ASPHALT MIX FATIGUE BEHAVIOR:

EXPERIMENTAL STRUCTURES AND MODELING

H. Oddon - Research Engineer G. Caroff - Head of Pavement Department

J.-C. Gramsammer - Research Engineer Sktauroute
Laboratoire Central des Ponts et ChaussCes 23, avenue du Centre
B.P. 19 Saint-Quentin-en-Yvelines
Bouguenais, 44340, France Guyancourt Cedex, 78286, France

Abstract. An extensive research program on asphalt asphalts of the same grade but from ditTerent origins
mix fatigue was conducted jointly by the LCPC led the LCPC to conduct a vast experimental program
(French transportation research laboratories) and examining asphalt fatigue on its circular fatigue test
Sc&auroute (French highway engineering agency) track between 1990 and 1994. The program sought to
between 1990 and 1994. answer the following questions :
The program included tests on the LCPC circular 0 Is the laboratory fatigue test of asphalt materials
fatigue test track and a theoretical analysis of the on which the French rational method of pavement
hehavior of structures. This article looks into the parts design is based, always rcalistic and representative of
of the program dealing with the fatigue tests and hehavior on the road ?
modeling. The aim of this program was to clarify the l Are there other fatigue tests which are more
behavior in the laboratory and on pavements of asphalt appropriate ?
mixes differing only in the type of asphalt. l What correlation coefficient should be adopted
Additionajly, it involved adding constituents to in the context of the French method for High Modulus
improve the design method in the case of high- Asphalt Mixes, developed with very hard binders ?
modulus asphalts. l What is the contribution of modificd binders in
On the fatigue test track more than 7 IO6 load the fatigue behavior of asphalt mixes ?
sequences were applied to twelve structures over the This program included three successive
course of three experiments. The experimental experiments, conducted in jointly with the ASFA
pavements revealed similar behavior in asphalt (Association of French motorway companies), with the
concrete of the same thickness, and the good involvement of the So&e des p&roles Shell in the
performance of high-mod&r.3 materials whenever first two tests and then EK4ntsr France. In
proper thicknesses were used. The model made it conjunction with these experiments on the test track a
possible to determine a correlection coefficient in long series of laboratory tests were carried out, in order
connection with rigid materials for the French to gain a better understauding of the me&misms
pavement design method and brought out the value of involved. The principal results of the observation and
controlled-force tests. the interpre-tation of this experimental series of tests
are the subject of this paper ; laboratory results are
Keywords. Fatigue bchavior - Experimental structures outlined in the article by de La Roche and Riviere
- Asphalt materials - Modeling. (1997).
After a brief description of the LCPC circular test
track and the methods used to monitor the
experimental pavements, the three successive
1 - INTRODUCTION experiments and their main results are presented. The
redts are then interpreted in the light of ah of test
DitTerences in behavior fatigue observed in the track findings and laboratory tests conducted at the
laboratory on asphalt concrete prepared with pure same time.
882

2 - TEE FATIGUE TEST TRACK, pavement structures on four sectors of 30 m length

THE EXPERIMENTAL PAVEMENTS AND each.
TEE METHODS USED The construction of the experimental pavements as
TO MONITOR PAVEMENT CHANGES web as the manufacture of materials are controlled by
the LCPC and its partners. At the time of
2.1 - The LCPC test track. The LCPC test track implementation, the LCPC instaLls several sensors in
(Photo 1) is a powerful simulator of heavy traffic. A the pavements which allow the evolution and behavior
summary of its features, which are described in detail of the pavements to be studied.
by Autret and al. (1987), is provided.
2.3 - Monitoring the evolution of the experimental
pavements. Evolution of the pavements is monitored
by sensors located in the pavements and by the surface
measurements by means of various inspection
equipment. In addition to this there is the daily visual
inspection and the degradation monitored by video ;
these inspections of degradation are expressed in terms
of level of cracking, which is the percentage of length
of cracked pavement over total length of pavement. A
transversal crack is counted as 0.5 m of cracked length.
The sensors located in the pavements are
essentially strain gauges. Vertical pressure sensors can
be combined with these and also dynamic deflection
sensors and of course, temperature gauges. This
instrumentation installed during construction is
Photo 1 : circular test track of the LCPC - possibly complemented later by devices allowing the
overview water conditions of the ground and the non-treated
materials (piezometers, strain gauges) to be monitored.
Figure 1 is an example of the instrumentation of one of
This circular test track of 40-m diameter, pavements tested here.
simulates the movement of heavy axles, single or twin, Surface measurements include the use of the
driven by four arms at speeds capable of exceeding 100 Benkehnan beam, the Falling Weight Deflectometer,
kmh. the inclinometer or the trausversoprotilograph.
The applied loads can be varied so as to cover the
range of axle loads that exist around the world : the
stresses applied by these loads are kept constant by
means of an intermediate rolling support.
With its adjustable radii of rotation and its 0
transverse wheel sweep mechanisms, the test track is 20
capable of applying loads to the entire surface of a 6-m
40
wide pavement.
The installation operates automatically night and 60
day and, if necessary, during weekends. This allows 80
tests capable of exceeding well over a million circuit
100
laps per month.
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
To date, the test track has simulated the passage of
more than 50 million heavy axles in about 150
experiments, conducted as in this case with one or
several partners (one experiment corresponds to one
pavement tested with a given load configuration).
Figure 1 : instrumentation example
2.2 - Experimental pavements. The pavements are of an experimental pavement
constructed according to the practices usually from the programme “Asphalt mix fatigue”
employed by the road construction companies. An
experimental track generally comprises four different
 H. ODEON ETAL. 883

