Download as pdf or txt
Download as pdf or txt
You are on page 1of 158

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011

Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
FATIGUE AND
D Y N A M I C TESTING OF
BITUMINOUS MIXTURES

A symposium
presented at the
Seventy-sixth Annual Meeting
AMERICAN SOCIETY FOR
TESTING AND MATERIALS
Philadelphia, Pa. 24-29 June 1973

ASTM SPECIAL TECHNICAL PUBLICATION 561


V. P. Puzinauskas, symposium chairman

List price $15.50


04-561000-08

AMERICAN SOCIETY FOR TESTING AND MATERIALS


1916 Race Street, Philadelphia, Pa. 19103

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
9 by AMERICAN SOCIETY FOR TESTING AND MATERIALS 1974

Library of Congress Catalog Card Number: 74-81156

NOTE
The Society is not responsible, as a body,
for the statements and opinions
advanced in this publication.

Printed in TaUahassee, Fla.


October 1974

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
Foreword

The symposium on Fatigue and Dynamic Testing of Bituminous Mixtures was


presented at the Seventy-sixth Annual Meeting of the American Society for
Testing and Materials held in Philadelphia, Pa., 24-29 June 1973. Committee D-4
on Road and Paving Materials sponsored the symposium. V. P. Puzinauskas, The
Asphalt Institute, presided as symposium chairman.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
Related
ASTM Publications

Fatigue of Compacted BituminousAggregateMixtures,


STP 508 (1972), $13.75 (04-508000-08)

ViscosityTesting of Asphalt and Experiencewith


Viscosity Graded Specifications,STP 532 (1973),
$8.75 (04-532000-08)

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
Contents

Introduction

Statistical Characteristics of Fatigue Damage Accumulation


in Flexible Pavements-J. I/. SOUSSOUand F. MOAVENZADEH 3
Probabflistic Methods 4
Closed Form Probabilistic Solution 4
Numerical Application 8
Summary and Conclusions 10
Influence of Laboratory Test Results for
Asphaltic Concrete-L. H. IRWINand B. M. GALLAWAY 12
Equipment and Procedures 18
Discussion 35
Conclusions and Recommendations 40
Techniques for Characterizing Bituminous Materials Using a
Versatile Triaxial Testing System-R. L. TERREL,
i. S. AWAD,and L. R. FOSS 47
Triaxial Test System 48
Material 54
Tests-Procedure and Results 56
Concluding Remarks 65
Summary of Complex Modulus Laboratory Test Procedures
and Results--M.W.W/TCZAKand R. E. ROOT 67
Effect of Test Variables 70
Variability of the Dynamic Modulus Test 75
Relationship Between Dynamic Modulus and Flexural
Stiffness Tests 87
Recommendations and Conclusions 90
Dynamic Response and Fatigue Characteristics of Asphaltic
Mixtures-C. L. SARAFand KAMRANMAJIDZADEH 95
Materials and Testing Procedures 97
Analysis of Results 103
Summary and Conclusions 113

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
Dynamic Testing of Bituminous Mixtures for Permanent
Deformation Response-J. MORRISand R. C. G. HASS 115
Design Approaches for Prediction of Permanent Deformation 116
Equipment Operational Requirements 117
Description of Apparatus and Test Methods 119
Conclusions 129
Material Characterizations for Rational Pavement Design-w. J. KENIS 132
Structural Subsystem (VESYS II) Interactions 134
Variability of Material Properties 136
Suggested Laboratory Method for Estimating Compliance
Variation 144
Limitations and Use of the VESYS II Program 146
Summary 150

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
STP561-EB/Oct. 1974

Introduction

Throughout their lives, pavement components are subjected to stresses and


strains of varying magnitudes and duration. These repetitive forces, cumulative-
ly, can, in time, decrease pavement stiffness and increase deflection and, in turn,
cause pavement failure manifested by excessive cracking and permanent
deformation.
These factors emphasize the importance of knowledge and measurements of
stress-strain relationships of either the pavement or laboratory-prepared paving
mixtures under imposed repetitive loadings. Such measurements can provide
indication of the levels of stresses and strains permissible for a given pavement
system. Furthermore, such measurements, when performed on the pavement
itself, can indicate the condition of the pavement at the time of the
measurement. The information gained from these measurements may then be
used in the design of new pavement structure or for the establishment of the
layer thickness needed to strengthen the existing pavement.
Current empirical paving mixture design methods, such as Marshall or Hveem
procedures, do not provide this information. Such tests are performed under
static loading conditions involving a single load application. Additionally, the
data obtained from these tests are expressed in arbitrary empirical units which
are difficult to correlate with the forces or deflections occurring in actual
pavements. Thus, new, more rational paving mixture design tests which could be
used alone or to supplement the old empirical tests are needed.
The papers presented at this symposium deal with testing procedures and
methods for the determination of dynamic modulus, fatigue, and creep
properties of paving mixtures. Laboratory or field compacted paving mixture
specimens are tested under a variety of loading conditions and at temperatures
normally encountered during the span of pavement life. Thus, it may be
expected that the tests and procedures described would contribute greatly to an
improved process for the characterization of pavement materials, and to a better
understanding of stresses and strains in pavements under repeated loading and
variable temperature conditions. This, in turn, could lead not only to a selection
of more suitable paving mixtures, but also to development of a more rational
design procedure for paving mixtures and for the entire pavement structure. It is
hoped that papers presented in this symposium will provide a sound basis
towards such goals.
K P. Puzinauskas
Research engineer,The
Asphalt Institute,
CollegePark, Md., 20740;
symposiumchairman.

