You are on page 1of 27

The version of records of this manuscript has been published and it is available in the

Journal of the Association of Asphalt Paving Technologists


Volume 74, Pages 97-137, 2005

1
Rational Fatigue Limits for Asphalt Binders Derived from Pavement Analysis

Rodrigo Delgadillo1 and Hussain Bahia2

Abstract

The Superpave specification has some critical gaps, most of them related to the performance
characterization of modified binders. This issue is generally recognized by the State Highway
Agencies, industry and researchers (1). In 2001 NCHRP 459 (1) report was published, offering a
revised system for testing and evaluating asphalt binders based on damage behavior. For fatigue
characterization of binders, a new specification parameter was proposed, called Np20. One of the
main advantages of the new parameter was the inclusion of the pavement structure in the
specification. This paper starts with a review of why G*sin has to be replaced by the new parameter
Np20.
The NCHRP 459 proposal for Np20 was, however, only conceptual and lacked the details
required for implementation, specific criteria and limits. The main purpose of this paper is to show
how pavement structural analysis can be used to implement the binder fatigue specification
parameter Np20. Pavement analysis using a multilayer elastic theory program was used to determine
the strain levels in the pavement for six different pavement structures typically used in practice. From
those strains, the fatigue life of 19 different binders was calculated. Known relationships between
mixture and binder strain, and between mixture and binder fatigue life, were used in the calculations.
The results allowed determining the suitability of the 19 binders for different traffic volumes, speeds
and pavement structures. Testing conditions (temperature and input energy) for the conditions of
Wisconsin and specification limits for Np20 are proposed.
Keywords: asphalt binder, fatigue, pavement analysis, specifications,

Introduction

One of the main conclusions of the NCHRP 9-10 project is that the current AASHTO MP1 and
MP1-a does not accurately rank modified asphalts according to their contribution to mixture damage
resistance (2). This was not a new finding but rather a confirmation of many concerns that State
Highway Agencies and the industry have expressed since the implementation of the Superpave
binder specification in 1993. The most important finding of the NCHRP project is in fact explaining
why the Superpave system fails to rank asphalt according to contributions to mixture behavior. The
findings point out that the reliance on linear viscoelasticity measured at small strains and a few cycles
cannot possibly estimate the damage behavior of asphalt binders, modified and unmodified. The
findings pointed out that the simplistic use of the dissipated energy concept in deriving G* sin for
fatigue was flawed because this energy dissipated represented mostly viscoelastic damping and not
necessarily energy dissipated in fatigue damage (3).
To replace the G* sin parameter, the same test is used but is conducted under a constant
stress or strain condition and is applied repeatedly to measure the change in G* and phase angle and
calculate the change in dissipated energy per cycle rather than the initial dissipated energy. The
1
Research Assistant, University of Wisconsin Madison
2
Associate Professor, University of Wisconsin Madison
The oral presentation was made by Mr. Delgadillo
Delgadillo and Bahia 2

initial dissipated energy is mostly reflecting the damping characteristics while the change in dissipated
energy can identify the number of cycles at which fatigue or instability of binders will start, which in
turn can be an indicator of fatigue life (3, 4).

Problems with the Current Superpave Fatigue Parameter

There are several studies published in the last 5-6 years that have indicated the short comings
of the current Superpave specification parameter for fatigue (G*sin ) (5, 6, 2). The problems can be
grouped into the following four areas: (a) it is good only for strain controlled conditions; (b) it is based
on the energy dissipated per loading cycle; (c) it is obtained after only a few cycles of loading, and (d)
it does not consider the type of pavement structure. The following sections briefly review why these
problems are critical.

Strain controlled conditions


The Superpave parameter G*sin is derived from the total energy (Wi) dissipated per loading
cycle under a strain controlled loading. Equation 1 shows the relationship for Wi under strain
controlled conditions.
Wi     02  G * sin ; (1)
where: Wi = total energy dissipated in cycle i
0 = constant strain applied
G* = complex modulus
 = phase angle

The concept in the specification is that fatigue is supposed to be minimized by controlling W i.


Lower values of G*sin would result in lower energy dissipated per loading cycle under strain
controlled conditions. There is thus a maximum limit enforced in the MP1 specifications. Fatigue is
known to be more critical in weak pavements, where the loading is believed to be strain controlled.
However, the limit between weak pavement and strong pavement is not very clear and has been
debated significantly through out the years in asphalt literature (7, 5). It would be more useful to have
a parameter that is independent of the mode of loading. Such a parameter would be applicable for all
types of pavement structures and would provide a more reliable and simple prediction for the fatigue
life of roads.

Energy Dissipated per Loading Cycle (Wi)


The energy dissipated per loading cycle Wi does not necessarily represents the tendency of
the binder to having a poor fatigue performance. Under cyclic, reversible loading the energy (Wi) can
be dissipated into two components: visco-elastic damping and irreversible damage. Thus, Wi
represents energy that is used by the time dependent behavior of the material or energy that is
dissipated in either viscous flow or in cracking. During cyclic reversible loading it is very difficult to
separate these mechanisms and thus the simple measure of Wi for a few cycles is not a measure of
cracking damage that could occur due to fatigue. This was proven experimentally by considering the
correlation between G*sin and the fatigue life of mixtures, which was found to be very low (1).
This problem with using total energy dissipated is not new and in fact in the field of asphalt
mixture fatigue, pseudo-strain concepts have been introduced to solve this problem by a number of
researchers (8, 9, 10). Although pseudo-strain can offer a solution, it requires rigorous curve fitting
and assumption of a constitute visco-elastic law for materials. The simpler solution to this challenge is
to consider the change in Wi values as a function of cycles because the relative change due to
Rational Fatigue Limits 3

repeated cycling is a good direct physical indicator of the initiation of the fatigue damage and the
progression. This will be shown later in this paper.

Numbers of Cycles of Loading


The value of G*sin is derived from Wi calculated after testing for only a few loading cycles.
The damage accumulation (fatigue) happens after many cycles of loading. This fatigue behavior can
be shown for example like a decrease in the G* value of the binder. Two asphalts can have very
similar performance after a few cycles of loading but they can defer greatly when thousands of cycles
are applied. Figure 1 shows the G* values for some binders as a function of the cycles of loading for
strain controlled conditions. It can be see that all the binders show a steady behavior in terms of G*
values for the first thousand cycles. However there is a certain point when G* starts to decrease. This
is a sign of damage occurring in the binder. It is also shown that binders that had similar G* value at
the beginning of the testing, start to deteriorate at different number of loading cycles. This means that
they have different fatigue lives.

Pavement Structure
Fatigue behavior in asphalts is highly dependent on the level of load applied to the binder.
Higher strains (or stresses) lead to earlier fatigue damage than lower ones (2). That is why fatigue
damage is usually more common in weak pavements, where the

1.6E07

1.2E07
G* (Pa)

8.0E06

4.0E06

0.0E00
100 1000 Cycles 10000 100000 1000000

Figure 1. G* v/s Loading Cycles for Strain Controlled Testing (2)

strain levels are higher. The current parameter G*sin does not consider the load applied to the
binder in the characterization of the fatigue behavior. By integrating pavement structure in the
measurement, the differences between weak structures and strong structures can be considered and
thus the inaccuracy caused by neglecting pavement conditions in the current specifications can be
corrected.