3 - TEE ASPHALT MIX FATIGUE TEST TRACK

PROGRAM

This program was conducted over a four-year

period, between 1990 and 1994. During this period, the
test track applied seven million loads in successive
stages (simulating the 130~kN axle, the maximmn
legal load per axle in France). These various
experiments had the following points in common :
l Configuration of the loads : dual wheel axle of
65 kN with Dunlop SP 321 tires inRated to 0.8 MPa
(imprint i&&rated in Figure 2). 1 2 3 4 5 6 7 8 9 10 1,
l Angular loading velocity : 10 rpm (the linear Position of load
velocity varies between tests 1 and 2 on the one hand
and test 3 on the other hand as the mean radius of
rotation goes from 19 m for the first two experiments
Dishnce to medium radius (m)
to 18 m for the last one).
l The transversal sweep of the loads was the same
for the three tests. It is ilhrstrated in Figure 2. Figure 2 : load imprint and
l Four structures were tested in each test. The wheels transversal sweeping histogram
aggregates used to make the asphalt materials a.ll have
the same origin (Noubleau quarry) and, in addition, all
the materials have the same O/14 grading (Table 1).
l The sub base is the same for all the pavements Screen (mm) 1 14 10 6.3 4 2 1 0.315 0.08
studied. It consists of 400 mm of well graded untreated Passing(%) 1 99 76 55 44 33 23 13 7.4
grauular material (called “GRH” in the text) from the
Man&h&es quarry of grading O/20 and O/14 (that Table 1 : grading curve of the granular
means that 0 is the minimal particle size and 20 or 14
material used in the asphaltic mixes
the maximal particle size, in cm).
l The subgrade is of poor quality (CBR between 5
and 10).
For each of the three experiments presented, the
A second objective was to compare the behavior
following wiU be sommarized in order : the aims of the
experiment, the characteristics of the four experimental under traiZc of an improved classical asphalt road base
material (,,,A”) and a High modulus asphalt mix
pavements, running the tests and the principal results
(“EME”) with a view to specifying the design methods
obtained.
of these materials, of more recent use.
3.1 - Experiment 1. The observations relating to
experiment 1 have been the subject of a publication for 3.1.2 - Structures studied. The stictures are
the TRB by de la Roche et al. (1994) ; the principal shown in Figure 3 (next page). The theoretical
thickness of the asphalt courses is 80 mm for the first
elements of this are summarized in the following in
order to provide a greater understanding of the overall three sectors with the following in order : “BB” B
program. (50170 asphalt), “BB” A (50170 asphalt) and the
“EME” (1 O/20 asphalt). The theoretical thickness is
120 mm for the fourth sector consisting of the “GBA”
3.1.1 - Aims of the test. The first objective of this
test was to study the relative behavior, under traffic (50/70 asphalt). These courses are supported on 400
simulated by the test track, of two asphalt mixes mm of GRH (O/20 well graded untreated granular
developed horn asphalts of the same grade but different material). It is also to be noted that all of the structures
have been covered with a “BBTM” (Very Thin Asphalt
origins, other things being equal. These two materials,
Concrete) of 25 mm theoretical thickness.
called “BB” A and “BB” B, yielded very different
results duriug fatigue tests iu the laboratory (sinusoidal
bending on trapezoidal sample with controlled 3.1.3 - Running the experiment. The pavements
displacement, without rest period), leading to a ratio of were built and instrumentation was fitted under good
6 between the theoretical service life of the mixes, to conditions. Compaction values obtained were in
the advantage of “BB” B, as indicated by de la Roche accordance with those expected. However, the
and Riviere(1997). thicknesses used for the asphalt materials were
884

somewhat diEerent to the theoretical thicknesses ; it For this experiment, two loading phases can be
was necem.ry to take these diEerences into account distinguished.
when the results were interpreted. The actual A first phase, referred to as “cold”, with 90 % of
thitiesses are summarised in Table 2. The presence the loads applied between 0” C and 15” C (temperature
is also noted of a zone of low thickness for a length of measured every hour at the middle of the asphalt layer,
a few meters for the “EME” of sector III. at a depth of 40 mm), took place between October and
December 1990 and comprised 1,165,OOO loads. In
terms of rain, this phase was characterized by strong
rain at the very beginning of the test and during the
Structure m structure Il last 200,000 loads. There was practically no rain
EMEC lo10 BBA5WO between these two periods. As previously agreed, this
4% 5,4% phase was stopped after the first appearance of
BBTM 25 nun BBTM 25 m degradation on sectors I (“BB” B) and II (“BB” A).
BMF.c somm
A second phase referred to as “hot”, during which
GRH 203nnn GRH ?GOmn 80 % of the loads were made between 15 and 20’ C
GBH 2OJmm GRH mmm (temperature measured every hour at the middle of the
asphalt layer, at a depth of 40 mm), was conducted
between May and July 1991 with the application of
1,565,OOO additional loads. This second phase was
stopped after a signilicaut state of degradation was
obtained on all sectors. In terms of rainfa4 there was
very little rain over the whole of this second phase,
except for the first 100,000 loads.
The tkst phase was preceded by an initial
measurement sequence called “as built” state.

a) “As built” state

Here, we shall only present the deflection
measurements with the Benkehnan beam, the
measurements of radii of curvature with the
inclinometer and strain measurements given by the
%Ncture Iv sblicture I gauges located in the pavements (Table 3).
GB A W70 BBBSWO
SI SII SIII SIV
GBA 12Omm
BBB BBA EMJ%C GBA
“As built” measurements
GRH 2OOmm GRH 2oOmm
d (11100 mm) 96 117 95 84
GRH 2comn GRH 2OOmm R&(m) 111 91 161 158
EKX-JG(@% 134 122 77 95
ETRAN.9ww 131 120 42 111
Flgure 3 : experimental structures 0 (“Cl 19.5 19 19 17.5
from the 1 st experiment
Evolution of deflection
d after
400 000 96 117 95 84
SI SII SIII SIV loadings
BBB BBA EMEC GBA da&r
1 165000 118 150 114 91
thmathmathmathmo
BBTM 21 4 11 2 102 54 loadings
h4B 61 9 73 3 77 9 100 12
GRH 431 8 422 I 420 13 436 28 d : mean deflection at 20°C
RoC:meauradiusofcurvatureat21”C

Table 2 : mean thickness and standard Table 3 : “as built” measurements results for
deviation of the testing 1 structures (mm) the experiment 1 - evolution of the average
deflections during the “cold” period
 H. ODEON ET AL,. 885

b) The “cold” phase

The evolution of the pavements during this cold
phase is summarized by the deflection measurements 160
(Table 3). It concerns average deflections at 20” C. 140
These deflections increased by 23, 28, 20 and 8 % on 120
the sectors I, II, III and IV respectively. It therefore 100
appears that sector IV (,,,A”) has evolved littIe with 80
64
respect to the others, even with respect to sector III
40
(“EME”) for which the average increase of deflections
m
is only attributable to au under-sized zone of a fav 0
meters. 0 500 1000 1500 2olM 2500 3000
The results recorded for cracking con&m this Numberof loadings (x 1000)
classification of the structures as the cracks appeared
the same day on sectors I (“BB” B) and II (“BB” A), Figure 4 : deflections evolution
with comparable levels, 45 and 34 % respectively. during the 1st experiment
Under the same state of traftlc, cracks appeared on the
small under-sired zone of sector III, mentioned above,
leading to a level of cracking of 11 % for the “EIvIE”.
No crack was detected on sector IV at the conclusion of Extent of emeking (%)
100
this “cold” phase of the first experiment of the
“Asphalt h4ix Fatigue” program. 80 .,.*.. S II (BB A)
60 --•--SIII(EMEC)
c) The “hot” phase
The loads were resumed after five and a half 40
months break with, of course, the same load
configurations, the same speeds and the same m
transverse prosle sweeps.
The evolution of the average deflections, at 20” C, 0
is shown in Figure 4 which prolongs the values 0 500 1000 1500 2mo 2500 3000
obtained during the “cold’ phase (it is observed that, as Number of loadings (x 1000)
is always the case, when a pavement is no longer
circulated, its deflections tend to reduce and then Figure 5 : extent of cracking
recover their previous level rapidly with the first at the end of the 1st experiment
loads).
With traffic and the worsening of the cracking, the
deflections increase consistently on sectors I (“BB” B)
and II (“BB” A). They increase more rapidly on sector 3.1.4 - Preliminary conclusions for this first
III, where they translate the more marked evolution of experiment. The “BB” A and “BB” B asphalt
the “EME”. However, they are practically stable on the concretes, have shown approximately the same
“GBA” of sector IV. behavior under traftic, aside from the fluctuations in
Cracking progresses markedly over this ‘hot’ thickness.
phase on ail sectors. It increases in a similar manner The high moduhis asphalt mix was very sensitive
on sectors I and II always with a certain shitl in favor to under-thickness and all the more so if it were on a
of sector II (“BB” A) which is less affected. On sector deformable mpponing course.
III (‘EhJE”), the cracking develops later, however, With equal thickness (thick EME), and on the
once it has appeared, it will increase rapidly. This is same supporting course, the life duration of the “EMI?
also the case for sector IV, but with a much greater appears to be approximately 50 % greater than
shift in the time. Figure 5 shows the evolution of the traditional asphalt mixes.
levels of cracking on the four sectors, throughout the Finally, through this test, it can be stated that on a
experiment. deformable supporting course (which does not however
If we take a level of 50 % as our reference, it is correspond to the usual conditions of use of this
noted that in order to reach this level, it is necessary to material), an “EME” of 75 mm thickness was greatly
apply 1,250,OOOloads on sector I (“BB” B), 1,750,OOO outclassed by a “GBA” of 100 mm thickness.
on sectors II (“BB” A) and III (“EME”), and 2,750,OOO
on sector IV (GB A). 3.2 - Experiment 2. The second experiment was
conducted on the same site as the tirst. It comprised
886

