Indian Journal of Engineering & Materials Sciences

Vol. 15, February 2008, pp. 29-35

Dynamic permanent deformation testing of asphalt mixes and deformation

waveform analysis
Mesut Tigdemir*
Department of Civil Engineering, Suleyman Demirel University, Isparta, Turkey

Received 17 May 2007; accepted 31 December 2007

In this study, different specimens of hot mix asphalt (HMA) have been constructed in the laboratory and tested under
repeated loading; first permanent deformation and later fatigue tests using Suleyman Demirel University Asphalt Tester
equipment (SDU-Asphalt Tester). This study reports evaluation of available information about relationship between
permanent deformation and fatigue cracking. Analysis of data shows that the fatigue life of the pavement may be modelled
in terms of data obtained from repeated loading axial permanent deformation test results. Based on the analysis of the
estimation model for fatigue life from permanent deformation testing, it is concluded that for the evaluation of HMA in
fatigue characterization, repeated load axial test for permanent testing can be used in a very satisfactory manner. To be able
to model the fatigue lives, in addition to conventional fatigue model parameters new parameters from repeated load
permanent deformation test are taken into consideration. Hence, only by examining the permanent deformation tests, it may
be possible to predict the fatigue lives of the specimens without carrying out time-consuming fatigue tests.

Keywords: Permanent deformation, Asphalt mixes, Fatigue life, Modelling

There are many reports that provide much detail on more of the HMA layers as well as in the unbound
the failure mechanisms for the various hot-mix materials underneath the HMA. Eisenmann and
asphalt (HMA) distresses. The purpose of this study is Hilmer2 found that rutting was mainly caused by
to evaluate available information about the deformation flow rather than volume change.
relationship between permanent deformation and Fatigue cracking often is called alligator cracking
fatigue cracking, that is, the emphasis is placed on because its closely spaced crack pattern is similar to
permanent deformation and fatigue cracking. the pattern on an alligator’s back. This type of failure
Rutting (or permanent deformation) results from generally occurs when the pavement has been stressed
the accumulation of small amounts of unrecoverable to the limit of its fatigue life by repetitive axle load
strain as a result of repeated loads applied to the applications. Fatigue cracking is often associated with
pavement. Rutting can occur as a result of problems loads that are too heavy for the pavement structure or
with the subgrade, unbound base course, or HMA. more repetitions of a given load than provided for in
Consolidation or lateral movement of the HMA under design. The problem is often made worse by
traffic causes permanent deformation in HMA. Shear inadequate pavement drainage, which contributes to
failure (lateral movement) of the HMA courses this distress by allowing the pavement layers to
generally occurs in the top 100 mm of the pavement become saturated and to lose strength. The HMA
surface1; however, it can occur deeper if satisfactory layers experience high strains when the underlying
materials are not used. Rutting in pavement usually layers are weakened by excess moisture and fail
develops gradually with increasing numbers of load prematurely in fatigue. Fatigue cracking also can be
applications, typically appearing as longitudinal caused by repetitive passes with overweight trucks or
depressions in the wheel paths sometimes inadequate pavement thickness because of poor
accompanied by small upheavals to the sides. It is quality control during construction3,4.
typically caused by a combination of densification Fatigue cracking can lead to the development of
(decrease in volume and, hence, increase in density) potholes when the individual pieces of HMA
and shear deformation and can occur in any one or physically separate from the adjacent material and are
___________ dislodged from the pavement surface by the action of
*E-mail: tigdemir@mmf.sdu.edu.tr traffic. Potholes generally occur when fatigue
30 INDIAN J. ENG. MATER. SCI., FEBRUARY 2008