1
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Copyright*1974byby ASTMInternational
Downloaded/printed www.astm.org
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
J. E. Soussou 1 and F. M o a v e n z a d e h t

Statistical Characteristics of
Fatigue Damage Accumulation
in Flexible Pavements

REFERENCE: Soussou, J. E. and Moavenzadeh, F., "Statistical Character-


istics of Fatigue Damage Accumulation in Flexible Pavements," Fatigue and
Dynamic Testing of Bituminous Mixtures, ASTM STP 561, American Society
for Testing and Materials, 1974, pp. 3-11.
ABSTRACT: A closed form probabilistic formulation is presented for the
determination of the fatigue deterioration of flexible pavements. This
formulation is based on the simple phenomenological models which use
Miner's law. The expected value and variances of the measure of damage are
related to the statistical characteristics of the loading patterns and materials
properties. A spatial coefficient of correlation is determined for the measure
of damage and is used to determine the typical size of the cracked areas. This
study emphasizes the need for obtaining more complete material characteriza-
tion procedures which do include measurements for spatial variabilities.
KEY WORDS: fatigue tests, flexible pavements, cracking, damage accumula-
tion, spatial variability, bitumens

In a given climatic environment and for a particular traffic, several distress


mechanisms develop and propagate within a pavement, either independently or
through interacting processes. They affect the structural integrity of the
pavement by affecting either its continuity (disintegration) or its deformation
(distortion).
A regression analysis presented in the American Association of State Highway
Officials (AASHO) Road Test [1] 2 relates a measure of the performance of a
pavement (its Present Serviceability Index) to two distress manifestations:
cracking and permanent deformation. Although the AASHO analysis indicates
that the Present Serviceability Index is not very sensitive to the degree of
cracking of the pavement, cracking may have, however, indirect effects on the
accumulation of permanent deformation. For instance, when the surface layer is
cracked increased stresses will be transferred to the subgrade so that the rate of
accumulation of deformation will become more important. Also, the possibility
o f moisture seeping into the subgrade and weakening it will increase. These
indirect effects are generally ignored in the mathematical models for pavement
analyses.

1Assistant professor of civil engineering and professor of civil engineering, respectively,


Massachusetts Institute of Technology, Cambridge, Mass. 02139.
2The italic numbers in brackets refer to the list of references appended to this paper.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Copyright*1974 by
Downloaded/printed byASTM International www.astm.org
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
4 TESTING OF BITUMINOUS MIXTURES

Cracking is either due to load repetitions or to environmental factors. This


study will be restricted to the load associated cracking which is also referred to
as fatigue cracking.
Several models have been suggested to represent fatigue cracking. They can be
separated into two types: phenomenological models and mechanistic models.
The phenomenological models [2,3] are essentially based on some form of
Miner's law [4] ; they have the advantage of simplicity and availability of data
for different materials. Their principal disadvantages are that they do not
account satisfactorily for the influence of the geometry and the materials
inhomogeneities and they do not provide a quantitative measure for the extent
of cracking in a pavement. The mechanistic models [3,5] are still impractical to
use, but they are more amenable to provide a quantitative description of the
degree of cracking in a pavement.

Probabilistic Methods
Because of the variabilities and uncertainties associated with the operations of
a pavement, the inputs and outputs of a fatigue model must be described in
terms of probabilistic distributions instead of single valued estimates. The meth-
ods of approach to the formulation of probabilistic models may be divided into
two types: simulation procedures and direct probabilistic procedures.
The simulation procedures are a brute force approach to the problem. They
are used when a closed form probabilistic formulation is not available. The basic
idea is to generate several sets of input variables from their known statistical
distributions. Each set is used in the deterministic model to generate a corres-
ponding set of outputs. If the number of sets is large enough it is possible to
obtain means, variances, and statistical distributions for the output functions.
The simulation procedure is a simple numerical approach to problems not ame-
nable to analytical procedures due to their inherent complexities and interacting
factors. The accuracy of such an approach is a function of the number of sets
which are analyzed. A good accuracy can be obtained at the cost of large
computation times.
The direct probabilistic procedures attempt to obtain a closed form solution
in terms of the means and variances of the input variables. This is often done
through approximate formulae. These approximations involve expansion of the
solution into series similar to the expansion of a function into Taylor or Fourier
series.