Deriving a New Binder Parameter for Fatigue

A new specification parameter for determining the fatigue resistance of binders was proposed
in the NCHRP 9-10 research program (1). The new parameter is obtained after testing the asphalt in
the DSR binder for a large number of cycles until damage is shown in the sample. The kind of loading
is the same applied in the current Superpave fatigue testing, but the number of cycles is considerably
higher. The damage is characterized by a decrease in the energy dissipated per loading cycle
Delgadillo and Bahia 4

(11,12). The point of failure is obtained from analyzing the variations in the dissipated energy per
loading cycle using a parameter called dissipated energy ratio DER (12). The proposed fatigue
parameter is named Np20 and its application overcomes the problems with the Superpave parameter
G*sin discussed in the previous section.
There are, however many questions to be answered before this new parameter can be
implemented.
 What number of loading cycles is needed and how to define failure?
 What loading mode should be used?
 What geometry should be used?
 Which pavement structure should be simulated?
 What temperature should be selected for testing? and
 What limits should be used to accept binders?
The following sections cover these questions and offer some recommendations for selecting of
binder acceptance limits.

Number of Loading Cycles


Figure 1 shows how binders with similar characteristics in terms of G* in the first cycles of
loading can present very different fatigue performance after many cycles of loading. During the
NCHRP 9-10 project it was found that the best way to characterize the point when fatigue damage
occurs is shown is by using the concept of change in the dissipated energy per cycle. A binder can be
subjected to cyclic loading for a determined number of cycles without showing fatigue damage.
During this stage, the energy dissipated per loading cycle remains constant. However, if the binder
continues to be loaded and unloaded, a point is reached when it does not dissipate the same amount
energy per cycle anymore. This is a sign of fatigue damage. This behavior can be clearly shown
using a parameter called Dissipated Energy Ratio (DER), defined in Equation 2.

W
n

DER  1 i
; (2)
Wn

where Wi = energy dissipated at cycle i


Wn = energy dissipated at cycle n

The variation in the DER versus cycles of loading for a typical binder under cyclic stress
controlled and strain controlled loading are shown in Figures 2 and 3 respectively. During the first
cycles of loading, when W i remains, the plot shows a linear trend with a 45 slope. When fatigue
damage is reached the trend line followed by the DER depends on the mode of loading. In the strain
controlled test the value of W n decreases when fatigue damage is reached, so the DER increases.
For the stress controlled test, however, the value of W n increases when fatigue damage is reached,
so the DER decreases.
The point where the trendline deviates from the linear slope corresponds to the point where the
fatigue cracks are initiating in the binder. This point is refereed as crack initiation point. However, after
this point, if the binder is unloaded and enough resting time is given, it is still capable of recovering.
After more cycles of testing the crack propagation point is reached. When the binder goes beyond
this point no more healing is possible and the fatigue damage is not reversible. Conceptually, the
Rational Fatigue Limits 5

fatigue life of the binder is represented by the number of cycles required to undergo the crack
initiation without reaching the crack propagation.

Independence from the Mode of Loading


The fatigue damage point for binders Np20 was defined as the point (cycle number n) where
the dissipated energy ratio DER deviates 20% from the 45º inclination straight line. This point was
selected because it provides independence of the mode of loading. If the initial input energy is the
same, the Np20 value obtained using stress controlled or strain controlled test will be the same. The
Np20 occurs after the crack initiation point and usually before the crack propagation point, so it
provides a reasonable criteria for defining the fatigue failure of binders.

Stress controlled was selected as the most suitable test. It provides a more clear measurement of the
Np20 value. It can be seen that both, the crack initiation and crack propagation are more clearly
shown in the stress controlled testing shown in Figure 2 that in the strain controlled one shown in
Figure 3.

Sample Geometry
The proposed parameter Np20 is obtained from analysis of testing results of the asphalt binder
in the DSR. Several researchers have expressed concerns about the suitability of parallel plate
testing for binder fatigue characterization (8, 13, 14). It has been claimed that in this kind of testing,
edge effects influence the results of fatigue life of asphalts. The problem was addressed in a study
carried out at the University of Wisconsin Madison in collaboration with FHWA and MTE (15). The
research, requested by the TRB Expert Task Group on binders, compared the fatigue life of binders
tested using parallel plate with the fatigue life of asphalt mastic cylinders using torsion testing. The
torsion cylinder test, which was introduced in the testing procedure developed by Little and coworkers
at Texas A&M University (8), was used to study the edge effects. Two unmodified and three modified
binders were tested at 25C and 10C. The results of the study, although still preliminary, showed that
the fatigue relationship developed from parallel plate and torsion cylinder ranks binders similarly. If
the trend is confirmed, one can assume that the edge effect is not a major factor in asphalt fatigue
testing in parallel plate. Figure 4 is a sample of the results presented recently at the 2004 Petersen
Asphalt Conference (15).

Pavement Structure
Weak pavements deform more under traffic loads than strong pavements. This generates
more energy input over the binder used in the weak pavement. Fatigue life of binders is highly
sensitive to energy input. This fact is clearly shown in Figure 5, where the different fatigue lives
shown by the same binder subjected to different stress levels is presented. This variable was
considered in defining the new parameter Np20, as it is shown in the following section.
Delgadillo and Bahia 6

Figure 2. Variation in the DER for Stress Controlled Test (4)

Figure 3. Variation in the DER for Strain Controlled Test (4)


Rational Fatigue Limits 7

Fatigue relationship
The determination of the Np20 value depends mainly on two parameters: the energy input and
temperature of testing. It has been proven that the fatigue life relationships vary with the temperature
of the material (16). For each of the required

0.0
y = 0.8226x - 0.4035
K2 for PP Geomtery -1.0
R 2 = 0.7895
-2.0
-3.0
-4.0
-5.0
-6.0
-6.0 -4.0 -2.0 0.0
K2 for TC geomtery

1.E+13
K1 for PP Geomtery

0.579
y = 992.03x
1.E+11 2
R = 0.8791

1.E+09

1.E+07

1.E+05
1.E+05 1.E+07 1.E+09 1.E+11 1.E+13

K1 for TC geomtery
Figure 4. Comparison of Fatigue Coefficients Estimated by the Parallel Plate and Torsion Cylinder
Geometries (15).
Delgadillo and Bahia 8

Figure 5. Fatigue Life versus Stress Level (2)

pavement temperatures, the binder has to be tested at different energy levels for determining the
fatigue relationships. When the energy concept is used, there is no need to test at different
frequencies. The effect of frequency can be included in the energy input, as it will be explained later
in this paper.
The relationship between the fatigue life of a binder (Np20) and the energy input (W i), for a
specific pavement temperature, can be approximated by Equation 3.
Np20 = K1  (1/wi)K2 ; (3)
When plotted in a log-log scale, the relationship shown by Equation 3 is linear. For determining
the parameters K1 and K2 two points are needed. This means that, for a fixed pavement temperature,
the binder has to be tested at least at two different energy levels in order to obtain the fatigue
relationship. After the parameters are calculated, the fatigue relationship can be used to calculate the
fatigue life at any input energy for the selected pavement temperature.

Initial Energy Input


The two input energies needed to determine the fatigue relationship were chosen in order to
represent the weak pavement and strong pavement conditions. It was assumed that the binder in the
strong pavement is loaded in the linear range. In the weak pavement, the binder is considered to be
loaded in the non-linear range. For determining the energy needed to take the binder to the linear and
non-linear regions, amplitude sweep testing was done. The amplitude sweep plot data for a typical
binder is shown in Figure 6(a). The low strain/stress (strong pavement) was defined as one half of the
linear limit. The high strain/stress (weak pavement), was picked up slightly above the linear limit.
Figure 6 (b) shows how the two energy input levels are used to construct the relationship between
Np20 and the Wi.