cracking is in the advanced stages and when relatively height of about 63 mm). The properties of aggregates
thin layers of HMA have been used. In the past, are given in Table 1 and the conventional properties
fatigue cracking was thought to initiate from the of bitumen are given in Table 2. The specimens were
bottom and to migrate toward the surface. These prepared at two different gradations, at 6 different
cracks began because of the high tensile strain at the bitumen contents (4%, 4.5%, 4.65%, 5%, 5.15%,
bottom of the HMA. Recently, fatigue cracks have 5.65%), and at 5 polymer contents for three different
been observed to start at the surface and migrate polymers (0%, 2%, 3.5%, 5%, 6%). The polymers
downward. The surface cracking starts because of were used to obtain different mixes in evaluating the
tensile strains in the surface of the HMA. fatigue and permanent deformation behaviour. The
Generally, it is believed that for thin pavements the properties of modified bitumen are given in Table 2.
fatigue cracking typically starts at the bottom of the The specimens used in this study were bituminous
HMA and for thick pavements, the fatigue cracking mixtures of 101.6 mm diameter. Each specimen has
typically starts at the HMA surface. Typically, fatigue been compacted by a 50-cycle standard Marshall
cracking is caused by a lack of adequate pavement compactor for each surface.
structure and is not caused by a lack of control of The Marshall test as per ASTM D 1559 (1989) was
HMA properties; however, these properties can used for the determination of the optimum binder
certainly have a secondary effect5,6. content in the mix. The optimum binder content was
Bhattacharjee et al.7 showed the effect of wheel found to be 5.15% by weight of aggregate gradation-1
load on fatigue performance of the pavement in terms and 4.5% by weight of aggregate gradation-2 for the
of strain history and cracking. In their works different unmodified asphalt mix for design air voids of 4%
phases of strain history have been identified and their and 4.5% respectively. The other bitumen contents
role on fatigue performance discussed. A rational have been chosen ±0.5% of optimum bitumen content
approach of calculation of strain corresponding to for each gradation.
failure load applications has also been presented.
Equipment and testing program
Experimental Procedure To determine the permanent deformation,
Materials characteristics of paving materials is employed a
In this study, a total of 33 specimens with different repeated dynamic load test for several thousand
characteristics have been prepared. Crushed limestone repetitions and record the cumulative permanent
and AC 60/70 bitumen were used to obtain Marshall deformation as a function of the number of cycles
asphalt concrete specimens (101.6 mm diameter and (repetitions) over the test period. This approach was
Table 1—Properties of aggregates employed by Monismith et al.8 in the mid 1970’s
using uniaxial compression tests. Several research
Properties Standard
Aggregate studies conducted by Witczak et al.9 used a
Coarse Fine Filler temperature of 100°F or 130°F, and at 10, 20, or 30
Abrasion loss
psi unconfined deviator stress level. A haversine pulse
(%) ASTM DC-131 26.9 - - load of 0.1 s and 0.9 s dwell (rest time) is applied for
(Los Angeles) the test duration of approximately 3 h. This approach
Frost action (%) results in approximately 10,000 cycles applied to the
ASTM C-88 1.91 - -
(with Na2SO4)
specimen. Brown and Cooper used a range of various
ASTM C-127 2.697 - -
Specific gravity levels of confining pressure for the repeated load
ASTM C-128 - 2.716 -
(g/cm3) test10. A stress level of 14.4 psi (100 kPa) was
ASTM D-854 - - 2.732
Table 2—Conventional properties of unmodified / modified bitumen

Bitumen
AC AC AC AC
Properties Standard AC
+ + + +
(60/70)
2%SBS 3.5% SBS 5% SBS 6% SBS

Specific gravity at 25°C ASTM D70 1.036 - - - -

Penetration, dmm, 100 g, 5 s, 25ºC ASTM D5 62 55 49 43 41
Softening point (°C) ASTM D36 47 52 59 63 65
 TIGDEMIR: DYNAMIC PERMANENT DEFORMATION TESTING ASPHALT MIXES 31