1,400,OOOloads between November 1991 and January std~ m Structure II

1992. Naturally, the same load conditions were GB A W70 BB B 5WO
435% 5,4%
applied.
GBA 1oOmm
3.2.1 - Aims of the experiment. The objectives of GRH 2OOmm
this second test were the same as those of the Iirst
experiment in order to contirm the conclusions that
resulted from the first.

3.2.2 - The structures studied. The four

structures are shown in Figure 6. These are the same as
those of the previous test with some changes however :
reduction of the theoretical thickness of the “GBA”,
hm 120 to 100 mm; the BBTM (very thin asphalt
concrete) had delayed the appearance of cracks on the
surface during the first experiment therefore this was
omitted for the second test. The positions of the “BB”
A and “BB” B courses on the one hand and the “GBA”
and “EME” courses on the other were interchanged
from one experiment to the other.
We should also point out that 200 mm thickness of
the base course GRI-I O/20 was cut back and replaced
by a GRI-I O/l 4 of the same origin in order to try to
improve the leveling. BB A 50/‘70
5.4%
The results obtained on the site are good in terms
of compaction and considerably less good in terms of EMEC somm BBA 8Omm
thickness. The actual thickuesses are summariztxl in GRH 26Omm GRH 22Omm
Table 4. It is to be noted that sector I (“BB” A) is
thicker than sector II (“BB” B) by about 15 %, that GRH 24Mn.m GRH 2oOmm
sector III (GBA) is thicker thau planned and above all
sector lV (“EMI?‘) is too thin, by 20 mm on average. Figure 6 : experimental structures
from the 2nd experiment
SI SII SIII SIV
BBA BBB GBA EMEC SI SII SIII SIV
thm o thm o thm o thm D BBA BBB GBA EMEC
MB 93 3 78 3 111 6 56 7
GRH 469 8 463 6 435 14 515 23 d(l/lOOmm) 107 142 109 120
Table 4 : mean thickness and standard ROC (ml 97 60 111 86
deviation of the testing 2 structures (mm) ELONG(Pi‘% 18 125 36 212
E-S (tikf) 99 150 153 264
0 (“C) 10 13 16 16
3.2.3 - The experiment
d : mean deflection at 20°C
a) “As built” state RoC : mean radius of curvature at 2 1“C
As in the previous test, an “as built” state is
carried out for the deflection measurements, radii of Table 5 : “as built” measurements results for
curvature and strain (Table 5). the experiment 2
Loghlly, it was observed that at the beginning of
the test, it was the structures with the thickest asphalt
courses which have the lowest deflections (sectors I -
BB A - and III - GBA). This is also found however Finally, the strain measurements give surprising
less distinctly, with the radius of curvature results, at least concerning the longitudinal strains that
measurements. are surprisingly low at the base of the “BB” A and “GB
A” courses.
 H. ODEON ETAL. 887

b) Running the test

In terms of weather, the Iirst 500,000 loads were
carried out in a period of heavy rainfall. However, the
remainder of the test was carried out in practically total
absence of rain. The average daily temperatures varied
between 0 and 12” C throughout the test. In this 100
ccnmection, there very similar temperatures to the
“cold” phase of the previous test. 50 -+-SIII(GBA)
The evolution of the deflections of the four
pavements are shown in Figure 7. The deflections do 0
not account well for the evohrtion of the wearing 1000 1250 1 500
0 250 500 750
course because they retlect the variable behavior of the Number of loadings (x 1000)
non-treated base courses subject to large water
fluctuations during the test. It is only at the very Figure 7 : deflections evolution
beginning of the test that the four wearing courses can during the 2nd experiment
be diITerent.iated. A sharp increase in the deflections is
observed on sectors II (“BB” B) and IV (“Em”) with
the lint loads : these increases are slower for the other
Extent of crnckiog (%)
materials. 100
The radius of curvature measurements are more
directly linked to the characteristics of the asphalt 80
matsials. For example, the average value of the radii
60
of curvature was divided by two between the initial
measurements and the measurements carried out, at the 40
same temperature, after 440,000 loads.
The evolution of the level of cracking is shown in 20
Figure 8. This is a combination of the trends -.*-- S III (GB A) -..-. S IV(EMFC
0
mentioned above with wet and dry periods and the
dilTerences in hehavior of the four sectors. Sectors II
0 200 400 Number
600 o%edi$:;x 10’;;
(“BB” B) and IV (“Eh4E”) have cracked very rapidly
reaching approximately 90 % level at 400,000 loads.
Sectors I (“BB” A) and III (“GBA”) reach 30 % and 50 Figure 8 : extent of cracking
% level respectively, with the same tra&. Then, with at the end of the 2nd experiment
the period of dryness, the cracking stabilizes up to
900,000 loads and then resumes at the end of the test,
on the less cracked sectors, owing to a lowering of the
temperatures and rain. 100 % level is therefore reached
on the “EME” sector (state already reached afler
900,000 loads), 95 % for the “GBA” and the “BB” B
and 85 % for the “BB”A. Photo 2 shows the state of
sector IV (“EME”) at de conclusion of this second
experiment.

3.2.4 - Preliminary conclusions for this second

experiment. The very much superior behavior of the
BBA in comparison with the BBB can be explained by
the fact that the BBA course is clearly thicker than that
of the BBB. This second test does not invalidate the
conclusions of the first. Photo 2 : state of sector IV
Concerning the conclusions about the relative at the end of the 2nd experiment
behavior of “GBA” and the “EME”, the fluctuations in
thickness do not allow a simple conclusion to be made.
In qualitative terms, there are no contradictions, 3.3 - Experiment 3. This last test was conducted on a
however, with the.results of the first experiment. site situated a hundred meters from the site having
served as a supporting course for the previous
experiments. The subgrade of this site is similar to that
of the previous site. The conditions of application of the rotations. These works consisted of clearing the
the loads are identical to those of the previous tests impaired zones and replacing them with maintenance
(with the exception of the mean radius of rotation, courses.
which was reduced from 19 to 18).
3.2 million loads were applied behveen November
1993 and April 1994.
st~~ture m structureII
EME D lo/20 BBS5MO
3.3.1 - Aims of the experiment. The main 42% 5,4%
objective of this last test was to complete the
observations necessary for specifying the design
methods of the “EME” mixes (high modulus asphalt
mixes). The second objective consisted of assessing the
advantages of using modified binders, compared with
using pure asphalt mixes, with regard to fatigue
resistance of asphalt materials.