Closed Form Probabilistic Solution


In its simplest form, Miner's law is given by the following equation
m
D= ~ ni <~1 (1)
i=1 N i

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
SOUSSOU AND MOAVENZADEH ON FLEXIBLE PAVEMENTS 5

where n i is the number of load applications at the state (i) and N i represents the
number of cycles to failure for that particular state (i). The state (i) may
represent a given combination of strain level and temperature value. It is
assumed that a total of m states are considered.
If D reaches the value of l, failure is said to have occurred. But if D = 0.5,
one can only state that half the life of the specimen (or the pavement) is used
up. It does not imply that half the final number of cracks are already formed or
that all the cracks have half of their final length. It is not possible to state
whether many small cracks will eventually form a few large ones.
It is generally assumed that the number of cycles to failure, Ni, is related to
the strain amplitude, Xi, by the following relation

N i = CXi a (2)
where C and a are materials properties. It can further be assumed that the load
spectrum is given by the following

n i = f (Xi) (3)

that is, the number of applied load n i is a function of the magnitude of the
strains induced by these loads Xi. Combining Eqs 1-3 yields

m f(X~) Xi -a (4)
D =
i=1 C

Expected Value
A first order approximation yields the following closed form solution for the
expected value of D

m f Xi -a
D=
/=1 C-"

m
+ 1 ~ a2D __ aXi 2
2 i= 1 aXi 2 Xi (5)

+ 1 a2D

+ 1 02D I oC 2 + ~2D Cov[a,C]


o -6 aaaC

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
6 TESTING OF BITUMINOUS MIXTURES

Equation 5 assumes that only variables a and C are statistical in nature.


Equation 5 can be reduced to

m f X i -a f -af ~(d+l) ]
D = T, { 1 + [ _ __+ _ aXi 2
i= l C 2f )(if 2Xi 2

(6)
+ 1 (lnXi) 20a 2 + 1__ 2

ln__X/ Coy [a, c] }


C
In Eq 6 the bars represent expected values and the dots represent differentiation
with respect to Xi; a2a represents the variance of its subscript.

Variance
Similarly, the formulation for the variance of D
-7-
m f a)2
0"02 = ~ Ri 2 ( 7 oXi 2
i=l f
m
+ (~ Ri l n X i ) 2 0 a 2
i=1
m R--i (7)
+ ( Z ): ac:
i=1 C

m m Ri
- 2 ( Z Ri In Xi) ( Z ) Cov[a, cl
i=1 i=1 C

where
fx/-=
(8)

The value of the variance contains both the uncertainties in the estimation of
D and the variabilities due to the changes in the operating system. A low value
of aD 2 indicates that different sections of pavement are likely to fail at the same
time. However, a large value of OD2 does not necessarily indicate that the failure
is taking place in the form of a few large cracks instead of a more uniform failure
(many small cracks). This will be evaluated in the spatial covariance of D.

Spatial Convariance
Consider a material property such as a modulus of elasticity, E, which is given

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
SOUSSOU AND MOAVENZADEH ON FLEXIBLE PAVEMENTS 7

by its statistical distribution, for example, E and oE 2 . It can be stated that


represents the average quality of the material, while ere2 is a measure of its
uniformity or the quality control during its fabrication. It is clear that two
samples of material taken from the same batch are more likeiy to be similar than
two samples taken from different batches. It is also clear that two samples of
material placed the same day (in the same environmental conditions and by the
same machines and crews) are more likely to be similar than two arbitrary
samples. Mathematically a measure of the closeness of a property of two batches
of a material (for example, their modulus of elasticity) can be represented by
their covariance.
Thus, it is possible to define

Cov[E(pl), E(P2)] = E ( p t ) E(p2)-E(p,) E(p2) (9)

If Pt and P2 are two different locations, it is clear that the covariance will
increase when they are close to each other (because the materials, due to their
proximity, have a higher likelihood of being of the same batch or being treated
similarly in the field) and will decrease when they are farther apart. The
coefficient of correlation is defined by

Coy [E ( p l ) , E (p2) ]
PE (Pl), E (P2) =
0"E2
(10)

where E(p 1) and E(p2) represent the same material property. When Pl = P2, P
has a maximum value of one. As A p = (p: - Pl) increases (which implies the
points of interest in the pavement are farther apart from each other), p may
decrease to some value between 0 and 1.
Assuming that such a coefficient of correlation may be defined for Xi, a and
C, it is possible to obtain the following approximation for D

Coy [D(p), D (p + Ap)] =


m [ ~ )2 2
Ri 2 ( PX Ox P x i
i=I ~-- ~/
rn
+ ( X Ri In ~'i) 2 OaOa 2
(1t)
i=1
,'n
+ ( Y" Pc oc 2
i=l C

m m
- 2 ( 2; Ri ln.,~i) ( 2;
i=1 i=1
•)Cov
C
(a, C)

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
8 TESTING OF BITUMINOUS MIXTURES

Px, Pa, and Pc are functions of the distance Ap. It is possible also to define a
coefficient of correlation forD

Coy [D (p), D (p + Ap)I


PD =
oD 2 (12)

The coefficient of correlation .0D represents the extent of spatial uniformity.


It is possible to choose an arbitrary critical value of PD in order to determine a
length Ap c associated with the pavement. This critical length Ape will represent
the average size of cracked areas.

