Professional Documents
Culture Documents
OF ASPHALT BINDERS
By
Hussain U. Bahia
HuachuD lhai
Karen Bonnetti
Sadi Kose
July 1998
Bahia. lhOl. 8/lnnelli. & K05t 2
Abslnd
Background
1000
,"
;;
900
800 .
•!!!• 700
BInder Type
i-+- LS
600
••
___ 01
• SOO __ 02
••• 400 -,
~" 300
-;O- P2
•
•" 200 -
-+-P
•
;;" 100
0
0 100000 200000 300000 400000 500000 600000
Mixlure Stlffness (KPa)@O.01"1.5trJin
Figure I : Relationsbip between Mixture Stiffness and Binder SliffDesJ for different
Mndified Binden at ODe Binder Strain Level
Binder (LS) is a PG 64-22 unrnooified asphalt. The other binders are PG76-22
grade with different modifications. Binders 01 and 02 are oxidized asphalts while the
others are polymer modified asphalts.
A thorough analysis of the results indicated that one of the impanan\ factors that
were nat carefully considered in the analysis was the Slrllin amplitude during the testing
of the binders and the mixtures. Rt-testing of the binders under controlled conditions to
separate the effects of frequency from strain amplitude indicated that at the temperaturts
the analysis was conducted (20 - 40 C), these modified binders afe highly sensitive to
/k,II/II. lhai. Bmmrlli. & KosI.' ,
strain amplitude. It v..-as observed that the strain amplitude used for the binder testing
pla)'s a significant role in defining the binder effect on mixture rhwlngical behavior (see
Figure 2).
700 r------------------,
~l %Slra)n
...... 4%Slfain
-"" 10% Strain
'"*"'20% Strain
•• 300 ..... 30%Slrain
••
•
;; 200
,"•" 100
• oL-~----------~
o 100000 200000 300000 400000
Mixture Stiffness (KPa)@O.01%Straln
These preliminary analyses showed that the role of the asphalt binder could be
bener defined if the strain dependency of the different binders is considered. It is.
however. necessary to define the actual range of strain amplitude in a binder used in a
typical mixture. Estimating the distribution of binder domains within a mixture is a
complex task. It is expected that the aggregate gradation. aggregate shapes and the
distribution of binder within a mixture have significant effects on the distribution of film
thicknesses. The strain amplitude within the binder is a function oflhe film thickness and
thus will show a distribution that is related to the distribution of film thicknesses. The
concept of estimating an average film thickness based on idealized aggregate shapes is
not factual. Due to the randomness of shapes and sizes of aggregates. it is expected to
find a wide distribution of asphalt film thickness within a typical mixture.
It is known that betausc of the large difference in stiffness between the binder and
the aggregates. most of the bulk strain will be in the binder domains. Because of the non-
uniformity of the binder domains. it is expected thai the distribution of strains is very
complu . II is also expected that the common procedure of estimating strain in binders
Bahia. Zhoi. Bonnel/i. " Kase s
-
o -.003168
. 041))6
.0 8584
There are few important limitations with the imaging techniques used for this
study. It is rather difficult to have an actual representation of the very thin films of binder
that are expected to exist between finer panicles that are smaller than the I-nun range.
Based on this Wldersranding, it is expected that within an actual mixture a wider range of
film thickness will exist and therefore a wider range or strain amplitude could be realized
within the binder domains. It is realistic 10 assume that strains within the binder films can
be as high as 1010100 times of the bulk strains of the 10lal mixture.
6
The current Supcrpave birKIer tesling protocols. developed as part of the SHRP
program, focus on the linear visco-elastic properties (2). There were several reasons (0
select (he linear range. Among them. the most important ones are
(I) Good pavement design should limit the defonnalion of the pavement layers.
(2) II is much simpler and more convenient \0 lest in the linear region because of
the reduction of the number of variables.
While these reasons are valid, the work in this study indicates clearly that this
simplification cannot be accepted, particularly for modified binders. The finding thaI
slrain amplitude within tile binder domains can be much higher than the bulk strain
within a mixture makes il necessary 10 re-evaluate the testing protocols used in the
current binder specification. The hypothesis that non-linear viscoelastic behavior is
necessary to establish bener relationships between birKkr and mixture behavior is a
departure from the existing concepts in Superpave system, which relies on the linear
visco-elastic properties of binders and mixtures to predict performance. There are several
important reasons to consider non-linear behavior:
(I) Asphalt binders that show similar behavior in the linear range can show highly
variable. non-linear behaviors. which will require direct cvaluation of the
properties in the non-linear region.