3.3.2 - The structures studied. These structures

are outlined in Figure 9. Again, there are four sectors
of equal length with the reference material (“BB” B)
on sector I. Sector II is covered witb a course of
material with moditied binder comprising “SBS” ; this
material will be subsequently referred to as BB S. The
two other sectors involve “EME”, with the binder this
time being a hard asphalt 10/20, of Werent origin to
that used in experiments 1 and 2 ; this new EME will
be denoted Eh4E D. Sectors I, II and III have a
theoretical thiclmess of 80 mm whereas sector IV has a
theoretical thickness of 100 mm.
The compaction values obtained on the site are in
accordance with the prescriptions. Concerning the EMI? D IQ’20 BB B So/70

mean thicknesses, this time they are close to the 6,296 I I 5.4%

theoretical values, as 87, 84 and 91 mm were obtained

respectively on the first sectors with standard
deviations of the order of 5 mm. However, on sector III
(thin EME), it was noted that a zone of 7 m in length
had thicknesses lower by 5 to 10 mm -pared with
the rest of the sector. For sector IV, the thickness was Figure 9 : experimental structures
intentionally increased by 15 mm to have a sign&ant from the 3rd experiment
difference with the previous sector. 115 mm average
thickness was effectively obtained with standard
deviation of 4.3 mm.
SI SII SIII slv
3.3.3 - Running tbe experiment BBB BBS EMEC CBA
a) “As built” state d(l/lOOmm) 108 110 78 68
The deflections, the radii of curvahue and the Rd: W 175 150 260 510
strains are shown in Table 6. On sectors I (“BB” B)
EJBNG (@o 100 102 72 61
and II (BBS), the same order of values is obtained
more or less. The measurements indicate better s-s ww 107 108 61 53
characteristics on sector III (thin “EME”) and, all the 0 (“(2 11 7.5 8 11
more so, on sector IV (thick “EME”).
d : mean deflection at 2O’C
b) Running the test RoC : mean radios of curvature at 2 1“C
There were several phases for this test, phases Table 6 : “as built” measurements results for
corresponding to works carried out on the damaged the experiment 3
sectors, aimed at ensuring the normal contin~tion of
 H. ODfiON ET AL. 889

Thus, sector I (“BB” B), entirely cracked and rest of sector III (thin “EME”) of 91 mm average
showing loss of material, was removed at one million thichess and thick “EME” of 115 mm thickuess, it is
loads and replaced by new materials. The same was noted that in order to reach a 50 % level of cracking, it
applied tn sectors II (BBS) and III (thin “EME”) at 2.2 is necessary to apply 2.5 limes as many loads on the
million loads. Only sector IV (thick “EME”) withstood thick sector as on the thin sector.
the 3.2 million passages relatively well.
During the first phase, from the beginning to I
Deflection (11100 mm)
million loads, there was heavy rainfd and a 180
signitlcant increase in the ground water level. From 1 160
million to 2.2 miIlion loads, the rainfhll was less. 140
There was little rair&dl at the end of the test which 120
meant aajficia.Ily watering sector IV (thick “EME”) 100
during the last 400,000 loads, in order to speed up the 80
evolution of the degradation. 60
40
In terms of temperatures, following a relatively 20
mild winter, there were no large seasonal variations, I -.A-. S IV(EMEDj
between November and April. There were, however,
0 500 1000 1500 2000 2500 3000 3500
large daily fluctoations throughout the test. To Numberof loadings (x 1000)
summatize we can say that 100 % of the loads of the
Crst phase were applied at temperatures between 3 and Figure 10 : deflections evolution
13” C ; 90 % of the remaining loads were applied
during the 3rd experiment
between 5 and 20°C.
The evolution of the average deflections, at a
temperature of 2O”C, is shown in Figure 10. The rapid
evolution of the “BB” B of sector I (+ 70 %) is Extent of cIacking (?h)
100
observed, an evolution which displays very well the
degradation of this pavement which will be total at one 80
miIlion loads. The deflections on sectors II (BBS) and 60 - l -mSlII(EMED)
III (thin “EME”) increase more slowly with the trafIic,
however the increase is large as the deflections 40
increase hm 50 to 65 % respectively after 2 million
loads. Finally, sector IV (thick “EME”) only sees its 20
average deflection increase by 35 % during the Grst
two million loads ; at 3.2 million loads the increase in 0
deflections of this sector will reach 75 %. 0 500 1000 1500 2000 2500 3000 3500
The evolution of radii of curvature is certainly Numberof loadings (I 1000)
going to take account of the initial evolution of the
pavements, since at 700,000 loads, they are divided by Figure 11 : extent of cracking
4, 2.5, 2 and 1.3 respectively on the sectors I, II, HI at the end of the 3rd experiment
and IV.
The levels of cracking of the four pavements,
expressed in relation to the traffic, are shown in Figure 3.3.4 - Preliminary conclusions for the 3rd
Il. These levels corroborate the various measurements experiment. These conclmions are relatively
and observations mentioned above. On the first three straightfonvard
sectors, the evolutions are sharp, but with discrepancies l On a deformable supporting course of 2 to 3
between them. The first cracks observed on sector III mm deflection, the use of a modified asphalt, BB S
(thin “EME”) appeared in a zone of a few meters type, increases the life duration very slightly compared
where the thickness varied around 85 mm. If one with an asphalt concrete with pure asphalt, all things
compares this small zone of sector III (thin “EIvW) of beingequal.
85 mm to sectors I (“BB” B) and II (BBS) of l On a deformable supporting layer, an “EME”
equivalent thickness, it is noted that in order to reach 10120 (high modulus asphalt mix) does not provided a
the same state of degadation on these three sectors, by significant improvement compared with an asphalt
choosing the traf& of sector I (“BB” B) as reference, it concrete 50/70, for courses of thickness of the order of
is necessary to apply traffics 45 % greater on the zone 80 mm. It appears that these thin EME courses display
of 85 mm thickness of sector III (thin “EME”) and 75 weak behavior.
% greater on sector II (BBS). If one now looks at the
890