Numerical Application

Let us assume that the fatigue curve is defined by C = 10-4, a = +2.6, (see Fig.
1) .and that only one mean strain amplitude is considered .~= 10-4 in./in, and m
1, and n] = constant = 10 4 cycles; in such case Eq 6 reduces to

O = (104) (10-4)2"6 t 1 + 2.08 Ox2+ 287 Oa2 + ~


(10-4) ~ -~2 ~
lnX (13)
Coy [a, C] }
C
If o2e -~ 0, the expected value of D reduces to
Oa 2
Ox ~ I (14)
D = (0.004) ~( 1 + 2.08 + 287

~o-t

N= o(')
""x~x ~ = I0 -4 o-r
io-2
xxx~\~, ~ = 2,6 cr o = 0.26

I- i 0 -'~
co
~
x

io-4
\ x.t:[ STANDARD DEVIATION ""x ". a = "5 + o"a

..\
io-5 ",. a = ~ - c r o
I I I I I I I I I I "I
0 I 2 3 4 5 6 7 8 9 I0
LOG N

FIG. 1-Fatigue curves,

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
SOUSSOU AND MOAVENZADEH ON FLEXIBLE PAVEMENTS 9

In Eq 14 it is clear that the expected value of D is sensitive to the variability


and uncertainty of X and a. For instance, if OxZ/X2 = 0.01,/~increased by more
than 2 percent; while if O2a/a2 = 0.01 (which is true for the values shown in Fig.
1), D, according to Eq 14, increases by almost 300 percent.
Similarly the variance of D is given by

Ox2 oa 2
oD 2 = (0.004) 2 { 6.7 ~2 + 570 a ~ } (15)

while the covariance is given by

Coy [D (p),D (p + Ap) ] = (0.004) 2

~x 2 ~a 2
{ 6.7 Ox X2 +570pa a ~ } (16)

Thus, the numerical values just given yield

= (0.004) [1 + 0.021 + 2.87] = (0.004) (3.891) (17)

and

OO = (0.004) X 5.77

Therefore, the variability ofx and a has a major influence on the values of D and
oD. The variations of D and (oD/D) as functions of the uncertainties ofx and a
are shown in Fig. 2.

5
(1

td
.J

I--
2
o.

O I I i I I
0 0.1 0 0.1
COEFFICIENT OF V A R I A T I O N COEFFICIENT OF V A R I A T I O N

FIG. 2-Influence of the coefficients of variations of (a) and (x).

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
10 TESTING OF BITUMINOUS MIXTURES

Next, the interval separating cracks may be estimated. Let the expected value
of damageD = 0.5 and the standard deviation aD = 0.25. If the distribution of D
is Gaussian, then one can say that the probability o f D >~ 0.09 is 2 percent. For a
given crack, the probability that 98 percent of its neighboring points having a
damage value larger than 0.9 will be achieved when

+ 20 OD -- 2 oo l ~ - - p 2 /> 0.09 (18)

With the assumed values just mentioned this will yield p/> 0.98. From the curve
relating O to distance one determines then the typical size of cracked area. This
method of approach remains valid even if the distributions are not Gaussian,
although the numerical values would be different. This method is depicted
schematically in Fig. 3.

Q DISTRIBUTION OF D AT A GIVEN TIME


FOR ALL THE PAVEMENT
OI
I

I
I
%D,STRISUT,ON OF 0
FOR, = 0.98 9
i
PDF o

0.5 \ DISTANCE
D \ 2 % OF PAVEMENT
IS CRACKED AVERAGE SIZE OF
CRACKED AREAS

FIG. 3-Method of determination of the average cracked areas.

Summary and Conclusions


A closed form probabilistic formulation was presented for the determination
of the fatigue deterioration of flexible pavements. This formulation relates the
expected value and variances of the measure of damage D to the statistical
characteristics of the loadings and materials properties. A spatial coefficient of
correlation is determined for D. This coefficient OD is used to determine the
average size of cracked areas. This study emphasizes the need for obtaining more
complete material characterization procedures which do include measurements
of spatial variabilities.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
SOUSSOU AND MOAVENZADEH ON FLEXIBLE PAVEMENTS 11

References
[l ] Highway Research Board, "The AASHO Road Test: Report 5," Special Report 61E,
Publication 954, National Research Council, Washington, D. C., 1962.
[2] Monismith, C. L., "Asphalt Mixture Behavior in Repeated Flexure," Department of
Civil Engineering, University of California, Berkeley, Report No. TE 70-5, Dec. 1970.
[3] Soussou, J.E. and Moavenzadeh, F., "On the Crack Development in Pavements,"
Highway Research Board, 1973 Annual Meeting, Washington, D.C.
[4] Miner,M. A., "Cumulative Damage in Fatigue," Journal of Applied Mechanics, VoL
12, 1945.
[5] Majidzadeh,K. and Ramsamooj, D. V., "The Mechanistic Approach to the Solution
of Cracking in Pavements," Highway Research Board, 1973 Annual Meeting,
Washington, D. C.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
L. H. Irwin I and B. M. Gallaway 2