(2) A consideration of pavement structure in the binder testing is necessary for the
selection orthe binders. Pavements can have weak (nexible) structures, such
as the large number of primary and secondary highways built on Wlstabilized
granular bases and subbases. Pavements can have strong structures, such as
the large number of highways that consist of concrete (rigid) layers with
nexible asphalt overlays. A nexible structure will allow more deformations
and thus high strains by design while 8 strong structure will limit these
deformations 10 small values. If binders are functioning in the nonlinear
region. strain level will be a major factor in defining their behavior.
(3) Non·linearity in many materials is indicative of damage. It is considered as
the stage closer to failure than the linear (safe) range. Due to the repeated
nature of traffic load applications, binders performing in the non·linear region
can accumulate damage, such as fatigue and runing, faster and thus show
performanct very different from whaltheir linear behavior is indicating.
(4) More modified asphalt are being used to accommodate the higher and heavier
traffic. Thc majority of these binders are multi-phase systems that are known
to show non-linearity different from the un-modified binders. which are more
homogenous in theiTcomposition.
(5) Linear behaviors of binders do nOI show good relationships to mixture
behavior orlhe same aggregate structure and volumetrics. Non·linearity might
explain some or the discrepancy.
Bahia, 'fhai. Bonnell/. & Kost 7
strain range that it orrers. The equations required to correct for the heterogeneous flow
have been derived and for moderate levels of strains «400 percent) and it has been
shown to give results comparable to tile cone and plate (6).
In this slIJdy since it is well recognized that there arc advantages aoo
disadvantages 10 almost all geometries, it was decided not 10 change the geometl)' of
parallel plate used in Superpave system and accept the enors resulting from the
heterogeneous flow field as part of the experimental error. A comparison of results from
parallel plate and cone/plate for few asphalts indicated thaI results are very similar and
differences are negligible. A Dynamic Sllear Rheometer made by Bohlin Instruments
(BDR) was used for all the testing. An c)(ample of stress-strain loops that can be
measured for an asphalt binder is shown in Figure 4. The figure shows the loops for a
number of strain levels and shows the irregularity of the loops with increasing strain.
Such data can be used to calculate the Hannonic Distribution Function (HDF) and
determine the effe<:ts of non-linearity on the modulus and phase angle. Although in this
study the analysis of the stress-strain loops were conducted, to simplify presentation of
the imponant concepts in this paper the analysis is presented in terms of the complex
modulus and phase angle calculated from the fundamental harmonic only. The 0 1' and
01 are used as a substitute to the commonly used a" and 0 measured within the linear
region.
30000 I
I
20000 I
10000
-+- Straw! 2%
~ a -.I-I---i- --l __ Strain 25%
! ""'Il-Strain 50%
a. . \0000 t
-20000
-30000
~oooo L _ _ _ _ _ _ __ --'
Figure 4: Torque versus Position loops Bosean PG70·22 ror differtDI5tnlin levels It
30 C with 1 Hz rrequency.
Bahia, lha;' Bonnel/i, & Kose 9
The results are presented in several sections each addressing one of the imponant
factors included in the experiment
Previous research has shown that strain dependency is related to the rigidity of the
material as measured by G" (9). In fact the current version of AASHTO TP5 .• Standard
Procedure for Testing with the Dynamic Shear Rheometer," includes an equation giving
the linear limit of the suain as a function ofG " (AASHTO 93). Since temperature and
frequency both affect the G" ya,lue of asphalt binders it was expected that there is an
interaction between the strain, frequency and temperature.
Binder Composition
The olher important factor 10 consider is the type of binder. Figure 9 includes a
comparison of the strain SWttpS for six Stlected modified binders of the same grade,
produced from the same base asphalt (3), It is clear that strain dependency is also a strong
function of the type of modifier UStd. The data shows that binders modified with some
elastomers, plastomers. and by oxidation show significant strain dependency that can be
BviTiv. ZIT/I!. !lrJnm:1I1. & Kose 10
as high as two orders of magnitude drop in rigidity as a result of changing strain from 2
percent to 50 percent. The drop in G' values can be as low as only 50 percent for other
binders. The trend of change is also important Some binders show a continuous
decrease in G' while others show a plateau region at low strains followed by a sudden
drop and another plateau region at high strains. These trends might be related to the
morphology of the binder and the types of bonds that exist in the binder.