l The life duration of an “EME” structure l the BB S asphalt is not very structured.
increases rapidly, and hence its economic value, with
its thickness. By increasing from 90 to 110 mm of 4.2 - Asphalt mixes. The tests were carried out on
EME multiplies the life duration by 2.5. samples taken firorn plates of materials removed tiom
the site, Tom non-circulated zones of the test track at
3.5 - Summary of the experimental results of the test the end of the experiment.
track. Experiments 1 and 2 have clearly shown that
the two asphalt concretes “BB”A and “BB”B, 4.2.1 - Complex modulus. The complex modulus
developed from asphalts of the same grade but dif&rent tests were conducted according to the standard NFP
origins, with very different fatigue characteristics 98-260-2 ; some characteristic values are shown in
measured in the laboratory, have similar behavior on Table 8.
pavement. If one compares experiments 1 and 2, a hardening
All of the experiments in the asphalt mix fatigue of the different materials in experiment is noted in
program have hi&lighted the rigid behavior of the comparison to those fabricated for experiment 1.
high modulus asphalt mixes, with the advantages aud However, the relative classification of the various
disadvantages that entailS. Concerning the materials remains the same.
disadvantages, this material does not withstand
excessive strains caused by an under-sizing or a 4.2.2 - Fatigue tests. Different fatigue tests were
supporting course that is too deformable. As the first canied out : with controlled displacement or force,
test showed, it is clearly preferable to use a more with and without rest periods, on trapezoidal or
classical asphalt road base material technique and with prismatic samples. These di&rent procedures, carried
a greater thickness. The rapid evolution of the out by different laboratories, are summarized in Table
degradation once the first cracks appear is also a major 9 (next page).
disadvantage. Concerning the advantages, if this
material is suitably designed, it offers a structural
contribution and resistance to rut development that is
distinctly superior to that obtained with classical 50/70A 50/70B SOl7OS 10/2OC lOl20D
asphalt techniques of the same thickness. 12 123 3 12 3
The third experiment has allowed the contribution
bef. aft. bef. aft. bef. a&. bef. aft. bef. aft.
of using asphalt materials with modified binders (SBS Pene 65 42 61 37 59 39 16 13 20 14
type.) to be verified. The fatigue behavior is slightly TBA 48 53 51 59 60 65.5 69.5 75 63 69.5
superior to that of a pure asphalt material, all things lppfeif -1.1 -0.9 -0.5 0.1 - - 0.3 0.8 -0.3 0.1
being equal.
Pene : penetration at 25°C (MOO mm) ; TBA : Ring
4 - LABORATORY TESTS
and Ball temperature (“C) ; IP Pfeif: Pfeifer
This article only presents the principal penetration index
characteristics of the materials tested. For more detail, Table 7 : characteristics of the used binders
refer to the article by de La Roche and Rivi&re (1997). before and after RTFOT
4.1 - Asphalt binders. The asphalt mixes have been
the subject of ordinary laboratory tests (penetrability, SI SD SIII SIV
ring and ball temperature, complex modulus test). The
main characteristics obtained on the binders of origin, Experiment 1 BBB BBA EMEC GBA
before and atIer RTFOT, are summarized in Table 7. 1o”c-1oHz 10300 12400 15000 13500
The tests carried out show that : 20% 1OHZ 5 100 5 900 9 800 7 100
l the asphalts used are conformed to French Experiment 2 BBA BBB GBA EMEC
staudards ; loOc-loHz 16100 12900 17500 18300
l asphalt A is more kinetically aud thermally 20%IOHz 7 800 5 800 9100 11600
susceptible than asphalt B ;
l asphalt B has thermal susceptibility which Experiment 3 BBB BBS EME D
increases over the course of the experiments (which lo”c-1oHz 12 400 11 300 18 800
reduces the difference with asphalt A) ; it is always less 20%1 OHZ 6 100 5 300 12 300
structured than asphalt A ; Table 8 : complex modulus values (MPa),
l the hard asphalts are of very similar kinetic and at 1 OHz, for 10°C and 20°C
thermal susceptibility ;
 H. ODBON ET AL. 891

Procedure 0 (“C) f VW Type of loading Phase Laboratory

---
Controlled displacement test

1 LCPC
3 levels of 8 samples

3 20 40 23 LCPC
3 levels of 8 samples 23 LPC Bordeaux
or 12 samples 2 Shell SRSA

4 20 40 a- 1:5
6 samples
2 Shell SRSA

5 20 40 -%- 1:lO
3 levels of 8 samples 3 LPC Bordeaux
or 6 samples 2 Shell SRSA

Controlled force test

20 40
3 levels of 8 samples 23 LPC Bordeaux
or 12 samples 2 Shell SRSA

20 25 Wh 1 LPC Bordeaux
3 levels of 8 samples

8 20 40 LPC Bordeaux
3 levels of 8 samples

9 20 40 /!I3 Q +1:5 2 Shell SRSA

6 samples

10 20 40 -%- 1:lO
3 levels of 8 samples 3 LPC Bordeaux
or 6 samples 2 Shell SRSA

11 20 40 1 Shell KSLA
6 to 8 samples

12 Shell KSLA

Table 9 : laboratory fatigue test procedures

892

4.2.3 - Comments. For the asphalt mixes, the theoretical life duration of the pavement, by using a
following is observed : relation born directly from the fatigue law of the
l the values of modulus measured by the different material in the laboratory ; in addition to de fatigue
laboratories are consistent, on the one hand between law, this relationship takes into accost the long term
materials when one successively compares the results bearing capacity of the gromd, the probability of
for each laboratory, on the other hand for the same failure of the pavement and comprises the shift factor.
material from measurements taken in the d%erent The approach adopted in interpreting these
laboratories ; the increase of the moduli during experiments is based on this logic. Firstly, using values
experiment 2 is due to a higher compaction of the of deflection and radius of curvature measured on the
materials sampled in place. surface, the equivalent structure is defined which
l the classification of materials according to their allows the calculation of similar strains to those
fatigue test characteristics depends on the test used, measured within the structure to be obtained. It is then
depending on whether one works with controlled possible to calculate the loading (stresses and/or
displacement or force, with or without rest periods. strains) which combined with the laboratory fatigue
. the value of 06 or ~6 increases for materials test results allow the theoretical service life of the
with the introduction of rest periods. structures to be determined according to the following
relationship :
5 - MODELING
lib
The work reported in this article follows that
presented at the 73rd TRE? by de La Roche and al
(1994). In this publication, the results were quoted for
the first experiment conducted on this subject. where ~4 : shin calculated in the structure
In this part, we first of all discuss the approach equivalent to circuit of one load ;
used in interpreting the results, as well as the principal ~$3) : straiu causing the failure of the sample
results obtained from interpreting experiment 1. Next, after lo6 applications of the loading
the results of interpreting experiments 2 and 3 will be stress/strain at a temperature of 0 “C ;
presented, and l?nally the summary of information b : slope of the fatigue curve ;
obtained after these three experiments. N : theoretical life duration of the structure ;
k: shift factor translating the difference
5.1 - Modeling of experiment No. 1 between theoretical method and
observations on actual pavement.
51.1 - Principle. Interpretation is based on the
French method of pavement design, described in the In the following, the shitt factor k of the method is
technical Guide ‘Conception et dimensiomrement des determined which allows the observed service life to be
stmctores de chaussees’ (“Structural Design of Road found on the test track ; a shitl factor of 1 translates an
Pavements”) (1994). It concerns a rational method interpretation that describes reality exactly; an
based on the following principle. identical shift factor for all the materials would
Stresses and strains induced by an axle in the indicate that the fatigue test adopted is representative
pavement are calculated at its initial stage. The effect of the relative behavior of the structures.
of the tragic and of its repeated loads is taken into
account with the help of the fatigue law, based on a 51.2 - Principal results. In interpreting
constant stress or strain level test. The pavement is experiment 1, sectors 3 and 4 were divided into hvo
assumed broken for a certain extent of cracking zones to take into account differences in thickness
depending on its service level. A shitl factor, based on (Table 10). The stresses and strains have been
observed performance, is introduced to predict the life calculated within the structure with circular imprint,
duration of the pavement. The method does not take conforming to those adopted in the French standard
into account the evolution of the bearing capacity of method, and with rectangular shaped imprint, closer to
the pavement (estimated by deflection for instance). real imprint and allowing a better approximation to the
To be more precise, the loading stress (strains strains measured. The principal results of the
and/or stresses) is calculated within the pavement calculation are listed in Table 10.
structure with the help of the elastic Burmister multi- From this analysis, it arises that :
layer elastic model ; the pavement is modeled by an l adopting rectangular imprint for the calculation
equivalent structure ; the load is approximated by improves the modeling (reduction of the difference
circular shaped imprints. The loading (strain a&or between measured and calculated strains) ;
stress) thus obtained is then used to determine the
 H. ODfiON ET AL. 893