Influence of Laboratory Test


Method on Fatigue Test Results
for Asphaltic Concrete

REFERENCE: Irwin, L. H. and Gallaway, B. M., "Influence of Laboratory


Test Method on Fatigue Test Results for Asphaltic Concrete," Fatigue and
Dynamic Testing of Bituminous Mixtures, ASTM STP 561, American Society
for Testing and Materials, 1974, pp. 12-46.
ABSTRACT: Fatigue characterizations obtained using three laboratory repeat-
edqoad, flexure test methods were compared. The test methods included uni-
axial stress fields (beam specimens), biaxial stress fields (plate specimens), full
stress reversal, and no stress reversal. In addition, specimens were sawn from
field-compacted asphaltic concrete, and test results were compared with those
from laboratory prepared specimens.
The degree of stress reversal and possibly the type of stress field to which
the specimens were subjected were found to influence the resultant slope and
intercept of the fatigue curve. The results of a comparison of fatigue character-
izations obtained with laboratory and field-compacted specimens were not
clear due to differences in the air void contents of the two mixtures.
An analysis to determine the minimum number of specimens required for
fatigue characteristics using the beam test method suggested (but did not
prove) that approximately ten specimens having 3 in. (76 mm) widths and
depths should be used.
KEYWORDS: fatigue (materials), fatigue life, fatigue tests, materials tests,
stresses, strains, compacting, stiffness, flexible pavements, bituminous con-
cretes, asphalts, aggregates, beams (supports), slabs

NOMENCLATURE
P Applied force on beam, F
M Real moment due to applied load, FL
V Real shear force due to applied load, F
m Virtual moment in beam, FL
F Virtual shear force in beam, F
X Distance along beam from supports, L
Y Distance across beam, L
z Distance above or below neutral axis, L

I Assistant professor and highway research engineer, Agricultural Engineering


Department, Cornell University, Ithaca, N.Y. 14850.
2professor and research engineer, Civil Engineering Department and Texas
Transportation Institute, Texas A&M University, College Station, Tex. 77843.

12
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Copyright9 1974 by
Downloaded/printed byASTM International www.astm.org
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
IRWIN AND GALLAWAY ON ASPHALTIC CONCRETE 13

0 Angle of rotation of plane cross section, radians


b Width of beam, L
h Depth of beam, L
L Distance between center of supports of beam, L
A =bh Cross sectional area of beam, L 2
I = bh3/12 Area moment of inertia of beam, L 4
7"* Shear stress due to virtual forces, F/L 2
7" Shear stress due to applied load, F/L 2
"7 Shear strain due to applied load, radians
Wo Deflection of center of beam due to applied load, L
U Work done by virtual forces due to deformation
by applied load, FL
E Elastic modulus of material, F/L 2
G Shear modulus of material, F[L 2
Poisson's ratio of material, dimensionless

In order to incorporate the effect of repetitive loading of the pavement


structure into rational pavement design procedures it is necessary to have a
fatigue characterization of the pavement materials. The objective of the fatigue
characterization is to define failure criteria (usually allowable stress or strain) in
terms of the number of cycles to failure.
In most instances it has been deemed necessary to develop a fatigue character-
ization for each separate paving mixture of asphaltic concrete. Numerous investi-
gations have been reported in the literature in recent years. Excellent compila-
tions of existing information have been published by Epps and Monismith
[1,2] ,3 Pell [3], Pell and Taylor [4], Kaltas and Puzinauskas [5], and Jimenez
[6].
To date, our understanding of the factors which influence the fatigue charac-
terization of asphaltic concrete is neither comprehensive nor clearly defined.
However, when we achieve better definition of these factors we may fred it
possible to greatly reduce, if not eliminate, the amount of laboratory repetitive
load testing necessary in order to obtain a fatigue characterization. To efficiently
arrive at a method whereby we can obtain a fatigue characterization without
extensive laboratory testing it will be necessary to combine and evaluate the data
from the many different investigatiofis already performed. Since the laboratory
testing methods used by different investigators have not been the same, it will be
necessary to determine whether the test method significantly influences the
resulting fatigue characterization, in order to properly evaluate the combined
data.
In order to provide additional perspective on the influence of test method on
the fatigue characterization of asphaltic concrete, this paper presents the results
of a comparative study of three test methods. The test series included: (1) beam

3The italic numbers in brackets refer to the list of references appended to this paper.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
14 TESTING OF BITUMINOUSMIXTURES
specimens, full wave sine loaded at the tldrd-points of the beam, (2) beam
specimens, half wave sine loaded at the thkd-points of the beam, and (3) circular
plate specimens, full wave sine loaded at the center of the plate. Also, to verify
whether a fatigue characterization obtained using laboratory prepared materials
accurately represents the behavior of field-compacted asphaltic concrete, an
additional series of full wave sine loaded beams was included using field-com-
pacted materials.
It is generally agreed that the laboratory method used for fatigue characteri-
zation should reflect the mode of loading of the materials in the field, namely,
repetitive flexure. A variety of laboratory techniques have been developed by
different investigators (Figs. 1-3) in order to attain repetitive flexure. The
University of California and The Asphalt Institute have been using a beam
specimen in third-point loading [7], while Pell at the University of Nottingham
has been using a rotating cantilever beam [3]. Other types of specimens have
included an oscillating cantilever [8] and a circular plate in biaxial flexure [6,
9]. Although the University of California and The Asphalt Institute use the same
type of specimen and test frame, the pattern of stressing differs as the University
of California uses a block pattern [2] and The Asphalt Institute uses a haversine
with delay [5] (Fig. 4).