1000 r<C~-------------,
-+-58 C
___ 46C
~ 34 C
.....:. ....,~
'....0011....101
10 ·_··"........ ,"*"" 22 C
-+-lO C
, l·~·~·~'~·~'~'~·~'~'~·~·~·~'~'~.~.~.~.~.~.~,~.~.~.~__J
0% 10% 20% 30% 40% 50"10 60%
S tra in (-Jo)
1000000 , ,
•
ii 100000
U
-----~'I'"
~
'000 C,_ _ _ _ _ _ _ _ _ _ __
0% 10'4 20'4 30'J1, ~O% 50'4 50%
Strain
1000
_._- --
-+- l %Slri ln
-
.•
~ '00
___ 4%Sb'aln
·-.- l O%Stlalfl
b -*-20% Sb'aln
...... 30%Slrain
'0 ----------------------~
0.Q1 010 1.00 to.oo 100_00
Frequency (Hz)
65
64
6J
62
~ "
60 ~ O . 15Hz
-+- 1.6 Hz
59
' ~15Hz
"
57
56
"" StrlII in
900 ----
BOO
700 -e-- Elastomer 1
- 600
~ 500
It
....... Elastomer2
~ Elastomer 3
;- 400 ...... Plaslomer 1
(!) 300
-+- Plastomer 2
200
-+- Oxidized 1
100 ---- ---
a
a 20 40 60
Slrain (%)
1800 ,- - - -- -- - - ---,
~tttttt*
1700
........ t .......
, ..... Elastomer 1
~
1600
....... Elastomer 2
1500
-+-Elastomer 3
• 1400
!!:.
b -+- Plastomer 1
-.... Plastomer 2
1200 ~ (hjdized 1
1100
Strain (% )
For many years healing of asph!llt mixtures have been researched and measured.
The significant works by researchers at the University of Texas A&M (12) and more
recently althe North Carolina State University (13) have shown that rest periods can have
a profound effects on recovery of damage in asphalt mixtures. Such behavior, although
generally recognized in asphalt binders, has not been measured and quantified before. In
this study since large strains resulted in significant reduction in 0 ' and marginal inc~
in 0, it was detided to investigate the effect of rest period on the damage (change) caused
by strain sweeps. Figure 11 is a sample of the effect of rest periods on strain damage for
an unmodified asphalL Figure 12 is another sample of data for a modified binder. The
results show that binders show significant healing with time. The results also show that
healing is a time dependent phenomenon, which depends on the composition of binder.
The recovery of the modified binder, which is produced from the same base asphalt
shown in Figure I I, shows a much rapid recovery than the base asphalt. This indicates
thaI modification tcchniques can alter the healing perfonnance. It also raises interest in
evaluation the effect of asphalt chemistry on healing of damage caused by higher strains.
In recent experiments, it is found that asphalt chemistI)' of UMlodified binders has an
imponanl effect. 111;s has been documenied in research on asphalt mixtures in recent
studies (1 4).
Bulnu. !.hm. Bvnnl'lIi. & Kuse 14
<SO
-4-0 Hr
"0
....... Retest (No Delay)
3"
-+- Retest {12 hrJ
300
•~• 150
•0 200
150
100
" 0
0% 10% 20% ". ".
Strain
so. '0%
Figure 11 : Effect of Rest Period 00 Stnin Sweep of BO~caD PGS8 af IT (1.6 Hz)
160000
140000
""*" Retest (1 Hr)
-+- Retest (3 Hr)
120000 - Retest(12 Hr)
'ii' 100Xl0
!'.
"
60000
60000
",m
20000
0
0 1000 2000 3000 4000 5000 6000
Cycles
Figurt 12: Effect of Rest Period on Stnin Sweep of PE Unslabilized Modified PGS8
al IT (1.6 Ht)
Bahia. Zhai, Bonnet/i. & KosI' IS
For many years fatigue of asphalt concrete mixtures has been the subject of
investigation. Although it is recognized that fatigue damage is mainly caused by cracking
or damage within the asphalt binders, very few studies have used binder testing to
evaluate fatigue of binders (\5). There is a significant lack of information about the role
of binder composition or rheological properties of binders in fatigue damage. In a
previous study the effect of mechanical working (fatigue) on rehological properties of
modified binders was evaluated by testing for 1000 cycles (3). The results indicated that
for small strains, there is minor amount of damage occurring in 1000 cycles. The testing
done using large sltllins stimulated the idea of considering fatigue testing at higher
strains, within the non-linear region. Based on the information collected for the strain
dependency a fatigue experiment was designed to include testing for a larger nwnber of
cycles at different strains levels, frequency levels, and for different modified and un·
modified binders. The testing also included rest periods to evaluate the effect of healing
on fatigue damage. The following sections present the results of fatigue testing under
various conditions within the non-linear region.