SI SH sm SIV l in the framework of the French approach for

BBB BBA max ’ design of new pavement structures (procedure 1
EMJX GE standard fatigue test), the shit? factor to be adopted for
the“EME”is 1.
Thickness (m)
Mat. bitx. 0.081 0.080 0.089 0.101 5.2 - Interpretation of experiment No. 2
GRH2 0.218 0.210 0.211 0.225
GRHl 0.218 0.210 0.211 0.225 5.2.1 - Determination of equivalent structures.
Experiment 2 aimed to contirm the results of
Number of cycles on the circular test track experilnent 1 ; its structures are similar to those of
(50% cracking) experiment 1 (Q 3.2)
(x1000) 1 100 1450 2000 2700 The Werences involve the upper course of GRH
which was replaced (because it was impaired by
Measured strain min/max (pdef) dismantIing experiment 1) before installing new
&lone MB 136 92 75 78 asphalt courses, as well as implanting different
%3.ns MB 124 89 41 87 materials which has been modified in order to remove
%wtGRH - 935 - - any possible disparities in bearing capacity of the
%ert sol 632 554 410 691 supporting layer (cf 3.2.2).
The determination of the moduli of sand and GRH
is achieved by inverse calculation from deflection and
Calculated E et ts at 15T - 1OHz
radius of curvature measurements (Table 11). A model
Et m-ix circ. 235 222 179 177 of 30 MPa is thus adopted for the sand, 90 MPa for the
rect 172 160 137 136 GRH and 150 MPa for the GRH2 (for the sector IV, a
otmax circ. 2.18 2.46 2.72 2.22 third course of GRH of low thickness (50 mm) is
rect 1.69 1.84 2.18 1.79 introduced, corresponding to the renewal of the support
on this sector which had to be rebuilt, for sector IV, the
Characteristics for calculation GRH3 has a module of 400 MPa). The values of the
~E*(15°C,10Hz)~ 7 700 9 125 12400 10300 sand moduli and the GRH adopted are lower than for
Proc.1 q 140 90 140 88 experiment 1, due to the high rainfhll between the
-l/b 6.3 5.2 5.3 4.3 experiments. The adjustment obtained is correct for the
Proc.7 06 0.67 0.52 1.38 0.67 deflection measurements but less good for the radius of
-lib 5.5 4.4 6.2 7.0 curvature measurements on structures II and III : the
Proc.11 Oh 1.19 1.13 1.91 - values calculated are respectively 16 and 70 % greater
-l/b 6.1 6.2 5.9 - than the values measured. These differences are due to
Proc.12 06 1.68 1.70 2.20 - differences in the stiffness of the four structures and
-l/b 6.9 9.0 3.4 - the nonlinear nature of the support which is not
represented in the model.
Back-calculated shift factor k
Proc. 1 circ. imp. 1.5 2.3 1.3 2.2 5.2.2 - Mechanical analysis
reeLimp 1.1 1.6 1.0 1.7
Proc.7 circ. imp. 2.7 4.1 2.0 3.2 a) Analysis of strains
rectimp 2.1 3.1 1.6 2.6 In order to validate the choice of the equivalent
Pro.lOcirc. imp. 1.5 1.8 1.4 - structures measured on the experimental pavements
rect. imp. 1.2 1.4 1.2 - and the values of strains calculated in identical
Pro.11 c&imp. 1.1 1.2 1.3 - conditions of temperature and frequency, for circular
rect. imp. 0.8 0.9 1.1 - imprint - r = 0.12 m usual case - and rectangular -
0.18 m x 0.30 m, shape much closer to real imprint
Table 10 : characteristics for calculations of (Table 11). The reduced mnnber of measurements and
modelling and results - Experiment 1 their disparity gives reason for caution : the
comparison is only usefhl for information purposes.
However, it is to be noted that the strain calculated
l adopting fatigue results obtained from from a rectangular imprint are lower than those
controlled force tests, with or without rest period calculated with a circular imprint, and that the
allows theoretical service life to be calculated that are longitudinal strains are greater than the transversal
comparable to those observed on the fatigue test track strains : this is corroborated by the appearance of
(absolute and relative) ; transversal cracks on the experimental pavements.
894