.'.v:',.., L(,o..'r~;.,.,,Z-.~ '.

. " " " "'.. '1"~'.;'~" ~ , " . : ~ "~'-" " ; "'


., . . . . . :

FIG. 1-Beam test geometries. (a) Simply supported beam and (b) beam on elastic
.foundation.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
IRWIN AND GALLAWAY ON ASPHALTIC CONCRETE 15

II.li

~~~. ~: .. .:.'b~.: '-.v';::.~:..--:

b
~;::2:"

FIG. 2-Orcular plate test geometries. (a) Plate with supporting pressure and (b) plate on
elastic foundation.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
16 TESTING OF BITUMINOUS MIXTURES

FIG. 3-Rotating cantilever test geometry.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
IRWIN A N D G A L L A W A Y ON A S P H A L T I C CONCRETE 17

o. FULL WAVE SINE LOADING

..... ..... /_

O
<I b. HALF WAVE SINE LOADING
o

C, HAVERSINE LOADING WITH DELAY

.. 8LOCK ,OADING

FIG. 4-Loading patterns.

The fatigue behavior of asphaltic concrete has been found to have a linear,
bilogarithmic relationship of the form

N/= kl ~'1 (1)


N: = ks e "2 (2)

where
N: = number of load repetitions to failure,
kl, k2 = regression constants,
17 = maximum stress in the extreme fiber, psi or MN/m2,
e = maximum initial strain in the extreme fiber, microstrain, and
F/l, //2 = negative regression constants describing the slope
of the regression line.

The value of the slope of the regression line, nl, in the above equations has been
reported by various investigators to lie between - 3 and - 7 for most asphaltic
paving mixtures. The value of n2 has generally been reported to lie between
-2.5 and -6.0. It has been shown by Pell [3], with further elaboration by Epps
and Monismith [I], that k2 is a function of the volume proportions of binder
and aggregate in the mix.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
18 TESTING OF BITUMINOUS MIXTURES

In a study conducted by Epps and Monismith [1], a comparison was made of


a fatigue characterization obtained by the University of California method
(third-point loaded beam, block loading pattern) with that obtained by the
University of Nottingham method (rotating cantilever beam, full wave sine
loading pattern). The aggregate gradation and the asphalt penetration of the
materials used by Epps and Monismith in the beam tests were similar to the
materials used by Pell [3] in the rotating cantilever tests, but the materials
differed in source, mineralogy, etc. Whereas PeU found the relationship

N[ = 1.6 X 10 -~1 e -6"1

Epps and Monismith obtained

iV/ = 6.1 X 10-8 e -3"4

Minor differences in the regression analysis techniques of the investigators did


not account for the difference in the regression constants. Further, a statistical
test performed by Epps and Monismith showed that a hypothesis that the slopes
were equal could be rejected at the 0.05 significance level. Since every effort was
made to assure similarity in the asphaltic concrete tested, it is most likely that
the differences in the fatigue behavior were due to differences in the test
methods.

Equipmentand Procedures
Test Equipment
The beam test frame was essentially similar to that described by Deacon [7],
except that it was modified to permit the use of specimens 3 in. (76 mm) in
width and depth and 15 in. (381 mm) long. The beams were loaded in
third-pdint flexure, with 1-in. (25-mm) damps located 4 in. (102 mm) on center.
The circular plate test frame was essentially similar to that described by
Jimenez and GaUaway [10] except that it was modified to apply a constant air
pressure to the bottom side of the specimens. The plates had a 144n.
(356-mm)-diameter dear span, and they were toaded over a 2 in. (51 ram)
circular area on the top center of the specimens.
Load was applied to the specimens using a closed-loop, servohydraulic testing
machine. The machine was operated in load feedback control for all tests, and it
was adjusted to shut off at specimen failure.
One test series of beam specimens was loaded with full stress reversal (full
wave sine loading, Fig. 4a) and a second series was loaded without stress reversal,
A block loading pattern (Fig. 4d) as used by Monismith, was not available on the
equipment used for this test series. A half wave sine loading pattern (Fig. 4b)
was selected for the second test series rather than a square wave, because it was
believed that the half wave sine more closdy resembled the block loading pat-

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
IRWIN AND GALLAWAY ON ASPHALTIC CONCRF.TE 19

tern in terms of rate of loading. Full stress reversal was achieved in the plate
specimens by applying a haversine loading (Fig. 4c) on the top center of the
specimens.
For the beam specimens, deflection was measured at midspan using a linear
variable differential transformer (LVDT). For the plate specimens, deflection
was measured using the testing machine displacement system. After completion
of testing, this method was found to be unsatisfactory due to the fact that the
test frame deformation was a large percentage of the measured deflection. As a