Testing Temperature
Testing of fatigue life of asphalt binders has proved to be dependent on the testing
temperature. Different fatigue relationships can be obtained for the same binder when different testing
temperatures are used (16). For this reason, the temperature of testing has to be selected prior to the
testing.
The testing temperature is selected as the average of the high and low pavement design
temperatures of the PG grading. In the current AASHTO MP1, the temperature for checking the
Rational Fatigue Limits 9

fatigue parameter G*·sin is chosen as the average plus 4°C. There is no need to add 4°C to the
average temperature in the determination of Np20, so the average should be used directly. For
example for a PG 58-28 grade, the testing is conducted at 15°C for the new parameter, instead of the
19°C used in the current specifications. Since PG 58-28 is the binder used in the most populated
areas of Wisconsin (south east region of the state), the average temperature of this PG grade was
selected (15C) for the purpose of the present work. Besides this temperature, another temperature
was also picked up for taking into account the fatigue that occurs in the pavement during the thaw
season, when the structure under the pavement is in its worse condition. The temperature selected
for representing this condition, considering the climatic conditions of Wisconsin, was 6C.
1.0E+06
4.0E+05 Np20
1.0E+05 -2.6274
3.0E+05 y = 1E+16x
Stress (Pa)

2
R = 0.9172
1.0E+04
2.0E+05

1.0E+05 1.0E+03

0.0E+00 1.0E+02
0.00E+00 4.00E-02 8.00E-02 1.20E-01 1.0E+03 1.0E+04 1.0E+05 1.0E+06

a) Strain b) Wi (Pa)
1b15-5 1b15-6

Figure 6. Amplitude Sweep Plot Data for a Typical Binder (a) and Np20 v/s Wi Relationship (b) (4)
Testing frequency
It has been proven that the testing frequency is a factor to consider in the determination of the
fatigue life of asphalt binders and mixtures (3, 16). The required frequency is related to the traffic
speed. For 60 mph the testing frequency should be 10 Hz. For slow speed (15 mph) the testing
frequency should be 2.5 Hz. Using the initial energy concept (W i), the effect of frequency on fatigue
life could be included in changing W i so testing at multiple frequencies is not necessary. Equation 1
and Equation 4 show the relationships for W i in a strain controlled and stress controlled testing
respectively.
1
Wi     02  ; (4)
G * / sin 
where: Wi = total energy dissipated in cycle i
0 = maximum stress applied
G* = complex modulus
 = phase angle
The values of G* and  vary with the testing frequency, so the effect of testing frequency can
be included in the value of G* and  for any of the modes of testing used. Considering this advantage
the testing is done at 10 Hz only, which corresponds to the fast traffic speed frequency. This way the
time required for testing is significantly reduced.
For slow traffic speed, the Wi values were corrected. For accounting the variation of W i due to
testing frequency, an estimation of the variation of the parameters G* and  with respect to frequency
is needed. It has been shown (17) that at the intermediate pavement temperature, G* decreases
Delgadillo and Bahia 10

around nine times when the frequency of testing is lowered from 10 hz to 0.1 hz (1% strain). This
results were used to approximate the variations in G* when the frequency is changed from 10 Hz to
2.5 Hz. Considering linear interpolation, the parameter would decrease approximately 30% when the
frequency decreases four times. For the same frequency change, the variations in  were shown to
be negligible.

Testing to Derive Limits for NP20

Experimental Design
A total of 18 modified asphalts and one base asphalt were tested using oscillatory shear
measurements under stress controlled conditions. The change in properties was measured using the
Dynamic Shear Rheometer (DSR) in accordance with the AASHTO Standard Procedures (TP5).
Each asphalt was tested after aging with the Pressure Aging Vessel Oven (PAV). G* and Phase
Angle () measurements were used to compute the Dissipated Energy Ratio, which in turn was used
to compare the results for each binder under different testing conditions. Some information about the
testing is listed below.
 Energy input: two levels of energy input were used. One in the middle of the linear range and one
slightly over the linear limit, as explained earlier.
 Testing temperature was chosen as 15C, which corresponds to the average of the high and low
temperature of the PG 58-28, as discussed earlier. An additional testing temperature was also
chosen for the tested binders of 6°C which accounts for the freeze and thaw effects
 The testing frequency used was 10 Hz, which correspond to the fast traffic speed, as discussed
earlier.

Summary of Results
The results of 19 binders tested for fatigue performance are shown in Table 1. The
measurements and calculations confirm the high dependency on the stress level of the binder fatigue
life. The stress sensitivity was very similar for the two testing temperatures. For both, 15C and 6C, an
average decrease in the fatigue life of 15 times was registered when the stress was changed from the
linear to non linear range. However, for 6C the variations were higher. For this temperature the
maximum decrease in fatigue life when moving from strong to weak structure was 108 times (D5) and
the minimum was two times (B9). For 15C a maximum decrease of 39 (B5) times and a minimum of
five times (D4) was shown.
From the test results, it is not possible to establish a clear trend in the temperature
susceptibility of the fatigue life of binders. Some binders increase their fatigue life about two to three
times (D6 and B9 for low stress; D5 high stress) when the testing temperature dropped from 15C to
6C. Other binders decrease their fatigue life more than five times (A1 for low stress; C5, B8, C4 and
A1 for high stress) for the same temperature change.
The measured values of G* and  shown very little variations with the stress level. The
maximum variations were around 20% for G* and 30% for . No trend could be established, since for
some binders the value of G* or  increased with increasing the stress level, and for other binder the
value of the parameters decrease.
Rational Fatigue Limits 11

Table 1. Results of Fatigue Tests at 6C and 15C (Part I)


PG Temp. Stress G* Phase Wi Np20
Code K1 K2
Grade [C] [kPa] [kPa] Angle [kPa] (x103)
B7 PG58-40 15 183 3,580 49.0 22 40 4.66E13 2.086
(Elvaloy) 137 4,482 46.3 10 234
6 430 12,178 43.5 33 23 1.48E21 3.722
344 14,117 41.1 17 248
B9 PG58-34 15 288 6,224 49.9 32 11 1.38E15 2.468
(Elvaloy) 200 7,200 46.9 13 102
6 659 24,332 40.0 36 28 8.48E10 1.422
527 24,146 39.4 23 53
D4 PG58-34 15 371 10,980 46.5 29 23 2.76E12 1.812
(SB) 253 12,317 44.3 11 123
6 861 37,057 36.8 38 20 5.77E21 3.817
689 38,696 35.3 22 147
D5 PG64-40 15 161 2,132 51.9 30 4 2.96E14 2.419
(SB) 97 2,833 48.0 8 116
6 336 7,682 46.2 33 2 6.64E19 3.651
202 9,265 41.9 9 221
C5 PG58-28 15 615 20,454 42.2 39 13 3.65E12 1.841
(Non 320 22,050 40.0 9 177
modif..) 6 984 61,117 31.1 26 82 9.34E14 2.280
782 61,650 30.5 16 248
B5 PG64-34 15 322 5,202 51.8 49 7 2.09E17 2.868
(Elvaloy) 198 6,646 48.0 14 282
6 630 20,747 42.6 41 28 6.79E16 2.689
485 24,318 40.4 20 194
D2 PG64-34 15 420 11,139 46.3 36 44 4.35E23 4.169
(SB) 336 11,736 45.0 21 387
6 840 32,186 38.1 42 115 4.63E12 1.643
673 38,384 35.4 21 352
A3 PG64-28 15 568 16,037 45.3 45 9 1.23E18 3.036
(SBS) 435 18,157 42.7 22 78
6 1264 44,837 43.3 77 17 6.20E21 3.597
993 50,151 41.3 41 161
B2 PG64-28 15 570 18,142 39.8 36 16 5.45E21 3.847
(Elvaloy) 395 20,152 37.3 15 500
6 1302 52,498 31.9 54 15 1.54E28 5.074
1023 58,535 29.8 28 422
D1 PG 64- 15 733 25,273 39.4 42 13 5.59E17 2.949
28
(SB) 500 27,707 37.0 17 184
Delgadillo and Bahia 12