SI SII SIII SIV calculated from equivalent structures with the help of
BBA BBB GBA EMEC an elastic model @mister for the circnlar imprints,
C&r-LCPC for the rectangular imprints) at 15’ C and
Thickness (m)
10 Hz. The temperature of 15” C is the equivalent
Mat. bitx. 0.093 0.078 0.108 0.058 temperature adopted in France for the calculation of
GRH3 - - - 0.050 design ; 10 Hz corresponds to the representative
GRH2 0.238 0.239 0.216 0.206 frequency of load passage. Calculations carried out for
GRHI 0.23 I 0.225 0.218 0.250 a temperature of 20°C and a frequency of 40 Hz
Number of cycles on the circular test track (identical to the conditions of several of the tests) have
(50% cracking) led to similar results (the moduli of the asphalt
(x1000) 1 000 315 375 360 materials being close to those at 15°C - 10 Hz) : the
results thus obtained will not be stated.
Measured strain min/max (pdef) It is to be noted that the vertical strains at the
&lone MEI 19 55/191 17/340 81/244 ground surface and the GRB surface are not
%ms MB 91/l 11 lW149 153 264 determining factors ; the life duration is estimated Born
Evert GRH - 948 1144 - the maximum of the longitndinal strain and the
Evert sol 755 - 147 - transversal strain (ditto for an analysis of stress). The
shift factor is determined from the number of passages
Calculated E et o at 15OC - 1OHz having led to 50 % cracked pavement (which is
ct max circ. 218 291 181 243 associated with 50 % risk of failure).
rect. 182 230 155 186 The loading stresses and strains calculated within
cstmax circ. 3.633 3.733 3.350 4.998 equivalent structures are also combined with the
rect 3.103 3.094 2.915 4.058 different test results available, with controlled
displacement or controlled force (Table 11). All the
Characteristics for calculation results are not reported here, neighboring laboratory
~E*(15°C,10Hz)~ 12000 9 300 13 300 14 900 results providing similar shitt factors.
Proc.1 &(Ti 90 122 82 122
-l/b 5.2 5.3 7.3 5.9 c) Conclusions
Proc.3 &A 95 134 87 130 Concerning the controlled displacement tests, it is
-l/b 4.6 6.7 4.3 5.9 noted that :
Proc.4 Eh 131 160 103 127 l de shift factor determined for the “BB” A is
-l/b 5.1 4.2 5.9 3.6 greater than that of “BB’B by 20 % approximately ;
Proc.6 crh 0.76 0.65 0.79 1.22 l the shift factor associated with the “EME” C
-l/b 5.2 6.0 4.8 5.6 remains much lower than those of the “BB” asphalt
Proc.8 06 1.15 1.01 1.09 1.62 concretes ;
-l/b 5.8 6.4 5.6 7.0 l that of the “GB” is also lower than those of the
“BB”‘s : this does not conform with the usual situation,
Back-calculated shift factor k
but no rational explanation could be offered ;
Proc. 1 circ. imp. 2.09 1.63 1.68 1.51 l the introduction of rest periods in the tests
rect. imp. 1.74 1.29 1.44 1.16 improves the estimation of life duration in absolute
Proc.3 circ. imp. 2.84 2.33 2.00 1.78 terms (reduction of the value of the shift factor) but
rect. imp. 2.37 1.84 1.71 1.36 does not signiticantly refine things in relative terms.
Proc.4 circ. imp. 2.06 1.76 1.79 1.63 These results contirm the observations made
rect. imp. 1.72 1.39 1.54 1.25 during experiment 1.
Proc.6 circ. imp. 3.87 3.72 2.87 3.01 For the stress analysis, it is observed that :
rect. imp. 3.30 3.09 2.49 2.44 l the shift factors of the two “BB” samples are
Proc.8 circ. imp. 2.55 2.43 2.14 2.35 close, irrespective of the test procedure ; again we find
rect. imp. 2.18 2.01 1.86 1.91 one of the conclusions of experiment 1.
Table 11 : characteristics for calculations of l the shi8 factors of the “GB” and the “EME” C
modelling and results - Experiment 2 are close, but lower than for the “BB” ’ s ;
l the introduction of rest periods causes a
reduction of the shitt factors of all the materials, which
b) Service life is positive, without bringing the values close together
The analysis approach adopted is the one (which would be indicative of a representative test of
described previously. The stresses and strains are real behavior).
 H. ODeON ET AL. 895

The conchrsions resulting from experiment 2 are SI SII SIII SIV

similar to those obtained from the interpretation of BBB BBS EMED EMED
experiment 1.
Thickness (m)
Mat. bitx. 0.086 0.084 0.091 0.111
5.3 - Interpretation of experiment No. 3
GF3-E 0.230 0.188 0.197 0.196
5.3.1 - Determination of equivalent structures. GRHI 0.200 0.200 0.200 0.200
The geometry of the structures is similar to those of the Number of cycles on the circular test track
structures of experiments 1 and 2, apart horn the (50% crack&)
subgrade (Q 3.3) ; the asphalt materials are supported (x1000) 600 1 100 1 250 3 500
on two successive courses of GRH (0.40 m) covering a
clay-sand equally as poor as that of experiments 1 and Measured strain min/max (Cldef)
2 (1.7 m). e1oneMB 84/106 87/115 63/73 51171
As in the first two experiments, determination of %ans MB 53158 64 35140 52
the moduli of the courses not linked is achieved by %A GRH 450/856 - - -
inverse calculation horn deflection measurements, Evert sol 5481978 _ _ _
carried out at different phases of the construction of the
pavements, and measurements of radii of curvature at Calculated E et o at 1YC - 1OHz
the surface. The modulus of the sand retained is 25 Et maX CirC. 210 238 129 112
MPa. As for the GRH courses, it is necessary to rect. 162 183 102 92
distingnish structures I and II for which a moduhts of at max circ. 2.659 2.613 2.164 2.424
125 MPa is retained for the lower GRH course and 300 rect 2.169 2.122 2.289 2.059
MPa for the upper course, and the structures III and IV Characteristics for calculation
- more rigid - for which 300 and 450 MPa respectively. (E*(15°C,10Hz)~ 9 300 8 100 15 700 15 700
It has not been possible to find a satisfactory unique
solution for linear elasticity. Proc.1 a6 138 175 141 141
-lib 5.8 6.1 6.1 6.1
5.3.2 - Mechanical analysis Proc.3 Eh 147 237 133 133
-l/b 5.3 5.5 5.3 5.3
a) Analysis of the strains hoc.5 Eh 227 292 187 187
The strains measured were compared with the -lib 6.2 8.9 7.4 7.4
strains calculated with the aid of equivalent structures, Proc.6 oh 0.62 0.82 - -
.”
for circular and rectangular imprints. A much better -I/D 5.5 8.8 - -
correlation is noted for the values (the calculated Proc.10 06 1.18 1.41 2.3 2.3
strains correspond to de average plus the standard -lib
7.9 9.2 8.8 8.8
deviation, approldmately, of the measured values, at
the temperature and frequency of measurement). The Back-calculated shift factor k
strains calculated due to the rectangular imprint are the Proc. 1 circ. imp. 1.21 1.18 0.86 0.89
closest to the values measured ; the longitudinal strains rect. imp. 0.93 0.91 0.68 0.73
are still greater than the transversal strains, which is Roc.3 circ. imp. 1.58 1.28 1.14 1.20
cmfixmed by the transversal cracks observed on the rect. imp. 1.22 0.99 0.90 0.98
pavements. Proc.5 circ. imp. 1.04 1.03 0.80 0.80
rect. imp. 0.80 0.80 0.63 0.65
b) Servtce life Proc.6 circ. imp. 3.21 2.57 - -
As in the previous cases, the life duration of the rect. imp. 2.62 2.08 - -
strnctures was calculated from the loading stresses Pro. 10 circ. imp. 1.73 1.49 1.10 1.08
(stresses or strains) calculated within the equivalent rect. imp. 1.41 1.21 0.91 0.92
structures and the results of fatigue tests. The loading
calculations were made for the following conditions :
Table 12 : characteristics for calculations of
15” C / 10 Hz and 20°C / 40 Hz, for circular imprints
(usual method) and the rectangular imprints. The modelling and results - Experiment 3
results are expressed in the form of shift factor
allowing the number of passages of axles leading to a
50 % level of cracking to be found on the experimental The vertical strains at the surface of the natural
structures of the test track. ground and the GRH courses indicate that the
pavement has slightly subsided under the effect of the
896

