (kPa)
F load applied by the actuator (kN)
b average width of the beam (m)
H average height of the beam (m)
L total length of the beam = 0.340 m
5.3 Maximum extension traction

The calculation of the extension shall be made from Equation (8), which depends solely on the
geometric characteristics of the beam, the tax shift () and the positions of application of loads.
From this observation it is clear the importance of geometric dimensions of the samples (Table 2).
where:
t maximum extension traction (%);
maximum deformation at the center of the beam (m);
L length of the beam between the claws outside, 0.34 m;
a spacing between the claws = L/3 = 0.38/3 = 0.127 m.
Note that fatigue testing of an asphalt mix is rarely performed to the same level as the in
service extension in a real pavement. The realization of a single test at that level of extension
necessarily entails the loss of several days or weeks because it is usually too low. Then comes the
practice, universally accepted, that the fatigue laws are obtained at higher extension levels and then
the results extrapolated from the testing levels to in service levels. In this study, the determination
of the fatigue life for each material was done performing two tests at HIGH extension (700 m)
levels and two tests at LOW (300 m) extension levels, which makes a total of four beams for
each mixture.
5.4 Flexural stiffness modulus and phase angle

The stiffness of an asphalt mix, considering a number of load repetitions, is calculated as the
quotient of tensile stress t by extension t, is determines according to equation (9):
Since the behavior of asphalt is visco-elastic (time lag in response time) and the loading is
repeated in time (frequency). Thus the complex modulus is characterized by two components: a real,
representing the energy stored in the material, and an imaginary component which corresponds to
the energy lost by internal friction within the material. The complex modulus can also be submitted
by its value in the module and its phase angle (), respectively according to the equations:
178
This is the most common form of presentation of the mechanical properties of an asphalt mix,
considering a given temperature and a given frequency of application of loads.
The calculation of the phase angle is provided from the following expression:
being:
phase angle, in degrees ( );
f frequency of the load in Hz;
t time delay between max and max , in seconds.
6 TEST RESULTS AND ANALYSIS
6.1 Basic results

In order to facilitate analysis of bending tests results, conducted on the four beams sawn slab
molded in the laboratory, the methodology for the analysis of test results that have been used
in other studies of this nature was adopted (CONSULPAV 2010b). The same principle for the
projection of all parameters monitored and calculated by the software ATS in different charts
showing their progress based on the number of cycles of load application. Moreover, this type of
representation allows making the detection of any anomalies that occurred during the test, and
simultaneously assess the validity of the chosen parameters. The whole test was analyzed in this
way, from where they withdrew the parameters determining for the further analysis of the fatigue
performance of the different mixtures.
Parameter values 100 , E100 , 100 , 100 , Nf /50 , among others, are compiled in Table 4.
6.2 Flexural stiffness modulus

The flexural stiffness modulus is determined from the graph of stiffness (E) versus Load Cycles (n).
From Equation (8), which gives the modulus of rigidity of the beam in asphalt, given the scale, the
force applied (via the load cell) and the resulting displacement (LVTD), the program automatically
calculates the ATS module stiffness at each cycle.
The progress of the modulus of rigidity thus calculated over the number of cycles of sinusoidal
loading on the beams of the SMA asphalt mixtures is presented in Figure 7.
As expected, the stiffness modulus decreases with the number of load applications and has
different values depending weather the test is performed at a low imposed extension (300 106 )
Table 4. Summary of test results obtained from simple bending AASHTO TP8.
179
Figure 7. Degradation of stiffness modulus of the SMA mixes during the loading cycles.
or a high one (700 106 ). The decrease in stiffness with number of load applications can be
represented as an exponential function:
where:
E0 module of the initial stiffness (kPa) (constant);
e exponential basis e=2.718;
n number of cycles;
constant dependent on the nature of the mixture.
This relationship allows, on the other hand, to calculate the number of cycles needed for the
asphalt to reach failure condition, that is, by definition when the stiffness decreases to half
Eruin = E0/2.
The number of cycles is chosen because, firstly, it is considered that this number of repeated load
is low enough not to affect the mechanical characteristics of the beam, and secondly it is considered
that the control parameters of test machine are all satisfied.
Figure depicts the relationship established between the module and extension of initial stiffness
corresponding traction. This relationship is established of the form: E = ln() + .
Parameters and were determined to quantify this pattern of behavior of the mixture, with
respect to changes in the flexural modulus of rigidity depending on the strain amplitude on the base
of the beam (or applied to the bituminous pavement).
In Table 5 the evolution of stiffness with strain level is presented so that the determination
of expected stiffness levels at lower stains can be ascertained. The stiffness modules at strain
amplitudes of respectively 80, 100 and 120 m are extrapolates. In the absence of the value of the
extension considered in a particular project, it is considered that all three are chosen sufficiently low
so as to be representative of any extensions that may appear at the base of an actual well designed
pavement.
180
Table 5. Modulus values of stiffness of SC SMA AMBS and Fibers, extrapolated to different strain
amplitudes.
Figure 8. Evolution of flexural rigidity modulus of bituminous mixtures SC SMA AMBS and Fibers, in the
range of 80 m extensions up to 700 m.
Plotted in Figure 8 is the best fit for the stiffness modulus values function of strain level obtained
for both bituminous mixtures tested. It appears that none of the results straight too much away from
the best fit, which gives some assurance of the validity of the results.
With respect to certain modules of rigidity, it is clear that the SC AMBS mix (5.5% of bitumen
3550) had higher values compared to the asphalt SC Fibers (6.0% of bitumen 3550). Moreover
the values extrapolated stiffness modules are relatively identical (about 6900 MPa for the AMBS
(iBind) SMA mixes and about 6200 MPa for the Fiber SMA mixes).
7 MODELS OF FATIGUE LIFE FOR BITUMINOUS MIXTURES
By definition, the fatigue strength of bituminous mixtures is the ability they provide to withstand
repeated bending stresses, without reaching failure. The fatigue cracking is a major mechanism of
distress of pavements and is manifested by the appearance of cracking caused by repeated loading
181
of traffic. The quantification of the fatigue strength of bituminous mixtures is one of the criteria in
the prevalent rational design of pavements.
This chapter is devoted to the presentation of the methodology of fatigue resistance of asphalt
mixtures studied. The law of fatigue obtained experimentally in the laboratory is faced with the
models for predicting the fatigue life developed by various institutions, among which are the
model of the Shell and Asphalt Institute (AI). These laws have been developing trying to relate mix
properties with actual fatigue life.
The model developed by Shell relates the fatigue life with the level of extension and stiffness of
asphalt, and as shown in Equation (14) as proposed by Claessen et al (1977).
where:
Nf Fatigue resistance (Number of cycles to failure);
t Strain amplitude in mm;
Smix Stiffness of asphalt mix in Pa (obtained, for example, by fatigue machine or
FWD test);
Vb Percentage by volume of bitumen Vb = 1.03
Pb
Specific Gravity, with Pb : percentage of bitumen.
The model proposed by the Asphalt Institute (1981) relates the fatigue life with the extension
level, the stiffness of the asphalt mix and level of compaction of asphalt. This model also allows
the definition of a conversion factor to transform laboratory results in predictable values of fatigue
life in service, as stated in Equation (15). In the case of the Shell model, the conversion factor is
already included in the coefficients of the model itself.
Being:
Nf Fatigue resistance;
t Strain amplitude (mm);
Smix Stiffness of asphalt, in psi; (1 psi = 145 MPa)
VFB Voids filled with asphalt mix of bitumen;
Sf Shift Factor to convert laboratory results in predictable outcomes on the pavement. The
recommended value for this factor is 18.4 for 10% area cracked.
It is noted that the models used for comparison are those used in most current design manuals,
so reliably uncontested.
In the case of the Asphalt Institute model, consider two cases for the coefficient Shift Factor,
the first Sf = 18.4 (proposed by AI), and Sf = 1 second, because:
a. The specimens tested simulate a new pavement as it suffered only one stage short term aging
(135 3 C for 4 hours 5 min) to simulate the manufacture of the mixture in the drum of
the central and not the aging long-term (85 3 C for 120 h 30 min), according to standard
AASTHO PP2-94 (Practice for short term and long aging of hot mix asphalt)
b. The definition of fatigue, which leads to the appearance of cracks in-situ results from repeated
application of traffic loads and temperature variations, is different from the definition of fatigue
obtained in the laboratory.
With regard to the fatigue life determined by laboratory experiment, it is defined as the point
at which the modulus of rigidity of the specimen reaches a percentage of the initial stiffness and
not the appearance of cracks (the module at the end of the fatigue life is set equal to 50% of initial
value).
182
For the test in controlled extension, the relationship between Fatigue Resistance, the level of
extension and stiffness of asphalt can be expressed by Equation (16), which represents the model
developed by Monismith et al (1985).
The charge cycle in which failure occurs is obtained by substituting the value of N in the previous
equation solving or simply:
where:
Ef ,50 stiffness in the state of failure or by definition 50% of initial stiffness
Ef ,50/A 0.5, by definition
Thus, the values obtained in laboratory for the fatigue life of the mixture analyzed in this
laboratory study, and those predicted by the two models described are presented in graphical form
in Figure 9 and 10.
From the observation of the figures above it can be inferred that the prediction model, that
most closely matches the values measured in the laboratory, is that of the Asphalt Institute (AI)
with a Shift Factor equal to 18.4. As such, it is reasonable to use this fatigue model for further
comparisons.
As it well understood, it is impossible to make samples from different mixes that are absolutely
equal in terms of air void content and binder content, and that are tested at the exact same strain
level. As such, the comparisons presented in Figure 8 represent just a first approximation of the
actual relation of the fatigue performance of those two mixes. To be a step closer to the actual
relative performance it was decided to investigate the ratio between the actual fatigue life obtained
from the test and that predicted for the same beam (considering actual air void content, strain level,
stiffness module and binder content) using the AI formula (Equation 15) for each mix. Those results
are presented in Table 6.
It can be observed that the actual results from Fiber SMA mixes are in average 61% of those
expected by the Asphalt Institute model. These differences are acceptable, given that the actual
binder source plays an important role in the actual performance. However using the same binder type
with AMBS SMA mixes, it can be observed that the actual ration between observed in laboratory
and predicted by the model is 83% indicating a better relative performance then those obtained by
the Fiber SMA mixes. As such, the AMBS (iBind) SMA mixes appear to have a fatigue life about
38% better than those with Fibers.
More glaring conclusions can be observed using the Shell predictive model (Equation 14), as
seen in Table 7. Based on the Shell model, using the actual beam properties (i.e. binder content,
stiffness, etc) the model would expect those beams to behave like if they add about 0.6% less binder
content then they actually have. This indicates very clearly that from a fatigue view point the extra
0.5% placed in SMA Fiber mixes are wasted inside the cellulose fibers. On the other hand the
AMBS mixes appear to behave as if they have about 1% more binder content then they actually
have.
Same trends, with respect to the difference in fatigue life of AMBS mixes vs. Fibers mixes, were
obtained in recent tests performed in China on SMA mixes. Similar fatigue tests were performed on
compacted beams using the 4P bending beam equipment, and analyzed using the Asphalt Institute
183
Figure 9. Comparison between the fatigue life determined the laboratory and models provided by Shell and
Asphalt Institute for AMBS and Fibers.
model. In this research, in SMA mixes with 6.0% bitumen, 15% of the bitumen was replaced by
AMBS, thus the result was 5.1% bitumen only. The results and analysis of the tests are summarized
in Table 8.
As can be seen, the ratio of actual fatigue life obtained from AMBS mixes was systematically
higher than those predicted from Fiber mixes. The ratio is between 30% and 110% more fatigue
184
Figure 10. Comparison between the fatigue life of AMBS and Fibers as determined in the laboratory.
Table 6. Summary of results of the comparison between actual and asphalt Institute predicted fatigue
life of the beams.
life, depending on the strain level. As such, it can be concluded that even with 5.1% of actual
binder content, AMBS mixes out performed Fiber mixes with 6.1% bitumen. This is an additional
proof that a substantial amount of the bitumen is absorbed and wasted inside the cellulose fibers
in conventional SMA mixtures.
185
Table 7. Summary of results of the comparison between actual and SHELL predicted fatigue life
of the beams.
Table 8. Summary of results of the comparison between actual and asphalt institute predicted fatigue
life of SMA beams in recent Chinese research.
8 MAIN CONCLUSIONS
The main goal of this effort was to investigate if AMBS (iBind) SMA mixes have the comparable
fatigue life as Fiber SMA Mixes. This is particularly relevant because AMBS SMA mixes are
generally designed and constructed with about 0.5% less binder content than Fiber SMA mixes. It
186
is argued that the extra 0.5% binder in Fiber SMA mixes are absorbed and wasted in the cellulose
and provide no additional benefit to the mix are then preventing draindown during the production
and construction phases.
These results, although only based on 4 beams produced for each type of mix, appear to indicate
that AMBS mixes do indeed yield an equal or better fatigue life then Fiber SMA mixes, even though
they contain less binder than Fiber SMA mixes. Actually, the ratio of actual fatigue life from beams
in laboratory, and that obtained from Asphalt Institute fatigue model (using actual beam properties),
indicates that AMBS mixes have between 30110% better longevity than the SMA Fiber mixes
they attempt to replace (based on the binder types used in this research). Using the SHELL Model
the results also indicate that the Fiber SMA mixes have a fatigue life similar to mixes with about
less 0.6% binder content. The IBIND SMA mixes appear to behave in the SHELL fatigue model
as if they actually have about 1% more binder content then they actually have.
Obviously with slightly less binder content the stiffness modulus is slightly higher as expected.
Both the AI and the SHELL models are very sensitive to beam stiffness, as such the fact that
AMBS mixes have a higher fatigue life with a stiffer modulus is making them rank very well in
those models.
ACKNOWLEDGEMENTS
This paper summarizes a research work financed by Desert Silica Industries (DSI). The major
laboratory work was performed by CONSULPAV Consultores e Projectistas de Pavimentos,
LDa, Portugal. The author would like to thank Mr. Andrey Vorobiev and Mr. Ronen Peled of DSI
for their support and innovative advices. Also, for Ms. Rossana Sousa, Mr. Francisco Silva, and
Mr. Pedro Nobre of CONSULPAV for their devoted laboratory work and laboratory data analysis.
REFERENCES
Claessen, A., Sommer, E.J, & Uge, P. 1977. Asphalt Pavement Design The Shell Method. Proceedings, 4th
International Conference Structural Design of Asphalt Pavements, An Arbor, Michigan.
CONSULPAV Lda. August 2010a. Study of European SMA 11 and South Carolina SMA 12.5 with PG-76
Bitumen iBind vs. Fibers. A Study Realized for DSI Innovative Chemicals.
CONSULPAV Lda. December 2010b. Study of South Carolina SMA 12.5 with PG 76-22 Bitumen
Determination of Stiffness Module and Fatigue Life, iBind vs. Fibers. A Study Realized for DSI Innovative
Chemicals.
Ishai, I., Sousa, J.B. and Svechinsky, G. 2011. Activated Minerals as Binder Stabilizer in SMA Paving Mixtures.
Compendium, 90th Annual Meeting of the Transportation Research Board (TRB), Washington DC, January
2011.
Monismith, C.L, Epps, L.A. & Finn, F.N. 1985. Improved Asphalt Mix Design. Proceedings, Association of
Asphalt Paving Technologists, Annual Meeting, San Antonio, Texas.
The Asphalt Institute. 1981. Thickness Design Manual (MS-1), 9th Edition, College Park, Maryland.
Watson, D. E. and Moore, J.R. August 2011. Evaluation of SMA Mixture with iBind and Fibers NCAT Report
11-04, National Center for Asphalt Technology (NCAT) at Auburn University Alabama.
187
Laboratory evaluation of fatigue and flexural stiffness of

warm mix asphalt
A.N. Mbaraga, K.J. Jenkins & J. Van den Heever

Stellenbosch University, Stellenbosch, Western Cape, South Africa
ABSTRACT: The quest for sustainability, improved performance and durability has led to the
consideration of alternative material types in lieu of conventional types. The past decade has seen
growing interest in Warm Mix Asphalt (WMA) use. WMA uses various WMA technologies to
reduce Hot Mix Asphalt (HMA) production temperatures. Growing interest and use of WMA is
attributed to both technical and economic benefits. Benefits such as reduced fumes generation,
low fuel usage and worker health safety among others, has led to its consideration in lieu of HMA.
This paper provides insights into the relative fatigue performance criterion and flexural stiffness
of WMA. Fatigue is a major structural distress in asphalt concrete pavements. Flexural stiffness
is a mix property linked to mix response and load spread-ability. The experimentation considered
HMA as a reference mix type in the evaluation of WMA. Laboratory results and analysis showed
that WMA compares well with HMA.
1 INTRODUCTION
Fatigue is a major structural distress in asphalt concrete pavements attributed to accumulated

damage under repeated load applications. In asphalt concrete pavement layers, fatigue cracking
arises from repeated tensile strains due to traffic loading. Its occurrence is either characterized
as a result of high tensile strains at the bottom of the asphaltic concrete layer (i.e. in thin layered
pavements) or at the surface of the pavement structure (i.e. in thick layered pavements).
Flexural stiffness is mix property linked to mix response and load spread-ability which is influ-
enced by the mix components, loading time and temperature. Characterization of flexural stiffness
is typified by mix mechanical properties (i.e. mix cohesive and adhesive properties). Cohesive and
adhesive properties influence mixture stiffness. Mixture stiffness plays a vital role in the deter-
mination of fatigue behavior and is predominately influenced as well by the mix components,
temperature and loading time.
Material fatigue failure (i.e. damage under repeated loading) is a result of flexural cracking
attributed to continuous repeated load applications (Adhikari et al. 2009). In this regard, fatigue
typifies flexural cracking, an aspect that defines material response as a result of the repeated
loading. Under repeated loading, fatigue damage starts at a molecular level progressively resulting
into micro-cracks and later into macro-cracks. The macro-cracks progressively propagate to a
critical size through the pavement forming fatigue cracks (and other forms of pavement distresses).
The micro and macro cracking mechanism is primarily influenced by mechanical properties (i.e.
cohesive and adhesive properties), loading time (or frequency) and temperature.
The mechanical properties are dependent on the stability of the aggregate, void content and
asphalt binder contained (i.e. mix components). Stability of the asphalt binder is primarily
influenced by temperature and loading time. Temperature and loading time play a vital role in
the characterization of the mix flexural properties and fatigue behavior.
The procedures of determining the flexural properties and fatigue behavior of asphaltic con-
crete mix types remain of interest to mix design procedures and probable mix application. In
189
the laboratory, a number of mix types are evaluated for quality with respect to fatigue and crack
resistance, flexural and load spread-ability, among other pertinent aspects.
Laboratory testing of fatigue is based on the accumulated damage concept as developed by
Miner M. A. (1945). Miners hypothesis stipulates damage as an inverse of the number of load
repetitions to failure. In pavement engineering, researchers have tested asphalt concrete mixtures
and formulated equations that are used to estimate the number of load repetitions (Nf ) corre-
sponding to strain () or dissipated energy level (wo ). Equations 1 and 2 below illustrate such
fatigue models. A number of such equations are either based on single pulse loading or con-
tinuous sinusoidal loading. In pavement engineering, dissipated energy approximation, Wohler
approach and even fracture mechanics technique are the common fatigue damage models normally
considered.
where i , i = experimentally determined coefficients; wo = dissipated energy; Nf = fatigue life.

During testing with a 4-point bending beam apparatus, an asphalt concrete beam is subjected to
repeated flexure. The maximum flexural tensile strain occurs at the outer surface fibre of the beam
specimen. The bending beam test considers Wohlers fatigue relationship (see Equation 3) which
relates the number of load repetitions (i.e. to failure) to the tensile strain.
where = strain at the bottom of the asphalt concrete layer; k1 , n = factors depending on the
composition and properties of the asphalt concrete mix
Dissipated energy is another approach to fatigue characterization and is independent of the
loading conditions during the test. Dissipated energy is related to fracture parameters in fracture
mechanics. For a flexural beam fatigue test, the cumulative dissipated energy to failure is the area
under the curve between dissipated energy and number of cycles.
Van Dijk W. et al. (1977) formulated a cumulative dissipated energy model that allows results
from different test types and test conditions of different mix types to be described by a mix specific
relationship (Equation 4).
where WN = cumulative dissipated energy to failure; A, Z = experimentally determined

coefficients.
The model correlates cumulative dissipated energy to the number of cycles to failure. This model
allows the prediction of fatigue behavior in the laboratory over a wide range of conditions using
a few test data. A and Z coefficients are determined by fitting a power law regression function
with the test data on a log scale. This fatigue relationship is independent of the type of test method,
loading frequency, mode of loading, rest periods, temperature and bitumen type.
In the laboratory, several test methods can be applied for evaluating mix fatigue criterion and
flexural characteristics. The 4-point bending beam test allows the evaluation of fatigue and flex-
ural stiffness with a wide range of test inputs and parameters. This paper provides insights into
the fatigue characterization and flexural properties of WMA compared to their equivalent HMA.
The paper makes a comparative analysis between two selected fatigue models (i.e. Wohler-type
fatigue relationship and Van Dijks dissipated energy approach).
190
2 OBJECTIVES AND SCOPE
The objective of this research was to evaluate flexural stiffness and relative fatigue performance
of WMA. The experimentation considered equivalent HMA mix types (as reference mix types)
during the evaluation of WMA mix types. The research focused on the evaluation of surface mix
types with different WMA technologies, reclaimed asphalt (RA) contents, asphalt binder types and
modifiers.
This paper discusses flexural stiffness and fatigue behavior of WMA mix types in relation to
their equivalent HMA mix types. Furthermore, analysis of the mix types in terms flexural
properties, rutting potential and fatigue criteria is made.
Flexural stiffness test data is used in the development of stiffness master curves. Master curves
are tools used to interpolate and extrapolate test data beyond the set test ranges in order to allow
further characterization of material. Flexural stiffness master curves are developed for the analysis
of the mix types in terms fatigue properties and rutting susceptibility at a reference temperature
of 20 C.
Fatigue is analyzed on the basis of two notable fatigue approaches as to define mix fatigue
characteristics and behavior. A formulation of mix fatigue models derived from the regression
equations is listed against the mix types. This allows comparison to be made among mix types,
WMA technologies and recycled asphalt (RA) content (i.e. with the cognizance of the mix variables
entailed). The paper makes notable recommendations pertinent to future use and specification of
WMA in addition to further research in material analysis.
3 DESCRIPTION OF MIX TYPES
The experimentation comprised surface mix types with two asphalt binder types, two asphalt binder
modifiers, three WMA technologies and two RA contents (see Table 1).
Asphalt binders (i.e. 60/70 pen or 80/100 pen) were combined with the respective WMA tech-
nologies in the production of WMA mix types. The 80/100 pen asphalt binder was modified
with either an asphalt elastomer (A-E2) or plastomer (A-P1) modifier. A-P1 modifier constitutes
ethylene-vinyl-acetate (EVA) copolymers. EVA was combined with the chemical additive WMA
technology (Rediset). A-E2 constituted styrene-butadiene-styrene (SBS); SBS was combined with
an organic additive WMA technology (Sasobit) to form Sasoflex. The 60/70 pen asphalt binder was
unmodified; this was combined with the WMA technologies during WMA production. RA stock-
piles comprised RA1 (16 mm) and RA2 (8 mm) with 4.7% and 5.9% asphalt binder content
respectively.
4 DESCRIPTION OF MIX TYPES
HMA grading specifications were used for WMA production (see Figure 1). Figure 2 illustrates
RA stockpiles gradation used. 10% and 20% RA were considered for this experimentation. WMA
Table 1. WMA and HMA mix types.
HMA Surface mix (Control mix) WMA Surface mix (Trial mix)
10% RA 60/70 (Control 1) 10% RA 60/70 Chemical Additive

10% RA 60/70 (Control 1) 10% RA 60/70 Foam
10% RA 60/70 (Control 1) 10% RA 60/70 Organic Additive
20% RA 80/100 Plastomer (Control 2) 20% RA 80/100 Plastomer and Chemical Additive
20% RA 80/100 Elastomer (Control 3) 20% RA 80/100 Elastomer and Organic Additive
191
production was dependent on type required to incorporate WMA technology to the mix. Foam tech-
nology required a slight modification of the asphalt plant while chemical and organic additive WMA
technology types didnt necessitate such modification. However, it was apparent that understanding
technology application for production was vital; these were blended with the asphalt binder.
It is noteworthy that production and compaction temperatures were a crucial factor to monitor
and comply with. All HMA mix types were produced at and above 160 C and compacted at and
above 135 C (unmodified mixes) and 145 C (modified mixes). WMA mix types were produced at
125 C (unmodified mixes) and 135 C (modified mixes) and then, compacted at 110 C (unmodified
mixes) and 120 C (modified mixes). The production and compaction temperatures of the WMA
mix types were decided based on the basis of previous experience attained from 1st and 2nd WMA
trials as well as from the mix laboratory analysis and evaluation.
4.1 Authors and abstract

The preparation of the slabs entailed adherence to set procedures such as monitoring of compaction
temperature in addition to evaluation of mix quality and compaction (i.e. as an activity). Three
sample slabs per mix type were prepared from three production lots at the asphalt plant site. For
each production lot, three or more lots per mix were prepared for field trials. Each mix lot was
monitored in terms of its compliance with set production and compaction temperatures besides
quality. The mix quality assessment entailed aggregate coating and overall mix suitability; this was
also conducted at the plant site laboratory.
Compaction followed the modified laboratory compaction method and procedure by Mbaraga
(2010). The method makes use of a metal mould (floor mounted) and small sized two-drum roller
Figure 1. Mix types gradation and HMA grading specifications (upper and lower limits).
Figure 2. RA 1 (16 mm) and RA 2 (8 mm) gradation.
192
(Bomag 30) as compactor. The compaction procedure entails placement of mix at the extreme
edges (i.e. for the creation of the on-ramp and off-ramp) before placing in the metal mould (see
Figure 3). It is essential that mix comply with set compaction temperature.
The method simulates field compaction criteria and recommends conducting of trial-runs in
order to ascertain mix compactive effort. 35 passes simulating field criteria was set as compactive
effort for all the mix types after a series of compaction trial-runs based on HMA mix type. From
Figure 3, the procedure and compaction is illustrated; bitu-guard was used to prevent sticking of
the mix on the roller drums.The suitability of the mix was assessed at the asphalt plant laboratory
in terms of the mix quality and viability. It was vital that all WMA mix types comply with HMA
specifications. Cores were obtained from the compacted slab and evaluated for compactability,
binder content and other pertinent factors of influence to mix quality and probable performance.
Table 2 lists the averages of the maximum theoretical density (Max. Theo.Density), bulk relative
density (BRD) and binder content (together with the HMA binder content specifications ranges).
4.2 Compaction method vs. field compaction