The first experiment conducted was \0 test for 5000 (or I 1O(0) cydes at 1 percent,
10 percent, 20 percent and 50 percent strain at multiple levels of G' . Figure 13 shows the
results for a modified asphal! and an un-modified asphah at a frequel"lCy of 1.6 Hz. The
results show that for both asphalts significant reduction in GO is observed as higher
strains are used in testing. The effects are highly signi ficant. At GOof 200 Kpa, both
binders are showing a drop of more than one oroer of magnitude (a drop to less than 20
Kpa) in the G' values within only 5000 cycles. In light of the discussion about the strains
within binder domains presented earlier in this paper, 20 percent strain is very plausible in
typical mixtures under normal traffic conditions. The drop in complex modulus is
expected to have significant consequences on fatigue damage in the pavement.
Figure 14 depiclS the effect of fatigue on the phase angle for the same binders.
Compared to the effects on the G' , the effect of fatigue on change in phase angle is only
marginal and limited to few degrees increase. This was observed for many binders tested
which indicates that phase angle is nol very sensitive to fatigue damage.
The strain levels represent the possible effect of pavement structure. Weak
pavement structures are expected to allow higher strains than strong pavemem SlruClure
under same level of traffic. The high sensitivity of fatigue to strain level observed in this
study is nOI surprising. It confirms the find ings of many research works that have related
fatigue damage to strain of asphalt mixtures (16-18). What it is showing, however, is that
these effects can be detected in binder testing and this can be included in a binder
specification systcm that is beller related to pavement performance.
Bahia. lAm. 8onlll'lli. & KQse 16
2SO
200
50
0
0 2000 4000 6000
Cye'"
Figure 13: G· vtrSus Cytles of Bosun PGS8 and PE Unslabiliud Modified PGS8 al
Diffmnl Strain Levels (1.61h:) (200Kpa)
70 ---
69
"
67
...... PE I.I"IITJ:XiIied 10%
~"
" -w.-PE unmodified 20%
..... Boscan 10%
64
~ Boscan2D%
62
61
60
0 2000 4000 6000
C",..
Figure 14: /) Vt!1U5 Cydes orBman PGS8 and PE Unslabilized Modified PGS8 at
Differenl Strain Levels (1.6Hz) (200Kpa)
Bahia. ZhiJi. Bonnelli. & Kase 17
Trame speed can vary significantly in the field . It is known that fatigue
performance of different materials, particularly visco·elastic materials, can be sensitive to
loading rate or frequency. Figure 15 includes the fatigue results for a modified asphalt at
three different frequencies (0.1, 1.0 and 10 Hz). All testing was done at 20 percent strain.
The results clearly show the significant effect of the frequency. At 10 Hz repeated
loading at 20 percent for approximately 3000 cycles resulted in reduction of the G" by an
order of magnitude. After only 2000 cycles the G- value at 10 Hz has reduced to the
value measuredat 1.0 Hz. At 1.0 Hz, 3000 cycles are sufficient to reduce the G" value to
the value measured at 0.1 Hz.
These significant effects of frequency and the increase io fatigue damage with
frequency were observed for many of the asphalts tested. The sensitivity to fatigue
however is different from ooe bioder to another and so is the innuence of other factors,
such as frequency, strain level, and number of cycles. The consequences of the fatigue
behavior in terms of pavement performance are very important. In the current Superpave
specification the grade of asphalt changes every six degrees interval. On 'average, six
degrees intervals result in increasing the G" by approximately 100 percent (one fold).
The effect of fatigue al moderately high strains appears to be much more signi ficant, for a
low number of cycles. A reduction of G" by 1000 percent (10 fo lds) after only 4000
cycles indicate that a small amount of traffic can change grade by at least 3 temperature
grade intervals. This might clarify the importance of this behavior and the need to
consider fatigue within the non·linear region.