traffic. However, this rut development of the The calculations of Mode Factor were carried out
supporting layer which is limited, has probably for each pavement from the equivalent structure, for a
disrupted the fatigue mechanism very little. We shall modulus of asphalt material at 15°C - 10 Hz, for which
assume that it does not intervene in the following a reduction of modulus of 50 % has been considered
analysis. (Table 13).
The analysis was conducted for the test results
obtained with controlled displacement and for those
obtaiued with controlled force. SI sn SIII SIV
Phase 1
c) Conclusions BBB BBA EMEC GBA
Irrespective of the type of test, it is noted that : O.d8 0.03 -O.Ol/-0.11 -0.13/-0.21
l the use of rectangular imprint leads to a lesser
dispersion of the values of the shi8 factor. Phase 2
l the “EME” had behavior consistent with its BBA BBB GBA EMEC
thickness: the shift factor is identical for sectors Ill -0.27 -0.13 -0.34 0.12
bhin “EME’Y and IV (thick “EME”) , ,: Phase 3
l the &ft factor to be retained for the “EME” is BBB BBS EMED EMED
always less than that of the asphalt concretes. 0.09 0.16 0.05 -0.06
Concerning the tests with controlled displacement,
it is observed that : Table 13 : Mode Factor values
. the shift factors obtained for the two asphalt
concretes “BB” B and “BB” S are identical, both for
the reference test at 10” C - 25 Hz and for the test at It is noted that the structures studied have a Mode
2O”C-4OHzwithrestperiod(l: 10); Factor close to 0, which explains that one can not rule
l estimation of the life duration of the structures precisely on their mode of ftmctioniug. It concerns
is improved by the use of the characteristics resultiug structures for low to medium trafEc; one might
from the tests with rest periods, in absolute terms ; but reasonably think that thicker structures would behave
the modeling is not significantly better iu relative iu a clearer fashion for the controlled stress mode.
terms; The Mode Factor does not therefore depend on the
Concemiug the tests with controlled force, it single surface asphalt course (example : phase 3,
results that : structure II having 9 1 mm of “EME”, functioning with
l these tests do not contribute a great improve- controlled displacement) : it is the rigidity of the
ment in terms of the choice of shift factors of the di@- structure overall (surface course and supporting layer)
rent materials (neither in terms of the estimation of which determines the type of functioning. This
real life duration of the structure - i.e. shi8 factor close mechanism can be visuahzed easily by plotting the
to 1 - nor in terms of the relative behavior of the mate- evolution of Mode Factor versus the rigidity of the
rials - identical shift factors for the d&rent materials); structnre (fig. 12) :
l the results obtained from the tests with rest
periods are better in absolute terms, but not in relative E asphalt 3
concrete X Hasoh& concrete
terms. St =
EGRH x r3
5.4 - Interpretation by using the Mode Factor. These
analyses were carried out experiment by experiment MF
and reveal a disparate ftmctioning of some structures 0.40
with respect to others. In order to clarify these apparent
0.20
differences in behavior between structures, the notion
of Mode Factor, introduced by Monismith (1971), is
0.00
used. The Mode Factor MF is obtained by calculating:
MF=(A-B/(A+B) -0.20
where A and B are the relative variations of
stresses and strains calculated within a structure for a -0.40
given reduction of the mod&s of the asphalt course. A 1.00 10.00 lco.00
value of MF equal to 1 (and by extension, positive) st VahIc
corresponds to behavior with controlled strain, MF = -
1 (or negative) corresponds to behavior with controlled Figure 12 : relationship
stress. between “Stiffness” and Mode Factor
 H. ODEON ET AL. 891

5.5 - Summary of the modeling. In conclusion of the layer model by Burmister and associated with the
interpretation of these three experiments, it can be controlled displacement fatigue test at lO”C, 25Hz :
stated that : 1. the shill factor to be retained for the high
l for these structures that are not very thick, the modulus asphalt mixes (“EME”) is 1 (as a reminder,
modeling is improved if one adopts the same that of the asphalt concretes is 1.3) : this value has
rectangular shaped imprints, close to the real imprints been conl%rmed by a study of actual pavements already
of tires. The order of magnitude of calculated loading built with these materials ;
stresses and strains and the ratios between longitudinal 2. the moditied asphalt concrete “BB” S confIrms
and transversal loading stresses are impoved. its good behavior on the road : at equivalent thickness,
. if one considers the usual French method of it has had a life duration practically double that of the
design which combines a linear elastic modeling and reference asphalt concrete.
the controlled strain test conducted at IO”C, 25Hz : l concerning the results from the modeling car-
a. the relative behavior of the asphalt concretes ried out from other fatigue tests, it is to be noted that :
“BB” B, “BB s” as well as the “GB” A (from the tirst 3. the controlled force tests improve the modeling
experiment) validates the present method ; of the relative behavior of the asphalts BB A and BB B,
b. the shift factor associated with “EME” and the more rigid materials GB A and EME ; estima-
structures is always less than that adopted for the tion of the life duration is less good in absolute terms ;
asphalt concretes ; 4. the introduction of rest periods improves the
c. it is not possible to explain with the aid of this modehng a little : the increase observed in the value of
usual modehng approach the di%rence of behavior E6(or of 06) allows the estimation of life duration to be
with the fatigue test observed between “BB” B and improved (reduction of shift factor), but it is not
“BB” A, a ditTerence which does not appear on real enough to explain the differences in behavior of the
pavements ; “soft” materials (asphalt concrete type) and rigid
. if one considers other fatigue tests carried out, materials (EME and GB A type) ;
with controlled displacement or force, with or without 5. Consequently, there are no obvious reasons in
rest periods, it is noted that : the short term to question the fatigue test currently
d. the controlled force tests allow the relative used in France (controlled displacement, lo”, 25 Hz).
behavior of the BB A and BB B during the first two . concerning the modeling, it is noted that the for
experiments to be found by modeling, as well as the structures with low rigidity such as those studied, the
relative behavior of the rigid materials EME and use of a rectangular impriut, close to that of a tire,
GBA ; the shift factors to be adopted remain different greatly improves the estimation of loading stresses in
however; these tests do not allow the duration of life of the upper courses of the pavement.
the BB S to be estimated correctly.
e. the tests with rest period, conducted for strain REFERENCES
and stress, improve the estimation of service life, bnt
do not allow the relative behavior of the difherent Au&et P. et al., “The Circular Test Track of the
materials to be explained either. Laboratoire Central des Ponts et Chaussees,
Names - First Results”, Proc. of the Sixth Znt.
6 - CONCLUSIONS Co@ on Asphalt Pavements, pp. 550-561, 1987
De La Roche et al., “Study of the Fatigue of Asphalt
The Asphalt Mix Fatigue experiments, conducted Mixes Using the Circular Test Track of the
between 1990 and 1994 on the LCPC circular fatigue Laboratoire Central des Ponts et C&au&es in
test track, consisted of testing, during three distinct Nantes, France”, Proc. ofthe 73rd TRB, no 1436,
experiments, twelve structures made of six different 1994.
materials : three asphalt concretes, two high-modulus De La Roche et al., “Fatigue Behavior of Asphalt
asphalt mixes and one conventional asphalt road base Mixes - Influence of test procedures on Laboratory
material. This program was accompanied by a long Fatigue Performances”, Proc. of the Eighth Int.
series of tests on asphalt materials, designed so as to Co@ on Asphalt Pavement, 1997
assess their behavior towards fatigue according to dif- Monismith C.L. and Salam Y.M., “Distress Characte-
ferent test procedures ; the results of this set of tests are ristics of Asphalt Concrete Mixtures”, Proc.
cited in the article by de La Roche and Riviere (1997). Association of Asphalt Paving Technologtits, Vol
Analysis of the behavior of the structures has 42, 1973
showu the following points : Guide technique Conception et dimensiomrement des
l in the framework of the French method of new chaussees (Technical Guide for the Structural
pavement design, based on the linear elastic multi- Design of Pavements), SETRA-LCPC, 1994