The suitability of the compaction method was assessed by obtaining cores from the compacted
slab and comparing them to field compacted cores. Figure 4 illustrates as representative results
attained for the 20% RA 80/100 pen mix types. 4.0% void content was set as target void content
for experimentation.
It was apparent that void content was difficult to control at both field and with the use of
the compaction method. Some mix types compacted better (i.e. lower than 4.0% void content)
while other mix types were difficult to compact. This infers that some WMA technologies offer
better compaction aid than others; an appropriate compactive effort should be selected per WMA
Figure 3. Modified laboratory compaction method (Mbaraga, 2010).
Table 2. Mix particulars and properties.
Max. Binder Content (%)

Theo. Density BRD (Specifications:
Surface Mix Type (kg/m3 ) (kg/m3 ) Min-Max)
10% RA 60/70 Control 1 2.474 2.376 5.1 (5.05.6)

10% RA 60/70 Chemical Additive 2.481 2.368 5.2 (5.05.6)
10% RA 60/70 Foam 2.477 2.376 5.1 (5.05.6)
10% RA 60/70 Organic Additive 2.473 2.374 5.3 (5.05.6)
20% RA 80/100 Plastomer Control 2 2.488 2.378 4.8 (4.75.3)
20% RA 80/100 Plastomer and Chemical Additive 2.482 2.363 4.8 (4.75.3)
20% RA 80/100 Elastomer Control 3 2.486 2.380 5.1 (4.75.3)
20% RA 80/100 Elastomer and Organic Additive 2.467 2.369 5.1 (4.75.3)
193
Figure 4. Field Compaction versus Compaction Method.
Figure 5. Beam specimens void content ranges.
technology in this regard. Cognizance of the void content variation was considered in the evaluation
of the mix types.
4.3 Beam specimens

The compaction method recommends that the middle zone of the over slab be selected for use in
the manufacture of beam specimens. The middle zone was sawn into beams using a diamond saw.
Dry sawing was preferred to wet. Four beams (measuring approximately 400 mm length 63 mm
width 50 mm height) were obtained per slab. The beams were also assessed for compactability.
Figure 5 lists (as representative) results of 10% RA 60/70 pen and 20% RA 80/100 elastomer mix
types.
It is illustrative that the 10% RA 60/70 organic additive required more compactive effort com-
pared to the rest of the mix types. Most beams portrayed a void content range at and about the
set target. The beam height is a crucial factor of influence to fatigue testing; refinement of beam
height and even width was essential. A fine-toothed blade that allows sawing off about 2 mm from
the specimen was used to fine-tune the beam height and width. Later, the sawn specimens were
stored in a controlled temperature room set at 10 C.
194
5 LABORATORY TESTING AND DATA ANALYSIS
The experimentation made use of the pneumatic 4-point bending beam apparatus for the evaluation
of fatigue and flexural stiffness testing. Fatigue testing was conducted at a test temperature of
50 C and a frequency of 10 Hz using sinusoidal strain (as loading mode) as stipulated in the
interim guideline for the design of HMA in South Africa (2001). The experimentation considered
300 (peak to peak) or 150 (peak) as benchmark strain. The extra strain regimes were selected
depending on the number of cycles attained at 300 . According to the interim guideline, low
strain regime range from 180 to 230 , medium strain regime range from 320 to 370 and
high strain regime range from 380 to 430 . The experimentation considered 50% reduction
of initial flexural stiffness as mix fatigue failure. The initial flexural stiffness was recorded after
50 cycles from the start of the test.
Test data acquired was used for the development of the log-log scale fatigue graphs. Tables 3
and 4 list the fatigue test data acquired. A total of nine beams (of known void content and height)
were tested for fatigue at three different strain levels (i.e. three beams per strain level).
Tables 3 and 4, list the initial flexural stiffness recorded, applied strain, average phase angle,
number of cycles and cumulative dissipated energy attained. The interpretation of data takes cog-
nizance of the variation in void content. It was noteworthy that high initial stiffness didnt infer
good fatigue life. It was deduced that other factors other than mix stiffness (i.e. cohesive and
adhesive properties) influenced mix fatigue life.
Figures 6, 7, 8 and 9 illustrate fatigue life based on the two fatigue approaches. Figures 6 and 8
are based on fatigue life as related to tensile strain. Figures 7 and 9 illustrate fatigue life as stipulated
by van Dijk et al. (1977). Analysis of the 10% RA 60/70 pen mix types (see Figures 6 and 7), it was
observed that all WMA mix types performed better than (and even equal to) their equivalent HMA
mix type (i.e. reference mix). From the tensile strain approach as well as the dissipated energy
approach, it is illustrative that the WMA mix types have a higher relative fatigue performance than
their equivalent HMA mix type.
A comparison of WMA technologies from the two fatigue approaches shows that WMA foam
has a higher overall relative fatigue performance than the rest of the 10% RA 60/70 mix types.
WMA chemical additive performed slightly similar to HMA mix type. Regarding mix sensitivity
to higher strain regimes, all 10% RA 60/70 mix types showed similar trend to high strain regime
sensitivity as their equivalent HMA mix type except for the WMA organic additive (see Figure 6).
The WMA organic additive was less sensitive to higher strain regimes but with lower fatigue life
compared to WMA Foam mix type.
A comparison of the fatigue approaches typifies a similarity in defining mix fatigue performance
criterion but a difference in terms of reliability of data. The dissipated approach provides a better
reliability of fatigue data points as exemplified by the high R-Sq.
Figures 8 and 9 illustrate fatigue life for the 20% RA 80/100 pen mix types in terms of tensile
strain and cumulative dissipated energy respectively. With the elastomer mix types, HMA elastomer
mix type is less sensitive to high strain regimes compared to WMA elastomer mix type (see Figure 8).
The more sensitivity to higher strain regimes exemplified by the WMA elastomer could be attributed
to use of WMA technology and production temperature. However, the WMA elastomer showed
a higher general fatigue life compared to its equivalent HMA mix type. Additionally, the HMA
elastomer illustrates lower R-sq. compared to WMA elastomer (see Figure 8); this could suggest that
other variables do influence the fatigue-tensile strain relationship. From Figure 9, less dissipated
energy was required to fatigue HMA elastomer to failure as compared to WMA elastomer. This
infers that the less aged binder in WMA elastomer (i.e. due to lower production temperatures) is
responsible for its improved fatigue life.
Regarding 20% RA 80/100 pen plastomer mix types, WMA plastomer shows lower general
fatigue life compared to HMA plastomer. It was illustrative the WMA plastomer was more sensitive
to higher strain regimes compared to the equivalent HMA mix type. It is noteworthy that the HMA
plastomer registered a lower R-Sq. value compared to WMA plastomer, (see Figure 8). This is
attributed to the fact that the model does not incorporate other variables within its mathematical
195
Table 3. 10% RA 60/70 mix types.
Applied Initial flexural Cumulative Av. Phase

10% RA 60/70 Strain Stiffness Number of Dissipated Angle
Mix Types () (MPa) Cycles Energy (MPa) (Degrees)
Control 1 300 15,215 105,890 11.021 6.227

300 17,114 203,060 24.357 6.372
300 15,868 171,850 23.028 8.201
230 15,808 925,500 55.062 5.764
230 15,956 403,260 29.492 7.538
230 15,887 627,700 40.159 6.304
180 16,650 2,470,600 102.396 4.517
180 14,447 1,696,780 66.916 5.800
180 16,255 3,460,730 134.309 5.992
Chemical 300 19,668 486,480 45.966 4.955
Additive 300 17,921 243,340 29.418 7.349
300 12,888 103,050 12.987 10.751
230
230 16,201 1,174,010 98.24 5.897
230 12,140 224,340 16.005 9.459
180 12,273 2,139,310 89.847 7.690
180 17,362 1,864,760 101.218 6.740
180 19,852 2,439,310 126.053 5.237
Foam 380 16,534 165,740 25.083 5.519
380 14,351 105,110 15.953 6.634
380 14,970 110,460 17.113 7.911
300 14,408 275,740 30.960 6.177
300 16,195 461,980 52.419 6.422
300 16,466 291,150 36.241 6.658
230 17,096 1,143,730 90.683 5.585
230 16,503 1,652,930 114.433 5.326
230 16,116 1,188,930 89.693 5.075
Organic 300 14,755 208,350 21.998 4.578
Additive 300 14,065 1,922,080 131.411 6.943
300 17,131 155,900 20.610 8.873
230 13,693 543,840 36.058 5.159
230 14,219 451,330 54.133 7.437
230 16,275 325,420 24.582 6.109
180 14,527 1,854,390 72.573 5.109
180 13,113 3,800,000 138.381 5.310
180 17,210 1,374,170 64.898 5.985
Note: indicates extrapolated fatigue point and Av. = Average Phase Angle
relationship. A similar implication is notable regarding the tensile strain relationship; lower R-Sq.
values are attained compared to the dissipated energy approach.
On the basis of the regression analysis, the cumulative dissipated energy versus number of load
cycles provided higher R-sq. values compared to the strain versus number of cycles approach.
This infers that either more test data points or incorporation of other variables pertinent to mix
fatigue performance are required when considering the tensile strain approach fatigue model. The
cumulative dissipated energy approach requires a few data points.
This experimentation considered controlled-strain in its fatigue testing. With controlled strain the
number of cycles has a greater increase as compared to controlled-stress. Dissipated energy predicts
196
Table 4. 20% RA 80/100 mix types.
20% RA 80/100 Applied Initial flexural Cumulative Av. Phase

Modified Strain Stiffness Number of Dissipated Angle
Mix Types () (MPa) Cycles Energy (MPa) (Degrees)
Control 2 380 16,329 1,596,380 155.026 4.336

Plastomer 380 16,835 315,240 49.471 4.947
380 11,563 830,290 104.545 5.925
300 16,808 493,380 59.140 5.384
300 16,466 1,555,010 154.33 5.497
300 16,567 661,690 72.667 5.953
250 18,240 3,800,000 314.092 5.661
250 16,787 1,032,700 94.066 5.383
250 19,238 4,200,000 329.01 4.753
Plastomer 380 17,207 674,300 79.660 6.289
Chemical 380 15,953 212,340 30.425 5.438
Additive 380 16,191 256,900 37.578 6.941
300 18,660 917,760 94.135 4.692
300 18,501 283,940 31.804 5.276
300 15,009 694,040 89.990 6.598
230 17,104 1,895,080 147.162 4.982
230 15,304 1,469,070 116.225 4.861
230 15,208 1,775,900 129.097 4.626
Control 3 380 13,041 49,530 9.699 9.191
Elastomer 380 13,648 41,710 8.423 8.309
380 11,593 29,730 5.974 9.105
300 11,097 134,450 13.148 9.396
300 11,509 91,040 10.172 8.791
300 19,393 870,960 95.701 6.045
250 17,864 3,234,640 281.234 5.288
250 13,192 579,680 48.739 7.779
250 12,671 622,260 51.374 8.686
Elastomer 380 14,784 223,880 27.59 7.100
Organic 380 15,626 143,680 26.769 7.140
Additive 380 14,300 90,080 16.275 9.028
300 14,131 406,590 50.632 8.478
300 17,864 421,390 58.984 8.538
300 18,076 295,400 38.393 6.264
230 15,070 1,493,490 120.885 7.074
230 16,945 1,893,000 167.271 6.957
230 16,447 1,891,530 112.913 5.084
Note: indicates an extrapolated point the cumulative dissipated energy result is an interpolated factor
fatigue behavior of mixtures in the laboratory. For this experimentation, cumulative dissipated
energy approach is considered for mix fatigue life comparison.
5.1 Fatigue mix models

The fatigue behavior of a specific mix is characterized by a slope of cumulative dissipated energy
versus number of load repetitions to failure. Table 5 lists fatigue models of the various mix types
under evaluation. The experimentally determined coefficients (i.e. A and Z) typify mixture quality
as well as fatigue properties. The R-Sq. values are also listed against the mix types. It can be
197
Figure 6. Strain Nf (10% RA 60/70 mix types).
Figure 7. Cumulative dissipated energy versus Nf (10% RA 60/70 mix types).
assessed that unlike with the tensile strain versus number of load repetitions, the R-Sq. value is of
a higher value.
5.2 Flexural stiffness test data and analysis of results

Flexural stiffness testing was conducted at one strain regime (i.e. 300 ) for 300 cycles per test
temperature at different frequency sweeps. Test temperature ranged from 50 C to 250 C (at an
interval of 50 C) and frequency sweeps of 0.5 Hz, 1 Hz, 2 Hz, 5 Hz and 10 Hz were applied per
198
Figure 8. Strain versus Nf (20% RA 80/100 modified mix types)
Figure 9. Cumulative dissipated energy versus Nf (20% RA 80/100 modified mix types)
beam for 300 cycles. Two beams with a void content proximity to 4.0% were tested for flexural
stiffness. Test data acquired was used for the development of flexural stiffness master curves.
Flexural stiffness test data was used in the development of a master curve at a reference
temperature of 200 C. This allowed the extrapolation of test data beyond the set test conditions.
The Arrhenius equation was applied for the superposition and extrapolation of the test data at the
reference temperature of 200 C. Figures 10 and 11 illustrate flexural stiffness master curves for
10% RA 60/70 pen and 20% RA 80/100 pen mix types respectively.
From Figure 10, it is deduced that 10% RA 60/70 pen unmodified WMA mix types are more
prone to rutting than the equivalent HMA mix type at a reference temperature of 200 C. This could
be attributed to the low production temperatures that results in a less aged asphalt binder than with
HMA mix type. In terms of fatigue criterion, all mix types illustrated similar fatigue criteria as their
equivalent HMA. From Figure 11, WMA elastomer mix type showed a lower flexural stiffness than
199
Table 5. Fatigue Models cumulative dissipated energy approach.
Fatigue Models (Dissipated Energy Approach)
Mix Type A Z R-Sq. WN = A (Nf )Z
10% RA 60/70 HMA Control 0.0095 0.6292 0.967 WN = 0.0095(Nf )0.6292

10% RA 60/70 Chemical Additive 0.0034 0.7159 0.942 WN = 0.0034(Nf )0.7159
10% RA 60/70 Organic Additive 0.0109 0.6249 0.994 WN = 0.0109(Nf )0.6249
10% RA 60/70 Foam 0.0050 0.7023 0.904 WN = 0.0050(Nf )0.7023
20% RA 80/100 Plastomer Control 0.0026 0.7698 0.983 WN = 0.0026(Nf )0.7698
20% RA 80/100 WMA Plastomer 0.0048 0.7159 0.963 WN = 0.0048(Nf )0.7159
20% RA 80/100 Elastomer Control 0.0014 0.8022 0.973 WN = 0.0014(Nf )0.8022
20% RA 80/100 WMA Elastomer 0.0066 0.6902 0.970 WN = 0.0066(Nf )0.6902
Figure 10. Master curves for 10% RA 60/70 mix types.
its equivalent HMA mix type. However, the WMA plastomer compared well with its equivalent
HMA mix type; this could be attributed to the plastomer.
With the plastomer modified mixes, there was no observable indication to rutting potential. How-
ever with the unmodified mix types and elastomer modified, this was clearly illustrated. This infers
that asphalt binder modification (i.e. with plastomer modifier) and the WMA technology improve
mix susceptibility to rutting even with 20% RA content and soft asphalt binder (80/100 pen).
6 RECOMMENDATIONS AND CONCLUSIONS
The increased need to conform to sustainable practices will demand future practice to adopt sustain-
able materials. The growing interest in WMA shall necessitate the need to accrue reliable data and
research regarding WMA application in addition to specifications. Laboratory characterization of
alternative material types such as WMA shall necessitate realistic approaches to assessing material
property and response to loading or performance criterion as well as environment. However, it is
200
Figure 11. Master curves for 20% RA 80/100 mix types.
of importance that analyzed test data acquired from testing the alternative material be of better or
equal performance criterion and of acceptable quality when compared to equivalent conventional
types. Additionally, test equipment and methodological procedure shall require simulation of field
criteria and practice.
WMA fatigue and flexural stiffness were tested using the 4-point beam apparatus. Specimens
were manufactured simulating field criteria. WMA mix types compared well with their equivalent
HMA mix types. Two fatigue approaches were applied to evaluate mix fatigue performance crite-
rion. Mix based fatigue models following Van Dijks formulation were stated; the experimentally
determined coefficient shown that mix types differed. Additionally, the development of master
curves based on the Arrhenius equation at a reference temperature of 200 C showed that, WMA
mix types were generally more prone to rutting than their equivalent HMA mix types. However,
with binder modification, the rutting potential in WMA can be counteract. This infers that use
of asphalt modifiers is likely to improve mix quality and probable performance. As expected,
the plastomer mix types recorded higher flexural stiffness compared to elastomer types. How-
ever, the elastomer mix types portrayed the elastic character exemplified by the slope of the
fatigue line.
Furthermore, use of asphalt modifiers such as EVA and SBS with WMA technologies is possible.
The low production temperature attributed to WMA production does make the mix susceptible to
rutting. However use of enhancers or mix modifiers could improve mix quality. It is recommended
that further analysis be undertaken to assess rutting potential and use of modifiers or mix and
property enhancers.
7 FUTURE RESEARCH
Further research study is required to correlate laboratory WMA fatigue models to field or in-service
fatigue life. Future research could focus on the development of WMA fatigue transfer functions
based on laboratory WMA testing and evaluation. The testing and evaluation could encompass
more WMA technology types, higher percentages of RA contents and different asphalt binder
types.
201
Rheological analysis on asphalt concrete mix and asphalt binder is required. This could encom-
pass the evaluation of WMA technologies in terms of binder type or grade, use of asphalt modifiers
as well as chemistry of the mix recipes. Such study will provide information on the feasibility of
WMA technology use relative to mix and binder type.
Additionally, mix rheological analysis at varying RA contents using different WMA technologies
shall initiate in-depth research analysis of the varying parameters in terms of performance. Such
research could encompass the WMA technology efficiency measurement in terms of wettability,
compactability and mechanical properties to note but a few. This analysis will initiate a break
through into the correlation of WMA technology use, RA content, binder type and applicability
as well as performance.
Lastly, an integrated of fatigue analysis based on fatigue models and fracture mechanics is vital
to characterize future alternative material types. It is the view of the authors that use of linear
cumulative functions does not fully account fatigue distress. Fatigue distress follows a distinctly
non-linear path. In this regard, it is believed that the sequence of loading exemplified by the linear
cumulative function does not typify the propagation of a crack within the material under load.
ACKNOWLEDGEMENT
The research was sponsored by the South African Roads Federation (SARF) through the Depart-
ment of Civil Engineering and the SANRAL chair at the University of Stellenbosch. The authors
appreciate the involvement of the Warm Interest Group in particular Krishna Naidoo, Tony Lewis
and Wyand Nortje. This research couldnt have been completed without the significant contribu-
tion of the parties mentioned and the authorities at the Institute for Transport Technology (ITT) at
University of Stellenbosch.
REFERENCES
Adhikari S., Shen S. and Zhanping Y. 2009 Evaluation of Fatigue Models of Hot-Mix Asphalt through
Laboratory. Transportation Research Record, Journal of the Transportation Research Board, No. 2127,
Transportation Research Board of the National Academies, Washington D.C., pp. 3642 .
MbaragaA.N. 2010 Master Curves and FatigueTesting ofWarm MixAsphalt, MastersThesis, Civil Engineering
Department, University of Stellenbosch, South Africa.
Medani T.O. 1999 A Simplified Procedure for the Estimation of the Fatigue and Crack Growth Characteristics
of Asphaltic Mixes, Masters Thesis, International Institute of Infrastructural, Hydraulics and Environmental
Engineering, Deft, Netherlands.
Miner M.A. 1945 Cumulative Damage in Fatigue. Journal of Applied Mechanics, Vol. 12, No. 9, 1945,
pp. A159A164.
Taute A., Verhaeghe B.M.J.A. and Visser A.T. 2001Interim Guidelines for the Design of Hot Mix Asphalt in
South Africa. Pretoria South Africa.
Van Dijk W. and Visser W. 1977 The Energy Approach to Fatigue for Pavement Design. Proceedings Association
of Asphalt Pavement Technologists San Antonio, TX, USA.
202
Comparisons with other tests
Comparison of fatigue properties using 2-point and 4-point bending

tests Czech experience
P. Hyzl & M. Varaus

Brno University of Technology, Veveri, Brno, Czech Republic
P. Mondschein, J. Valentin & V. Soucek

Czech Technical University, Thakurova, Prague, Czech Republic
ABSTRACT: In this paper two different methodologies of fatigue testing as used in the Czech
Republic have been compared. Fatigue tests on selected Asphalt Concrete mixtures (AC) have been
done using by two Point Bending (2PB) and four Point Bending tests (4PB) using trapezoidal and
prismatic beams respectively. The methodologies of the above stated tests are known for several
years. The fatigue characteristics of experimental asphalt mixes have been compared. However,
the interpretation of the test results according to the EN 12697-24 standard is according to Czech
experience still not completely clear. The paper takes over the results from the 2nd Workshop on
Four Point Bending, which took place in Portugal, where stiffness modules were compared on the
same mixes.
1 INTRODUCTION
In the Czech Republic three test methods are used for the determination of fatigue characteristics.
However, only two of these test methods fulfill the conditions and specifications of the accepted
procedures according to EN 12697-24. For this investigations the same devices were used which
were applied for the assessment of stiffness modulus on the same mixes as presented e.g. in (Varaus,
2009).
Fatigue life in this research has been defined as the number of load repetitions Nf ,50 at which the
stiffness modulus has decreased to half its initial value. The initial value is defined as the stiffness
modulus in the 100th cycle. This is a standard approach applied in the Czech Republic.
Fatigue resistance, as specified in the Czech Republic, is the strain 6 at which the fatigue life
Nf ,50 equals 106 load repetitions. In order to get a fair or good estimation of this strain value at
least 3 strain levels should be tested. One strain level should lead to an Nf ,50 value of more than
106 repetitions in order to avoid extrapolations (Novotn, 2005).
Also at each strain level at least 6 specimens should be tested to get equivalent repeatability of
the test. The mode of loading in the fatigue tests is controlled deflection. It is adopted that during
the fatigue test this implies more or less a constant strain. With respect to assessed asphalt mixes
and earlier determined values of stiffness modules at 15 C this paper follows the results presented
in (Soucek, 2008) and (Varaus, 2009).
The standard test conditions used in the Czech Republic for 2PB test is 10 C. For 4PB test
usually 20 C or 30 C are applied, whereas the temperature range for dynamic modulus testing is
between 0 C and 40 C. The test conditions given actually according to Czech standards for fatigue
testing are described in the paragraph 3. Generally sinusoidal signal is applied and predominately
the loading mode with constant deflection (strain) is preferred.
In the past several research projects and comparative studies have been done. Project focused
on asphalt fatigue assessment and improvement of used methodology has been finished at CTU
205
Prague (Novotn, 2005). The 2 PB test method has been compared for various temperature (10
20 C) and frequency conditions (515 Hz) with ITT fatigue test used predominately in the Czech
Republic until 2005. Within RILEM an interlaboratory testing has been organized to compare 11
different test methods for utilizing and assessment of fatigue characteristics (Di Benedetto, 2004).
Stress and strain modes have been applied always with same conditions for used frequency (10 Hz)
and temperature (10 C). 2PB and 4PB test method has been used and compared also during the
research done in the Netherlands which was based on the dissipated energy concept (Pronk, 1997).
Fatigue tests have been done in strain controlled mode. Temperature of 20 C and similar frequency
has been applied for both methods (29.3 Hz for 4PB and 30 Hz for 2PB). The set strain levels were
in the interval of 100170 microstrain which does not correspond to the levels used for testing
described by this paper.
2 SPECIFICATIONS OF FATIGUE CHARACTERISTICS IN THE CZECH REPUBLIC
In the Czech Republic the results of fatigue tests (, Nf ,50 ) are plotted in a Whler diagram with
Nf ,50 on the horizontal axis and the applied strain on the vertical axis. Instead of strain the initial
load (for controlled deflection mode) or the applied load (for controlled force mode) is used in
the Whler diagram. However, outside the Czech Republic (and for 2PB and 4PB tests) it is more
common to plot the (initial) strain on the horizontal axis and the fatigue life on the vertical axis.
For some mixes (e.g. cold recycling, high stiffness asphalt concrete mixes) it is recommended to
apply four strain levels instead of three.
The results of the fatigue tests in a Whler diagram (log-log scale) can be characterized or fitted
by a straight line of which the formula is given by equation 1.
The parameter B represents the slope of the straight line in the Whler diagram.
3 USED FATIGUE TEST METHOD IN THE CZECH REPUBLIC
3.1 General
In the Czech Republic currently only two laboratories are able to perform fatigue tests and related
fatigue behavior assessment of asphalt mixes. The Czech Technical University in Prague (CTU)
is equipped with pneumatic test device for four-point beam test (4PB). In parallel it is possible to
use Nottingham Asphalt Tester (NAT) where quasi-fatigue test ITFT (indirect tensile fatigue test)
can be done. This test method has been used for several years as The second equipment for fatigue
testing is located at Brno Institute of Technology (TU Brno), where test device for two-point test
(2PB) is available. TU Brno at the same time developed modified test equipment similar to NAT,
nevertheless none fatigue or quasi-fatigue tests including the ITFT are done on this test apparatus.
Because of the testing history in the Czech Republic and the application of ITFT and 2PB tests
in the past some comparison studies were done during last 15 years to assess correlations between
2PB-TR and ITFT test methods. These tests confirmed relatively good applicability of the ITFT
method. Nevertheless based on ongoing research and recent findings (e.g. Di Benedetto 2004)
today test methods compared in this paper are preferred.
The objective of comparative testing executed between both laboratories was to assess results of
fatigue characteristics on same asphalt mixes gained by different test method and calculated on the
same way. The used asphalt mixes represent well different applications in the pavement structure
as well as the influence of different mix grading and used binder. Mixes ACO 11S, ACL16+
and ACP22+ have been used and are described in Table 1. These asphalt mixes can be basically
206
Table 1. Basic description of evaluated AC mixes with binder.
Asphalt concrete type Denomination Bituminous binder
AC for surface course AC 11S PMB 45/80 50