subsequently adopted as the standard for their tests. tests. The loading was repeated until the specimen
The test was conducted at 104°F (40°C). failed in the fatigue tests.
The RLAT test is undertaken to establish the In permanent deformation tests cumulative
resistance of a mixture to permanent deformation. A permanent deformation curve of each specimen were
number of load pulses are applied to the specimen, on plotted against loading cycles. Since one loading
the flat (cut) surface, with the resulting deformation period is simulated as 20 intervals for loading and
recorded. The axial strain obtained at the end of the reading of deformation using transducers, a
test is a valuable measure of the specimen’s resistance deformation waveform has 21 steps and some of these
to permanent deformation. deformations were used in the estimation model
In this study 33 specimens were tested. The (Fig. 2). Figure 2 shows the cumulative permanent
samples have been prepared convenient to the deformation curves for different specimens.
Marshall stability tests; hence the diameter of sample
is 101.6 mm. First permanent deformation tests have Results and Discussion
been carried out for all specimens and this test was For statistical evaluation General Linear
followed by fatigue tests. The results have been Programming (GLM) model was developed. In this
evaluated to obtain available information about the model, in addition to loading, specimen height, and
relationship between permanent deformation and density of asphalt mixes, deformations (Wi) at specific
fatigue cracking. steps of loading waveform, differences between peak
The testing programme has been carried out using values of 100th , 200th and 300th loading waveform
repeated-loading indirect tensile test equipment, values are presented. Depending on their contribution
SDU-asphalt tester developed by Tigdemir11. This to the model, some of the variables affecting the
testing device has been used in different research fatigue life were omitted from the step-wise analysis
projects by researchers and their results have been whereas some were kept.
reported elesewhere12-14. Although the stress state Figure 3 shows the deformations (Wi) at specific
within the specimen is complex, critical stresses and steps for a loading duration (1000 ms). In Fig. 4, the
strains are readily computed if linear elastic behaviour 100th, 200th and 300th deformation waveforms are
is assumed. A biaxial state of stress exists along the given and difference between the peak deformation
vertical load axis. Along this axis, the horizontal values of 300th and 200th and of 300th and 100th
tensile stress is reasonably constant while vertical loading were calculated for modelling process.
compressive stress varies more significantly15. The characteristics of asphalt cylinders (101.6 ×
In the permanent deformation tests, loading stress HAve mm) for the test specimen and test results of
of 100 kPa axial stress was repeated 3700 times repeated loading axial test and indirect tensile fatigue
(7400 s) and the first 100 loadings were the test are shown in Table 3. Model equation (1) was
preconditioning stage. The loading and rest period linearly modelled by step-wise analysis. The
were 1 s hence frequency is 0.5 cycles/s. The coefficients of the independent variables in the model
permanent deformation tests have been carried out at and the coefficient of determination in the equation
were calculated using GLM. Moreover, SYSTAT
40°C. The samples are preconditioned to the testing
statistical software was utilized in the analyses. In
temperature overnight.
The fatigue test was carried out at 25°C. After
being prepared the samples were kept in normal room
conditions for an overnight then they were placed into
a cabinet in which temperature is controlled and
conditioned to the test temperature. The samples were
mounted into the testing frame for fatigue tests and
for permanent deformation test (Fig. 1).
Fatigue tests under conditions of controlled stress
were carried out under the frequency of 30
cycles/min. The loading and rest periods were 1 s
respectively in both of the fatigue tests. The loads Fig. 1—Test configuration of permanent deformation and
applied were 0.70, 0.80, and 1.10 kN in the fatigue fatigue testing
32 INDIAN J. ENG. MATER. SCI., FEBRUARY 2008

Fig. 2—Cumulative deformation curves of different specimen groups

Fig 3—A deformation waveform and steps (Wi) used in model for
loading duration

Fig 5—Plot of Residuals against predicted values

Tables 4 and 5, results of step-wise analysis and

analysis of variance can be seen. It may be noted that
these analyses include coefficients of the independent
variables that are used in the model.

InN = Const + a1 *W12 + a2 *W15 + Air Voids%*

(a3 *W4 + a4 *Load + a5 *D200-100 + a6 *W4 *W15)
+ a7*load*Total (Wi) … (1)

Figure 5 is the plot of residuals against predicted

values of model organised with permanent
Fig. 4—100th, 200th, and 300th permanent deformation waveforms deformation waveform variables.
 TIGDEMIR: DYNAMIC PERMANENT DEFORMATION TESTING ASPHALT MIXES 33

Table 3—Characteristics of asphalt specimens and permanent deformation waveform

Have W4 W6 W8 W12 W15 D Load Density Air Ln(N)

(mm) microns microns microns microns microns 200-100 (kN) (kN/m3) voids% (N cycs)