Table 1. Results of Fatigue Tests at 6C and 15C (Part II)


PG Temp. Stress G* Phase Wi Np20
Code K1 K2
Grade [C] [kPa] [kPa] Angle [kPa] (x103)
D1 PG64-28 6 1490 75,414 29.1 45 11 1.84E24 4.347
(SB) 1185 80,425 27.7 26 128
D6 PG70-34 15 289 5,194 52.4 40 17 1.38E17 2.806
(SB) 208 6,578 48.9 16 237
6 650 17,503 44.8 53 11 7.19E33 6.310
580 18,913 43.4 38 85
B8 PG70-34 15 336 6,491 49.1 41 9 1.10E16 2.621
(Elvaloy) 210 8,007 45.3 12 210
6 620 24,567 39.5 31 51 1.29E31 5.874
540 24,651 39.1 23 275
C4 PG64-22 15 880 37,045 36.6 39 21 1.60E17 2.804
(SBS) 609 39,900 34.5 17 237
6 1390 101,987 25.8 26 133 6.27E13 1.966
973 114,387 26.6 12 637
C2 PG70-28 15 563 15,833 43.5 43 11 1.04E21 3.659
(SBS) 450 19,122 40.5 22 143
6 1355 52,609 33.2 60 13 2.52E17 2.784
996 58,071 30.6 27 112
A1 PG70-28 15 573 18,360 41.0 37 4 2.34E20 3.670
(SBS) 458 19,725 39.2 21 32
6 1141 56,841 30.6 37 26 8.94E23 4.283
913 57,905 29.4 22 216
B4 PG70-28 15 545 15,813 41.8 39 11 1.35E22 3.935
(Elvaloy) 436 16,960 40.1 23 98
6 1251 50,963 32.6 52 8 3.65E199 41.487
1042 44,256 38.8 48 182
B6 PG76-34 15 295 5,421 49.6 38 22 5.90E14 2.276
(Elvaloy) 236 6,591 46.5 19 104
6 675 23,694 39.7 39 16 1.70E17 2.843
540 24,320 38.9 24 63
C6 PG76-28 15 509 14,148 43.4 40 5 6.27E24 4.597
(SBS) 407 17,248 41.0 20 110
6 1097 44,856 34.0 47 10 7.61E26 4.896
894 47,041 32.3 29 117
B3 PG76-28 15 588 16,130 42.2 45 9 3.45E21 3.772
(Elvaloy) 452 17,506 39.6 23 114
6 1257 52,292 32.4 51 29 1.74E20 3.352
898 57,289 30.2 22 464
Rational Fatigue Limits 13

Computer Analysis with Multi-Layer Elastic Theory Program

To truly estimate the effect of pavement structure on the fatigue life of roads, a good estimate
of the pavement strains under traffic loads is needed. A computer program based on multilayer elastic
theory was used to calculate the strains in the asphalt layer. The pavement strain estimation allows
calculating the input energy that defines the fatigue life of the binder. Correlations from previous
research were used to estimate the mixture fatigue life from binder fatigue life, as explained in the
following sections.

Traffic Information
The six levels of traffic volume recognized on Wisconsin highways were used for deriving the
fatigue specifications. The traffic levels considered are shown in the second column of Table 10, in
accordance to the Wisconsin PG Binder Selection Criteria (18).
Two different traffic speeds were considered. High speed equal to 60 mph (10 Hz testing
frequency), which is 5 mph lower than the maximum allowable traffic speed in Wisconsin. Low speed
equal to 15 mph (2.5 Hz testing frequency), for taking into account the slow movements of traffic in
urban areas.

Selected Pavement Structures


Six different pavements structures were analyzed, three for each traffic speed. The pavement
structures selected represent typical structures used in Wisconsin for different traffic levels and
speeds (19). Two seasons were considered for the calculations of the material properties to include
the low subgrade support that results from the melting season.
The three selected pavement structures for high traffic speed and low traffic speed are shown
in Tables 2 and 3 respectively.

Table 2. Selected Structures for Fast Traffic

Structure Thickness
Layer Material [in]
Surface HMA 8
Base Rubbilized PCC 9
Major Highway
Subbase Crushed Aggregate 6
Subgrade Poorly sorted --
Surface HMA 4.5
Medium
Base Crushed Aggregate 12
Highway
Subgrade Poorly sorted --
Surface HMA 3
Minor Road Base Crushed Aggregate 6
Subgrade Poorly sorted --
Delgadillo and Bahia 14

Table 3. Selected Structures for Slow Traffic

Structure Thickness
Layer Material [in]
Surface HMA 7
Base Crushed Aggregate 18
Major Urban
Subbase Fly Ash Stab. 6
Subgrade Poorly sorted --
Surface HMA 4.5
Base Crushed Aggregate 6
Medium Urban
Subbase Fly Ash Stab. 6
Subgrade Poorly sorted --
Surface HMA 3
Minor Urban Base Crushed Aggregate 6
Subgrade Poorly sorted --

Resilient Modulus for the Granular Layers


The resilient modulus of the granular layers and subgrade was selected based on standard
values used in pavement design and given in the bibliography (20). Table 4 shows the values for the
resilient modulus and Poisson’s ratio.

Table 4. Resilient Modulus Granular Layers and Subgrade

Material Mr [psi] Poisson’s Ratio


Rubbilized PCC 50,000 0.35
Crushed Aggregate 25,000 0.35
Fly Ash Stab. 25,000 0.35
Poorly sorted 5,000
Subgrade 0.40

Seasonal variations were considered for the subgrade Resilient Modulus (Mr). A value of 1,250
psi was used in the thaw season of April and May.

Resilient Modulus of Asphalt Layers


Pavement Temperatures
The pavement temperature can be estimated from the air temperature. The relationship
between the mean pavement temperature T p and the mean monthly air temperature Ta is given can
be calculated using the following formula presented in Equation 5 (20).
 1  34
T p  Ta 1   6; (5)
 z  4  z 4

where: Ta and Tp in [F]


z = depth from pavement surface in [in]
The average temperature of the asphalt layer can be estimated using one third of the layer
thickness as the depth z. Table 5 shows the monthly average pavement temperatures for the
selected pavement structures. The air temperatures correspond to the average monthly temperatures
of Madison Wisconsin (21)
Rational Fatigue Limits 15

Table 5. Monthly Average Pavement Temperatures

Month Mean Air Mean Pavement TC


Temp. C H = 8 [in] H = 7 [in] H = 4.5 [in] H = 3 [in]
Jan -9 -7.2 -7.3 -7.5 -7.7
Feb -6 -3.7 -3.8 -4.0 -4.1
Mar 1 4.3 4.3 4.3 4.3
Apr 8 12.4 12.4 12.6 12.7
May 14 19.3 19.4 19.7 19.9
Jun 19 25.0 25.2 25.6 25.9
Jul 22 28.5 28.6 29.1 29.5
Aug 21 27.3 27.5 27.9 28.3
Sep 16 21.6 21.7 22.0 22.3
Oct 9 13.5 13.6 13.8 13.9
Nov 2 5.5 5.5 5.5 5.5
Dec -6 -3.7 -3.8 -4.0 -4.1
Average 7.6 11.9 11.9 12.1 12.2

Two different seasons were considered for the analysis. The thaw season, corresponding to
the months of April and May, was separated from the rest of the year. The average pavement
temperatures for the two seasons considered are shown in Table 6.