AC for binder course AC 16+ 50/70
AC for base course ACP 22+ 50/70
considered as typical representatives for flexible pavements as are applied in the Czech Republic.
There is only one exception which is related to none comparison of specific high stiffness asphalt
concrete mixes (VMT).
To explain the denomination characteristics S and +, these are used according to the Czech
national annex to distinguish different quality classes of asphalt mixes according to expected level
of deterioration and designed traffic intensities. For the abbreviation can be stated that S stays
for premium mix, where cylindrical test specimens are compacted by 2 75 Marshall blows and
+ stays for conventional higher quality mix where 2 50 Marshall blows are required.
3.2 Two-point test 2PB (TU Brno)

Innovated Czech pavement design method, which is given in the specifications TP170 of the
Ministry of Transportation, already respects performance based assessment of stiffness and fatigue
characteristics for bitumen bound materials according to standards EN 12697-26 and EN 12697-24.
In the case of fatigue test the characteristics are determined at 15 C and with frequency of 10 Hz.
3.3 Four-point beam test 4PB (CTU Prague)

Development in use of this test method can be followed in the Czech Republic during last four
years. The test is performed on different test conditions than are specified for two-point test. The
test temperature is according to CSN EN 12 697-24 set at 20 C, nevertheless national appendix
of standard CSN EN 13108-20 even prescribes 30 C. The reason for this difference was never
logically explained and will be changed during the periodical review of technical standards and
use of national parameters. Generally it has been found during previous tests done at the CTU in
Prague, that reaching reasonable fatigue characteristics for temperature of 30 C was more or less
impossible. At the same time clams in many cases crowded into the asphalt specimen declining
the test results. Increased test temperature defined in the CSN EN 13108-20 might be related to
initial testing done in the past for different temperature levels and frequencies. The frequency used
is according to the national appendix of CSN EN 13108-20 defined by 30 Hz, what has been found
during last few years as a better parameter than lower frequency like for stiffness.
Nevertheless because of stiffness modules gained previously for the same mixes at 20 C the
final conditions of the fatigue testing were 20 C and 30 Hz. The reason for this modification of set
conditions was in possible and easier comparison of results between fatigue and stiffness testing.
For the boundary conditions always at least 3 different strain levels have been selected for each
asphalt mix which varied in the range 100500 microstrain depending on the mix type. There was
always one strain level at which the asphalt mix specimens reached more than 1 mil. At the same
time an individual target has been set to reach for each strain level at least 10,000 cycles. Therefore
the minimum and maximum level varies for each mix resulting in at list 1 mil. cycles for the lowest
level, at least 10,000 cycles for the highest level and then a value set in the middle of the min-max
interval. Loading cycles before the stiffness level decreased below 50%. The test has been always
done on at least 18 specimens.
In case of the four-point beam fatigue test, as available at the CTU in Prague, prismatic specimens
with dimension of 50 50 400 mm are used. These specimens are exposed to a repeated bending
207
Figure 1. Test device for fatigue test according to the 2PB-TR method in EN 12697-24.
Figure 2. Test device for fatigue test according to the 4PB-PR method in EN 12697-24.
in four points with expected free rotation and horizontal shift in all loading points and supports. In
the modeling one clamp is always defined as fixed to secure the balance of the testing scheme. The
bending is caused by the motion of central loading points in vertical direction which is orthogonal to
the horizontal axis of the specimen. The vertical position of both end-points stays unchanged. The
activated repeated strain is symmetric on both sides. At the same time it is necessary to control, that
the sinusoidal amplitude stays constant in time. Sinusoidal strain signal is applied. The specimen
is loaded by defined loading and frequency, which can be changed similarly like the strain which
should be reached during the repeated cycles.
For odd asphalt mixes following intervals of strain levels were applied: ACO 11S 250450 m/m,
ACL 16+ 150500 m/m and ACP 22+ 110250 m/m.
4 EXPERIMENTAL RESULTS
Tests have been performed according to the settings described in the previous paragraph. In the
Figures 3 and 4 results of 2PB-TR test method is shown. The testing was executed always on 15
specimens with different strain levels, which does not correspond to a standard test procedure
which is usually applied according to EN 12 697-24. The accuracy of the test, as following the
standardized Czech procedure, is further affected by none repeatability for each of the 15 strain
level. Generally it might be statistically correct to have only 3 strain levels with at least 5 repetitions
208
Figure 3. Whlers diagram for ACL 16+ mix, 2PB-TR test; (R2 = 0.8789).
Figure 4. Whlers diagram for ACP 22+ mix, 2PB-TR test; (R2 = 0.8789).
209
Figure 5. Whlers diagram for ACL16+ mix, 4PB-PR test; (R2 = 0.7982).
for each strain level. Figures 5 to 7 present resulting Whlers diagram for mixes tested by 4PB-PR
method. For these results the basic regression equation is given as well.
Following the gained test results it can be stated that in case of 4PB-PR for ACL 16+ not entirely
suitable strain levels have been selected. Because of the chosen setting for none of the strain levels
106 loading cycles have been reached and the fatigue parameter 6 was in this case determined
by extrapolation. This approach of fatigue characteristics determination is however not suitable,
because it is not possible with sufficient accuracy estimated fatigue behavior of an asphalt mix.
Generally interpreting gained results it should be stated that higher 6 value means better fatigue
life-time. Similarly higher value in the exponent part of the regression equation in Figures 57 or
lower value of characteristic B, can be explained by more resistant material against fatigue.
Comparison of calculated fatigue characteristics gained from testing of both compared test
method is summarized in following Table 2.
From the results it can be stated, that fatigue testing executed by 2PB-TR and 4PB-PR method
with simile temperatures but different frequencies the 6 values are more or less same. From
this point of view the test methods seems to be comparable. Following the remaining fatigue
characteristics as used by the Czech pavement design method, the results show more significant
difference, especially for a parameter and general for both parameters if evaluating asphalt mix
ACL 16+.
Based on the stated findings only the mix ACP 22+ might be relevant for a comprehensive
comparison of fatigue parameters gained by both compared test methods. In this case all measure-
ments have been done according to valid methodologies. For the mix ACO 11S the test was not
done on two-point beam test. For ACL 16+ mix the fatigue parameters were in case of 4PB-PR
determined by extrapolation. The slope of fatigue curves deviates in case of ACP 22+ minimally.
The resulting 6 parameters correspond well with bitumen content used in the mixture and with the
maximum size of aggregate used for this type of asphalt concrete. On the other hand the results
have shown that for determination of comparability of both methods used in the Czech Republic
210
Figure 6. Whlers diagram for AC0 11S mix, 4PB-PR test; (R2 = 0.8362).
Figure 7. Whlers diagram for ACP22+ mix, 4PB-PR test; (R2 = 0.8498).
211
Table 2. Calculated fatigue characteristics.
2PB-TR test method 4PB-PR test method
Mix Par. A Par. B 6 Par. A Par. B 6
ACO 11S 2.601 0.170 240.5

ACL 16+ 2.797 0.167 159.8 2.673 0.187 159.7
ACP 22+ 2.838 0.172 134.0 2.871 0.168 132.1
it will be necessary to continue with further testing and specification of boundary conditions.
The first comparison has at least shown that there is a good correlation of both test methods.
ACKNOWLEDGMENT
This paper has been supported by the research projects GACR 103/09/0335 Rheology and experi-
mental determination of functional characteristics of warm asphalt mixes, TA02030639 Durable
acoustic asphalt pavement courses with utilization of bituminous binders modified by rubber
microfiller including innovative technology of rubber milling and TA02030549 The most
effective utilization of reclaimed asphalt pavement layers for production of new asphalt mixes.
REFERENCES
CSN EN 13108-20 Asphalt mixtures Specification for materials Part 20: Type testing
Di, Benedetto, H., La Roche, C., Baaj, H., Pronk, A., Lundstrm, R. 2004. Fatigue of bituminous mixtures. In
RILEM Materials and Structures, Vol 37, April 2004, pp. 202216, Paris.
EN 12697-5 Bituminous mixtures Test methods for hot mix asphalt Part 5: Determination of the maximum
density.
EN 12697-6 Bituminous mixtures Test methods for hot mix asphalt Part 6: Determination of bulk density
of bituminous specimen
EN 12697-24 Bituminous mixtures Test methods for hot mix asphalt Part 24: Resistance to fatigue
EN 12697-26 Bituminous mixtures Test methods for hot mix asphalt Part 26: Stiffness
EN 12697-33 Bituminous mixtures Test methods for hot mix asphalt Part 33: Specimens prepared by roller
compactor
Novotn, B., Luxemburk, F. 2005. Fatigue of asphalt mixes and pavement design optimization. Final report of
the research project GACR 103/02/0396, CTU in Prague.
Pronk, A.C. 1997. Comparison of 2 and 4 point fatigue tests and healing in 4 point dynamic bending test based
on the dissipated energy concept. Ann Arbor.
Soucek V. 2008. Fatigue properties of asfalt mixes. Master thesis (in Czech), CTU Prague.
Varaus, M., Hzl, P., Vacn, O., Soucek, V. 2009. Comparison of stiffness moduli using 2-point and 4-point
bending tests. 2nd Workshop on Four Point Bending, Pais (ed.), University of Minho.
212
Hungarian experience with different bending devices
Z. Puchard & A. Gorgenyi

Colas Hungary Co., Budapest, Hungary
ABSTRACT: The central laboratory of the Hungarian group of Colas has tested the stiffness
modulus and fatigue of asphalt mixes with various methods since 2005. The laboratory is equipped
with two-point and four-point devices capable of determining stiffness modulus and fatigue, and a
device designed to determine indirect tensile strength.
We generally test stiffness modulus and fatigue under the temperature and frequency conditions
stipulated in EN 13108-20 standard. We have also performed comparative tests using 2-point and
4-point testing equipment on the same asphalt mixes applying the same temperature and frequency
as well as diverse temperatures and frequencies for modulus and fatigue tests.
The presentation summarises the findings of the tests we have carried out until now, i.e. stiffness
modulus tests with 2-point, 4-point and ITT equipment, and comparative fatigue tests with 2-point
and 4-point equipment.
Keywords: 4-point equipment, 2-point equipment, stiffness modulus, fatigue, comparative tests
1 INTRODUCTION
The development laboratory of the Hungarian group of Colas (Colas Hungary Ltd. Central Labo-
ratory) has tested stiffness modulus and resistance to fatigue testing since 2005. We have a 2-point
and a universal fatigue tester, which is also appropriate for performing 4-point and ITT testing.
We purchased both units (the 2-point and universal testers) from Cooper Research Technology Ltd.
(England). Our 2-point equipment is suitable to perform the 2PB stiffness modulus test according to
EN 12697-26:2005 Annex A and fatigue testing according to EN 12697-24:2004+A1:2008 Annex
A. The universal fatigue tester equipment is suitable to perform stiffness modulus tests according to
EN 12697-26:2005 Annex B (4 PB) and Annex C (ITT), as well as to determine fatigue according
to EN 12697-24:2004+A1:2008 Annex D. At both 2 PB and 4 PB stiffness modulus test methods
are the cycle 100 and the strain level 50 m/m. At ITT test the rise time, measured from when the
load pulse commences and which is the time taken for the applied load to increase from zero to
maximum value 124 ms. Both 2 PB and 4 PB resistance to fatigue tests are performed in controlled
strain and the end of test (fatigue life) is until the stiffness modulus has dropped to 50% of the
initial value.
EN 12697-26 describes the different stiffness modulus measurement methods. The methods can
be found in this standard shown Figure 1. In standard EN 12697-24 can be found resistance to
fatigue test methods. This standard defines 2 PB-TR (2 PB), 2 PB-PR, 3 PB-PR, 4 PB-PR (4 PB)
and ITT test methods.
Hungarian road specifications on asphalt mixes are based on the European product standards.
Hungarian specifications (UT) opted for the fundamental requirements of asphalt concrete mixes,
this way stiffness and fatigue are requirements in Hungarian specifications pursuant to Point 5.4.
of EN 13108-1:2006 Hungarian requirements on the type of asphalt concrete mixes include high
modulus asphalt mixes. Their required stiffness and fatigue values are determined by the specific
requirements illustrated in Table 1.
213
The test temperatures and frequencies for the requirement values indicated in Table 1 are identical
with the requirements stipulated in standard EN 13108-20:2006.
According to Hungarian requirements, stiffness modulus and fatigue values, in addition to high
modulus asphalt mixes, have to also be tested for bituminous concrete mixes that can be used for
heavy traffic, however, there are no requirement values set for such mixes as yet. At present, the
tests performed serve to provide data for the requirements to be specified in the future.
Due to the above, we perform stiffness modulus and fatigue tests routinely. We have performed
hundreds of tests recently. In most cases, we applied the ITT method since it is fast and easy to
perform to carry out the stiffness modulus tests, and applied the 2-point method again, due to its
fastness to perform the fatigue tests (our equipment has two test frames). In order to see how the
requirement values can be realized in case of test results produced through different methods and
boundary conditions, we used diverse methods to perform comparative tests on the same asphalt
mixes. We represent the results of these tests in the following.
Figure 1. Stiffness modulus measuremet methods in EN standard.
Table 1. Stiffness and fatigue requirements of AC 16/22 binder (NM) asphalt mixes.
Test method Temperature, C Frequency, Hz Requirement
Stiffness, S
2 PB 15 10 min. 11000 MPa
4 PB 20 8 min. 7000 MPa
ITT 20 124 s min. 7000 MPa
Fatigue, 6
2 PB 10 25 min. 115 microstrain
4 PB 20 30 min. 115 microstrain
214
2 EQUIPMENTS AND TEST SPECIMENS
2.1 Universal fatigue tester

Figure 2 shows the universal fatigue tester unit.
There are two test frames in the temperature controlled cabinet. One test frame can perform the
indirect tensile test (ITT), the other can perform the four point bending test (4 PB). The actuator
is capable of applying up to at least 10 kN the maximum load will depend upon the air supply
pressure. With repeated or cyclic loading the maximum load will depend upon the flow rate of the
air supplied. A servo-valve connected to the inlet of the actuator is powered by a 24 volt supply
and controlled by a 0 to 10 volt signal witch determines the direction and speed of travel of the
actuator and the load applied. By varying the voltage signal to the servo-valve the load applied to a
specimen can be varied between tension and compression at frequencies up to 50 Hz. The S type
load cell transducer is a train gauge bridge device which can measure forces up to 20 kN which is
twice the capacity of the actuator.
Specimen dimension in the case of a 4-point test frame may be 50 50 400 mm. We use
a precision cutting machine to produce the prismatic beams from slabs which was made with
roller compactor according to EN 12697-33:2003+A1:2008 point 7.2.4.2. In Hungary we use
the Marshall mix design procedure therefore we compact the slabs until they reach the Marshall
free void content. We do not use the 4-point test at mixes with 32 mm maximum grain size,
because the cross-section of the test specimen does not satisfy the requirements stipulated in EN
12697-24:2004+A1:2008 and EN 12697-26:2004 Chapter D3.
We produce the cylindrical specimen used for the ITT tests by 2 50 hits using a Marshall
compactor. The diameter of the specimen tested is usually 100 mm with a height of 60 mm. At
Figure 2. Universal fatigue tester.
215
testing mixes with a maximum 32 mm grain size, we are using specimen of 150 mm in diameter
and 60 mm in height, produced in a gyrator.
2.2 Two point bending tester

The two-point equipment is illustrated in Figure 3.
The equipment comprises:
a rigid frame made from welded steel section

a three phase electric motor
an adjustable eccentric fitted to the drive shaft of the electric motor
a proxistor is mounted beneath the eccentric to provide a means of counting cycles and
determining motor speed
a lever arm assembly hinged on leaf sprongs that gives a 1:5 reduction in horizontal movement
two horizontally mounted loading arms that trasmit the reduced movements from the lever arm
through force transducers to the tops of two trapezoidal test specimens
a rigidly mounted LVDT displacement transducer measures the horizontal movement of the
loading arm assembly
the test space is enclosed within an insulated stainless steel cabinet with sliding, double glazed
doors that provide access to test space
Figure 3. A general view of two point bending tester.
216
Table 2. Test temperatures and frequencies.
Test method Temperature, C Frequency, Hz
Stiffness, S
2 PB 15 10
4 PB 15 10
4 PB 20 8
ITT 15 124 s
ITTT 20 124 s
Fatigue, 6
2 PB 10 25
4 PB 10 25
4 PB 20 30
an interface unit contains signal conditioning for transducers. The two force transducers are
connected to the interface unit via charge amplifiers
a temperature conditioning unit is connected, via supply and return ducts, to the insulated cabinet.
We use a precision cutting machine to produce the trapezoidal shaped specimen from slabs
which was made with roller compactor according to EN 12697-33:2003+A1:2008 point 7.2.4.2. We
compact these slabs as well up to reaching the Marshall free void content. Test specimen dimensions
correspond to the dimensions stipulated in EN 12697-24:2004+A1:2008 and EN 12697-26:2004
in Table 1 of Chapter A3.
3 TEST RESULTS
We performed stiffness modulus and fatigue tests on the following five different asphalt mixes to
compare their respective results.
Mix 1 AC 11 wearing (F) 50/70 the binder was regular bitumen with penetration range 50/70,
the aggregates fraction used was crushed andesite, maximal grain size 11 mm (F: means an
asphalt type for heavy traffic, and produced with normal bitumen)
Mix 2 AC 16 binder (mNM) 40/80-60 the binder was relative soft modified bitumen with pene-
tration range 40/80, the aggregates fraction used was crushed andesite, maximal grain size
16 mm (mNM: means a high modulus asphalt NM , produced with modified bitumen)
Mix 3 AC 22 binder (F) 35/50 the binder was regular bitumen with penetration range 35/50, the
aggregates fraction used was crushed andesite, maximal grain size 16 mm (F: means an
asphalt type for heavy traffic, and produced with normal bitumen)
Mix 4 AC 22 binder (mNM) 25/55-65 the binder was modified bitumen with penetration range
25/55, the aggregates fraction used was crushed andesite, maximal grain size 22 mm (mNM:
means a high modulus asphalt NM , produced with modified bitumen)
Mix 5 AC 22 binder (mNM) 10/40-65 the binder was modified bitumen with penetration range
10/40, the aggregates fraction used was crushed andesite, maximal grain size 22 mm (mNM:
means a high modulus asphalt NM , produced with modified bitumen)
We carried out the fatigue and modulus tests at the temperatures and frequencies indicated in
Table 2.
The dimensions of the specimens used for the tests satisfied the requirements stipulated in
standard EN 12697-26:2004 for modulus tests and in standard EN 12697-24:2004+A1:2008 for
fatigue tests. We ran the modulus tests on four specimens of each mix. The following figures show
217
Table 3. Mean deviation of modulus.
Type of test Mean deviation of stiffness modulus, %
4 PB ITT (in % of 4 PB-PR) 23

2 PB 4 PB (in % of 4 PB-TR) 34
2 PB ITT (in % of 2 PB-TR) 43
Figure 4. Comparison of stiffness modulus test result measured according EN standard.
the mean values calculated from the four test results of each mix. At fatigue tests, we examined 18
specimens of each mix at three different strain levels (which is six specimens at each strain level).
We calculated the fatigue lines and the 6 values from the mean of the six test results at the different
strain levels.
3.1 Stiffness modulus test results

The stiffness modulus values of the five asphalt mixes measured at standard temperature and
frequency are illustrated in Figure 4.
Figure 4 shows that the four point test and ITT always produce lower values than the two point
test. The rate of average difference is indicated in Table 3.
The rate of difference is the highest for the 2 PB and ITT tests, followed by the 2 PB 4 PB
tests, and it is the lowest for the 4 PB ITT tests.
If we compare the values obtained through the modulus tests performed using various methods
with the requirements of the road specifications (Table 1), the results of the two point test satisfy said
requirement with respect to all five asphalt mixes tested. Mix No. 2 fails to satisfy this requirement
when subjected to the ITT and 4-point tests.
Figure 5 shows the comparison of the modulus values measured at 15 C using three different
methods.
It is clear from Figure 5 that usually the 4-point test provides the highest modulus value. This
is followed by the modulus measured by the 2-point test. Lastly, the ITT test provides the lowest
modulus value we could measure. There is an average 9% difference between the modulus values
measured by the 4-point and the 2-point test. The difference between the modulus values measured
by the 4-point and the ITT test is 21% on average.
Figure 6 shows the comparison of the modulus values measured at 15 C and 20 C using the
4-point method.
The modulus values measured at 20 C are always lower than those measured at 15 C. This is
primarily caused by the difference in the test temperature. The average difference is 48% concerning
the test values measured at 20 C.
218
Figure 5. Stiffness modulus of different asphalt mixes at 15 C temperature.
Figure 6. Stiffness modulus of different asphalt mixes measured with 4 PB.
Table 4. Fatigue test results with different test conditions at Mix 1.
Method Strain amplitude [m/m] Trend of line (p)
2PB at 10 C 25 Hz 131 0.146

4PB at 10 C 25 Hz 136 0.162
4PB at 20 C 30 Hz 143 0.153
3.2 Fatigue test results

The stiffness modulus values of the five asphalt mixes measured at standard temperature and
frequency are illustrated in Figure 7.
Figure 7 shows that the four-point method always produced higher fatigue results than the
two-point method. The average difference is 11% with respect to two-point results.
When comparing the test results with the specification values (Table 1), we can establish that all
five asphalt mixes examined requirement very well.
We also carried out the 4-point test for Mix 1 at 10 C and 25 Hz. The fatigue lines are shown in
Figure 8, while the elongation and trend of line values are displayed in Table 4.
Figure 8 shows that the three fatigue lines are virtually parallel. The elongation value at 10 C
temperature and 25 Hz frequency is 4% higher for the four-point method than the results of the
two-point method. The results of the four-point method at 20 C and 30 Hz are 5% higher than the
results measured at 10 C and 25 Hz.
219
Figure 7. Comparison of fatigue test result measured according EN standard.
Figure 8. Fatigue lines with different test conditions at Mix 1.
4 CONCLUSIONS
We applied various methods stipulated in standards to test the stiffness modulus and fatigue of five
typical asphalt mixes used most frequently in Hungarian road construction (stiffness: 4 PB-PR,
2 PB-PR, ITT fatigue: 4 PB-PR, 2 PB-PR). After comparing and analysing the test results, the
conclusions are as follows:
The four-point and ITT methods delivered lower modulus values than the two-point method
when performing the stiffness modulus tests according to the requirements stipulated in standard
EN 12697-20:2006. This reason is primarily the difference in test temperatures (4 PB and ITT at
20 C, 2 PB at 15 C). This finding is also substantiated by the lower modulus values registered
at 20 C test temperature than at 15 C test temperature for the modulus values determined by the
four-point method.
When we performed the stiffness modulus test at 15 C with all three methods, it was the four-
point method that produced the highest modulus values. The cause of the difference may lie with
the differing operating principles of the test equipment as well as the test specimens of different
shapes and dimensions. This finding is also backed by that the difference between the results
achieved by the four-point and ITT methods is around 20% (21% difference between the results
in the case of 15 C test temperature, and 23% difference in the case of 20 C test temperature).
When we performed the fatigue test under the conditions stipulated in standard EN 12697-
20:2006, the four-point method produced higher values than the two-point method. The cause
220
of the variance is again primarily lies with the difference in test temperatures (4-point method
at 20 C, 2-point method at 10 C). The average difference between the values the two methods
display is 11%.
We found very little difference (4%) when comparing the results of the two-point and four-point
tests performed at an identical temperature (10 C) and frequency (25 Hz). This goes to show
that the difference (different principle, shape and dimensions of the specimen) between the test
methods used for fatigue tests has little impact on the test results.
Considering the above comparisons, we can establish that the test temperature, frequency and
different test methods themselves (testing principle, shape and dimensions of the specimen)
have a much greater impact on the stiffness modulus results than on the fatigue test results.
REFERENCES
EN 12697-20:2006 Bituminous mixtures. Material specifications. Part 20: Type testing