1000
10 "-- - - - - - - - '--""
o 2000 4000 6000
Cycles
'000000
--- - -
...... TPG5810%
" .0000
.....-- B PG58 20%
....- B PG70 10%
~ BPG7020%
o sooo .0000
Cycles
Figure 16: Timt Sweep ofTwo Nut Asph.11(2 Gndes) at DiffereDt Stu in LeYcis
(200K,.)
1000000 ,- - -_ _ _ _,
1000 L _ _ _ __ -----'
o 5000 10000 15000
Cycles
l000000 r--------~
~ _Bosca n 58
.
~ 100000
~-'- SBS Radial
""'*"" SBS Unear
~ SBDiblock
b ..... EVA
~PE unstabilized
-+-EMA
10000 L _ _ _ _ _ _----.J
Modification by process is becoming well used because of the low cost associated
with it. The issue of oxidized asphalts of equivalent grades 10 other modified binders is
still being widely debated. The fatigue behavior of a set of oxidized binders produced
from the same base asphalts is shown in Figure 19. As seen in the figure the fatigue
disturbance of the oxidi7.ed asphalts are mostly smaller than thai of the base aspbahs.
What is important to notice, however, is the o:o:idized asphalts that vary significantly in
their behavior despite the fact that same base asphalt is used in their production. In
addition, it is observed that polymeric modification is offering much beller resistance to
fatigue disturban<:e that some of the oxidation te(hniques. This finding is important
because it puts the modification by polymeric additives at a different class than
modification by process. It certainly sheds more light on the effect of polymeric additives
and raises some questions about modi fi cations by oxidation. Figure 20 depicts a
comparison between the base asphalts, the same asphalt modified with an elastomeric
modifier, plastomeric modifier, and by controlled oxidation, The results show the
dramatic differences in fatigue perfonnance,
The resuits concerning the somce of asphalt and the type of modification,
although show very important trends, need to be confinned with a larger set of binders. II
is important to ootice that the testing in this study included only three base asphalts. How
much improvement can modifiers offer and what are the critical chemical properties that
can improve resistance of base asphalts to fatigue distmbance is yet need to be
understood.
Healing effects, as detennined by allowing rest periods, have been discussed with
regard to strain sweeps. It is also expetted that healing will have significant effect on
fatigue disturbance. Selected samples were tested after various rest periods. Figure 21 is
prepared to show the effect of re·testing of an wunodified binder within 30 minutes after
completing faligue testing for 5000 cycles and after a rest period of 12 bours. The results
confinn thai healing has a significant effect on recovery from fatigue damage and that
recovery is a time dependent phenomenon, Figure 22 shows the results for a binder
modified with a plastomeric polymer. [t appears that this type of binder can recover
within 3 hours of the initial test. What is more imponant is how much this disturbance is
recoverable. After waiting for 12 hours and retesting the same sample, the trend is very
similar to the initial fatigue curve. This indicates thai no pennanent damage is occurring
and thai the binder can fully recover to its original status, This finding is consistent with
what has been observed in previous research for asphalt concrete mixtures (14), It has
signi fi cant consequences regarding the effcct of the amount of traffic and the frequency of
the traffic on fatigue disturbance of asphalt pavements.
Bahia, lha;. Bonnelli. & Kose 22
1000000 I-~~~~~~~~--'
-_ .._-
-+- T Oxidized1 10%
100000
'"'""'*- T Oxidized1 20%
~ T Oxidized2 10%
~
• -e- T Oxid ized2 20%
b -+- 8 Oxidized1 10%
........ a Oxidizedl 20%
10000
........ 80xidized210%
-+- 8 Oxidized2 20%
1000
o 5000 10000
Cyele.
1000000
100000 -+-LS
•
-
0.
•
Cl
, ___ E1
I ~P
10000 1-+-01
.~-.-
1000
o 5000 10000
Cycles
Figure 20. Effects of Modifier and Process on Fatigue (20 percent Strain 37C)
Bahia, Zhai, Bonnetti, & Kase 23
-+- 0 II
250 ...... Retest (30 min)
ii 200 """*- Relesl (12 II)
~
l!O
•
0
150
100
50
0
0 1000 2000 3000 <4000 5000 6000
Cycles
Figure 21 : Errttl of Rtst Period on BOSCln PG58 at IT 1,6 Hz 20 percent Strain
250000 , - - - - - - -- -- - ,
200000
.