Table 6. Seasonal Average Pavement Temperatures

Month Mean Pavement Temperature C


H = 8 [in] H = 7 [in] H = 4.5 [in] H = 3 [in]
March - April 8.3 8.4 8.4 8.5
May - February 12.6 12.7 12.8 13.0

The values from Table 6 were rounded and temperatures equal to 8C and 13C were used for
the thaw season and the normal season respectively. These values were used for all the pavement
structures in the multilayer analysis.
Resilient Modulus of Asphalt Mixtures
The resilient modulus of the asphalt layer was estimated from extensional dynamic modulus
E*m relationship developed by Christiansen and coworkers (22). The relationship is based on Hirsch
model and it is shown in Equation 6.

1
  VMA   VFA  VMA  1  VMA / 10 2 VMA 
E * m  Pc 4.2  10 6 1  2   3 G * b     1  Pc   
  10   10 4   4.2  10
6 VFA  3 G * b 

Pc 
20  3 G * b
 VFA / VMA 0.58

; (6)
650  3 G *  VFA / VMA
0.58
b

Where: E*m = mixture dynamic modulus [psi]


Pc = Aggregate contact volume
VMA = voids in the mineral aggregate [%]
VFA = voids filled with asphalt [%]
Delgadillo and Bahia 16

G*b = Shear modulus of the binder [psi]

The VMA value depends on the aggregate size. The Wisconsin DOT Standard Specifications
for Highway design and Construction (23) specify nominal aggregate sizes of 19 mm and 2.5 mm for
the lower pavement layer and upper pavement layer respectively. The minimum VMA values
corresponding to these sizes are 13% and 14%. A VMA value of 14% was chosen for determining the
E*m of the asphalt mixtures in the present work. The VFA values recommended in the same
specifications are between 65% and 75% for most levels of traffic. A VFA value of 70% was selected
for the present calculations.
The G*b values for the selected pavement temperatures of 8C and 13C were interpolated from
the measured values at 6C and 15C, which are shown in Table 1. Two values of G* were measured
for each temperature, one for low stress level and one for high stress level. Consequently two values
of E*m were obtained for each binder and temperature. The difference between those two values
proves to be 4% as an average, which does not have a significant effect in the pavement strains and
fatigue life. For this reason, and in order to simplify data and calculations, it was decided to use the
same E*m for both stress levels. The average value between the low stress and the high stress was
chosen as the most appropriated one. The calculated E*m for each binder and temperature are
presented in Table 7.

Analysis
Strains
The strains in the asphalt layer were calculated using the Kenlayer computer program (24). For
the analysis, the following assumptions were made:
 Materials linear elastic. Each layer was assumed linear elastic with the modulus of elasticity
presented previously
 Two periods per year. A regular season and a thaw season were considered in the analysis. This
allows for evaluating the effect of the decrease in the subgrade support when it thaws.
 One single axle load group. The load considered is the resulting of a contact pressure 80 psi
single axle with a contact radius = 6 in.
 Strains are calculated at the bottom of the asphalt layer, for each pavement structure.
No damage analysis was carried out with the program. Only strains were taken from the
program output. The fatigue life was calculated in a separate spreadsheet.
Rational Fatigue Limits 17

Table 7. Resilient Modulus Asphalt Layers


E* Mix [ksi]
Code PG Grade Pav.Temp.[C] E* Mix [ksi] (60mph) (15mph)
B7 PG 58-40 13 950 800
(Elvaloy) 8 1250 1100
B9 PG 58-34 13 1200 1050
(Elvaloy) 8 1550 1350
D4 PG 58-34 13 1450 1250
(SB) 8 1800 1600
D5 PG 64-40 13 800 650
(SB) 8 1050 900
C5 PG 58-28 13 1750 1550
(Non
Modif.) 8 2050 1850
B5 PG 64-34 13 1150 1000
(Elvaloy) 8 1500 1300
D2 PG 64-34 13 1400 1250
(SB) 8 1750 1550
A3 PG 64-28 13 1600 1450
(SBS) 8 1950 1750
B2 PG 64-28 13 1700 1500
(Elvaloy) 8 2000 1800
D1 PG 64-28 13 1850 1650
(SB) 8 2200 2000
D6 PG 70-34 13 1100 950
(SB) 8 1400 1250
B8 PG 70-34 13 1200 1050
(Elvaloy) 8 1550 1350
C4 PG 64-22 13 2050 1850
(SBS) 8 2350 2200
C2 PG 70-28 13 1650 1450
(SBS) 8 2000 1800
A1 PG 70-28 13 1700 1500
(SBS) 8 2000 1850
B4 PG 70-28 13 1600 1400
(Elvaloy) 8 1950 1750
B6 PG 76-34 13 1150 1000
(Elvaloy) 8 1550 1350
C6 PG 76-28 13 1600 1400
(SBS) 8 1900 1700
B3 PG 76-28 13 1650 1450
(Elvaloy) 8 2000 1800
Delgadillo and Bahia 18

Damage analysis
The fatigue life of the binder was calculated using as spreadsheet program. Classic theory
assuming that the fatigue crack initiates in the bottom of the asphalt layer was considered. The
following steps describe the calculations carried out in the spreadsheet for determining the traffic
volume that each binder is suitable for. The results are presented for each pavement structure and
traffic speed.
 Asphalt mixture shear strain. The vertical and horizontal strains at the bottom of the asphalt layer
are used to calculate the mixture shear strain. The output shear strain from Kenlayer is null, which
indicates that the horizontal and vertical planes are principal strains planes. The maximum shear
strain () in any plane at that point is equal to half the difference between the vertical strain (v)
and the horizontal strain (r), as shown in Equation 7.
 = (v - r) / 2; (7)
 Binder shear strain. The shear strain demand on the binder is calculated from mixture shear strain
given by Equation 7. Previous research estimated that an appropriate average estimate for binder
strain 0 is between 30 and 100 times the mixture strain (25). A value of 50 times was chosen for
the present analysis.
 Input energy Wi, is calculated as shown in Equation 1. The values of G* and  were measured for
each binder at 6C and 15C, as it was presented in a previous section. The G* and  values at the
average seasonal temperatures (8C and 13C) were interpolated from the measured values. 0 is
the binder shear strain.
 Binder fatigue life Np20. The fatigue relationship defined in Equation 3 was used to calculate
Np20. The K1 and K2 parameters were obtained at 6C and 15C as presented in a previous
section. It was assumed that the fatigue life relationships at these temperatures were good
estimates of the fatigue life at the average seasonal pavement temperatures considered (8C and
13C). For each binder and pavement structure, two Np20 values were calculated, one for each
season.
 Mixture fatigue life Nmix. Mixture fatigue life was correlated with binder fatigue life (Nbinder) during
the NCHRP 9-10 project (2). Nine different binders and four kinds of aggregates were used to
establish the relationship. The beam fatigue test was applied to the mixtures (AASHTO TP8). For
binders and mixtures, the test was carried out at the intermediate grade temperature of the binder.
The frequency used was 10 Hz for asphalt and mix. The fatigue life was defined as the number of
cycles required for the stiffness to decrease 50%. The strain level used in the binder test was 40
to 120 times the strain level in the mix test, which is similar to ratio used in the present work (50
times). The relationship obtained is shown in Equation 8. The correlation for the fitting was equal
to 0.84.
Nmix = (Nbinder + 52808) / 4.9724; (8)
 Allowable traffic volume (ESALs). The number of ESALs for each binder and pavement structure
was calculated from the two seasonal Nmix values. The Nmix results were weighed according to the
length of the corresponding season. The thaw season lasts only two months, so it was weighed as
1/6 of the year. The regular season lasts 10 months so it was weighed as 5/6 of the year.