EN 12697-24:2004+A1:2008 Bituminous mixtures. Test methods for hot mix asphalt. Part 24 Resistance to
fatigue
EN 12697-26:2005 Bituminous mixtures. Test methods for hot mix asphalt. Part 26 Stiffness
EN 12696-33:2003+A1:2008 Bituminous mixtures. Test methods for hot mix asphalt. Part 33. Specimen
prepared by roller compactor
EN 13108-1:2006 Bituminous mixtures: Material specifications. Part 1: Asphalt Concrete
T 2-3.301-1:2010 Bituminous Mixtures for Road Constructions. Asphalt Concrete
221
Development of quality control test procedure for characterizing

fracture properties of asphalt mixtures
S. Saadeh & O. Eljairi

Department of Civil Eng. and Const. Eng. Mgmt., California State University, Long Beach, CA, US
B. Kramer
Twining Laboratories, Long Beach, CA, US
E. Hajj
Western Regional Superpave Center, Reno, NV, US
ABSTRACT: The main objective of this study is to investigate the use of Semi-Circular Bend
(SCB) test as a Quality Assurance/Control (QA/QC) measure for field construction. Comparison
of fracture properties from SCB test and Fatigue Beam Test (FBT) was conducted. Asphalt slabs
were specially fabricated from two mixtures containing PG64-10 and PG58-22 binders. The SCB
Jc and the FBT Nf values for the PG64-10 and PG58-22 mixtures on dry and wet conditions were
determined. The SCB Jc and FBT Nf indicated better fracture properties for the dry mixtures
compared to wet mixtures. Similar to SCB Jc , the ANOVA on the FBT Nf parameter indicated
that there is significant effect of the condition type (dry Vs. wet) on the measured Nf and SCB
Jc values. In addition, the FBT Nf parameter indicated that there is no significant effect of the
mixture type (PG64-10 Vs. PG58-22) on the measured Nf values. In general, the SCB Jc and the
FBT Nf had similar ranking of both mixtures at dry and wet conditions. The FBT Nf had higher
variability (16.548.9% coefficient of variability) compared to SCB Jc (033.2% coefficient of
variation). The SCB test has a great potential as QA/QC test of fracture properties of asphalt
mixtures.
1 INTRODUCTION
Fatigue life resistance of asphalt concrete (AC) is defined as its ability to resist the repeating traffic
loading without significant cracking or failure, Harvey et al. (1995). Fatigue cracking is primary
distress in asphalt pavement due to repetitive stresses and strains caused by both traffic loading and
environmental factor such as temperature differences. The fatigue resistance of AC is investigated
by a number of fatigue testing.
Roque et al. (2004) proposed the Dissipated Creep Strain Energy (DCSE) limit as one of
the most important factors that control crack performance and hence durability of AC mixtures.
Roque studied 22 field mixtures that have been in service for more than 10 years in the State of
Florida. In order to determine this parameter, two laboratory tests conducted on the same spec-
imen. These tests are: the indirect tensile resilient modulus test (ITMr) and the indirect tensile
strength (ITS) test. Both tests were conducted at 10 C on 150 mm diameter and 50 mm thick spec-
imens. From the strength test, failure strain (f ), tensile strength (St ) and fracture energy (FE) were
determined.
The indirect tensile strength test is the most traditional test method used to evaluate the fracture
properties of AC mixture in laboratory, and has been used in some cracking characteristics study
223
as well, Roque (1999) and Baoshan (2004). During ITS test, the specimen is loaded till failure and
the load and deformation is continuously recorded, and then the indirect tensile strength and the
corresponding strain are computed.
Wagoner et al. (2005) had introduced the disk-shaped compact tension (DCT) test for AC
material. The test uses specimen modified from the ASTM E399 standard for compact tension
testing of metals.
Zhou et al. (2007) presented the Texas overlay test (OT). The researchers at Texas Transportation
Institute developed this test to assess the fatigue cracking prediction. The device test AC specimens
that are glued to two steel plates, with half of its length rest on each plate. One of the steel plates
is fixed and the other moves horizontally to simulate the opening and closing of the crack under
overlays. The AC mixture specimen has a standard dimension of 150 mm long by 75 mm wide by
38 mm high. The specimen can be fabricated from Superpave Gyratory Compactor (SGC) or from
field cores.
The Strategic Highway Research Program (SHRP) Project A-003A included the development
of a flexural beam test for measuring fatigue of AC mixtures and an analysis system based on
multilayer elastic theory to estimate in-situ performance, Deacon et al. (1995). In California, the
Asphalt Research Program performed a five year study on the fatigue performance of AC mixtures,
Harvey et al. (1995). The fatigue beam test (FBT) developed during these studies well predicted the
asphalt mixture fatigue performance. However, the fatigue beam test requires extensive analysis
and great deal of machining which is expensive and difficult to accomplish.
A quality control test that is simple to perform and well correlated to the fatigue beam test
is needed to support the frequently required quality assurance/control testing. The Semi-Circular
Bend (SCB) test is being proposed as a QC test because it is simple to perform (it can be conducted
using Hveem stabilometer that is used in the mix design), inexpensive (one compacted specimen
makes four SCB specimens), and simple to analyze (the output parameter is indicative of the
dissipated energy during the crack propagation).
The SCB test is still unknown to most Asphalt technologists in the US even though it was
introduced to asphalt community in US by Europeans and South African researchers in the late
1990s and early 2000s. The origin use of SCB was to characterize the fracture resistance in rock
mechanics Boashan (2005). Lim et al. (1994) used the SCB for determining the characteristics
of a water-saturated synthetic mudstone. It was found that the SCB can provide reliable results
that are comparable to other methods that are used to compute the Mode I fracture testing. SCB
test was employed to determine the Mode I for mudstone. In their research, they recommended
the use of the semi-circular specimens under three point loading which were developed by the
Chong et al. (1984). Chong and his group proposed to use a semi-circular core specimen with
a single edge notch, subjected to three point loading. Recently, SCB test has been introduced to
pavement material area, and many research projects have applied this method to investigate the
fracture property of different AC mixtures (Molenaar et al. (2002), Mull et al. (2002), Mohammad
et al. (2004), and Mohammad et al. (2003)). The SCB can be a potential test to further explore the
material fracture properties.
SCB has many advantages over other tests in characterizing AC mixtures. The major important
advantage is that different notch depths can be made easily on the semi-circular test specimen.
The notch is made by a special saw blade of 3.0 mm thickness. SCB test can be performed and
many specimens can be made with one core sample. Moreover, SCB specimens can be prepared
directly from cylindrical samples obtained from standard core prepared in the superpave gyratory
compactor (SGC) or taken from field.
The analysis of this test is based on an elasticplastic fracture mechanics concept of critical strain
energy release rate, which is named critical value of J-integral (Jc ). Some studies have shown Jc
value to be sensitive to a number of variables such as type of binder, type of aggregate size and way
of compaction, Wu (2005). All these factors are significant to fracture properties of AC mixtures.
This study investigated the use of Semi-Circular Bend (SCB) test as a quality assurance/control
(QA/QC) measure for field construction.
224
2 OBJECTIVES
The main objective of this study is to investigate the use of SCB test as a QA/QC measure for field
construction. In particular, the objectives include the following:
Compare the fracture properties of AC mixtures using the SCB test to those of the FBT.
Investigate the influence of moisture sensitivity on the fracture properties as evaluated by SCB
and FBT.
3 SCOPE
Two binder types, PG64-10 and PG58-22, were selected for evaluation in this study. The job mix
formulas for each mix type considered were identical except for the binder type. The asphalt binder
met California specifications. Granite was the predominate aggregate used in the AC mixture types
considered.
There was one day of production for the SCB and FBT mixtures. Viscosity and dynamic shear
rheometer, aging, and bending beam rheometer measurements were used to determine the rheolog-
ical properties of both binder types. The AC mixture fracture properties were determined using the
SCB and FBT tests. Both mixtures were compacted to 5 1% air void in the beams and core testing
specimens. Cores for SCB testing and beams for FBT were sawed from the same slab to ensure
consistency among specimens and reduce the variability in the results due to changes in specimens.
The specimens were tested in both dry and wet condition to evaluate moisture sensitivity.
4 MATERIAL PROPERTIES AND MIXTURE DESIGN
4.1 Asphalt cement binder

PG64-10 and PG58-22 were the two binder types considered in this study since they are commonly
used in California, meeting the AASHTO M320 specifications. Binders were supplied by Blue
Diamond Materials.
4.2 Aggregates
The aggregates used in both binder course mixtures were produced at the Lehigh Hanson Irwindale,
California plant (SMARA No. 91-19-0025) are derived from alluvium of the San Gabriel Mountains
as deposited in the San Gabriel River basin. A petrographic examination performed in accordance
with ASTM C 295 indicated the aggregate materials are composed of igneous, metamorphic, and
sedimentary rocks and minerals. These materials are principally granitic in origin with minor
amounts of schist, diorite, gabbro and basalt.
4.3 Mixture design

The Hveem mix design procedure was used to determine the optimum asphalt content of the AC
mixtures following the California Standard Specifications for Roads and Bridges. The final
aggregate structure for the AC mixtures was determined using the petrographic examination. The
mixture included 15% RAP which had a RAP binder content of 5% by total weight of mix. The
maximum density was 2492.4 kg/m3 (155.6 lb/ft3 ) for both mixtures. The maximum theoretical
specific gravities were 2.474 and 2.473 for the PG58-22 and PG64-10 mixtures, respectively.
The optimum asphalt binder content for both mixtures was found to be 4.3% by dry weight of
aggregate. Table 1 presents the job mix formula (JMF), mixture properties, and gradation analysis
for the binder course mixtures that were used in this study.
225
Table 1. JMF, mix properties, and gradation analysis.
Maximum Density = 2492.4 kg/m3 (155.6 lb/ft3 )

Rap Binder Content = 5%
Sieve Combined MinMax
38.1mm (1 ) 100 100100

25.4 mm (1 ) 100 100100
19.05 mm ( ) 100 100100
12.7 mm ( ) 97 95100
9.5 mm (3/8 ) 82 7288
No. 4 55 4660
No. 8 40 2842
No. 16 30
No. 30 21 1527
No. 50 13 1020
No. 100 7
No. 200 3.9 27
Figure 1. a) Slab mold fabrication b) Mixture spreading c) Compaction of the PG64-10 Mixture and
d) Compaction PG58-22 mixture.
4.4 Slab compaction

Vibratory steel wheel rollers were used to compact the fabricated AC mixtures specimens to
required density. A conventional steel wheel rolling protocol was followed, i.e. breakdown, inter-
mediate, and then finished roller. The established rolling pattern was maintained the same for
both mixtures (Figure 1). The average AC plant temperature was 165.5 C (330 F) for both mix
types evaluated. The initial compaction temperature for the mixtures was approximately 148.8 C
(300 F). A frame was prepared to fabricate the slabs. It was covered by a special red rosin paper
226
Figure 2. a) Coring SCB specimens and b) Sawing FBT specimens.
as shown in Figure 1a to 1d. The asphalt was then placed on the frame, Figure 1b. Vibratory steel
wheel roller compactor was used to compact the slabs. On the next day the sides of frame were
removed in order to obtain the desired specimens.
4.5 Fabrication of AC mixture specimens

Specimens were prepared according to the specific requirements of each test. Cylindrical 127 mm
(6 in.) diameter specimens were cored for SCB test and rectangular beams were sawn for FBT at
50 mm thick 63 mm wide 380 mm long. (2.0 thick 2.5 wide 15 long), figure 2.
5 FUNDAMENTAL MATERIAL CHARACTERIZATION
Asphalt binders were tested and characterized by performance grading tests including rotational
viscometer, dynamic shear rheometer, bending beam rheometer, rotation thin film oven, and pres-
sure aging vessel. The AC mixtures fracture properties were characterized using the Semi-Circular
Bend (SCB) test and Fatigue Beam test (FBT).
The following sections outlines the test methodology used for fundamental material characteri-
zations of the asphalt cement binders and the AC mixtures used.
5.1 Asphalt cement binder rheology

Asphalt binders were tested and characterized according to AASHTO R29, Grading or Verifying
the Performance Grade (PG) of an Asphalt Binder.
5.2 Moisture conditioning

The moisture conditioning procedure outlined in Jones et al. (2008) was used for this purpose.
1. The specimen was placed in the vacuum container supported above the container bottom by
a spacer. The container was filled with water so that the specimen was totally submerged in
the water and a vacuum of 635 mm (25 in.) of mercury was applied. After 30 minutes the
vacuum was stopped and the sample was removed and the saturated surface dry mass according
to AASHTO T-166 was determined. The volume of absorbed water and the degree of saturation
was determined. A saturation level of 7080% was targeted for various specimens.
2. The vacuum-saturated specimen was then placed in a water bath at a pre-set temperature of
60 C. The specimen was supported on a rigid, flat (steel or wood) plate to prevent deformation
of the specimen during conditioning. The top surface of the specimen was about 25 mm below
the water surface.
227
Table 2. SCB test factorial.
Specimen Name
Condition Mix Before Cutting After Cutting
Dry PG64-10 (1-10 SCB [1]) (1-10 SCB [1]) A

(1-10 SCB [1]) B
(2-10 SCB [1]) (2-10 SCB [1]) A
(2-10 SCB [1]) B
(2-10 SCB [2]) (2-10 SCB [2]) A
(2-10 SCB [2]) B
PG58-22 (5-22-SCB [2]) (5-22 SCB [2]) A
(5-22 SCB [2]) B
(6-22 SCB [1]) (6-22 SCB [1]) A
(6-22 SCB [1]) B
(6-22 SCB [2]) (6-22 SCB [2]) A
(6-22 SCB [2]) B
Wet PG64-10 (3-10 SCB [1]) (3-10 SCB [1]) A
(3-10 SCB [1]) B
(4-10 SCB [1]) (4-10 SCB [1]) A
(4-10 SCB [1]) B
(4-10 SCB [2]) (4-10 SCB [2]) A
(4-10 SCB [2]) B
PG58-22 (Extra 22 SCB [1]) (Extra 22 SCB [1]) A
(Extra 22 SCB [1]) B
(Extra 22 SCB [3]) (Extra 22 SCB [3]) A
(Extra 22 SCB [4]) (Extra 22 SCB [4]) A
3. After 24 hours, the water bath was drained and refilled with cold tap water. The water bath
temperature was set to 20 C. It took two hours for the water bath to reach an equilibrium
temperature.
4. The specimen was then removed from the water bath, and the saturated surface dry mass was
determined.
5. The specimen was then wrapped with Parafilm to ensure no water leakage.
6. The bonded nut (for the beam) was checked. When loose, it was removed and re-bonded with
epoxy resin.
7. A layer of scotch tape was applied to the areas where the specimen contacts the clamps of the
fatigue machine. This would prevent adhesion between the Parafilm and the clamps.
8. The fatigue test of the conditioned beam was started within 24 hours.
5.3 Testing Factorial

Both SCB and FBT specimens had a measured air voids of (5 1%). Specimens that had air
voids out of this range were discarded. Tables 2 and 3 present the test factorial for SCB and FBT,
respectively. A description of each test is provided below.
5.4 Semi-Circular Bend (SCB) test

This test was used to characterize the fracture resistance of asphalt mixtures based on a fracture
mechanics concept. The critical strain energy release rate, which is called the critical value of
J-integral or Jc was determined for each semi-circular specimen that tested. Three notch depths of
228
Table 3. FBT factorial.
Condition Mix Specimen Name
Dry PG64-10 1-10 FBT 1]

2-10 FBT [1]
2-10 FBT [2]
PG58-22 5-22 FBT [1]
5-22 FBT [2]
6-22 FBT [1]
Wet PG64-10 3-10 FBT [1]
4-10 FBT [2]
Ex-10 FBT
PG58-22 6-22 FBT [2]
7-22 FBT [2]
8-22 FBT [2]
Figure 3. Set-up of semi-circular bend test.
(25.4-, 31.8- and 38.0 mm (1-, 1.25-, and 1.5-in.) were used in this test. The tests were conducted
at 20 C. A semi-circular specimen was loaded monotonically until fracture under a constant cross-
head deformation rate of 0.5 mm/min in a three-point bend load configuration, Figure 3. The load
and deformation was continuously recorded and the critical value of J-integral was determined
based on the following equation:
where: b is sample thickness, a is the notch depth, and U is the strain energy to failure.
5.5 Fatigue Beam Test (FBT)

This test was conducted according to AASHTO T-321. The test temperature was 20 C and stain level
of 350 micro-strain. FBT test of asphalt mixture predicts the field performance of asphalt pavement
229
under the similar conditions of asphalt such as traffic, periods between traffic stresses, multi-axle
loading and environmental condition. The parameter that is considered for fatigue cracking of
pavement fatigue life is the number of load cycles to failure. Failure is considered as the 50%
reduction in initial stiffness which is measured at cycle 50.
From a slab compacted in accordance to the method described above, six AC mixture beam
specimen were prepared for each mix. The beams tested were 2.0 thick 2.5 wide 15 long
(50 mm thick 63 mm wide 380 mm long).
6 DISCUSSION OF RESULTS
6.1 Asphalt cement binder rheology

Tables 4a and b present the asphalt binder rheology results for both asphalt cement binders evaluated
in this study. It can be seen that the asphalt cement binders considered met the specifications
according to ASTM 6373 and AASHTO M320. Also, It should be noted that the PG64-10 had a
higher mid-range temperature (28 C) compared to PG58-22 (22 C).
6.2 Semi-Circular Bend (SCB) test

A total of 24 SCB tests were performed, 2 mixtures (PG64-10 and PG58-28), 2 conditions (Dry and
Wet), 3 notch-depths (11.251.5 in.) and 2 replicates each. Figure 4 presents the load-displacement
curves for the three notches of PG64-10 and PG58-22 mixtures for the dry and wet conditions.
Table 4a. Asphalt cement binder PG58-22 rheology.
Test Result Test Method Specification
135 C,
Rotational Viscosity, Pa.s 0.261 AASHTO T316 3.0 Maximum
Flash, COC, C 298 AASHTO T48 230 Minimum
Solubility in TCE, w% 99.7 AASHTO T44 99.0 Minimum
Dynamic Shear Rheometer, 58 C
Complex Viscosity, , Pa.s 163.5 AASHTO T315
G , kPa 1.64
Phase Angle, 88.1
G /Sin , kPa 1.641 1.00 Minimum
RTFO-Aged
Mass Change, w% 0.115 AASHTO T240 1.00 Maximum
G , kPa 4,285 AASHTO T315
Phase Angle, 84.9
Ductility, 25 C, cm 150+ AAHSTO T51 75 Minimum
PAV-Aged, 2.1 MPa, 20 Hrs, 100 C AAHSTO R28
G , kPa 6015 AASHTO T315
Phase Angle, 47.2
G Sin , kPa 4416 5000 Maximum
Bending Beam Rheometer, 12 C
Stiffness, MPa 139 AASHTO T313 300 Maximum
m-value 0.333 0.300 Minimum
Specific Gravity, 60 F 1.018 AASHTO T228
API Gravity, 60 F 7.5
230
Figure 5 presents the strain energy curves for both mixtures at dry and wet conditions. The peak
load and the displacement values were determined after each SCB test in order to calculate the
critical Jc using equation (1).
The SCB Jc values for PG64-10 and PG58-22 mixtures for the dry and wet conditions are shown
in Table 5. The SCB Jc values are normally distributed as can be seen in Figure 6. The Jc value
ranged between 8.7 and 2.0 lb/in for all the specimens in the study. The PG64-10 mixture achieved
higher Jc values compared to PG58-22 mixture for both dry and wet conditions.
A regression analysis was used, considering two variables, namely binder type and specimen
condition. The analysis of variance (ANOVA) results are presented in Table 6.
The ANOVA indicated that there is no significant effect of the mixture type on the measured
SCB Jc , however, there is significant effect of the condition (dry Vs wet) on the measured SCB Jc
as indicated by the P-value (P-value < 0.05 indicates that the parameter is significant).
6.3 Fatigue Beam Test

A total of 12 FBT were performed, 2 mixtures (PG64-10 and PG58-28), 2 conditions (dry and wet),
and 3 replicates each. The test was performed according to AASHTO T-321 at 20 C (68 F) and
stain level of 350 micro-strain. The Initial Dissipated Energy and the cycles to failure (Nf ) were
determined. Table 7a represents the FBT test results for the PG64-10 and PG58-22 AC mixtures,
at both conditions (dry and wet), evaluated in this study. It can be seen that the results had a high
coefficient of variation (%CV) (27 to 93%). The authors decided to examine and reduce the results
by eliminating the outlier sample to reduce variability of the results. Table 7b presents the reduced
FBT test results for the PG64-10 and PG58-22 AC mixtures, at both conditions (dry and wet).
Table 4b. Asphalt cement binder PG64-10 rheology.
Test Result Test Method Specification
135 C,
Rotational Viscosity, Pa.s 0.345 AASHTO T316 3.0 Maximum
Flash, COC, C 310 AASHTO T48 230 Minimum
Solubility in TCE, w% 99.8 AASHTO T44 99.0 Minimum
Complex Viscosity, , Pa.s 150 AASHTO T315
G , kPa 1.503
Phase Angle, 88.4
RTFO-Aged
Mass Change, w% 0.190 AASHTO T240 1.00 Maximum
G , kPa 3,874 AASHTO T315
Phase Angle, 85.5
Ductility, 25 C, cm 150+ AAHSTO T51 75 Minimum
PAV-Aged, 2.1 MPa, 20 Hrs, 100 C AAHSTO R28
G , kPa 3180 AASHTO T315
Phase Angle, 53.8
G Sin , kPa 2568 5000 Maximum
Bending Beam Rheometer, 6 C
Stiffness, MPa 47 AASHTO T313 300 Maximum
m-value 0.454 0.300 Minimum
Specific Gravity, 60 F 1.022 AASHTO T228
API Gravity, 60 F 7.0
231
Figure 4. Load-deflection curves for a) Dry PG64-10 b) Dry PG58-22 c) Wet PG64-10, and d) Wet PG58-22.
232
Figure 5. Strain energy curves for a) Dry PG64-10 b) Dry PG58-22 c) Wet PG64-10, and d) Wet PG58-22.
233
Table 5. Critical value of J-integral (Jc ) for all mixtures.
Jc
Specimen
Mixture Jc (lb/in) Jc (kN/m) Mean %CV Rank
Dry PG64-10 8.7 1.524 8.70 0.0 1

8.7 1.524
PG58-22 9.31 1.630 8.55 12.7 2
7.78 1.362
Wet PG64-10 3.42 0.599 4.06 22.3 3
4.7 0.823
PG58-22 1.53 0.268 2.00 33.2 4
2.47 0.433
Figure 6. Probability plot for of Jc .
Table 6. ANOVA for the SCB Jc .
Source DF Seq SS Adj SS Adj MS F P
Condition 1 62.552 62.552 62.552 73.66 0

Mixture 1 2.453 2.453 2.453 2.89 0.15
Error 5 4.246 4.246 0.849
Total 7 69.251
The %CV was reduced to 16.5 and 48%. The data are normally distributed as can be seen in
Figures 7a and b.
A regression analysis was performed on the effect of mixture type and the condition of the
mixtures on the initial dissipated and the cycles to failure (Nf ). The FBT initial dissipated energy
234
Table 7a. Fatigue beam test results-original.
Initial Dissipated Energy, psi Cycles to failure, Nf
Mixture Specimen Avg. %CV Rank Specimen Avg. %CV Rank
Dry PG64-10 0.01 0.01263 19.3 1 97,829 76,300 27.4 2

0.0148 75,000
0.0131 56,070
PG58-22 0.0073 0.0097 23.0 4 139,133 129,954 64.1 1
0.0101 208,293
0.0117 42,436
Wet PG64-10 0.0095 0.00905 7.0 3 14,262 10,597 48.9 4
0.0086 6,932

PG58-22 0.008 0.00917 15.9 2 6,850 16,724 93.0 3

0.0087 34,657
0.0108 8,664
The beam broke during the initial stage of testing.
Table 7b. Fatigue beam test results modified.
Initial Dissipated Energy, psi Cycles to failure, Nf
Mixture Specimen Avg. %CV Rank Specimen Avg. %CV Rank
Dry PG64-10 0.01 0.0124 27.4 1 97,829 86,415 18.7 2

0.0148 75,000
PG58-22 0.0073 0.0087 22.8 4 139,133 173,713 28.2 1
0.0101 208,293
Wet PG64-10 0.0095 0.00905 7.0 3 14,262 10,597 48.9 3
0.0086 6,932
PG58-22 0.008 0.0094 21.1 2 6,850 7,757 16.5 4
0.0108 8,664
ANOVA results are presented in Table 8. The ANOVA indicated that there is no significant effect
of the mixture type and condition on the measured initial dissipated energy.
The FBT Nf ANOVA results are presented in Table 9. Similar to the SCB Jc , the ANOVA
indicated that there is no significant effect of the mixture type on the measured Nf , however, there
is significant effect of the condition (dry Vs wet) on the measured Nf .
6.4 Comparison between SCB Jc and FBT

The SCB Jc parameter and the FTB Nf ranked the dry mixtures higher than the wet mixtures for
PG64-10 and PG58-22 mixtures. In addition, SCB Jc and FBT Nf had similar ranking for the wet
mixtures. For the dry mixtures the Jc had ranked the PG64-10 mixture higher than the PG58-22
mixture, however the FBT Nf ranked the PG58-22 higher than the PG64-10.
The FBT initial dissipated energy had ranked the mixture different from FBT Nf and the SCB Jc .
This is expected as this parameter does not reflect the fracture properties of the mixture.
In terms of coefficient of variability (%CV) the FBT Nf exhibited higher variability than those
of SCB Jc .
235
Figure 7. a) Probability plot for initial dissipated energy and b) Probability plot for cycles to failure Nf .
In general, it can be seen that the SCB Jc had similar ranking to the FBT Nf . A regression analysis
to assess the relation between the SCB Jc and the FBT Nf was conducted. The model is presented
in equation (2).
The P-value of the statistical model is (0.023) and R2 = 60%.
236
Table 8. ANOVA for the FBT initial dissipated energy.
Condition 1 0.0000035 0.0000035 0.0000035 0.63 0.464

Mixture 1 0.0000056 0.0000056 0.0000056 1 0.363
Error 5 0.000028 0.000028 0.0000056
Total 7 0.0000371
Table 9. ANOVA for the FBT Nf .
Condition 1 29227212651 29227212651 29227212651 21.67 0.006