consequence the actual specimen deflection for the circular plates was not
determined with sufficient precision.
All load and deflection data were recorded on a dual-pen strip-chart recorder.
Temperature in the testing room was controlled at 70 -+ I ~ (21 + 0.5~

TABLE 1-Prope, ties of asphalt used in East Texas crushed stone asphaltic mixtures.

Field Mix Laboratory Mix


Unageda Recoveredb Unagedc RecoveredW

Penetration, 77"Fe 73 34 60 32

Softening point, "Ff 123 142 128 139


Viscosity, poises
77"F (sliding plate)g 1.5 X 106 7.2 X 106 3.0 X 106 10.6 X 106
140"F (capillary)h 2700 14000 2700 7200
275"F (capillary)h 4.9 7.9 4.8 6.3

Specific gravityi . . . . . . 1.03 1.08

a Data from a sample taken at the batch plant, Jan. 1971.


b Recovered by Abson process/from an uncompacted sample of hot mix taken at the job
site, Jan. 1971.
c Data from a sample sent to Texas A&M from American Petrofina, Dec. 1971.
d Recovered by Abson process/from an uncompacted sample of hot mix prepared at
Texas A&M University, Jan. 1972.
e ASTM Test for Penetration of Bituminous Materials (D 5-71) [11 ].
fASTM Test for Softening Point of Asphalts and Tar Pitches (Ring-and-BaliApparatus) (D
36-70) [11].
g Test Method No. Calif. 348-A, Determination of Viscosity of Bituminous Materials by
Means of the Microvisometer [12].
h ASTM Test for Absolute Viscosity of Asphalts (D 2171-66) [11 ].
i ASTM Test for Specific Gravity of Semi-Solid Bituminous Materials (D 70-70) [11].
] ASTM Test for Recovery of Asphalt from Solution by Abson Method (D 1856-69)
[111

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
20 TESTING OF BITUMINOUS MIXTURES

Materials

The AC-20 asphalt and crushed aggregate used in the tests were from the
same sources as those used in experimental pavement sections in a heavy duty
haul road at the Big Brown Steam Electric Station, Fairfield, Tex. Approximate-
ly one ton of asphaltic mixture was prepared in a portable batch plant for use in
laboratory-compacted test specimens. The results of comparative tests on the
asphalt, aggregates, and hot-mix asphaltic concrete for the field and laboratory
mixed materials are given in Tables 1, 2, and 3. In general there was excellent
agreement between the properties of the field and the laboratory materials.

TABLE 2-East Texas crushed stone aggregate properties.

Fielda Laboratory
Min Max

Gradation, percent passingb


3/4 in. 100 100 98
1/2 in. 100 100 97
3/8 in. 98 98 93
No. 4 86 87 83
No. 8 71 74 74
No. 16 61 64 67
No. 30 56 60 63
No. 50 53 56 59
No. I00 15 21 21
No. 200 (wash)c 3.3 6.2 7.6
Bulk specific gravityd 2.52

Appaxent specific gravityd 2.62

Absorption, %d 1.48

a Data obtained from four cores and an uncompacted sample of hot mix.
b ASTM Test for Particle-Size Analysis of Soils (D 422-63) [11 ].
c ASTM Test for Amount of Material in Soils Finer than the No. 200 Sieve (D 1140-54)
[I~1.
u ASTM Test for Specific Gravity and Absorption of Fine Aggregate (C 128-68) [11 ].

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
IRWIN AND GALLAWAY ON ASPHALTIC CONCRETE 21
TABLE 3-East Texas crushed stone asphaltic mixture data.

Field Mixa Laboratory Mix

Extracted asphalt content, b


% by weight of total mix 6.8 7.0
Maximum specific gravity (Rice
Method)C 2.38 2.38
Hveem (average of three specimens)d
Stability 37 30
Density, lb/ft 3 132.6 135.0
Air voids, % 10.8 8.9
Marshall 50-blow (average of three
specimens)e
Stability, lb 1110 1560
Flow, 0.01 in. 7 10
Density, lb/ft 3 130.7 133.2
Air voids, % 12.0 10.2
Marshall 75-blow (average of three
specimens)e
Stability, lb 1900 1970
Flow, 0.01 in. 9
Density, lb/ft 3 131.2 i34.9
Air voids, % 11.7 9.1
Cores (one core from each of four
sections Min Max
Density, lb/ft 3 125.7 129.9 ...
Air voids, % 11.2 14.8 ...
Extracted asphalt content, % 6.5 7.1 ...

a All data except that on cores obtained from an uncompacted sample of hot mix taken at
the construction site, Jan. 1971.
b ASTM Test for Quantitative Extraction of Bitumen from Bituminous Paving Mixtures
(Method A) (D 2172-67) [11 ].
e ASTM Test for Maximum Specific Gravity of Bituminous Paving Mixtures (D 2041-71)
[q].
" Test Method Tex-208-F, Stabflometer Value of Bituminous Mixtures [13].
e ASTM Test for Resistance to Plastic Flow of Bituminous Mixtures Using Marshall
Apparatus (D 1559-71), [11].

Specimen Preparation

The beam specimens and the circular plate specimens were compacted in
three 1-in. (25-ram) lifts to an average density of ~ 126 lb/ft 3 (616 kg/m3).
This density was selected as being on the low end o f the range of densities
obtained from cores taken from the field pavement. The beams were fabricated
to the dimensions 3 by 3 by 1588 in. (76 by 76 by 387 mm) and the plates to
1789 in. diameter by 3 in. thick, (445 m m diameter by 76 mm thick).
Preliminary studies were made to determine the appropriate distribution of
compaetive effort in order to obtain uniform density throughout each specimen.
Kneading compaction, using a 2-7/8 by 1 in. (73 by 25 ram) tamping foot was

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
22 TESTING OF BITUMINOUS MIXTURES