Results

Allowable ESALs were calculated for each binder at six different pavement structures (three for
high speed and three for low speed). An example of the results for binder D2, using fatigue for high
frequency of loading, for three types of pavement structure is shown in Table 8. Table 8 includes the
Rational Fatigue Limits 19

mixture strains (radial, vertical, and shear) calculated from Kenlayer, which are used to estimate the
maximum strain in the binder. This estimate is used with the G* value and delta value of the binder,
measured at the high frequency testing rate in the DSR, to calculate the maximum input energy. The
Np20 in binder can be estimated using the input energy Wi and the fatigue coefficients K 1 and K2 (as
shown in Equation 3). The allowable Np20 of the binder is converted into the mixture allowable cycles
(Nmix) using Equation 8. The total allowable ESAL value shown at the bottom of Table 8 for each
type of highway is obtained from the Nmix for both periods.
The same type of analysis is used to calculate the allowable ESALs for low speed for the same
binder as shown in Table 9. As can be seen the speed can have a very important effect that is
dependent on highway type (pavement structure).
To see if this analysis is logical and could be applied for the binders used in practice today,
the binders ranked according to the current traffic levels specified by the Wisconsin DOT. The total
allowable ESALs determined for each of the binders were compared to the six traffic volume levels
considered in the design of Wisconsin highways (18). Table 10 presents the binders that could be
used for the different pavement structures and traffic speeds.

Table 8. Allowable ESALs for D2 and High Speed


Medium
D2
Wi 
Major Highway
   0
2
 G
Highway
*  si
Minor Road
n 

Period 1 Period 2 Period 1 Period 2 Period 1 Period 2


Mixture Radial -6.6E-05 -6.7E-05 -1.5E-04 -1.5E-04 -2.8E-04 -2.5E-04
Strain Vertical 7.3E-05 7.6E-05 1.7E-04 1.7E-04 3.1E-04 2.8E-04
Shear 6.9E-05 7.1E-05 1.6E-04 1.6E-04 2.9E-04 2.7E-04
Input Binder 0 3.5E-03 3.6E-03 7.9E-03 8.1E-03 1.5E-02 1.3E-02
Energy G* [kPa] 32462 17658 29986 16737 27509 15816
sin 0.61 0.68 0.63 0.69 0.64 0.70
Wi [Pa] 743 481 3711 2375 11842 6255
Fatigue K1 -1.643 -4.169 -1.643 -4.169 -1.643 -4.169
Life K2 4.63E12 4.35E23 4.63E12 4.35E23 4.63E12 4.35E23
N Binder 8.87E07 2.88E12 6.31E06 3.68E09 9.38E05 6.50E07
N Mix 17.8 E6 578 E9 1.3 E6 741 E6 199 E3 13.0 E6
Total = 107 E6 ESAL 7.6 E6 ESAL 1.2 E6 ESAL
Delgadillo and Bahia 20

Table 9. Allowable Esals for D2 and Low Speed


Major Urban Medium Urban Minor Urban
D2 Wi
    0
2  G *  si

Period 1 Period 2 Period 1 Period 1 Period 2 Period 1


n 

Mixture Radial -8.4E-05 -9.1E-05 -1.6E-04 -1.6E-04 -3.0E-04 -2.7E-04


Strain Vertical 9.3E-05 1.0E-04 1.8E-04 1.8E-04 3.3E-04 3.0E-04
Shear 8.8E-05 9.6E-05 1.7E-04 1.7E-04 3.1E-04 2.9E-04
Input Binder 0 4.4E-03 4.8E-03 8.6E-03 8.7E-03 1.6E-02 1.4E-02
Energy G* [kPa] 22723 12360 20990 11716 19256 11071
sin 0.61 0.68 0.63 0.69 0.64 0.70
Wi [Pa] 845 612 3032 1926 9554 4980
Fatigue K1 -1.643 -4.169 -1.643 -4.169 -1.643 -4.169
Life K2 4.63E12 4.35E23 4.63E12 4.35E23 4.63E12 4.35E23
N Binder 7.18E07 1.05E12 8.80E06 8.82E09 1.33E06 1.68E08
N Mix 14.5E6 211E9 1.78E6 1.77E9 279 E3 33.8E6
Total = 86.7 E6 ESAL 10.7 E6 ESAL 1.67 E6 ESAL

Table 10. Applicable Binders for the Selected Pavement Structures


Traffic
Volume Applicable Binders
Speed (million Strong Medium Weak Structure
ESAL) Structure Structure
B7, D5, A3, D1, B8,
0 – 0.3 D4, C5 C2, A1, B6
B5, D2, B2, D6, B4,
0.3 - 1.0 B7, B9, D5, C4 B3
D2, A3, D1, B8,
Fast 1.0 - 3.0 D4, C5 B6 C6
(60
mph) 3.0 - 10 B9 B5, C2, A1
10.0 - 30.0 B7, C4 B2, D6
All other
> 30.0 binders B4, C6, B3
B7, B9, D5, D1, B8,
0 – 0.3 D4, C5 C4, A1, B6
0.3 - 1.0 B9 B5, D2, A3, C2
B7, D5, B8, C4,
Slow 1.0 - 3.0 B9, D4, C5 B6 B2, D6, B4, C6, B3
(15
mph) 3.0 - 10 D2, A3, D1, A1
B7, D5, D2,
10.0 - 30.0 C4, B6 B5, D6, C2
All other
> 30.0 binders B2, B4, C6, B3
Rational Fatigue Limits 21

In Table 10, a binder is applicable for a specific traffic level (at a specific pavement structure
and traffic speed) only if the allowed ESALS calculated were equal to or higher than the highest value
of the traffic range. For example, for high speed, medium structure, D2 has 7,682,645 allowable
ESALs, as it is shown in Table 8. That is why in Table 10 the binder is ranked in the 1.0 to 3.0 million
ESALs level. The binder, however, could also be used for any of the lower traffic volume levels for the
same speed and pavement structure. A special case was considered for the highest traffic range (>30
million ESALs). Only binders that showed outstanding fatigue life were assigned to this traffic
category. These are the binders that showed a fatigue life larger than 100 million ESALs. These
binders are very unlikely to ever show fatigue failure during service at the corresponding pavement
and speed conditions. As shown in Table 10, the results of this approach appears logical since there
are binders that are applicable to all the traffic and pavement condition combinations with only a few
exceptions. The condition of weak structure and high volume of traffic are very extreme combination
and there appears to be no binder to meet the demanding conditions.