Mixture 1 3566619111 3566619111 3566619111 2.64 0.165
Error 5 6743118760 6743118760 1348623752
Total 7 39536950522
The SCB test has a great potential as QA/QC test of fracture properties of asphalt mixtures.
However, the results of this pilot study need to be further investigated and confirmed. The up
coming project will address the SCB and FBT for nine mixtures used in California.
7 CONCLUSIONS
The main objective of this study was to investigate the use of SCB test as a quality assurance/control
(QA/QC) measure for field construction. Comparison of fracture properties from SCB and FBT
was conducted. The Jc values and the FBT Nf for the PG64-10 and PG58-22 mixtures on dry and
wet conditions were determined. The following summarizes the findings of this limited evaluation:
The SCB Jc and FBT Nf indicated better fracture properties for the dry mixtures compared to
wet mixtures.
The dry PG64-10 mixture achieved similar SCB Jc values to PG58-22 mixture. However, the
wet PG64-10 mixture achieved higher SCB Jc values than PG58-22 mixture.
The dry PG58-22 mixture achieved higher FBT Nf values than PG64-10 mixture. However, the
wet PG64-10 mixture achieved higher FBT Nf values than PG58-22 mixture.
The FBT initial dissipated energy was not consistent with the SCB Jc and the FBT Nf values.
This is expected as this parameter does not reflect the fracture properties of asphalt mixtures.
In general, the SCB Jc and the FBT Nf had similar ranking of both mixtures at dry and wet
conditions.
The FBT Nf had higher variability, %CV (16.5 48.9%), compared to SCB Jc , %CV (0
33.2%).
The ANOVA on the SCB Jc parameter indicated that there is a significant effect of the condition
(dry Vs. wet) on the measured Jc values. In addition, the SCB Jc parameter indicated that there
is no significant effect of the mixture type (PG64-10 Vs PG58-22) on the measured Jc values.
Similar to SCB Jc , the ANOVA on the FBT Nf parameter indicated that there is significant effect
of the condition type (dry Vs. wet) on the measured Nf values. In addition, the FBT Nf parameter
indicated that there is no significant effect of the mixture type (PG64-10 Vs. PG58-22) on the
measured Nf values.
A regression analysis on relation between the SCB Jc and the FBT Nf indicated that the model
is significant with R2 = 60%.
The results of this pilot study indicate that the SCB test has a great potential as QA/QC test of
fracture properties of asphalt mixtures.
237
ACKNOWLEDGMENT
The authors would like to thank the United States Department of Transportation, California
Department of Transportation, and METRANS for their interest and provision of grant support
to make this project possible. The contents of this paper reflect the views of the authors, who are
responsible for the facts and the accuracy of the information presented herein. The contents do
not necessarily reflect the official views or policies of the State of California or the Department of
Transportation. This report does not constitute a standard, specification, or regulation.
REFERENCES
AASHTO T 166 Bulk specific gravity of compacted mixture using saturated surface-dry specimens.
Baoshan, Huang, Egan, Brian K., Kingery, William R., Laboratory Study of Fatigue Characteristics of HMA
Surface Mixtures Containing RAP TRB 2004 Annual Meeting CD-ROM.
Chong K. P. and Kuruppu M. D., New Specimen for Fracture Toughness Determination for Rock and Other
Materials, International Journal of Fracture 26 (1984), R59R62.
Deacon, J A, Leahy, R B and Monismith, C L., (1995) Mix testing and analysis systems resulting from SHRP
contract A003A, Proc. ASCE Transp. Congress, Vol. 1, pp. 69171.
Harvey, J.T., J.A. Deacon, B.W. Tsai, and C.L. Monismith, Fatigue Performance of Asphalt Concrete Mixes
and Its Relationship To Asphalt Concrete Pavement Performance in California, Asphalt Research Program:
CAL/APT Program, Institute of Transportation Studies, University of California, Berkeley, October 1995.
Huang, Baoshan. Comparison of Semi-Circular Bending and indirect Tensile strength tests for HMA
mixtures. Geotechnical special publication, 130 (2005), 177188.
Jones, David , Rongzong Wu, Bor-Wen Tsai, Qing Lu, and John T. Harvey, Warm-Mix Asphalt Study: Test
rack Construction and First-Level Analysis of Phase 1 HVS and Laboratory Testing, report UCPRC-RR-
2008-11).
Lim, I.L. Fracture Testing of a Soft Rock with Semi-circular Specimens Under Three-Point Bending. Part 1-
Mode I Rock Mech. Min. Sci & Geomech. Abstr. Vol. 31. No, 3, 185197, 1994.
Mohammad, L. N., Negulescu, I, Wu, Z., Daranga, C., Daly, W.H and Abadie, C., Investigation of the Use of
Recycled Polymer Modified Asphalt Binder in Asphalt Concrete Pavements, Journal of the Association
of Asphalt Paving Technologists- Proceedings of the Technical Sessions, v 72, Asphalt Paving Technology
2003, pp. 551594.
Mohammad, N. L., Wu, Z., Aglan, M., Characterization of Fracture and Fatigue Resistance on Recycled
Polymer-Modified Asphalt Pavements, The 5th RILEM conference on Pavement Cracking, France, 2004
(In press).
Molenaar, A.A.A., Scarpas, A., Liu, X., Erkens, S.M.J.G. Semi-Circular Bending Test: Simple but Use-
ful? Association of Asphalt Paving Technologists-Proceedings of the Technical Sessions, v 71, 2002,
pp. 794815.
Mull, M. A., Stuart, K. and Yehia, A.,Fracture resistance characterization of chemically modified crumb
rubber asphalt pavement, Journal of Materials Science, Vol. 37, pp. 557566, 2002.
Roque, R., Birgisson B., Drakos C., Dietrich B. Development and Field Evaluation of Energy-Based Criteria
for Top-down Cracking Performance of Hot Mix Asphalt, Journal of the Association of Asphalt Paving
Technologists, Vol. 73, Baton Rouge Louisiana, 2004.
Roque, R., Zhang, A., and Sankar, B., Determination of Crack Growth Rate Parameters of Asphalt Mixture
Using the Superpave IDT, Journal of the Association of Asphalt Paving Technologists, Vol. 68, 1999,
pp. 404433.
Wagner, M. P., Buttlar, W. G., and Paulino, G. H. Disk-shaped Compact Tension Test for Asphalt Concrete
Journal of Experimental Mechanics, v 45, n 3, June, 2005, pp. 270277.
Wu, Zhong. Fracture resistance characterization of Superpave mixtures using the semi-circular bending test.
Journal of ASTM International 2.3 (2005):135149.
Zhou, F., Hu, S., Scullion, T., Chen, C., Qi, X., and Claros, G. (2007) Development and Verification of the
Overlay Tester Based Fatigue Cracking Prediction Approach Journal of the Association of Asphalt Paving
Technologists, 76.
238
Fatigue resistance: Is it possible having a unique response?
C. Maggiore & G. Airey

University of Nottingham, UK
A. Collop
De Montfort University, Leicester, UK
G. Di Mino & M. Di Liberto

University of Palermo, Palermo, Italy
P. Marsac
IFSTTAR, Nantes, France
ABSTRACT: The mechanical characterisation of the asphalt concrete in terms of both the fatigue
resistance and the stiffness modulus is necessary to use any design method of the flexible road
pavements.
Different kinds of test are usually used in experimental work such as bending tests, uniaxial
tests, etc., but sometimes they do not give the same answer.
In this paper mechanical characterization was carried out by means of fatigue tests undertaken
with two most used testing machines for asphalt material: Two Point Bending (2PB) test at IFST-
TAR in Nantes (France) and Four Point Bending (4PB) test at University of Palermo, in Palermo
(Italy).
Different strain controlled tests were undertaken for the same material under the same loading
conditions, frequency and temperature (15 Hz and 20 C), according to the European standard 12697
part 24 and 26.
The first results of this interlaboratory activity are showed in this paper.
1 INTRODUCTION
Flexural fatigue is one of the main failure modes in asphalt mixtures and flexible pavement struc-
tures. This means good prediction of a pavements fatigue life will help to develop and improve
pavement design procedures.
Fatigue resistance and stiffness are two required parameters for pavement design necessary to
dimension the pavement structure (layer thickness).
European standards (EN 12697 part 24 and 26) specifies the methods for characterizing stiffness
and fatigue of bituminous mixtures by different kind of tests, including bending tests (2PB and
4PB) and direct and indirect tensile tests. Usually tests are performed applying a sinusoidal loading
(stress or strain) to a different kind of specimen (trapezoidal or prismatic), depending on the mode
of loading.
1.1 Fatigue
Flexural fatigue due to repeated traffic loading is a process of cumulative damage.
From a mechanical point of view, the mechanism of fatigue can be divided into two parts:
the first one is the occurrence of tensile stress/strain in the base layer; the second one is the
239
Figure 1. Load and strain amplitude for 2PB and 4PB.
repetitive occurrence of such tensile stress/strain under traffic repetitions. The repetition of the
tensile stress/strain causes the accumulation of micro damage in the bottom of the base layer that,
over time, results in the break between the aggregate and the binder, thus generating more or less
deep cracks. In other words if a beam is subjected to load, the beam would tend to assume a
convex downward shape, with tensile stress/strain in the bottom part and compressive stress/strain
in the top one. Since the asphalt pavement has viscoelastic behaviour, it recovers when the load
is removed. At the end of this first cycle there is part of the strain that is recovered and a small
part that is permanent. Under the next load the pavement undergoes the same cycle. Ultimately the
pavement will fail due to damage accumulation (Rajib B. et al. 2009).
Physically speaking, micro crackings originate at the bottom of an asphalt concrete layer caused
by horizontal tensile strain; this compromises the contact between the aggregate skeleton and the
binder (particle-to-particle contacts). Furthermore the water trapped in the cracks and the repeated
loading leads to a decrease in the strength of the mixtures and micro crackings starting to propagate
towards the layers above and leading to pavement collapse. This phenomenon is called Bottom-Up
Cracking (Thom N., 2008).
According to European standards, fatigue can be evaluated by means of bending tests or direct
and indirect tests. This paper focuses on bending tests, in particular on 2 Point Bending (2PB)
and 4 Point Bending (4PB) tests. The two type of tests are widely used for measuring fatigue
resistance and stiffness for asphaltic paving material. Usually a sinusoidal loading is applied to the
specimen, but geometry of the specimen and loading mode are different in the two tests. In the
2PB, trapezoidal specimen is glued between to plates (at the top and at the bottom) and the fracture
usually happens at 1/3 of height, where bending moment is maximum. In the 4PB, instead, the
specimen is not glued and the fracture happens in the middle part of the beam characterized by a
constant maximum value of bending moment (see Figure 1).
1.2 4 Point Bending test (4PB)

The 4 Point Bending test is the most used fatigue test in the United States.
The tests were performed for the experimental work with servo-hydraulic testing system at
University of Palermo (see Figure 3). Prismatic beams were manufactured with dimensions of
380 mm in length, 50 mm in height and 63 mm in width.
The specimen is restrained at four points by means of four clamps: the two outside remains
static (they can only shift horizontally), the centrals deflect according to the strain applied. The
deformation of the specimen is measured at the bottom between the two central clumps.
240
Figure 2. Configuration of 4PB test.
Figure 3. Beam Fatigue testing machine at University of Palermo.
This test simulates very well a pavement fatigue failure under traffic loading because repeated
loading causes tension in the bottom zone of the specimen, cracking will initiate and propagates to
the top zone until failure; failure usually occurs in the area of uniform bending moment between
the two inner clamps. (see Figure 2).
In this type of test free lateral translation are permitted in order to prevent internal stresses
developing in the specimen.
In the 4 point bending test, initial stiffness is usually chosen between the 50th and the 100th load
application. Conventionally, fatigue failure is the moment when the stiffness has decreased to half
of its initial value (SHRPA-404, 1994, Di Benedetto et al, 2004, Shen et al, 2007).
241
Figure 4. Trapezoidal specimen for 2PB.
Figure 5. 2 Point Bendng Test.
1.3 2 Point Bending test (2PB)

The 2 Point Bending test is more used and widely spread in Europe.
The tests were performed for the experimental work at IFSTTAR in Nantes. The methodol-
ogy consists of applying a sinusoidal continuous waveform at the top of a trapezoidal specimen.
Schematic of trapezoidal specimen is represented in the Figure 4. Usually four specimen for each
tests were used (see Figure 5).
The trapezoidal specimen is mounted as vertical cantilever as seen in Figure 5. Sinusoidal
constant displacement is applied at the top of the specimen, while the bottom base is fixed.
As mentioned before, fracture usually occurs at 1/3 height from the bottom because that area is
the most stressed in the specimen as shown in the Figure 6.
As in the 4PB, in the 2PB, initial stiffness is usually chosen between the 50th and the 100th load
application. Traditionally, fatigue test ends when stiffness has decreased to half of its initial value
(Rowe, 1993, SHRPA-404, 1994).
2 LABORATORING TESTING
2.1 Material
A 10 mm Dense Bitumen Macadam (DBM) was chosen for the experimental work. 100 Pen Binder
was chosen for the mixture. The aggregates type selected was a crushed limestone. The aggregate
242
Figure 6. Fracture in trapezoidal specimen.
Figure 7. Aggregate gradation curve.
gradation curve is shown in Figure 7; four lines are presented: the upper, the lower and the middle
point curve from the British Standards.
3 TEST RESUTS
3.1 4 Point Bending test Procedure

In this study, the fatigue tests were carried out applying a continuous sinusoidal cyclic waveform,
in controlled displacement (strain) mode at constant temperature (20 C degrees) and at constant
frequency (15 Hz) (EN-12697-24-2007, EN-12697-26-2007).
Strain levels between 140190 were chosen for the testing.
243
Figure 8. Strain waveform.
Figure 9. Stress waveform.
Typical strain waveform applied to the prismatic specimen and typical stress waveform recorded
are presented in Figure 8 and 9.
For each test stiffness modulus, phase angle and dissipated energy are calculated.
Figures 10 shows the typical trend of the stiffness modulus. It is calculated from the measured
force, displacement and phase lag between force and displacement, considering the mass of the
beam and the clamps. As seen in the figure, a two stage evolution process is recorded during a
fatigue test and it could be associated to two phenomenological aspects in flexible pavement: crack
initiation and crack propagation (forming a network of micro and macro cracks), usually followed
by the failure of the specimen (Pronk, 1999, Di Benedetto et al, 2004).
Also for the phase angle, it is possible noticing the same behaviour as the stiffness modulus.
After a rapid increasing, the phase angle shows a regular behaviour (see Figure 11).
Dissipated energy usually changes during a fatigue test, due to the beginning of microcrack
during the fatigue process; it usually decreases in strain controlled mode, it increase in stress
controlled mode testing (see Figure 12).
The evolution of dissipated energy is shown in Figure 13. In the first stage of the test, the
hysteresis loop is well defined. After this the loop changes shape. It starts to rotate and the area
becomes smaller (for strain controlled test mode). At the end of the tests, dissipated energy is
usually characterised by an irregular hysteresis loop. Thus, variation in dissipated energy should
be a good parameter to describe fatigue phenomenon in asphaltic material and it is considered the
starting point for the development of the fatigue model.
244
Figure 10. Stiffness evolution during a fatigue test (160 e).
Figure 11. Phase Angle evolution during a fatigue test (160 e).
Figure 12. Dissipated Energy versus Number of cycle during a fatigue test (170 e) fitted with power law.
245
Figure 13. Evolution of hysteresis loop during a fatigue test.
Figure 14. Strain waveform.
3.2 2 Point Bending test Procedure

Same loading conditions were considered for 2 Point bending tests in Nantes. Fatigue tests were
undertaken in controlled displacement mode at 15 Hz and 20 C (EN-12697-24-2007, EN-12697-
26-2007).
Strain levels between 120190 were chosen for the 2PB tests.
Typical strain waveform applied to the trapezoidal specimen and typical stress waveform recorded
are presented in Figure 14 and 15.
The applied and the recorded waveform look symmetrical around the point zero.
As for the 4 PB tests, stiffness modulus, phase angle and dissipated energy are calculated.
Figures 16 shows the typical trend of the stiffness modulus. It is calculated considering the
recorded stress divided by the applied strain. As it can be seen, a three stage evolution process
is recorded during a fatigue test. After a rapid evolution of stiffness (phase I), due to the internal
heating phenomenon, stiffness decrease seems more regular (phase II). Fracture occurs in the final
stage (phase III) and it is characterised by an acceleration of stiffness drop (Pell, 1967, Hopman
et al, 1989, Baburamani, 1999, Bankowski et al., 2007).
Phase angle, after a rapid increase, shows a constant behaviour during the test. As can be seen
its value is 40 degrees during the fatigue test (see Figure 17).
246
Figure 15. Stress waveform.
Figure 16. Stiffness evolution during a fatigue test (160 ).
Figure 17. Phase Angle evolution during a fatigue test (160 ).
247
Figure 18. Dissipated Energy versus Number of cycle during a fatigue test (170 ) fitted with power law.
Figure 19. Evolution of hysteresis loop during a fatigue test.
Figure 18 shows the decrease in dissipated energy for controlled displacement mode undertaken
with the 2PB. Figure 19 shows the evolution of the hysteresis loop (dissipated energy) during a
fatigue test.
4 CONCLUSION
An interlaboratory study between IFSTTAR and University of Palermo have been developed to
better understand the fatigue phenomenon by means of different testing machine.
Mechanical characterization was carried out by means of fatigue tests undertaken with two most
used testing machines for asphalt material: two point bending (2PB) test and four point bending
(4PB) test.
248
The first results of this interlaboratory activity seem to underline some difference between the
two kinds of testing machines. Fatigue tests are undertaken on the same material, at the same
loading conditions, same frequency (15 Hz) and same temperature (20 C); however differences are
observed in the determination of stiffness, phase angle, number of cycle to failure and dissipated
energy.
Several reasons may explain those behaviours:
different loading conditions: constant bending moment in the middle section of the prismatic
specimen and no shear stress in 4PB; only one point in the trapezoidal specimen is characterized
by the highest bending in 2PB and shear stresses exist in the specimen.
different boundary conditions (fixed trapezoidal specimen that is glued at the top and at the
bottom between two plates in 2PB; free prismatic specimen, horizontal translation and rotation
are allowed in 4PB);
different size sample (Di Benedetto et al. 2004);
presence of shear forces in the 2PB bending test; no shear forces in 4PB (Pronk, 1999);
other phenomena such as heating can influence the results differently in the two kind of tests
(Di Benedetto et al. 2004).
The most important point is understanding which testing machine simulates better the real
fatigue phenomenon; which machine gives us the right response of the material. Those are not easy
questions, thus further analysis within a wider spectrum of temperatures and frequencies, will be
needed to answer to them.
Furthermore, dissipated energy (depending on strain, stress and phase angle) is a good parameter
to describe fatigue behavior of bituminous materials and the key point to better understand the
mechanical response of the same material from different kind of test.
REFERENCES
Baburamani, P. 1999. Asphalt fatigue life prediction models:a literature review. Research Report ARR 334,
ARRB Transport Research
Bankowski W. and Sybilsky D., 2007. Energetic method as an alternative for conventional method in fatigue
life analysis of bituminous mixtures. 4th International SIIV Congress Palermo (Italy), 1214 September
2007.
Di Benedetto H., De La Roche, C., Baaj, H., Pronk, A. and Lundstrom, R. 2004. Fatigue of bituminous
mixtures. RILEM TC 182-PEB Performance testing and evaluation of bituminous materials. Matrials
and Structures, Vol. 37, April 2004.
EN-12697-24-2007. Bituminous mixtures Test methods for hot mix asphalt Part 24: Resistance to fatigue
EN-12697-26-2004. Bituminous mixtures Test methods for hot mix asphalt Part 26: Stiffness
Hopman P.C., Kunst P.A.J.C. and Pronk A.C. 1989. A renewed interpretation method for fatigue measurement,
verification of Miners rule. 4th Eurobitume Symposium Madrid 1: 557561.
Pell, P. S. 1967. Fatigue of Asphalt Pvement Mixes. Proceedings of the Second International Conference on
the Structural Design of Asphalt Pavements, Ann Arbor, Michigan
Pronk, A.C. 1999. Fatigue lives of asphalt beams in 2 and 4point dynamic bending tests based on a new
fatigue life definition using the dissipated energy concepts. Road and Hydraulic Engineering Division,
Rijkswaterstaat, The Netherlands.
Rajib B., Mallick and El-Korchi, T. 2009. Pavement Engineering: Principle and Practise, CRC Press.
Rowe, G. M. 1993. Performance of Asphalt Mixtures in the Trapezoidal Fatigue Test. Proceedings of the
Association of Asphalt Paving Technologist 62.
Shen, S. and S. H. Carpenter. 2007. Dissipate Energy Concepts for HMA Performance: Fatigue and Healing.
Departement of Civil and Environmental Engineering. Urbana, Illinois, University of Illinois at Urbana-
Champaign.
SHRPA-404. 1994. Fatigue Response of Asphalt-Aggregate Mixes. Washington, DC, Asphalt Research Pro-
gram Institute of Transportation Studies University of California, Berkeley. Strategic Highway Research
Program National Research Council.
Thom, N. 2008. Principles of pavement engineering, Thomas Telford Limited.
249
Non-asphaltic materials evaluation
A laboratory study of the influence of multiple axle loads

on the fatigue performance of a cemented material
M.A. Moffatt
ARRB Group, Melbourne, Victoria, Australia
Department of Civil Engineering, Monash University, Melbourne, Victoria, Australia
G.W. Jameson
ARRB Group, Melbourne, Victoria, Australia
W. Young
Department of Civil Engineering, Monash University, Melbourne, Victoria, Australia
ABSTRACT: A laboratory investigation of the effects of different multiple axle group loads on
the fatigue performance of a cemented material was conducted. The study made use of a pre-
existing laboratory test process for assessing the flexural modulus and fatigue characteristics of
cemented materials, extending it by simulating multiple axle loads (single, tandem, triaxle and
quad-axle groups) using a range of load pulse shapes designed to broadly simulate truck axle
groups. The study developed relationships between fatigue performance and initial strain level
for each simulated load shape. The flexural modulus of the cemented material used was found to
be highly variable, with the majority of samples between 15 000 and 22 000 MPa. The variable
nature of the samples hampered the drawing of conclusive findings. A relationship was determined
between the fatigue life, the initial strain level and the combination of the peak load applied and
the load type.
1 INTRODUCTION
1.1 Background
One of the key components of a possible future Australian mass-distance heavy vehicle charging
scheme is the assessment of the effects of different heavy vehicle axle group types and loads on
infrastructure wear. Feasibility studies for an incremental road pricing scheme have highlighted the
need for improved knowledge about the relative damage of axle group types.
This paper summarises a laboratory study of the effects of different multiple axle load groups
on the fatigue performance of a cemented material. The study reported is part of a larger study
that aims to use a combination of existing knowledge, new laboratory characterisation techniques
and field trials as the basis for the development of a framework that can be used to quantify the
pavement damage due to different multiple axle groups. Moffatt (2011a) documents the progress
to date in the overall study.
Moffatt (2011b) contains more detailed descriptions of methods, equipment, sample preparation,
software development, and all primary data and testing results, for this cemented material study.
1.2 Cemented materials

In the context of this paper, cemented materials are a combination of crushed granular material,
water and cementitious binder, mixed and compacted to form a pavement layer that develops
significant tensile strength. Typical binder contents of between three and seven percent, by mass, are
253
used. Contents lower than this lead to materials that are considered to be modified granular materials
with little tensile strength, and higher binder contents yield mixtures that are more susceptible to
shrinkage cracking during curing and approach concrete in characteristics. The alternative term
cement treated materials is interchangeably used when referring to materials that use a Portland
cement based binder.
In the Austroads flexible pavement design process (Austroads 2008) the performance of
cemented materials is expressed as their resistance to flexural fatigue due to application of repeated
loads. The flexible pavement design process considers the fatigue performance of these materials
to be more sensitive to changes in load magnitude than other materials used in flexible pavement
structures.
Following the occurrence of fatigue cracking of a cemented pavement layer, further degradation
of the layer may occur with additional loading, resulting in a reduction of the stiffness of the layer
to one similar to an unbound granular material. The Australian pavement design processes can
consider this post-cracking phase of life. The focus of this paper, however, is fatigue performance
of the cemented material, and so consideration is not given to the post-cracking phase of life.
2 METHOD
2.1 Overview
The laboratory study reported here had the following broad characteristics:
use of pre-existing (controlled load) laboratory test processes for assessing the flexural modulus
and fatigue performance of cemented materials
simulation of multiple axle group loads using a range of load shapes designed not to precisely
simulate truck axle groups, but rather to allow examination of the underlying characteristics of
the load shapes linkable to material performance
the development of the relationship between fatigue performance and initial flexural strain for
each simulated load shape, and subsequent comparison of those relationships.
2.2 Laboratory flexural testing methods

Yeo (2008a) describes the development of a suite of laboratory test methods for assessing the
behaviour of cemented materials. Of particular note were methods for assessing the flexural resilient
modulus and the flexural fatigue of large sized cemented material specimens. Yeo also documents
the apparent good match between the laboratory observed performance of two cemented materials
and the performance of test pavements, comprised of those materials, when subjected with full-scale
accelerated pavement testing using the Accelerated Loading Facility.
The multiple axle laboratory study followed these established test methods, with the exception
that the standard haversine loading pulse was replaced with a series of pulse shapes simulating
multiple axle loads.
The methods are based on the use of rectangular test beams whose square-cross-sectional dimen-
sions may vary from 80 80 mm upwards, with a typical dimension of 100 100 mm. The methods
describe the manufacture of samples in the laboratory, and also the use, as in this study, of beams
prepared from pavement beds constructed in the field.
The beams are supported by apparatus used in concrete testing (Standards Australia 2000), shown
in Figure 1. A closed-loop control system is used to apply a controlled haversine pulse load to the
specimens. For each loading pulse the displacement of the test sample is measured at the vertical
mid-span (i.e. between the two upper load rollers) and this displacement is then used, assuming
simple beam theory, to determine the resilient modulus.
The standard flexural fatigue test defines the fatigue life as the number of cycles to reduce the
flexural modulus of the sample to a value equal to half the initial value. The initial modulus of the
sample is defined as the average of the modulus values determined in the first 50 cycles of the test.
254
Figure 1. Flexural beam roller supports and load rollers.
2.3 Sample preparation