62.53 21.3 27.66 24.33 10.66 4 41 0.80 2.386 4,50 3.53

62.27 28 35.33 27 11 4.66 22 1.10 2.404 3,77 2.40
63.7 18 21.3 19 8.6 5.6 28 0.70 2.384 4,58 5.14
61.43 28.3 33 24 11.6 4 31 0.70 2.448 2,01 3.61
61.7 29 32 25.3 12.3 6.3 52 1.10 2.443 2,21 2.83
62.37 33 36 27.3 11.6 4.6 54 1.10 2.431 2,69 2.83
62.63 48 54 40 19 6 32 0.70 2.406 3,69 3.66
61.97 43 52 38 17 6 36 0.70 2.400 3,94 3.61
61.43 38.6 44.3 33 16.6 7 39 1.10 2.407 3,65 1.79
61.93 35.6 42.3 29 13 5.3 43 0.70 2.434 2,57 3.89
61.57 34.3 40.6 30.6 16.3 7 31 0.70 2.419 3,17 3.69
61.73 38.6 43.6 37.3 18 9.30 52 1.10 2.443 2,21 1.61
64.2 28 32 30 13 6 45 1.10 2.384 4,93 1.79
62.43 31 34 31 10 3 34 0.70 2.402 4,21 4.25
62.73 32 40 31 14 6 23 0.70 2.434 2,94 4.92
63.63 35 41 33 14 6 39 0.70 2.432 3,02 4.65
63.27 36 42 31 18 8 35 1.10 2.411 3,86 2.64
64 45 49 42 17 7 15 0.70 2.380 5,09 3.40
63.83 36.3 45.3 31.3 14.6 6 22 1.10 2.400 4,29 1.10
64.53 34.6 41.3 31.6 12.3 5 17 0.70 2.352 6,21 3.61
64.47 32 38 30 13 4 32 0.70 2.375 5,29 4.94
63.7 33 38 27.6 11 3 21 1.10 2.374 5,33 2.20
63.57 43 48 41 13 3 17 0.70 2.390 4,69 4.93
63.6 32 39 27 10 3 6 0.70 2.404 4,13 4.13
63.67 29 36 24 9 2 20 1.10 2.392 4,61 2.83
63.97 28 32 30 10 2 18 0.70 2.365 5,69 4.01
64.56 34 39 27 13 5 24 0.70 2.380 5,09 4.70
63.63 40 46 40 16 4 36 1.10 2.368 5,57 2.08
61.73 33 38 27 11 4 38 0.70 2.413 3,78 4.97
62.6 35 42 33 16 7 41 1.10 2.407 4,01 2.20
63.73 43 48 37 15 9 44 0.70 2.388 4,42 4.48
63.1 35.6 43.3 28.6 11.3 3.6 28 0.70 2.415 3,33 4.32
63.6 34 41 29.6 12.3 7 29 1.10 2.371 5,10 2.20

Table 4—Step-wise analysis

Effect Coefficient Std Error Std Coef Tolerance t P(2 Tail)

CONSTANT 5.040 0.853 0.000 . 5.911 0.000

W15 0.711 0.197 1.151 0.039 3.600 0.001
W12 -0.181 0.051 -0.443 0.253 -3.547 0.002
Air voids*W4 0.021 0.007 0.833 0.053 3.052 0.005
Total (Wi)*Y -0.0002 0.000 -0.867 0.178 -5.822 0.000
Air voids *Y -0.008 0.001 -0.878 0.384 -8.639 0.000
Air voids *D200-100 0.011 0.002 0.421 0.606 5.209 0.000
Air voids *W15*W4 -0.003 0.001 -1.021 0.033 -2.942 0.007