Analysis of Results

In the fatigue analysis presented in the previous section, three factors were considered: traffic
speed, seasonal effects, and pavement structure. To understand the relative importance of these
factors, each is analyzed separately.

Importance of Traffic Speed on Fatigue life of Pavements


The traffic speed affects the fatigue life mainly by influencing the binder and mixture modulus.
Higher speed results in a higher modulus; higher modulus will result in a lower strain for a fixed
loading application. The effect on fatigue life however is not simple and could depend on binder
sensitivity to loading rate and fatigue dependency on energy input. Some binders are more sensitive
than others in terms on modulus change but are less sensitive in terms of fatigue change with energy
input. Figure 7 is a summary of the ratio of fatigue life, defined by the allowable EASLs, when slow
speed (15mph) is used relative to high speed (60 mph). What is interesting to notice is that the ratios
are always greater than 1.0, which means that slow speed is better for all binders when a medium
pavement structure is used in the analysis. It is also noticed that the changes are rather small and in
most cases they are within a factor of 2.0. As will be discussed next other factors might have much
more important effects.

Seasonal Effect in the Fatigue life of Pavements


The seasonal effect on the strain in the asphalt layer is based on two factors. First, the lower
temperature of the thaw season compared with the normal season results in a higher modulus of
elasticity of the asphalt layer in the thaw season. Second, the modulus of the subgrade is lower
during the thaw season due to high moisture. These two effects act in opposite ways. The higher
asphalt modulus in the thaw season tends to
Delgadillo and Bahia 22

Figure 7. Sensitivity of Fatigue Life to Traffic Speed

decreases the strains in the pavement, compared with the normal season. On the contrary, the lower
bearing capacity of the subgrade tends to increase the strains in the thaw season, compared with the
normal season. The resulting strains for all the binders did not show great variations between the two
seasons. The effects of the two factors seem to cancel out.
Despite the fact that the strains did not vary significantly between the two seasons, the fatigue
life of the pavements showed important variations between the two periods. The reason for these
variations is the dependency on temperature for the binder fatigue life relationship. Some binders
showed higher fatigue in the thaw season than in the normal season. Some other binders presented
a longer fatigue life in the normal season than in the thaw season. Figures 8 and 9 were prepared to
show the ratio of fatigue life estimated for the pavement during normal season (dry season, average
temperature) to the fatigue life estimated for the thaw (wet and cold) season. Figure 8 shows the
binders with a ratio lower than 1.0 and Figure 9 shows the binders with ratios higher than 1.0. It is
important to observe the great variations in the ratios for the different binders. What these data
indicate is that the seasonal effects are very important and should be seriously considered in fatigue
selection criterion. The detailed analysis also indicates that the most influencing factor in the seasonal
effect is the fatigue relationship represented by the fatigue parameters K1 and K2. The

Figure 8. Seasonal Effect on Pavement Fatigue Life (Binders with Higher Fatigue life in the Thaw
Season)
Rational Fatigue Limits 23

Figure 9. Seasonal Effect on Pavement Fatigue Life (Binders with Higher Fatigue life in the Normal
Season)

same binder has different fatigue relationships depending on the testing temperature, as it was
discussed previously in this paper. The susceptibility to temperature and to strain in pavements is
very asphalt specific.

Influence of the Pavement Structure


Pavement structure has a direct impact on the strain demand on the binder. The strains in the
binder were between three

Figure 10. Sensitivity of Fatigue Life to Pavement Structure

and four times higher in the weak structure than in the strong structure for all binders. Due to this
difference, the input energy in the weak structure was between six and 19 times higher in the weak
structure than in the strong structure for the binders considered. Differences of this magnitude in Wi
result in larger differences in the fatigue life because the relationship follows a power law. Figure 10 is
prepared to show the ratio of the fatigue life of mix (Nmix) of strong pavement structure to a weak
pavement structure. As can be seen, the fatigue life of weak structures range from two to five orders
of magnitude (100 to 100000) lower than the life for the strong structure. The strain sensitivity of the
binders showed a wide range and could not be linked to specific grades or type of additive. Some
asphalts like B9, C5 and B6 showed a decrease of only two orders of magnitude in the fatigue life for
weak pavement. Some other asphalts like B8 and D6 had a decrease of four to five orders of
magnitude for the weak structure. What these results show is that pavement structure is a very
important factor that must be included in rational binder selection criteria.
Delgadillo and Bahia 24

Proposed Criteria

The analysis carried out in the previous section represents a mechanical-empirical approach to
determine the fatigue life of binders. The steps required to carry out the analysis are three: testing the
binder at the corresponding Wi and temperatures

Figure 11. Wi values, Major Highway, Thaw Season

Figure 12. Wi values, Minor Road, Thaw Season

expected in the typical pavement structures; conducting a pavement analysis with multilayer elastic
theory program to estimate specific Wi values, and from the binder fatigue life, estimate the allowable
EASLs for the pavements.
From the analysis results, it is possible to derive some recommendations for the testing
conditions and required limits for accepting binders for specific pavement conditions. The testing
conditions are represented by the input energy that represents each pavement structure at the
appropriate pavement temperature. Figures 11 and 12 shows the Wi values in the thaw season (8C)
that represent the major highway (strong structure) and minor road (weak structure) for all the
binders. The group of binders represents a good sample of the binders currently used in the State of
Wisconsin. Therefore, the average Wi values from all the binders for a specific structure and
temperature were used as an estimate of the required Wi for testing. Table 11 presents the estimated
Wi values for the selected structures and temperatures.
Rational Fatigue Limits 25

Table 11. Testing Conditions for Selected Structures and Temperatures

Testing Input Energy for Testing [Pa]


Temperature Major Medium Minor Major Medium Minor
[C] Highway Highway Road Urban Urban Urban
8 800 4000 13000 800 3000 10000
13 500 2500 6500 500 2000 5000

As shown in Table 11, the input energy for testing recommended for the major highway (high
speed) is lower than the Medium and Minor highway because the structure considered for Major
highways is stronger than the other types of highways. This means that deformations (strains) are
lower in the first one which is reflected in the testing conditions used. Also it shows that the fast speed
requires more Wi for testing, as it was explained earlier. Only for the major highway there are no
differences in the Wi for fast and slow speed, which means that for the strong structure the speed
effects are minimized.
Proposed Limits for Np20 for the Structures and Temperatures Considered
The estimated limits for Np20 are a function of the traffic volume. For the six levels of traffic
volume considered, the minimum Np20 can be obtained from the fatigue relationship between mixture
and binder that was used previously in the analysis. When the design traffic is input in the
relationship, the minimum Np20 for that traffic volume is obtained. The testing conditions (Wi,
temperature) vary for each binder depending on the pavement and weather conditions (see Table
11). But the Np20 limits required are function of the traffic volume only. Table 12 summarizes the
minimum Np20 values required for each traffic range. Further validation is needed to confirm if the
relationship used between binder fatigue life and mixture fatigue life is applicable directly to the
pavement in the field. Combining Table 11 and Table 12 can serve as a system of testing and
accepting asphalt binders to meet specific conditions.