The samples used in this laboratory study were extracted from an untrafficked section of a con-
structed pavement. Yeo (2008b) describes the construction of the 150 mm thick cemented base
pavement, which was intended for subsequent testing with the Accelerated Loading Facility (ALF).
The pavements were constructed inside the large ALF shed (54 m in length by 18 m wide).
The cemented material was sourced from the Boral Para Hills quarry, South Australia, and met
the South Australian Department for Transport, Energy and Infrastructures specification for a
20 mm PM2/20QG Class 2 quarry produced stabilised material (DTEI 2008). The source rock was
a siltstone/quartzite. The material was bound with 4%, by dry mass, of general purpose cement,
with an additional retarding agent, Daratard, added at the manufacturers recommended rate of
600 ml per 100 kg of cement. The addition of the retarder, whilst uncommon in Victoria (where
the pavement was placed) is common in South Australia. The retarder slowed the cement hydration
process, and provided an increased working time of up to four hours, increasing the likelihood of
achieving a uniform construction.
The key characteristics of the material, including Atterberg limits and grading of the unbound
material, and the unconfined compressive strength of the bound material are shown in Table 1.
The location used for sourcing the cemented base material for the laboratory study was not
trafficked by ALF in the subsequent trial and shrinkage cracking had not been observed at this
location during the construction and curing phases. Additionally, the section was relatively uniform
in thickness and density, and exhibited uniform surface deflections when measured with a Falling
Weight Deflectometer (FWD)
The sample size for flexural modulus and fatigue specimens typically used in the ARRB test
method is 100 (width) 100 (depth) 400 mm (length). Prior work had demonstrated that a skilled
operator using a diamond bladed concrete saw could accurately cut 100 (width) 400 mm (length)
samples from an in situ pavement, provided that accurate guidelines were painted on the surface to
be cut. Material cut in this manner requires only the removal of surfacing material and trimming
to the specified depth to yield test samples of the required dimensions.
On 20 December 2006, exactly two years after construction of the cemented base, a concentrated
mass of 270 samples were cut using a diamond bladed concrete saw lubricated with water. The
samples were carefully extracted the following day, and were placed, packed with damp sand, into
255
Table 1. Characterisation of cemented siltstone.
Parameter Value
Maximum dry density(1) 2.070 t/m3

Optimum moisture content(1) 8.0%
Plasticity index (PI) 6
Plastic limit 15
Liquid limit 21
UCS at 96% relative compaction(2) , at 28 days cure(3) 6.5 MPa
Particle size distribution:
% passing 19.0 mm sieve 99
(1) In accordance with Standards Australia (1997) using modified compactive effort
(2) Relative compaction = (density)/(maximum dry density)
(3) In accordance with Standards Australia (1996), samples soaked for four hours prior
to testing
storage bins. Despite the action of water and the saw blade the cut faces of the samples did not
show signs of significant erosion of material.
Before laboratory testing, samples were removed from the storage bins and trimmed to an even
100 mm depth using a water lubricated saw in the laboratory. Again, no significant erosion of the
cut faces was evident. Once cut to final dimensions, the samples were stored indoors on rigid
shelves, wrapped in moist hessian fabric.
In accordance with the test procedure the samples were transferred to the moist atmosphere of
a fog room at least two days before flexural testing.
An alternative to the use of samples extracted from a pavement would have been the manufacture
of samples in the laboratory. This approach was not adopted for the following reasons:
Sample manufacture in the laboratory is labour intensive, and the manufacture of a large number
of samples would be have been both costly and take several months to undertake.
After manufacture of samples a minimum period of six months would be needed to cure and age
the samples to a condition approximating field conditions.
Concerns regarding the difference between compaction levels achieved in the laboratory not
matching field conditions would be negated if field samples were used.
2.4 Test equipment

Two identical 14 kN Universal Testing Machines (UTS-14P) manufactured by IPC Global were used
to conduct the tests discussed in this paper (Figure 2). The UTS-14P is a closed-loop testing system
that incorporates a 14 kN pneumatic actuator on a large capacity load frame with an integrated
load cell. The system can utilise a range of transducers and loading jigs. Control of the system is
achieved through software. For this study, specimen support and loading jigs were manufactured
in accordance with Figure 1, and two calibrated Linear Variable Differential Transformers (LVDT)
were attached to the system to measure the mid-span displacement during testing. These LVDTs
were anchored to the specimen by a custom-built support frame resting over the support rollers.
The high speed of the loading pulses, necessitated that the LVDT frame be rigid and not susceptible
256
Figure 2. Loading frame with sample.
to induced vibrations caused by the loading, this led to the tall walls of the support frame visible
in Figure 2.
The IPC supplied flexible programmable test software UTS019 User Programmable Test (IPC
Global 2007) was used to conduct the testing. This software was programmed to conduct both
flexural resilient modulus testing and repeated flexural testing under a variety of different load
shapes. In line with the standard method, the software was programmed to run controlled load
tests.
A variety of load shapes are built into the UTS019 software, including the haversine shape used
in this study. The software can also import ASCII text files containing user defined pulse shapes.
Imported load shapes are defined as a series of 512 integer numbers. The software automatically
scales the integers so that the minimum number represents no applied load, and the maximum
integer represents the application of the maximum load defined elsewhere in the control software.
By scaling user defined pulse shapes in this manner, the software allows a single pulse definition
to be used for varying maximum load magnitudes. This pulse import feature was used for all axle
group simulation load shapes discussed below.
2.5 Load pulse shapes

The standard flexural modulus and fatigue test procedures discussed above are controlled load
tests, i.e. the equipment control system controls the load applied during the test, ensuring that it
follows a user defined shape and magnitude. This controlled load approach was retained for the
multiple axle laboratory study, with different load shapes developed to simulate different multiple
axle groups. The standard tests use a haversine load shape to simulate a single axle, and haversine
shapes were used as the basis for the multiple axle load shapes. All of the load shapes used in the
study resulted in the flexural response of the specimens representing the transverse response in a
pavement structure (i.e. in a direction perpendicular to the direction of travel).
257
Figure 3. Assumed relationship between load pulse shape and axle spacing.
Figure 3 shows a time-based load shape representing a tandem axle, and shows some of the
underlying assumptions that were made when generating the load shapes:
The peak-to-peak spacing of the shape (a in Figure 3) matches the axle spacing at a given travel
speed.
The load rise at the start of the pulse (a/2 in Figure 3) added to the load drop at the end of the
pulse (a/2) is the same as the spacing between axles (a).
Using these assumptions a relationship between the vehicle travel speed and the width of loading
pulse used in the simulation was determined (Equation 1). Using a typical spacing between axles
within a group of 1.25 m, the standard tests use of a haversine representation of a single axle with a
pulse width of 250 ms corresponds with a travel speed of approximately 18 km/h. This same pulse
width also corresponds with a quad-axle group travelling at 72 km/h. Trial-and-error determined
that the pneumatic equipment could repeatedly apply a quad-axle shape in this time period, but was
unable to do so at faster test speeds, i.e. shorter pulse widths. Thus, the pulse width of 250 ms was
fixed for the quad-axle group.
where p = axle group load pulse width (ms); n = number of axles in axle group; a = spacing of
axles within axle group; s = vehicle speed (km/h)
Given the long term nature of fatigue testing and a finite number of samples, the testing program
was limited as follows:
One travel speed was simulated (72 km/h).
One axle spacing within multiple axle groups was simulated (1.25 m).
The peak loads of all axles within a load shape were the same (i.e. the load applied by each
simulated axle within a group was equal).
The pulse width was varied between axle groups to ensure a single travel speed was simulated
for all axle groups (250 ms for a quad-axle down to 250/4 = 62.5 ms for a single axle).
A rest period equal to the difference between the 250 ms and the pulse width was built into the
load shape.
An additional rest period of 250 ms was added to all axle group shapes.
The varying rest period was incorporated in order to keep the travel speed constant for all load
shapes, and the practical constraints of the test control software. Figure 4 shows the combination of
258
Figure 4. Load pulse shapes showing rest periods used in fatigue testing.
Figure 5. Interaction between axle peaks within load pulse shape.
the pulse width and rest period for the standard fatigue test, and the pulses used to simulate single,
tandem, triaxle and quad-axle axle groups.
In the tandem axle simulation shown in Figure 3 the load is seen to drop off between the two
axles. Figure 5 demonstrates the two extremes of load behaviour that could occur between axles
in a group. Using the term interaction to describe the degree to which the two haversine shapes
interact with each other, it can be seen that full interaction, i = 1, results in a load where a single
sustained load is applied, whereas no interaction at all, i = 0, results in two distinct loads.
Figure 6 shows the theoretical static tensile strain generated at the bottom of a cemented pavement
layer loaded with a tandem axle, based on the following:
CIRCLY (Mincad 2008) linear-elastic modelling using the Austroads pavement design approach
was used (Austroads 2008)
The half-tandem axle was modelled, spaced at 1.25 m, with single tyres
Three thicknesses of cemented base were modelled, each with a modulus of 20 000 MPa (i.e.
similar to the samples tested during this study)
The subgrade was modelled with a vertical modulus of 70 MPa.
The interaction between the two axles can be clearly seen in this figure, varying between 0.3 for
the 100 mm cemented base and 0.8 for the 400 mm base.
259
Figure 6. Theoretical shapes of tensile strains at bottom of three cemented bases loaded with tandem axle
group.
Table 2. Description of load pulse shapes.
Shape name Axle group simulated Interaction between axles within group
Haversine Standard tests pulse shape

1_00 Single axle
2_40 Tandem 40% (i = 0.4)
2_80 Tandem 80% (i = 0.8)
3_40 Triaxle 40% (i = 0.4)
3_80 Triaxle 80% (i = 0.8)
4_40 Quad-axle 40% (i = 0.4)
4_80 Quad-axle 80% (i = 0.8)
In order to determine how much this interaction affected the performance of the samples, two
different levels were used for all multiple axle load simulations. The lowest level of interaction that
could be reliably achieved by the pneumatic equipment was i = 0.4. Lower levels could be achieved
with longer pulse widths (i.e. slower travel speeds), however it was decided that it was important
to maintain a reasonably high simulated travel speed to ensure that the tests were representative of
in-service pavements. The other level of interaction was set at i = 0.8.
Table 2 and Figure 7 show the load shapes used in the study. Flexural modulus tests were
conducted using all load shapes, including the standard test haversine shape, and flexural fatigue
tests were conducted using all shapes with the exception of the haversine shape. These synthetic
shapes were generated by combining haversine functions; full details are documented by Moffatt
(2011b).
3 TESTING PROGRAM
3.1 General
Full details of the testing program, and all results obtained can be found in Moffatt (2011b).
3.2 Test sequence

For each sample, the following test sequence was followed:
The sample was prepared for testing.
The flexural modulus was determined using all of the eight load shapes.
Flexural fatigue testing was conducted using a single load shape at a selected load level.
260
Figure 7. Load pulse shapes.
3.2.1 Sample preparation

The saw cut samples were removed from their storage in moist hessian and plastic wrapping and
were placed in a laboratory fog room for a minimum of two days of preconditioning prior to
testing. A sample ready for testing was then removed from the fog room and left standing whilst its
dimensions were measured and recorded. The mass of the beam was measured to enable subsequent
determination of the moisture content of the sample.
The sample was then wrapped in thin plastic cling wrap to minimise any moisture loss during
testing. The relatively loosely wrapped sample was placed in the loading rig and the LVDT support
frame placed and held onto the sample by use of rubber bands anchored to the lower support
rollers. Incisions were made in the cling wrap to ensure that the LVDT probes rested directly upon
the sample. Flexural modulus testing was then conducted.
3.2.2 Flexural modulus test

The modulus testing applied sequences of 100 cycles of each of the load shapes shown in Figure 7,
with each pulse shape having a width of 250 ms and followed by a 750 ms rest period. The 100
cycles of each load shape were separated by a five second rest period prior to application of the 100
cycles of the next shape. Each load shape sequence used the same maximum load level, which was
set by the operator at a low enough value to ensure that the beam was not damaged by successive
cycles, but high enough to ensure that displacement transducer output was sufficient to overcome
any signal noise. This displacement was typically 5 micron (i.e. 0.005 mm). Fatigue testing of the
sample was commenced immediately after the modulus tests were conducted.
261
3.2.3 Flexural fatigue test
Fatigue tests were conducted as load-controlled tests using a single load shape and load level.
The pulse width of the applied load shape was the same 250 ms used in the modulus test, but the
rest period that followed the load application was decreased to 250 ms in line with the standard
procedure.
Rather than conduct tests for a single load shape before progressing to the next load shape,
tests for all load shapes were interleaved so that the fatigue relationship for each shape could be
developed progressively. This allowed for progressive monitoring of results for all shapes, and also
acted to neutralise the effects of any unrecognised progressive systematic errors that may have
occurred over the course of the testing.
During the fatigue testing of some samples it was apparent that the load being applied was below
the fatigue threshold as the modulus was not reducing over time. In some very long term tests,
and in spite of the cling film wrapping, the modulus of the sample was seen to gradually increase
with the slow drying of the sample (i.e. the sample was slowly stiffening and the applied load level
was no longer damaging). In both of these cases the testing was halted, the load level increased,
and testing restarted. These cases were flagged for subsequent examination, and in the majority
of cases they were excluded from subsequent analysis. The only restarted tests that have been
retained were those for which examination of the raw data showed a reasonably smooth continuous
transition between phases of the test. These cases were limited to those where the increase in load
after restarting the test was very minor, but enough for the sample to start to exhibit deteriorating
modulus with cycles. Excluded data is not reported in this paper.
Following flexural fatigue testing the sample was weighed and then oven dried to determine
the moisture content and dry density of the sample. This was done using the dry mass and the
volume of the beam, a method which would yield spurious results if the samples surface had been
eroded during saw cutting. Determination of the sample volume by immersion in water would
have overcome this issue, but was not considered warranted given the limited number of samples
affected by erosion.
3.3 Flexural modulus data

The data collected during each modulus test was carefully examined for the following cases:
A degradation of the sample was observed during the modulus test (i.e. the displacement was
seen to increase with successive applications of the same load level): these cases were noted and
the sample discarded.
The two LVDTs provided significantly different results, with one sensor obviously anomalous:
one case was found, in which it was clear one LVDT had become stuck, and the other was
recording appropriate results.
The two LVDTs provided significantly different results, with no way of determining which sensor
was at fault: in all cases where this occurred there was evidence that the sample had either been
dropped or mishandled, and a note was subsequently made and the sample discarded.
A sudden increase in displacement, as measured by both LVDTs, leading to a drop in calcu-
lated modulus occurred during the testing: this phenomenon was attributed to the expansion
of pre-existing micro-cracks within the sample, in such cases a note was made and the sample
discarded.
The range of modulus values measured under haversine loading across the samples is shown in
Figure 8. It can be seen that the modulus varied over a very wide range, 4000 to over 30 000 MPa,
with the majority of samples exhibiting a modulus between 15 000 and 22 000 MPa. This wide
range of modulus values was unexpected given the uniform nature of the field collected density
and field deflection data, as was the magnitude of the modulus.
262
Figure 8. Distribution of flexural modulus of samples (haversine pulse).
3.4 Test geometry and sample size

The initial loading jig used had a span of 300 mm between the support rollers (dimension L in
Figure 1) and 100 mm between the upper load rollers (dimension l in Figure 1). As the study aimed
to use fatigue testing to develop a relationship between fatigue performance and tensile strain for
each load shape, fatigue testing using a range of strain levels was required. After some modulus
and flexural fatigue tests had been conducted it became apparent that the loading frame was unable
to produce high enough tensile strains at the bottom of the unexpectedly high stiffness material to
generate a significantly wide enough range of strain levels. Whilst the actuator in the pneumatic rigs
was rated as having a capability of applying a 14 kN peak load, it was found that application of the
complex multiple axle load shapes in a 250 ms cycle was only reliably and sustainably achievable
at about 5 kN peak load.
As a result it was decided that a simple change to the test geometry would enable the generation
of higher strains in the samples, without compromising the ability to also conduct tests at lower
strain levels. As the extracted samples were of 400 mm length, a final span of 375 mm was selected
for the revised geometry (i.e. L equal to 375 mm, and l equal to 125 mm in Figure 1). A revised
test jig was designed and two units created, one for each of the two pneumatic systems used in the
study. The revised jig allowed spacings of 375 mm and the original 300 mm. Once these systems
were put into operation, all subsequent testing was conducted at 375 mm spacing.
The increased span of the loading jig allowed the generation of higher strain levels, but not
high enough to produce flexural fatigue of stiff materials in 10 000 to 50 000 load cycles. Having
increased the span length of the sample to the practical limit and rejected changes in pulse width,
the only other recourse was to reduce the cross-sectional area of the samples. Given the maximum
aggregate size in the material was 20 mm, it was determined that 80 mm width by 80 mm depth was
the smallest cross-sectional area that could be achieved without compromising the homogeneity of
the material within the sample.
3.5 Flexural fatigue data

Over 1400 hours of testing time was taken to generate the data represented. In addition to this
testing, many hundreds of hours of compromised data was collected but has not been reported.
This compromised data included the following cases:
tests which did not fatigue the sample

tests which were prematurely halted by electrical power failure
263
tests which were prematurely halted by computer or operator error (unfortunately, these tests
were often long term tests, with the error occurring after more than 500 000 cycles had been
applied)
tests on 80 mm 80 mm beams that had previously been subjected to repeated fatigue test cycles
when sized 100 mm 100 mm
tests that were halted and restarted at higher load levels
fatigue tests which ran for less than 500 cycles
fatigue tests for which the modulus determined at cycle 50 was less than 10 000 MPa (as the
majority of samples exhibited much higher moduli, it was considered that modulus values
lower than 10 000 MPa indicated a substantially different material probably as the result of
micro-cracking within the sample).
4 ANALYSIS
4.1 Flexural modulus

A wide range of flexural moduli was obtained between the samples tested. The samples came from
a small area of constructed test pavement approximately 12 m 1 m in size, from pugmill-mixed
material, placed to a relatively even thickness and compacted with the same compactive effort.
In situ density readings taken with a nuclear density meter and FWD measured deflections, taken
prior to extraction of the samples from the pavement, did not indicate widely varying material
characteristics. After fatigue testing of each sample the density of the sample was determined,
and it was found that the spread of modulus data was not readily matched by corresponding
variations in sample density. The modulus tests were conducted at similar load levels, and so
the variation could be directly attributed to a modulus dependency on load level (a dependency
was observed, see below, but the dependency was not significant enough to explain the range
of data).
Given the lack of alternative explanations, the range of modulus values observed was attributed
to natural variation in material and cement binder composition and distribution, and on the likely
presence of varying amounts of micro-cracking within the samples caused by shrinkage related
stresses.
The modulus of each sample was determined using the standard test haversine pulse shape and
each of the seven multiple axle load shapes shown in Table 2. Statistical analysis, using a series
of paired t-tests and ANOVA, demonstrated that the moduli determined under the different load
pulse shapes were statistically different for each shape. The practical significance of this difference,
however, was extremely minor. At most the modulus variation with load pulse shape was found to
be 5%, and for the vast majority of samples tested the variations were less than 2%. Given the wide
range of material behaviour discussed above, such small variations can be considered to be of little
practical effect.
The flexural modulus determined during the early stages of the fatigue testing of the samples
was found to be consistently 11% to 17% lower than the modulus determined during the modulus
tests (a drop of approximately 2500 MPa). This is demonstrated in Figure 9.
For each sample, flexural fatigue testing was commenced immediately after the modulus tests
were completed, and close examination of the modulus test data indicated that the samples did
not demonstrate a reduction in modulus over the length of the modulus test. Accordingly, the
observed difference in modulus between the modulus test and the flexural fatigue tests could only
be attributed to the difference in the loading regime (i.e. pulse width and load magnitude) used in
the two tests.
Select tests were conducted in order to determine how the differences in loading regimes between
the modulus and fatigue tests may have affected the modulus value measured. Figure 10 shows the
modulus determined for two samples, using haversine loading and a range of pulse widths. The effect
on modulus due to changes in load pulse width was also assessed using samples of two different
264
Figure 9. Difference between modulus during modulus test and start of fatigue test.
Figure 10. Effect of load pulse width on flexural modulus.
cemented materials prepared for a separate study, with similar behaviour observed (Figure 11). The
samples were cut from laboratory compacted and cured samples of the following materials:
basalt crushed rock with 3% cement binder (samples MA and MB in Figure 11)
calcrete limestone crushed rock with 5% cement binder (samples RA in Figure 11).
Both the field constructed and laboratory prepared samples demonstrated moderate drops in
flexural modulus with increased width of the load pulse shape, however the magnitude of the drop
was considered insignificant in comparison to the absolute value of modulus, and the observed
differences shown in Figure 9.
265
Figure 11. Effect of load pulse width on flexural modulus (additional materials).
Figure 12. Effect of strain level on flexural modulus.
The effect of load magnitude, and thus the strains generated, was found to be more significant
than pulse width. The flexural modulus of three samples was tested at different load levels, ranging
from low non-destructive loads to levels where degradation of the samples was evident. Figure 12
shows the results of this testing. By expressing the load level as the strain generated rather than the
absolute force applied, the effects of differences between sample dimensions have been removed.
It can be seen that the modulus value dropped significantly with increasing load level, and that the
difference between the moduli measured at strain levels used during the routine modulus tests with
those measured at typical flexural fatigue test levels is comparable with the difference observed in
Figure 9.
This testing was also repeated using the two additional materials described above and showed
similar results, although the range of strain levels used in this additional testing was limited by the
need to ensure that damaging load levels were not applied, as the undamaged samples were to be
used as part of a separate study.
266
Figure 13. Typical change in flexural modulus during fatigue test.
In summary, it was concluded that the observed difference in flexural modulus between the
modulus and fatigue tests was primarily the result of the different load levels used in the two tests.
4.2 Flexural fatigue

4.2.1 General
Full results of the flexural fatigue testing are documented in Moffatt (2011b). A very wide range
of scatter was observed in the flexural fatigue results. This might be expected given the range of
flexural modulus data discussed above. The scatter, unfortunately, limits the types of analysis that
can be conducted and yet provide meaningful findings.
The test method, apparatus, equipment, control and analysis software used for this study have
produced more uniform data when testing laboratory prepared and cured samples as part of subse-
quent research work, and so the scatter observed here could be attributed, at least in part, to the use,
and possibly the handling, of field extracted samples. Despite uniform construction and the absence
of trafficking, the test pavement would have been subjected to shrinkage related stresses during its
initial curing period, leading to the possible generation of micro-cracks within the material.
Figure 13 shows the reduction in flexural modulus that occurred during a typical fatigue test. The
rate of modulus reduction is very high during the early cycles of loading, before the rate decreases
to a more gradual and steady reduction which is maintained for the majority of the test duration.
Near the end of the test the rate of modulus reduction increases again until the sample breaks. By
defining the initial modulus as the average of the first 50 cycles, the standard method uses data
that is usually contained in the initial rapid reduction phase of the test. During this study it was
found that some samples showed an extremely rapid drop in modulus in only the first few cycles,
and others a less dramatic reduction. The magnitude of the initial reduction seemed to have little
bearing on the gradual rate of deterioration of the sample during the main phase of the tests.
The variety of behaviour of the samples during the initial cycles of fatigue testing was attributed
to the micro-cracked nature of the field samples, with those samples with more micro-cracking
demonstrating a more rapid drop in modulus in the first few cycles of loading than less cracked
samples.
4.2.2 Definition of initial strain level

In order to mitigate this effect, alternative definitions of initial modulus were explored. The
assumption was made that by the 100th cycle of the test that the initial effects of differences
in micro-cracking had occurred, and that the sample was deteriorating in a uniform manner.
267
Figure 14. Comparison of strain at cycle 1 and cycle 100 for all fatigue tests.
Figure 15. Comparison of strain at cycle 50 and cycle 100 for all fatigue tests.
The following figures compare the strain determined during the 100th cycle with different
definitions of the initial strain level. The data for all load shapes and all fatigue tests has been
combined in these figures. Figure 14 shows, by the amount of scatter in the data, that the strain
determined during the first pulse was a poor indication of the strain that would be measured during
the 100th cycle of the test. In comparison, the strain determined during the 50th cycle (Figure 15)
provided a much better indication of the strain level at the 100th cycle. The initial condition in the
standard method, the average of the first 50 cycles, is shown in Figure 16 to be better than the first
pulse alone, but considerably worse than using the 50th cycle.
As a result, the definition of the initial modulus condition of each sample for this study was
defined as the modulus determined during the 50th loading cycle.
268
Figure 16. Comparison of mean of strains of cycles 1 to 50 and cycle 100 for all fatigue tests.
Figure 17. Flexural fatigue results for all axle load shapes.
4.2.3 Definition of fatigue life

The standard method defines the fatigue life of the sample, i.e. the loading cycle at which fatigue
failure is deemed to have occurred, as the cycle at which the modulus has dropped to a value equal
to or less than half the initial value. During the testing program it was observed that breakage of
the sample occurred at relatively the same time as the modulus reached half its initial value. In
some cases, as in the typical response shown in Figure 13, the sample broke before the modulus
was able to drop to half its initial value. In such cases the fatigue life is deemed to be the cycle at
which the break occurred.
4.2.4 Fatigue results

Figure 17 presents the number of cycles to reach fatigue failure, N , as a function of the initial strain
level applied. The parameter N corresponds to the loading cycle, or axle group being simulated,
269
Table 3. Description of load pulse shapes.
Shape name Regression equation R2
1_00 log10 (N) = 29.942 12.91 log10 () 0.55