Dep Var: LNN N: 33 Multiple R: 0.949 Squared multiple R: 0.901

Adjusted squared multiple R: 0.873 Standard error of estimate: 0.409

Table 5— Analysis of variance

Source Sum-of-squares df Mean-Square F-ratio P

Regression 38.055 7 5.436 32.518 0.000

residual 4.180 25 0.167
34 INDIAN J. ENG. MATER. SCI., FEBRUARY 2008

Fig 6—Scattering between predicted data and experimental data

Comparison between predicted fatigue life value fatigue experiment processing by crushing the
from the permanent deformation waveform model specimen can be eliminated.
with the specimen properties and experimental fatigue Analysis of data showed the fatigue life of the
life value is shown in Fig. 6. The correlation is pavement may be modelled in terms of strain data
extremely satisfying. Model is able to predict the obtained from repeated load axial permanent
fatigue experiment results deterministically using the deformation test results. Based on the analysis of the
specimen and experiment of permanent deformation estimation model for fatigue life from permanent
parameters (variables of permanent deformation deformation testing, it is concluded that the repeated
waveforms of 100th, 200th and 300th, density, load axial test for permanent testing can be used very
specimen height, and loading with a coefficient of satisfactorily for evaluation of HMA for fatigue
determination value of 0.901). characterization. To be able to model the fatigue lives,
in addition to conventional fatigue model parameters,
Conclusions new parameters from repeated load permanent
This study evaluated the results of repeated load deformation test were taken into consideration.
axial test for permanent deformation and repeated Hence, only by examining the permanent deformation
loading indirect tensile test for fatigue on hot mix tests, the first three hundred loading, it may be
asphalt (HMA) samples and provided remarkable possible to predict the fatigue lives of the specimens
information about fatigue and permanent deformation without carrying out time-consuming fatigue tests.
endurance, which are crucial ingredients of pavement
design. Standard models use mixture parameters and References
loading conditions, whereas this study used extra 1 Nevelt G & Thanfold H, Proc Assoc Asphalt Paving
permanent deformation waveform variables. By Technol, 57 (1988) 320
extensive parametric study of this model with an 2 Eisenmann J & Hilmer A, Proc Sixth Intl Conf Structural
Design of Asphalt Pavements, 1 (1987) 392
inverse approach, time-consuming fatigue
3 Roberts F L, Kandhal P S, Brown E R, Lee D Y & Kennedy
experiments can be eliminated. Repeated load axial T W, Hot Mix Asphalt Materials, Mixture Design, and
test measurements are relatively easy and time saving Construction, (NAPA Education Foundation, Lanham, Md.),
processes. Furthermore, the waveform variables used 1996
in modelling stage belong to the first 300 loading of a 4 McGennis R B, Anderson R M, Kennedy T W &
total of 3600 loadings in permanent deformation Solaimanian M, Background of Superpave Asphalt Mixture
Design and Analysis, (FHWA, U.S. Department of
testing. Model parameters are calculated from these Transportation), 1994
waveforms and this data is entered into the model. 5 Brown E R, Kandhal P S & Zhang J, Performance Testing
Such an approach helps predicting the fatigue lives for Hot Mix Asphalt, (NCAT Auburn, Ala.), 2001
without carrying out specimen-crushing fatigue 6 Brown E R, Kandhal P S & Zhang J, Performance Testing
experiments. Thus, in addition to saving time, cost of for Hot-Mix Asphalt, (TRB, Transportation Research
 TIGDEMIR: DYNAMIC PERMANENT DEFORMATION TESTING ASPHALT MIXES 35

Circular E-C068: New Simple Performance Tests for Asphalt Maryland, College Park, Maryland), 1998
Mixes), 2004. 10 Brown S F & Cooper K E, J Assoc Asphalt Paving Tech, 53
7 Bhattacharjee S, Gould J S, Mallick R B & Hugo F, Int J (1984).
Pavement Eng, 5 (2004) 61 11 Tigdemir M, Karasahin M & Kalyoncuoglu S F, J Test Eval,
8 Monismith C L, Ogawa N & Freeme C, Permanent 29 (2001) 387
Deformation of Subgrade Soils Due to Repeated Loadings 12 Tigdemir M, Kalyoncuoglu S F & Kalyoncuoglu U Y,
(Transportation Research Record 537, Transportation NDT&E Int, 37 (2004) 597
Research Board, National Research Council, Washington, 13 Kalyoncuoglu S F, J Transport Eng-ASCE, 131 (2005) 725
D.C.), 1975 14 Ahmedzade P, Tigdemir M & Kalyoncuoglu S F, Constr
9 Witczak M W & Kaloush K E, Performance Evaluation of Build Mater, 21 (2007) 626
Asphalt Modified Mixtures Using Superpave and P-401 Mix 15 Fatigue response of asphalt-aggregate mixes, (Strategic
Gradings, (Technical Report to the Maryland Department of Highway Research Program. SHRP-A-404. Washington,
Transportation, Maryland Port Administration, University of DC), 1994.