~
•
150000 - -- ---
____ 10% 01'1'
-+- 20% ot-t
o~ ________ ~
Figure 22: Erred ofRtst Period on PE Unstllbilized Modified PG58 at20e 1.6 H'l20
percent Strain
Bahiu, 7Jwi, BO/met/i. & Ka.le 24
Su mmary of Findings
This study was conducted to evaluate the behavior of asphalt binders in Ihe non-linear
viscoelastic region. The following points summarize the findings from this study.
1. Strain dependency of asphalt binders can playa major role in defining relationships
between rheology of asphalt binders and rheology of asphalt mixtures.
2. Based on computer simulations and finite element analysis. it is shown that asphalt
binders can be subjected 10 strain levels much higher than the bulk strain to which
mixtures are subjected. Depending on size distribution of aggregate, binders can be
subjected to strains 10 times higher than the bulk strains of mixtures. It is therefore
necessary to consider large strain testing (non-linear) to accurately estimate the role of
binders in mixture performance.
1 Asphalt binders, particularly modified binders, vary significantly in their strain
dependency. Some binders are morc sensitive to strain level than othcrs due to their
composition or microstructure.
4. Strain dependency is a strong funct ion of temperature and loading time. Lower
temperatures and/or higher frequenc ies result in higher sensitivity 10 strain leveL
Some binders are so sensitive to strain that complex modulus reduces by as much as
10 folds when strain changes from 2percent 10 SOpercen\. Strain levels 20percent to
SOpercent in (he thin films of asphalt are plausible in typical mixture used for paving.
S. There is a significant interaction between strain levels and time loading (frequency).
The data collected for binders show that the interaction is not simple and cannot be
represented by simple shift factors. This complex interaction can have significant
affects on construction of rheological master curves and time temperatW"e shift
functions.
6. SlIain dependency trends are not uniform for all binders. Some binders show simple
transition from small to large strains while other binders show specific transition
regions. The strain dependency behavior can give some indications of the integrity of
binders and the compatibility of additives.
7. Strain level has a very significant affect on classical fatigue behavior of asphalt
binders. Fatigue damage, as defined by reduction in rigidly (a· ) increases rapidly
with increasing strain.
8. Because reduction in a· values and increase in is can have different consequences in
terms of pavement performance. It is proposed that the term Mechanical Work
Disturbance (MWD) be used to replace classical strain-controlled or stress-controlled
fatigue. Effects of Mechanical Work Disturbance (MWD) on a" is much more
pronounced than its effects on phase angle (6). Reduction of a· by as many as 100
times (2 orders of magnitude) is observed as a result of only 10,000 cycles at
moderate strains of 20 percent.
9. There is an affect of loading time (frequency) and temperature on fatigue disturbance.
The effects arc howtver much smaller than those of strain Icvel. Strain level is by far
the most important factor controlling fatigue disturbance.
10. Based on the limited binders tested, fatigue disturbance appears to be a strong
fu nclion of asphalt composition and types of modification. Polymeric modifiers (both
Bahia. Zhoi. Bonnetti. & Kose 25
elastomers and plastomers) are observed to have a profound effect on limiting fatigue
disturbance and resist reduction in 0' values.
11 . Rest periods show significant effect on recovery from shear thinning (strain
dependency) and from accumulation of fatigue damage. Healing is the most probable
mechanism for their recovery. Based on test results collected so far, it appears that
healing is a function of composition of base asphalt and type of modifiers.
Aeknowledgmenl
This work is part of the FHWA Project "Superpave National Training Center II".
The project main contractor is the Asphalt Institute. The authors gratefully acknowledge
the support of FHWA . The authors would like to acknowledge the support of the Project
Officers, Mr. Juhn Bukowski and Mr. John D'Angelo. The support of Dr. Richard May
and Mr. Mike Anderson of the Asphalt Institute are also acknowledged.
DiS(:Jaimer
The opinions and conclusions expressed or implied in the report are those of the
research agency. They are not necessarily those of the Transportation Research Board, the
National Research Council, the Federal Highway Administration, the American
Association of State Highway and Transportation Officials, or of the individual states
participating in the National Cooperative Highway Research Program.
Iklhia. Zhai. Bunnclli. & KoSI' 26
References
Preprint
Tuesday, 9:00 AM
Presiding,
Gerald Huber