2_40 log10 (N) = 59.149 27.95 log10 () 0.56
2_80 log10 (N) = 23.586 9.82 log10 () 0.60
3_40 log10 (N) = 36.989 16.44 log10 () 0.63
3_80 log10 (N) = 20.032 8.04 log10 () 0.80
4_40 log10 (N) = 41.007 18.64 log10 () 0.81
4_80 log10 (N) = 16.302 6.13 log10 () 0.84
Figure 18. Summary of flexural fatigue relationships (cycles of axle group).
and so both a pulse of the single axle simulation and a pulse of, say, the triaxle simulation would
correspond to a value of N of one. Linear regression were conducted for each load shape type, and
the resulting relationships are shown in Table 3.
At initial peak strains of about 100 microstrain the load shape, including the number of peaks
within it, would appear to have little effect on the fatigue life. The fatigue life at these strain levels,
however, is very low: approximately 10 000 cycles. With decreasing strain levels, the load shape
has a much more significant effect on fatigue performance.
Examining the relationships using an interaction of 0.8, it can be seen that for a given level of
peak strain the quad-axle simulation (shape 4_80) was more damaging than the tri-axle simulation
(shape 3_80) which in turn was more damaging than the tandem axle (shape 2_80). This matches
intuitive expectation; additional axle repetitions at the same load level would be expected to fatigue
samples at a higher rate. The high variability of the data, and the resulting poor fit of the regression
lines, make a similar comparison difficult for the shapes with an interaction of 0.4.
A distinct difference between the fatigue performance of samples loaded with the same number
of simulated axle groups but with different levels of interaction of the load peaks was observed. For
a given peak strain level the pulse shapes with an interaction of 0.8 have a lower fatigue life than
the corresponding shapes with an interaction of 0.4. A comparison of the slopes, however, suggests
that the fatigue performance of samples loaded with an interaction of 0.4 was more sensitive to
changes in load level. Taking the differences between the two quad-axle simulations as an example,
it can be inferred that fatigue performance is not solely related to either the number of peak loads
within the pulse shape (as both the 4_40 and 4_80 shapes have four peaks) or to the pulse width
(as both shapes were 250 ms wide).
270
Table 4. Areas under load shapes relative to area under
single axle shape.
Shape name Relative area
1_00 1.000
2_40 2.389
2_80 2.786
3_40 3.789
3_80 4.586
4_40 5.178
4_80 6.371
It was hoped that the area under the load pulse shape, or the area under the resulting strain
shape, may provide some explanation for the difference in observed performance. Table 4 lists the
area under each load shape relative to the area under the single axle load shape. As discussed in
Section 2.4, the load shapes were expressed as a series of integer numbers that were scaled by the
control software so that the peak(s) of the shapes corresponded with the application of the desired
peak load level. The absolute area under a specific load shape, say the single axle simulation
shape 1_00, will therefore also be a function of the load level applied. Multiple regression studies
incorporating the relative area and peak loads applied during testing were conducted, and the
relationship shown in Equation 2, with an R2 of 0.53, was found to be significant (all parameters
at the 5% level).
Regressions that included modulus were not found to be significant. In these controlled load
tests the initial strain calculated is, of course, itself a function of the initial modulus both being
estimates based on measured applied loads and displacement responses during the tests.
Equation 2 shows a clear relationship between the fatigue life, the initial strain level and the
combination of the peak load applied and the load type.
where N = number of cycles of axle group load to fatigue; = initial strain (microstrain);
P = peak load applied (kN); RA = relative area under load shape (Table 4).
5 CONCLUSIONS
5.1 General
The wide range of flexural moduli obtained from the samples tested was unexpected. The samples
were all extracted from an untrafficked 12 m2 area of a pugmill-mixed material, placed and cured in
a controlled manner. The variability of the pavement area (thickness, density, deflection response,
etc.) was considered to be as low, if not lower, than that obtained in routine pavement constructions.
The range of moduli values observed has been attributed to the likely varying and uncontrolled
presence of micro-cracking in the material. The test pavement would have been subjected to shrink-
age related stresses during its initial curing period, leading to the generation of micro-cracks within
the material.
As the material is subject to confining pressures whilst in situ, this micro-cracking would probably
not be detectable in either the in situ density readings or the deflection data collected after the
construction process. Due to the lack of sample confinement and bending of simply supported
beams, it would be expected that flexural tests would be particularly sensitive to the presence of
micro-cracking.
271
After conclusion of the testing program reported in this paper, the same test equipment and
software were used, following the standard methods, to determine the flexural moduli and fatigue
characteristics of four different cemented materials. These tests demonstrated markedly less vari-
ability in flexural modulus, and much less scatter in the data used to develop fatigue relationships.
Significantly, all of the samples were prepared in the laboratory using a BP slab compactor.
At the commencement of this study, the use of samples extracted from a pavement was considered
the most practical and timely way in which to proceed with the testing, and it is unfortunate that
this has probably resulted in a varying data set that hampers drawing significant conclusions.
Nevertheless, the data does support some conclusions and hint at others.
5.2 Flexural modulus

The flexural modulus of cemented materials was found to be, for practical purposes, independent
of the duration and shape of the load pulse.
The flexural modulus was demonstrated to be dependent upon the magnitude of the load applied
during the test. The modulus dependence on load for unbound granular materials, under confine-
ment and subject to compressive loads, produces increasing moduli with increasing load level. The
flexural modulus of the cemented material tested showed lower modulus readings with increasing
load level.
5.3 Flexural fatigue performance

At initial peak strains of about 100 microstrain the load shape, including the number of peaks
within it, would appear to have had little effect on the fatigue life. The fatigue life at these strain
levels was very low, approximately 10 000 cycles, and so would most likely not be encountered
in a properly designed pavement. With decreasing strain levels, the load shape has a much more
significant effect on fatigue performance.
The fatigue data resulting from load shapes with an interaction of 0.8 showed that the quad-axle
simulation was more damaging than the triaxle simulation, which in turn was more damaging
than the tandem axle, as would be expected. Additionally, the difference in fatigue performance
between these axle group simulations was not simply a function of the number of peak strain levels
within the shape (i.e. the number of axles simulated within the group).
For a given peak strain level the pulse shapes with an interaction of 0.8 have a lower fatigue life
than the corresponding shapes with an interaction of 0.4. A comparison of the slopes, however,
suggests that the fatigue performance of samples loaded with an interaction of 0.4 were more
sensitive to changes in load level.
Despite the wide amount of scatter in the flexural fatigue results a statistically significant rela-
tionship was determined between the fatigue life, the initial strain level and the combination of the
peak load applied and the relative area under the load shape.
The data indicated that, for a given peak load level, a single axle caused less damage than that
same load level applied on each axle of a multiple axle group with an interaction of 0.8, but caused
more damage than that same load level applied on each axle of a multiple axle group with an
interaction of 0.4. This is unexpected, unexplained and should be investigated further. Rather than
developing an entire fatigue relationship, additional testing could focus on a single strain level,
and see if the relative differences between fatigue performances due to each simulated axle group
can be repeated. Additionally, testing could be repeated at a different (slower) load rate in order to
determine whether the loading rate used in the study had an unforseen effect on the results obtained.
REFERENCES
Austroads 2008. Guide to pavement technology: part 2: pavement structural design. AGPT02/08, Austroads,
Sydney, NSW.
272
Department of Transport, Energy and Infrastructure 2008. Master specification: part 215 supply of
pavement materials. DTEI, Adelaide, SA, viewed 30 July 2008, http://www.transport.sa.gov.au/caps/
division2roadworks_/part215pavementmaterials/part215pavementmaterials.doc.
IPC Global 2007. UTS019 User programmable test software reference manual. January 2007, IPC Global
Ltd, Boronia, Vic.
MINCAD Systems 2008. CIRCLY 5 software reference manual. MINCAD Systems, Richmond South, VIC.
Moffatt, M. 2011a. The influence of multiple axle loads on pavement performance: interim findings. Austroads
Technical Report AP-T184/11, Austroads, Sydney, NSW.
Moffatt, M. 2011b. A laboratory study of the influence of multiple axle loads on the performance of a cement
treated material interim report. Austroads Technical Report AP-T185/11, Austroads, Sydney, NSW.
Standards Australia 1996. Methods of testing soils: method 51: unconfined compressive strength of compacted
materials. AS 114.51.11-1996, Standards Australia, Strathfield, NSW.
Standards Australia 1997. Methods of testing soils for engineering purposes, AS 11289-1997, Standards
Australia, Strathfield, NSW.
Standards Australia 2000. Methods for testing concrete: method 11: determination of the modulus of rupture,
AS 1012.11-2000, Standards Australia, Strathfield, NSW,
Yeo, R. 2008a. The development and evaluation of protocols for the laboratory characterisation of cemented
materials, Austroads Technical Report AP-T101/08, Austroads, Sydney, NSW
Yeo, R. 2008b. Construction report for cemented test pavements: influence of vertical loading on the
performance of unbound and cemented materials, Austroads Technical Report AP-T103/08, Austroads,
Sydney, NSW.
273
Four point bending tests on cement treated materials
G. Gaarkeuken
Ballast Nedam Asfalt, Nieuwegein, The Netherlands
B.W. Sluer
MNO-Vervat, Nieuw-Vennep, The Netherlands
M.M.J. Jacobs
BAM Wegen, Utrecht, The Netherlands
ABSTRACT: The fatigue and stiffness properties of cement treated granular base layers have
a significant influence on the design of a pavement. Ballast Nedam Asfalt has investigated the
possibility of using four point bending tests to determine the functional properties of conventional
AGRAC (cement treated reclaimed asphalt as specified in the Dutch Standard, with approximately
2% cement). Four point bending tests on material cut from a pavement and on laboratory made
AGRAC have been done. Influences of testing procedure, dimensions of the prismatic beams, strain
value, clamping and others are analyzed. Comparable tests have been carried out by BAM Wegen
and Ballast Nedam Asfalt on laboratory made cement treated sand.
The European Specifications for bituminous mixtures EN 12697-24 (Resistance fatigue) and
EN 12697-26 (Stiffness) are used for the tests. The research has resulted in a protocol for testing
and processing of the functional properties from the test results.
1 INTRODUCTION
Especially on intensively and heavily loaded asphalt concrete pavements cement treated base
courses are often used. On heavily loaded pavements the load conditions result in a large asphalt
thickness with granular base materials. Cement treated base layers lead to a reduction of the
construction costs due to smaller asphalt thicknesses. To reduce the risk of crack reflection of a
pavement with a cement treated base layer usually a minimum asphalt thickness is specified in
the design. Furthermore failure of the cement treated base due to an instant exceed of the bending
strength is normative for the design of the pavement. Usually fatigue failure of the cement treated
base is not the leading design criterion.
Figure 1 shows the much used fatigue line of a cement treated sand of SAG-VNC (SAG, 1996). As
a comparison the fatigue line of a cement treated reclaimed asphalt with 3.5% of cement (CROW,
2005) is added. Although the composition of the showed cement treated re-claimed asphalt is
slightly different from the Dutch Standard (approximate cement percentage is 2) it visualizes the
difference in fatigue behavior.
When the load conditions (number of axle loads and axle load distribution) are known, for both
bound base materials a fatigue design criterion can be used. However, the fatigue line of a cement
treated sand is very steep and only a single exceptionally heavy wheel load can lead to instant
failure. That is why instant failure due to exceed of the bending strength is used as the normative
design criterion. Cement treated reclaimed asphalt has a higher bending strength compared to
cement treated sand and this increases the resistance of a cement treated reclaimed asphalt base
layer to instant failure due to an accidental high overload. For cement treated reclaimed asphalt
275
Figure 1. Fatigue characteristics of cement treated sand an cement treated reclaimed asphalt.
the resistance against fatigue is leading. For cement treated reclaimed asphalt a design based on
exceed of the bending strength leads to unnecessary large pavement thicknesses.
Based on the previous analysis one can conclude that the relevant mechanical parameters for the
determination of the service life of a pavement with a cement treated base are:
The stiffness of the cement treated base;

The bending strength of the cement treated base;
The fatigue characteristics of the cement treated base.
This paper handles the determination of stiffness and fatigue characteristics of a cement treated
soil/sand/fine granular mix (by BAM Wegen), cement treated sand (by Ballast Nedam Asfalt) and
a cement treated reclaimed asphalt (by Ballast Nedam Asfalt) with the four point bending test.
The test protocols EN 12697-24 and NEN 12697-26 for the determination of fatigue and stiffness
characteristics are used. These protocols are specified for asphalt concrete mixtures. Remarks are
made for the applicability of these protocols for cement treated materials. To make the protocols
applicable for cement treated materials some adjustments to the protocols have been made. Also
some remarks are made for the processing of the test results.
2 CEMENT TREATED SAND AND CEMENT TREATED SOIL/SAND WITH ADDITIVE
Ballast Nedam Asfalt has tested a cement treated sand with 8% (m/m) of cement. Stiffness and
fatigue test have been executed on prismatic beams 450 50 50 mm in accordance with EN
12697-24 and EN 12697-26. BAM Wegen executed fatigue tests on a cement treated soil/sand/fine
granular mix with 810% cement and an additive. The additive is used for immobilization of
polluted particles in the base material. At the end thirteen prismatic beams have been tested on
fatigue by BAM Wegen, but only six tests were successful. Ballast Nedam Asfalt performed nine
fatigue test but only five were successful. The possible causes of this high failure percentage in the
fatigue test are:
The inhomogeneity of the base material and its aggregate sizes in relation to the dimensions
of the beam. Prismatic beams of 450 50 50 mm are used (BAM Wegen), whereas the base
material has a diameter up to 16/22 mm;
276
Figure 2. Frequency sweep cement treated sand.
The lack of experience with performing fatigue tests on base materials. This makes it difficult
to assess the necessary strain level (BAM Wegen);
Applied initial strains at the moment the beams are clamped (Ballast Nedam Asfalt and BAM
Wegen)
Brittle behavior of the cement treated base materials.
Ballast Nedam Asfalt performed six frequency sweeps to determine the stiffness of the cement
treated sand. A stiffness of approximately 10000 MPa is measured with a very small variation.
There is no frequency dependency of the stiffness observed, see Figure 2. Also the phase angle is
very small. In fact, the material behaves elastic.
The results of the fatigue tests are summarized in Table 1 (BAM Wegen) and Table 2 (Ballast
Nedam Asfalt).
Based on the results the following conclusions are made:
1. De phase angle of the cement treated sand is almost equal to zero. The mechanical behavior of
the material is considered to be elastic. The initial phase angle of the cement treated mix with
an additive is about 3 to 5 . Although the additive has increased the phase angle compared to
the mixture without additive, the phase angle is still very small. The material behaves almost
elastic. The energy loss due to visco-elastic behavior is negligible.
2. The cement treated sand without additive has a much higher stiffness. The lower stiffness of the
soil/sand/fine granular mix is mainly caused by the relatively high content of fines in the mix.
Also the additive is supposed to give the material some flexibility which might result in a lower
stiffness.
3. The variation in stiffness in the cement treated sand is small. There are six stiffness tests per-
formed with an average stiffness of 10186 MPa and a standard deviation of 332 MPa. The average
initial stiffness from the fatigue test was 9543 MPa with a standard deviation of 168 MPa. The
variation in the initial stiffness of the cement treated mix with additive is large. This is probably
due to the inhomogeneity of the components, which in fact is waste material.
4. The variation in obtained fatigue life (Nfat ) for the cement treated mix with additive is large
compared to the cement treated sand. Again the inhomogeneity of the constituent materials
might be the main cause. Also the size of the cross section of the prismatic beam can be of
influence.
277
Table 1. Fatigue results cement treated sand with additive (BAM Wegen).
Strain Initial stiffness Initial phase angle Nfat Phase angle

Sample [m/m] [MPa] [] [] [ ]
P2B3 143 7609 3.6 3.08 106 3.3

P2B4 152 5406 5.2 5.62 102 15.4
P2B5 165 3375 5.3 3.32 103 8.0
P3B2 143 2381 8.0 4.58 106 4.3
P3B3 145 5164 3.7 4.54 106 3.3
P3B4 144 6220 3.9 3.28 106 3.3
P1B3 200 Direct failure after start of the test
P2B1 Failed during frequency sweep
Log(k1) 121.96
k2 53.70
r2 0.52
6 146
This is an estimated value obtained by linear extrapolation of the available data.
Determined values are inaccurate. Only six tests were performed successfully with four extrapolated results.
Table 2. Fatigue results cement treated sand (Ballast Nedam Asfalt).
Strain Initial stiffness Initial phase angle Nfat Phase angle

Sample [m/m] [MPa] [] [] [ ]
Beam1 160 9264 0.5 8.80 102 14.4

Beam3 140 9641 0.4 2.75 105 5.3
Beam10 150 9539 0.6 4.32 104 6.8
Beam2 130 9572 0.5 >4 106 5.5
Beam4 135 9701 0.4 >4 106 5.4
Beam5 150 Direct failure after start of the test
Log(k1) 97.476
k2 42.80
r2 0.953
6 137
Determined values are inaccurate. Only five tests were performed successfully.
5. A very steep fatigue line for both mixtures is obtained (k2 = 53.70 and k2 = 42.80). This
means that the fatigue life is very sensitive to the applied strain level. In fact one must conclude
that instant failure due to exceed of bending strength or strain at break instead of fatigue is the
decisive design criterion. For instance: a safe failure criterion could be a maximum strain level
of 125130 m/m with additive and 115120 m/m without additive, which corresponds with
approximately 109 load cycles.
6. In the results of a fatigue test the brittle behavior of the cement treated sand is visible. Instead
of a gradual decline of the stiffness, the stiffness decreases very abrupt, see Figure 3. Instant
278
Figure 3. Fatigue test cement treated sand.
failure occurs instead of fatigue failure. In contrast to a fatigue test on asphalt concrete a cracked
beam is the final result, even if the fatigue criterion of 50% stiffness reduction is applied.
7. Fatigue testing on cement treated sand and cement treated soil/sand/fine granular mix specimens
is difficult to execute. A strain level of for instance 145 m/m can result in a fatigue life of 106
load cycles whereas a strain level of 150 m/m results in almost direct failure.
3 CEMENT TREATED RECLAIMED ASPHALT
An extended research is carried out by Ballast Nedam Asfalt on the behavior of cement treated
reclaimed asphalt as specified in the Dutch Standard. The following mixture has been examined:
6% water
80% (m/m) reclaimed asphalt
20% (m/m) sand
2% (m/m) cement (Hoogovencement CEM IIIB 42.5N)
This mix composition is used in a test section on Highway 12 near Zoetermeer in the Netherlands.
The following material characteristics of the mixture are obtained:
Compressive strength on cylinders 150 mm taken from the test section;
Indirect tensile strength on cylinders 150 mm taken from the test section;
Stiffness of the mixture by 4 point bending test on prismatic beams 70 70 450 mm cut from
the test section and on laboratory made prismatic beams of the same mixture;
Fatigue characteristics of the mixture by 4 point bending test on prismatic beams
70 70 450 mm cut from the road and on laboratory made prismatic beams of the same
mixture;
Bending strength of the mixture by 3 point bending test on prismatic beams 40 40 160 mm
cut from the test section.
As indicated before the tested material is taken out of a test section of a real pavement and from
laboratory made material. Furthermore the influence of the number of load repetitions on stiffness
and fatigue characteristics and the influence of the density on the material properties is examined.
279
Table 3. Frequency sweep cement treated reclaimed asphalt, highway section.
Frequency Strain level E Phase angle

[Hz] [m/m] [MPa] []
0.1 46.6 3025 11.6

0.2 46.4 3250 11.6
0.5 47.7 3572 11.8
1.0 47.6 3847 11.8
2.0 47.2 4149 11.6
5.0 46.7 4563 11.5
8.0 46.5 4785 11.3
10.0 46.3 4883 11.2
20.0 46.0 5235 10.8
30.0 46.1 5436 10.7
0.1 46.4 2985 11.5
Average (8 Hz) 4785
St. dev. (8 Hz) 945
85% reliability (8 Hz) 3805
3.1 Compressive strength and indirect tensile strength

The Dutch standard requires a compressive strength of 1.0 MPa seven days after production and a
compressive strength of 2.0 MPa 28 days after production of the cement treated reclaimed asphalt.
From the highway section 40 cylinders 150 mm are cored and taken to the laboratory. At 28 days
after production 20 cylinders are used to determine the compressive strength (20 C). An average
value of 2.9 MPa with a standard deviation of 0.9 MPa is determined. Additionally for 20 cylinders
the indirect tensile strength (20 C) is measured (speed of deformation of 50 2 mm/min). An
average value of 0.56 MPa with a standard deviation of 0.11 MPa is determined.
For all the measured cylinders the dry density is compared with the measured strength values.
For as well as the compressive strength as the indirect tensile strength a very small dependency of
the density is visible. Higher densities lead to higher strength values.
These tests are repeated on laboratory made samples of the same mixture. At 28 days after
production an average compressive strength of 2.5 MPa with a standard deviation of 0.4 MPa is
determined. At 100 days after production the average value increases to 3.3 MPa with an standard
deviation of 0.3 MPa. At 28 days after fabrication an average value for the indirect tensile strength
of 0.42 MPa with a standard deviation of 0.10 MPa is measured. At 100 days after fabrication the
average value increases to 0.59 MPa with a standard deviation of 0.12 MPa. Based on the test results
one can conclude that the mixture fulfills to the Dutch specifications for cement treated reclaimed
asphalt.
3.2 Stiffness and fatigue

From the highway section a slab of the cement treated reclaimed asphalt base layer is cut. In the
laboratory 18 prismatic beams 70 70 450 mm are sawn from this slab. For these prismatic
beams the stiffness and fatigue properties are determined in accordance with EN 12697-24 and
EN 12697-26. Tests are carried out approximately 28 days after production and compaction of the
cement treated reclaimed asphalt. In Table 3 the average results of the frequency sweep with respect
to the stiffness and phase angle of the prismatic beams are presented.
Despite the cement treatment there is a frequency dependency of the stiffness visible (also
Figure 4). Compared with the results of the cement treated sand the phase angle is higher. Some
visco-elastic behavior is observed. The influence of the bitumen on the mechanical behavior of
the cement treated reclaimed asphalt mixture cannot be neglected. The mechanical behavior of the
mixture is also expected to be dependent on the temperature.
280
Figure 4. Frequency sweep cement treated reclaimed asphalt.
Table 4. Stiffnesses of laboratory and in situ samples.
Highway samples 28 days Laboratory 100 days Laboratory 28 days
Average stiffness 4785 4544 4524

St. dev. 945 241 393
85% reliability 3805 4294 4117
The stiffness tests performed on the laboratory prepared samples show similar results. The tests
are performed 28 days and 100 days after production. Table 4 compares the 8Hz stiffness results.
For all the prismatic beams (from the highway test section as well as the laboratory made samples)
the fatigue characteristics are determined. The results are shown in Figure 5.
The fatigue lines of the 28 days old prismatic beams and the 100 days old prismatic beams
compare very well. The fatigue line of the cement treated reclaimed asphalt from the highway test
section is a bit lower. The stiffness of the 100 days old laboratory beams is lower than the 28 days old
laboratory beams. This was not expected. Also the fatigue line of the 28 days old prismatic beams
is steeper. When the dynamic properties of a beam are compared with its density the conclusion
is drawn that the mechanical behavior is dependent on the density. A higher density leads to a
higher stiffness and a steeper fatigue line as a consequence (Figure 6). With an aimed density of
2050 kg/m3 one can conclude that the highway section was overcompacted.
When the variation of the results is included in the characterization the dynamic properties of
the cement treated reclaimed asphalt for road design are enumerated in Table 5.
3.3 Bending strength

The bending strength is determined on 40 40 160 mm prismatic beams sawn from the test
section on Highway 12. The EN 196-1 is used as test protocol. In the 3-point bending test a dis-
placement speed of 0.085 mm/s is used. An average bending strength of 1.13 MPa is measured with
a standard deviation of 0.35 MPa. With an assessment of the stiffness of approximately 4000 MPa
this corresponds with a failure strain of 250300 m/m. Instant failure due to an exceed of the
bending strength or strain at break by a single large wheel load is not very likely to happen. The
service life of the cement treated reclaimed asphalt seems to depend on fatigue.
281
Figure 5. Fatigue characteristics cement treated asphalt concrete.
Figure 6. Stiffness versus density cement treated reclaimed asphalt.
Table 5. Characteristic properties of cement treated reclaimed asphalt.
Road type Reliability [] Fatigue Initial Stiffness [MPa]
Primary road 85% Log(N) = 17.3765.865.log() 4230

Secondary road 75% Log(N) = 17.4895.865.log() 4480
Low order road 70% Log(N) = 17.5365.865.log() 4590
Dutch categories of roads based on road order.
Corresponding reliability for road design.
282
4 TEST PROCEDURE 4 POINT BENDING TESTS ON CEMENT TREATED MATERIALS
The experiences of BAM Wegen and Ballast Nedam Asfalt with the execution of 4 point bending
tests on cement treated materials lead to certain preconditions for sampling, for the test protocol
and for processing of the data.
4.1 Production of prismatic beams

The EN 12697-24 and EN 12697-26 prescibes a maximum aggregate size dependent dimen-
sion of the prismatic beam. With a maximum aggregate size of 16 mm a prismatic beam of
450 50 50 mm fulfills these requirements. With larger aggregate sizes (up to 32 mm) the dimen-
sions should increase to 450 70 70 mm. In the Dutch application of the EN protocols for testing
asphalt concrete specimens all the beams have dimensions 450 50 50 mm. To decrease the
influences of inhomogeneities and variations in bonding and stone size it is recommended to use
the maximum dimensions.
For the manufacturing of the beams there are different options:
Mixing of the mortar in plant combined with manufacturing of the beams in the laboratory;
Mixing of the mortar and fabrication of the beams in the laboratory;
In situ sampling of material from a pavement.
The type of manufacturing will have an influence on the results. The mechanical properties
determined from in situ samples compare best with the expected mechanical behavior in practice.
The degree of bonding by the cement is dependent on the storage time of the beams. For the
determination of the compressive strength a standard storage time after manufacturing of 7 and/or
28 days is used. For four point bending tests it is recommended to use a storage time of 28 days.
In contrast with the compressive strength there is no empirical relation between 7 days and 28 days
stiffness and fatigue characteristics available.
4.2 Clamping
In comparison with asphalt, which is highly viscous with large loading times, a cement treated
material shows brittle behaviour. Stresses introduced by small deflections due to inaccurate posi-
tioning of the beam before closure of the hydraulic or mechanical clamps will relax in an asphalt
concrete beam. On cement treated materials however the stresses and displacements introduced
by in accurate positioning before clamping can result in exceeding the tensile strength and instant
failure immediately after clamping. For instance with the following assumptions
E = 4000 MPa
A = 140 mm (distance between inner clamps)
b = 70 mm
h = 70 mm
a displacement offset of 0.2 mm will result in a force of 1.13 kN and a tensile strain of 347 m/m
with direct failure as a consequence. Very precise positioning of the prismatic beam before closure
of the hydraulic or mechanic clamps is absolutely necessary. The intitial stresses and strains after
clamping can be verified by checking the force directly after clamping. A precise adjustment of
the position of the inner clamps by hand is necessary to prevent or minimize initial force offset at
the start of the test.
4.3 Test protocol stiffness