Table 12. Np20 Requirements for Traffic Levels

Traffic Volume Minimum


(million ESAL) Np20
0 - 0.3 1500000
0.3 - 1.0 5000000
1.0 - 3.0 15000000
3.0 - 10 50000000
10.0 - 30.0 150000000
> 30.0 500000000

Summary of Findings

The calculation of strains from multilayer elastic analysis of typical pavement sections was used to
estimate the strain demands on asphalt mixtures due to actual loads on the pavement. The mixture
strains were converted in binder strains and used to estimate the input energy for the fatigue testing
of binders. A set of asphalt binders was tested at the range of strains to evaluate the possibility of
these binders satisfying the anticipated traffic volumes. The results were used to propose a set of
testing conditions and criterion limits for binder fatigue. The finding of the study could be summarized
as follows:
Delgadillo and Bahia 26

 Temperature, energy input, and loading rate can have significant effects on binder fatigue life
measured in the lab. The effects are highly binder type specific. Analysis of the relative effects, in
the context of pavement layered systems, shows that loading rate could have only a minor effect
compared to the pavement structure (strain in binder) and temperature.
 Pavement structure is found to be a very important factor in selecting binder for better resistance
of the fatigue life of pavements. Fatigue life of the same binder was between two to seven orders
of magnitude lower when a weak structure was considered compared with a strong structure. It is
therefore essential to include the pavement structure in future binder specifications for fatigue.
 Temperature has a dual effect on fatigue of asphalt. It could change binder and mixture modulus,
thus changing the strain input, and it could also change fatigue coefficients (K1 and K2). It is thus
essential that an effective fatigue testing temperature be selected that also includes critical
seasons of the pavement, such as thaw or wet seasons.
 There are some weak connections (assumptions) made in this study and require more work. The
effect of testing geometry is important and needs to be resolved. The relationship between binder
fatigue life and mixture fatigue life used in this analysis is based on only one study (NCHRP report
465). The allowable ESALs from each pavement are based on theoretical analysis and will need
to be validated with field studies. The Wisconsin DOT and the University of Wisconsin Madison
are currently conducting a field validation project for evaluating the results of this work and other
performance characteristics of binders.

Acknowlegement

This study was partially sponsored by a grant from the Wisconsin Department for
Transportation(WisDOT) through the Wisconsin Highway Research Program (WHRP). The support of
both WHRP and WisDOT are gratefully acknowledged. The feedback and guidance received from Mr.
Scott Shwandt (WHRP), Ms. Judie Ryan (WisDOT), Mr. Tom Brakow (WisDOT) and the help from
Mr. James Bongard (WisDOT) are acknowledged. The assistance of Dr. Kitae Nam (Washington
State University) in the binder testing is also gratefully acknowledged. The findings of the study
represent the opinions of the authors and are not necessarily supported nor endorsed by the
Wisconsin DOT or the WHRP.

References

1 J. D’ Angelo. “Performance Specification for Asphalt Binders in the U.S.”. Plenary


lecture at the 39th Annual Peterson Asphalt Research Conference. Laramie, Wyoming, July 15-17,
2002.
2 H.U. Bahia, D.I. Hanson, M. Zeng, NCHRP Report 459, “Characterization of Modified
Asphalt Binders in Superpave Mix Design”, National Academy Press, Washington, D.C., 2001.
3 H. U. Bahia, H. Zhai, M. Zeng., Y. Hu, and P. Turner, “Development of Binder
Specification Parameters Based on Characterization of Damage Behavior,” Journal of the Association
of Asphalt Paving Technologists, Vol. 70, (2001).
4 K. Bonnetti. K. Nam and H.U. Bahia. “Measuring and Defining Fatigue Behavior of
Asphalt Binders”, Transportation Research Records, 2002.
5 C. L. Monismith, J.A. Epps and F.N Finn, “Improved Asphalt Mix Design,” Proceedings,
Association of Asphalt Paving Technologists, Vol.54 1985.
6 R. Reese. “Properties of Aged Asphalt Binder Related to Asphalt Concrete Fatigue life”.
Journal of the Association of Asphalt Paving Technologists, Vol.66 1997
Rational Fatigue Limits 27

7 G. M. Rowe and S.F. Brown. “Validation of the Fatigue Performance of Asphalt Mixtures
with Small Scale Wheel Tracking Experiments”. Journal of the Association of Asphalt Paving
Technologists, Vol.66 1997.
8 Y. Kim, D.N. Little and R.L. Lytton. “Use of Dynamic Mechanical Analysis (DMA) to
evaluate the fatigue and Healing Potential of Asphalt Binders in Sand Asphalt Mixtures”. Journal of
the Association of Asphalt Paving Technologists, Vol.71 2002
9 Y. Kim, D.N. Little and R.L. Lytton. “Fatigue and Healing Characterization of Asphalt
Mixtures”. Jounal of Materiaals in Civil Engineering, ASCE. Vol 15, Jan-Feb2003.
10 Y. Kim, D.N. Little and R.L. Lytton. “Effect of Mineral Fillers on Fatigue Resistance and
Fundamental Material Characteristics”. Transportation Research Record, 2003.
11 W. Van Dijk and W.Visser. “The Energy Approach to Fatigue for Pavement Design”.
Journal of the Association of Asphalt Paving Technologists, Vol.44 1975.
12 A. C. Pronk “Evaluation of the Dissipated Energy Concept for the Interpretation of
Fatigue Measurements in the Crack Initiation Phase”. Road and Hydraulics Engineering Division, the
Netherlands, 1995.
13 D.A. Andersen, M.Y. Le Hir, M. Marasteanu, J-P. Planche, D. Martin and G. Gaauthier.
“Evaluation of Fatigue Criteria for Asphalt Binder”. Transportation Research Record, 2001.
14 H. Soenen and B. Eckmann. “Fatigue Testing of Bituminouns Binders with a Dynamic
Shear Rheometer”. Proceedings of the 2nd Eurasphalt and Eurobitumen Congress, 2000.
15 W. Martono, H.U. Bahia, J. D’Angelo. “Comparison of Asphalt Binders’ Fatigue
Measured using Parallel Plate Geometry with Torsion Cylinder Geometry”. 41st Petersen Asphalt
Research Conference, June 2004.
16 W. Martono and H.U. Bahia. “Defining Asphalt Binder Fatigue as a Function of
Pavement Temperature and Pavement Structure”. Accepted for presentation and publication on TRB
2003.
17 M. Zeng, H. U. Bahia, Huachun Zhai, M.R.Anderson and P. Tuner. “Rheological
Modeling of Modified Asphalt Binders and Mixtures”. Journal of the Association of Asphalt Paving
Technologists, Vol. 70 2001
18 The Wisconsin Department of Transportation, “The PG Binder Selection Criteria for the
State of Wisconsin”, February 2002.
19 Private communication with Mr. Scott Schwandt, WAPA
20 Y. Huang, “Pavement Analysis and Design”, Prentice Hall, 1993.
21 www.weather.com
22 D. W. Christiansen, T. Pellien and R. F. Bonaquist. “Hirsch Model for Estimating the
Modulus of Asphalt Concrete”. Journal of the Association of Asphalt Paving Technologists, Vol.72
2003
23 The Wisconsin Department of Transportation. “Standard Specification for Highway and
Structure Construction, 2000 Edition”, 2000.
24 Y.H. Huang. “Kenlayer Computer Program”. Department of Civil Engineering University
of Kentucky, Lexington, 1992.
25 S. Kose, M. Guler, H.U. Bahia and E. Massad. “Distribution of Strains within Asphalt
Binders in HMA Using Imaging and Finite Element Techniques.” TRB1728, Journal of the
Transportation Research Board, pp. 21-27, 2000, Transportation Research Board, Washington, D. C.

You might also like