In the EN 12697-26 for asphalt concrete a strain level of 50 m/m is used. This strain level is high
enough for an accurate measurement of the stiffness without any risk of damaging the prismatic
beam. The same prismatic beam can be used for a fatigue test. Although with a cement treated
283
Figure 7. Typical asphalt concrete fatigue curve.
material the risk of damage is higher, a strain level of 50 m/m is still safe. In case of a frequency
sweep it is recommended to use the same frequency at the beginning and the end of the test. By
comparing the first and last stiffness value one can conclude if damage has occurred in a test sample
or not.
In contrast with asphalt concrete the stiffness of cement treated materials is less dependent on
the temperature. It is recommended to use enough time for acclimatization of the beams to get to
a homogeneous temperature in the test samples and to use a constant temperature during the tests.
A temperature of 15 C or 20 C can be used.
4.4 Test protocol fatigue

In the EN 12697-24 three strain levels are used. For each strain level six prismatic beams are
tested which results in 18 fatigue tests. The stiffness of the prismatic beam after 100 load cycles is
measured. The fatigue life is defined as the number of load cycles to reduce the stiffness to half
of its initial value. Although 50% reduction of the stiffness as a criterion is an arbitrary choice, for
the fatigue life of an asphalt concrete it is defendable. Usually when half of the initial stiffness of
an asphalt concrete is reached there is a sharp decline of the stiffness per load cycle (Figure 7).
In a fatigue test on cement treated materials however, the form of the stiffness versus load cycle
curve can be significantly different. In test performed on cement treated reclaimed asphalt when
half of the initial stiffness is reached there is no sharp decline of the stiffness (Figure 8). Further
continuation of the test did not result in any visual cracks. This is in contrast with the fatigue test
on cement treated sand where instant failure occurred (Figure 3).
It is interesting to observe that, although the stiffness decreases extensively, the phase angle is
almost constant. Reclaimed asphalt will have a bitumen content of approximately 45% by mass
at which 2% of cement by mass is added. Because of the aged bitumen in the reclaimed asphalt,
the penetration of the bitumen will be low. The frequency sweep (Table 3) showed a frequency
dependency on the stiffness of the cement treated reclaimed asphalt. Combined with the gradually
decline of the stiffness (Table 3) the conclusion is made that bitumen has an influence on the
functional behavior of the cement treated reclaimed asphalt. In fatigue tests on asphalt mixtures
with hard graded bitumen a phase angle of 1015 is quite common.
284
Figure 8. Typical cement treated reclaimed asphalt fatigue curve.
With the properties stiffness, fatigue and bending strength of a cement treated base a structural
design of a pavement is possible when truck intensities and axle load distribution are available.
Instant failure by a maximum axle load or fatigue due to repeated axle loads is normative. For
cement treated mixtures with steep fatigue lines (k2 < 10) usually instant failure is normative.
The determination of the fatigue properties of a cement treated mixture with a steep fatigue line
(for instance cement treated sand) is very difficult. The use of a slightly too low strain level results
in an infinitely long test, whereas the use of a slightly too high strain level results in instant failure
of the beam. For these materials instant failure instead of fatigue will be normative in the design
and the difficult fatigue test can be skipped. That is why it is important to have an estimation of
the brittleness (or flexibility) of the material in advance. It is recommended to do the stiffness
test with a frequency sweep. If the stiffness is relatively independent on the applied frequency and
temperature, one can conclude that the material is brittle and instant failure will be normative. A
fatigue test is not necessary.
Except stiffness, fatigue and bending strength there are some additional characteristics that are of
importance (Jacobs, Sluer, 2010). For instance sensitivity for shrinkage, moisture and frost are well
known problems for cement treated materials. Although the determination of those characteristics
was not a part of this investigation, the following tests are recommended:
Density; As Figure 6 showed the stiffness and fatigue properties are very dependent on the
density of the mixture. High compaction normally results in better mechanical properties.
Semi Circular Bending Test (EN 12697-44); The test leads to a failure strain for the mixture
instead of a failure stress (bending strength EN 196-1). The SCB tests has the advantage that it
can be performed on cored cylinders. Additional tests on carved specimen will give information
about the sensitivity to crack reflection (Paris Law).
Moisture sensitivity; Due to moisture absorbance some bounded granular materials have a
tendency to expand (for instance slags). The sensitivity can be determined with the help of the
French test NF P 98-121.
Frost sensitivity; With freeze thaw tests (for instance EN 12371) the sensitivity for cracking by
temperature loads can be determined. Approximately 12 freeze thaw cycles are enough;
285
Shrinkage; The binding of the aggregate is usually accompanied with a decrease of the volume
of cement treated granular materials. The level of shrinkage determines the risk of shrinkage
cracks. At the Delft University, Faculty of Civil Engineering a shrinkage test is developed which
can be used.
REFERENCES
EN 12697-24, 2011. Bituminous mixtures Test methods for hot mix asphalt Part 24: Resistance to fatigue.
EN 12697-26, 2004. Bituminous mixtures Test methods for hot mix asphalt Part 26: Stiffness.
Jacobs, M. & Sluer, B. Infradagen 2010, Voorstel Typeonderzoek gebonden funderingsmaterialen.
Van Gurp, C.A.P.M., CROW, april 2005, Achtergrondrapport Keuzemodel wegconstructies.
Vereniging Stabilisatie Aannemersgroep (SAG) en Vereniging Nederlandse Cementindustrie (VNC), 1996.
Dimensioneren met Zandcement.
286
Repeated load flexural testing to characterize lightly stabilized

granular materials
D.K. Paul & C.T. Gnanendran

School of Engineering and Information Technology, University of New South Wales at ADFA,
Canberra, Australia
ABSTRACT: This paper presents a new and improved flexural beam testing setup for characteri-
zation of granular base materials lightly stabilized with cement-flyash binder for pavement design.
All the samples were prepared at their Optimum Moisture Contents (OMCs), cured for 28 days and
then tested under cyclic loading conditions at different stress levels. The newly developed setup
allowed accurate measurement of the mid-span vertical deflection of beam specimens from their
neutral axis. The test results indicated that the proposed cyclic load flexural testing apparatus can be
used to adequately characterize lightly stabilized granular base materials. It has also been observed
that resilient modulus (Mr ) of such materials is dependent on the level of stress application.
1 INTRODUCTION
From pavement structural viewpoint, stabilized base layer is subjected to flexure due to the wheel
load on the top/surface layer. As a result, tensile stress develops at the bottom and compressive
stress at the top (Figure 1). Thompson (1966) has shown that compressive and shear strength
properties are not the limiting factors for the application of chemically stabilized materials as base
or sub-base materials. It is the tensile properties based on which a stabilized base layer is designed
in mechanistic design of road pavement.
There are several testing methods, such as direct tensile, indirect tensile and flexural, for assessing
tensile characteristics of stabilized materials. Of them, flexural beam testing is preferred by many
pavement engineers because the stress conditions it produces closely represent those experienced in
the road pavement under wheel loading (e.g. AUSTROADS 2004; NCHRP 2004). This testing also
offers many advantages over other commonly used tensile testing methods. It produces a uniform
tensile stress at the outer bottom fiber and, unlike a direct tension test, does not require any special
Figure 1. Multilayer flexible system and inputs required for mechanistic design.
287
sample gripping arrangement which could induce stress concentration and consequently affect the
test results.
There is a number of literature regarding flexural testing heavily (normally) stabilized materials
(Moffat and Yeo 1998) and foamed asphalt mixtures (Fu el al. 2009). However, to the best of the
authors knowledge, there is very little literature regarding the characterization of lightly stabilized
materials by flexural beam testing, particularly under cyclic loading conditions. Several researchers
(Foley et al. 2001; Gnanendran and Piratheepan 2008; Gnanendran and Piratheepan 2010) reported
possible practical difficulties in preparing and handling a beam with a low amount of binder,
and therefore used/suggested other testing methods (i.e., indirect diametrical tensile testing in
particular) to determine the tensile properties of lightly stabilized materials.
In this study, a laboratory investigation was carried out to characterize lightly stabilized granular
materials in terms of their flexural properties. A typical granular base material was stabilized
with 13% cement-flyash binder and dynamically tested using a newly developed equipment.
Details of the material and binder used, sample preparation, curing technique, testing setup and
analysis of test results are presented in this paper.
2 EXPERIMENTAL INVESTIGATION
In this experimental investigation, a typical lightly stabilized granular material was stabilized
with 1%3% binder content. Dry density and molding moisture content are two important factors
which significantly affect the performances of pavement materials. Therefore, the Standard Proctor
Compaction test was carried out to establish the dry-density-moisture content relationship according
to AS-1289.5.1.1. (2003) which is similar to ASTM D698 (2007). The MDD of the parent material
without any binder was found to be 2088 kg/m3 whereas the addition of 3% cement-flyash increased
this value to 2157 kg/m3 with its OMC remaining almost constant at 9%. Therefore, all the samples
stabilized within this small binder range (1%3%) were prepared with a fixed moisture content of
9% throughout this study.
2.1 Materials
The parent material selected for this investigation was a typical freshly quarried crushed rock which
was obtained from the quarries located in Canberra, Australia. It was classified as well graded
sandy gravel with some fines according to Unified Soil Classification System. However, to get
consistent samples from such a crushed rock materials is very difficult and unreliable proportions
of particle in the samples are common. To overcome this potential inconsistency, reconstituted
material with an unchanged (or consistent) material grading shown in Figure 2, was adopted. This
reconstituted material, of consistent grading for the entire test program, was achieved by sieving
a large batch of parent material, separating them into different particle series and then remixing
them at suitable weight proportions. Therefore the adopted grading for the reconstituted sample,
hereafter referred as the parent material, was essentially the same for all the samples that were
tested in this experimental investigation.
As indicated earlier, the stabilizers used in this experimental study were general blend cement
and flyash. The cement and flyash were used in the ratio of 75% to 25% by dry weight. The parent
material was stabilized with 1.53% cementflyash for this laboratory investigation.
2.2 Sample preparation and curing

Flexural beam specimens of dimension 285 mm 76 mm 76 mm were prepared according to
ASTM D 1632-07 (2007). These specimens were compacted by applying static load through a
hydraulic compression machine. It is noted that all the prepared samples achieved 99.5% Standard
Proctor maximum dry density at the OMC. The prepared specimens were wrapped with polythene
288
Figure 2. Particle size distribution of parent material.
Table 1. Physical properties of parent material.
Property Value
Liquid limit (%) 18
Plastic limit (%) 15
Plasticity index (%) 3
Linear shrinkage (%) n/a
Optimum moisture content (%) 9.0
Maximum dry density (kg/m3 ) 2092
For fine grained portion of the material (i.e. material <425 m)
and put in a fog room at 23 2 C and 95 5% humidity for curing. After 28 days, the specimens
were taken away from the fog room and tested immediately in flexure.
2.3 Testing using newly developed flexural testing setup

In this investigation, a cyclic load flexural testing program has been carried out to determine the
mechanical properties of lightly stabilized materials by means of improved experimental setup. In
conventional flexural testing, mid-span vertical deformation is generally measured externally which
may include some extraneous deformations due to particle crushing at specimen support and/or
because of movement in the machine frame. As a result, the load-deformation curve obtained may
not be representative. To eliminate such erroneous vertical movements, a new and sophisticated
experimental setup has been developed in this study to monitor the net mid-span vertical deflection
of the beam.
The new setup, as shown in Figure 3, consisted of two horizontal bars aligned along the neutral
axes of the beam specimens. A pair of C-shaped aluminum brackets was used to support the
horizontal bars. These bars was pin jointed at one end and roller supported at the other (i.e. at
one end, the bars was held by bolts around which they could rotate but not move horizontally or
vertically with respect to a point on the neutral axis of the sample and at the other end they were
289
Figure 3. Photographic view of experimental setup used in dynamic testing.
rested on another bolt so that they could slide horizontally but not move vertically). At the middle
of each horizontal bar, a LVDT was mounted through a hole and tightened using a plastic bolt. The
LVDT tips touched the L-section aluminium angles which were attached to the beam specimen with
the help of glue as showed in the Figure 3. Therefore, it was possible to capture more accurately
and reliably the center deflection of the beam from the neutral axis with this newly developed
measurement setup.
A set of six specimens were prepared for each binder-parent material combinations Three speci-
mens were tested for determining modulus of rupture (MOR) without any deflection measurement.
The other three were tested for determining resilient modulus (Mr ) by applying different stress
levels. This was stress/load controlled testing and the frequency of the sinusoidal-type loading was
set at 3 Hz. Since the minimum cyclic load for a cycle equal to zero could lead to rocking action
and the range of the stress/load would also have an effect on the dynamic properties of the material
(Murdock and Kesler 1958), a small amount of constant load (e.g., 10% of the MOR) was kept
as a seating load. The applied stress levels for Mr test were chosen according to the MOR of the
specimen of each test, as shown in Table 2.
The maximum cyclic stress levels, referred to hereafter as the sub-maximal cyclic stress ratios,
were calculated to be 5070% for 1% binder content due to their low strength values and 3070% for
the rest of the specimens. It should be noted that the applied cyclic flexural stress ratios (SRs) were
essentially 10% less than the sub-maximal stress ratios. Thus, each sample was tested at several
stress ratios for a certain number of load cycles (i.e., generally 200 cycles for each sequence).
Before testing for the stiffness modulus, each specimen was pre-conditioned by applying 1000
cycles at the minimum stress ratio adopted for the particular material. This sequence was important
as it eliminates any permanent deflection accumulation and/or bedding error during the stiffness
testing (Elliott and Lourdesnathan 1989; Mohammad et al. 2006; Pan et al. 2006). Details of the
dynamic testing procedure are presented in Table 2.
290
Table 2. Sequence used for dynamic flexural testing.
Sequence Sub-maximal Cyclic Number of

Case number flexural stress flexural stress cycles
MOR < 300 kPa Pre-conditioning 0.5MOR 0.4MOR 1000

(i.e., with 1% 1 0.6MOR 0.5MOR 200
binder content) 2 0.7MOR 0.6MOR 200
MOR > 300 kPa Pre-conditioning 0.3MOR 0.2MOR 1000
(i.e., rest of 1 0.4MOR 0.3MOR 200
specimens) 2 0.5MOR 0.4MOR 200
3 0.6MOR 0.5MOR 200
4 0.7MOR 0.6MOR 200
The Mr for each stress sequence was calculated from Equation 1 suggested by Irwin and Gallaway
(1974) which includes the center deflection due to both the bending moment and shear under third-
point loading. It should be noted that the expression inside the bracket is the correction factor for
the shear deflection associated with the four-point bending test. It varies with the square of h/L and,
for higher values of h/L, the deflection due to shear should not be ignored (Pronk and Huurman
2009). For the given aspect ratio (i.e., 0.333) of beam specimens utilized in this study, contribution
to the total deformation by shear should be approximately 10%.
where Mr = Stiffness modulus in MPa and P = load in N, = repeated deflection in mm and

= Poissons ratio of the lightly stabilized material. The value of Poissons ratio of lightly stabilized
materials was assumed to be 0.2 as per AUSTROADS (2004) recommendation.
3 PRESENTATION AND DISCUSSION OF RESULTS
3.1 Modulus of rupture

The modulus of rupture, MOR, of soil-cement beam is determined by using Equation 2 provided
that the fractures occurred within the middle third of the beam span length.
where MOR = modulus of rupture in MPa, P = maximum applied load in N, L = span length in mm,
b = average width of specimen in mm and d = average depth of specimen in mm.
Figure 4 shows the variation of MOR of lightly stabilized granular materials with respect to
binder content. It is clear that the MOR increased almost linearly with the increase in binder
content. For example, for the amount of binder added from 1% to 3%, the MOR increased from
0.1 to 0.5 MPa. Difference of individual MOR values from its mean for different combinations of
binder was also shown in terms of error bar in Figure 4. As can be seen, the deviation was very
reasonable (less than 10%) and hence, the MOR can accurately be determined using this flexural
beam testing setup.
291
Figure 4. Variation of MOR with binder content.
Table 3. Mr of lightly stabilized materials at different SR.
Mr
1% binder 2% binder 3% binder
Applied SR Mean (MPa) CoV (%) Mean (MPa) CoV (%) Mean (MPa) CoV (%)
0.3 4652 19.1 6458 14.6

0.4 1255 8.1 4376 14.9 6221 15.0
0.5 971 14.2 4179 18.7 5903 11.5
0.6 909 7.0 4080 24.4 5644 8.7
3.2 Mr in flexure
The variations of Mr at different stress levels, determined using Equation 1, are presented in Table 3.
As expected, the Mr increased with the increase in stabilizing agent. For instance, the Mr at 50%
SR increased from just 971 MPa to 5903 MPa when the binder content increased from 1% to 3%.
The data scatter of Mr in terms of coefficient of variation (CoV) is also shown in Table 3. The CoV
was generally found to be less than 15% except for those with 2% binder content.
The influence of SR on the Mr of lightly stabilized materials was also studied. As depicted in
Figure 5, the magnitude of Mr decreased as the SR increased. This trend was consistent for all
three cases of binder content. For example, the Mr of specimens stabilized with 2% binder content
decreased from 4652 MPa to 4080 MPa when the SR increased from 0.3 to 0.6. This decrease in
Mr varied from 3% to 23% per 0.1 increases in SR, normally being more sensitive at lower binder
content.
292
Figure 5. Variation of Mr with SR.
4 CONCLUSION
Flexural beam testing with newly developed internal displacement measurement has been carried
out in this study to characterize lightly stabilized granular materials in terms of their MOR and Mr .
The sample were prepared from a typical granular material stabilized lightly with 13% cement
flyash, cured for 28 days and tested under cyclic loading at different stress sequences. It was found
that the proposed testing arrangement can measure the mid-span vertical deflection more reliably
and therefore, can determine the Mr and MOR more accurately. It was also found that both MOR
and Mr increased with the increase in binder content. Besides, the Mr is dependent on the stress
level and it decreased as the SR increased.
ACKNOWLEDGEMENT
The authors would like to thank Mr. David Sharp, Mr. Jim Baxter and Mr. Mathew Barret for their
technical assistance during the experimental work reported in this paper.
REFERENCES
AS-1289.5.1.1. 2003. Methods of testing soils for engineering purposes. Method 5.1.1: Soil compaction
and density testsDetermination of the dry density/moisture content relation of a soil using standard
compactive effort. Standards Australia, Sydney, Australia.
ASTM D698. 2007. Test Method for Laboratory Compaction Characteristics of Soil using Standard Effort
(12,400 ft-lbf/ft3 (600 kN.m/m3 )). Vol 04.08, ASTM International, Conshohocken, PA, USA.
ASTM D1632-07. 2007. Standard Practice for Making and Curing Soil-Cement Compression and Flexure
Test Specimens in the Laboratory, ASTM International, Conshohocken, PA, USA.
Austroads 2004. Guide to Stabilisation in Roadworks, Sydney, AUSTRALIA.
Elliott, R. P., & Lourdesnathan, D. 1989. Improved Characterization Model for Granular Bases. Transportation
Research Record, Journal of the Transportation Research Board, Washington, D. C., 1227, 128133.
293
Foley, G., & Group, A. S. E. 2001. Contract Report Mechanistic Design Issues for Stabilised Materials.
RC91022-3, AUSTROADS.
Fu, P., Jones, D., Harvey, J. T., & Bukhari, S. A. 2009. Laboratory test methods for foamed asphalt mix resilient
modulus. Road Materials and Pavement Design, 10(1), 87212.
Gnanendran, C. T., & Piratheepan, J. 2008. Characterisation of a lightly stabilised granular material by indirect
diametrical tensile testing. International Journal of Pavement Engineering, 9(6), 445456.
Gnanendran, C. T., & Piratheepan, J. 2010. Determination of Fatigue Life of a Granular Base Material Lightly
Stabilized with Slag Lime from Indirect Diametral Tensile Testing. Journal of Transportation Engineering,
136(8), 736745.
Irwin, L. H. & Gallaway, B. M. 1974., Influence of laboratory test method on fatigue test results for asphalt
concrete. STP 561, ASTM, Conshohocken, PA, USA.
Moffat, M. A., & Yeo, E. E. Y. 1998. Relationship between unconfined compressive strength and flexural
modulus for cemented materials. ARRB TR Working Document No. R-98/024.
Mohammad, L. N., Herath, A., Rasoulian, M., & Zhang, Z. J. 2006. Laboratory Evaluation of Untreated and
Treated Pavement Base Materials Repeated Load Permanent Deformation Test. Transportation Research
Record, Journal of the Transportation Research Board, Washington, D. C., 1967, 7888.
Murdock, J. W., & Kesler, C. E. 1958. Effect of range of stress on fatigue strength of plain concrete beams.
ACI Journal, 55(2), 221232.
NCHRP 2004. Final Report Guide for Mechanistic-Empirical Design of New and Rehabilitated Pavement
Design Structures Part 2 Design Inputs Chapter 2: Material Characterization. ERES Division of ARA
Inc., Champaign Illinois, USA.
Pan, T. Y., Tutumluer, E., & Anochie-Boateng, J. 2006. Aggregate Morphology Affecting Resilient Behavior
of Unbound Granular Materials. Transportation Research Record,Journal of the Transportation Research
Board, Washington, D. C., 1952, 1220.
Pronk, A. C. & Huurman, M. 2009. Shear deflection in 4PB tests, 2nd Workshop on Four Point Bending, The
University of Minho, Portugal, pp. 1926.
Thompson, M. R. 1966. Shear strength elastic properties of lime-soil mixtures. Highway research records,
Highway Research Board, National Research Council, Washington, D.C., 139, 114.
294
Cracking is recognized as one of the main causes of pavement
Four Point Bending

deterioration, and is the primary cause of the need for maintenance
and rehabilitation. Researchers around the world are working on the
problem of cracking in asphalt pavements, with the goal of developing
better understanding of the mechanics of cracking, creating test
methods for assessing the risk of cracking for different materials and
designs, and implementing these results into improved design methods
and specications. This Third Conference on Four-point bending held
at the University of California, Davis, USA, follows two successful
previous conferences, held at the Delft University of Technology, The
Netherlands, in 2007, and at the University of Minho, Portugal in 2009.
The primary objective of these conferences is to provide an exchange
of ideas and experience and to disseminate that knowledge among
researchers, government and private agencies and consultants, about
the use of the four-point bending test to evaluate stiffness and fatigue
resistance of bituminous mixtures. These proceedings include 23 papers
from 15 countries that have been subjected to peer review by a scientic
committee composed of experts in asphalt materials, design and testing.
Themes of the papers cover a range of topics, including modelling of
the four-point beam test, applications to mechanistic design, asphaltic
materials evaluation, comparisons with other tests and non-asphaltic
materials evaluation. Editors: J. Pais and J. Harvey
Four Point Bending is of interest to academics and professionals
interested in pavement engineering.
Editors
Pais
Harvey
an informa business
4PB_def.indd 1 09-08-12 20:40

Four Point Bending

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Four Point Bending

Caricato da

Copyright:

Formati disponibili

Cracking is recognized as one of the main causes of pavement

Four Point Bending

4PB_def.indd 1 09-08-12 20:40

No claim to original U.S. Government works

International Standard Book Number-13: 978-0-203-07306-3 (eBook - PDF)

Modeling of the Four Point Beam test

Applications to Mechanistic Design

Asphaltic materials evaluation

Comparisons with other tests

Non-asphaltic materials evaluation

John Harvey, University of California, Davis, USA (Chairman)

John Harvey, University of California, Davis, USA (Chairman)

A detailed FEM simulation of a 4-point bending test device

Figure 2. Overview of the 4PB device and its model representation.

2 FOUR POINT BENDING MACHINE

Table 1. Properties of steel and aluminium.

E [MPa] [] Specific mass [kg/m3]

Aluminium 70000 0.25 2700

3.1 Application of the standard

3.2 Shear effect

Poissons ratio 0.00 0.15 0.25 0.35 0.45

3.3 Clamp effects

Where Cclamp = correction factor for clamp effect [].

3.4 Clamp movements

3.5 Hinge stiffness

Actuator displacement Central deflection Applied force

Amplitude 0.03531 mm 0.03883 mm 111.77 N

4.2 Shear effect & Clamp effects

4.3 Clamp movements

4.4 Hinge stiffness

Description of a procedure for using the Modified Partial Healing

i,j = dummy variables (counter)

4 WEIGHING OVER A CROSS AREA

5 WEIGHING ALONG THE BEAM LENGTH L

6.2 Deflection Vb on the interval 0 < x < A

6.3 Deflection Vb on the interval A < x < L/2

The amplitude of the maximum deflection at x = L/2 is given by equation 13.

The differential equations are:

For the constant A2 /(6L3 )

The overall weighted modulus is calculated assuming a homogenous modulus distribution

0501 6.15 0.338 1.35 61.5 77

10 ENDURANCE LIMIT AND WHLER CURVE

11 DISCUSSIONS & CONCLUSIONS

ANNEX I POSSIBLE ITEMS FOR FURTHER RESEARCH

A first order approximation for equation 27 is given by equation 28.

Assessment of different flexure fatigue failure analysis methods

M.I. Souliman, W.A. Zeiada, K.E. Kaloush & M.S. Mamlouk

3.1 Dissipated energy based methods

3.2 Carpenter method

3.3 Pronk method

3.4 Hopman method

3.5 Stiffness degradation based methods

3.6 Rowe method

3.7 ASU method

3.8 Franken model fitting

Design asphalt Target air

Conventional PG 64-22 5.4 7

Y = 1 [A*XB + C*(expDX 1)], where, Y = stiffness; X = number of loading cycles; A, B,

4 TEST RESULTS AND ANALYSIS

4.1 Mixtures characteristics

4.2 Determination of Nf using all methods

Figure 3. Nf estimation for AR mixture at two temperatures.

Figure 4. Nf estimation for Fiber-Reinforced mixture at three temperatures

Y = 1 [AXB + C(expDX 1)], where, Y = stiffness; X = number of loading cycles; A, B,

PBA6A 76 (3 ) Varied 0.35 5.0 6.0