used for the beam specimens. Vibratory compaction using a device similar to
that described by Jimenez [6] was used for the circular plate specimens. After
they had cooled to room temperature each specimen was weighed in air and
water to determine density, and stored at 70 +- I~ (21 -+0.5~ until test.
In addition to the laboratory mixed and compacted specimens, a number of
large blocks of asphaltic concrete were obtained from the pavement of the Big
Brown haul road. These were subsequently sawn into 3 by 3 in. (76 by 76 mm)
beam specimens. The block samples were obtained outside of the wheel path,
and they had a somewhat lower average density than that found in the main
roadway. The average air void content for the field beams was 18.8 percent,
while the average air void content of the laboratory plates was 15.5 percent.
Cores taken from the wheel paths in the roadway ordinarily had an air void
content between 14.0 to 14.8 percent.

Test Procedures
After clamping in the test frame each specimen was held for a period of up to
89h to permit the relaxation of residual stresses. All specimens were tested in a
controlled stress mode of loading at a frequency of 100 cpm.
Each laboratory-compacted beam specimen was oriented in the test frame in
such a manner that the load was applied perpendicular to the direction of
compaction. This was done in order to offset any possible planes of weakness
which may have been introduced during compaction of the specimens. The load
was applied in the direction of compaction for the field-compacted beams.
For the full wave sine loaded beam specimens, regular adjustment of the
testing machine was made to reduce zero drift and plastic flow of the specimens.
For the half wave sine loaded specimens, carefully machined shims were inserted
in the test frame to prevent the beam from traveling below the at-rest position.
Due to inertial effects and a possible slight bending of the test frame, the center
of the beam did deflect below the at-rest position (dashed line, Fig. 4b).
However the amplitude of the negative deflection was much less than five
percent of the positive deflection, and it was neglected from further considera-
tion. In all cases for the beam tests, the number of cycles to rupture of the beam
was considered the number of cycles to failure (Nf).
For the plate specimens the supporting pressure (in psi) was adjusted such
that it was about one percent of the peak load (in pounds) which resulted in a
minimum of plastic flow in the specimens due to zero drift. For each plate
specimen, deflection was plotted versus number of cycles (log.log plot). A
gradually inclined straight line relationship continued to a transition point,
whereafter the line curved upward, as has been reported by Jimenez [6]. As
rupture was not experienced with the plate specimens, the number of cycles at
the point of departure from the straight line on the deflection plot was defined
as the number of cycles to failure (Nf). This defmition is consistent with that
used by Jimenez.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
IRWIN AND GALLAWAY ON ASPHALTIC CONCRETE 23

Data A n a l y s i s
Calculations of the maximum stress in the extreme fiber, o, maximum initial
strain in the extreme fiber, e, and initial resilient modulus of dastidty, E, were
made for each test specimen in accordance with the following formulae. For the
third-point loaded beams
PL
( 7 - bh 2 (3)

23PL3 (1 + 216h2 (1 +/1) )


E = 108 w o b h a l15L 2 (4)

(7
e = - (5)
E
and for the centrally loaded circular plates

c 2 a
3(l+tt) [P ( + 4s rra 2] (6)
o- 87rh ~ aT c)-q

E - 3(1 - / s 2) [P(4a 2 - 3 c 2 - 4c 2 s a_a_)]


16 lr w o h 2 c (7)

o
e--~ (l-u) (8)

where

amplitude of the dynamic load applied to the specimen, lb,


L = beam span between reaction points, in.,
b = beam width, in.,
h = beam or plate height, in.,
a = plate radius, in.,

c = radius of the dynamically loaded area, in.,

W o -~ center deflection of the beam or plate after 200 load applications, in.,
q = support pressure on the plate, psi,
# = Poisson's ratio, and
s = natural logarithmic operator.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
24 TESTING OF BITUMINOUS MIXTURES

Several comments should be made regarding the formulae used. The term in
parentheses in Eq 4 represents a correction to the standard strength of materials
solution which accounts for shear deformation in the measured deflection of the
beam. The correction was derived using an energy approach, suggested by Popov
[14], as outlined in Appendix I. It shouM be emphasized that the shear
deformation correction is a function of the specimen dimensions and geometry
of loading, and, except for Poisson's ratio, it is independent of the properties
(strength, density, void content, etc.) of the material tested.
It can be seen that as the depth to span ratio, h/L, becomes small, the term in
parentheses in Eq 4 approaches unity. However, for a beam 3 in. (76 ram) high,
loaded at the third-points over a span of 12 in. (305 ram), for a Poisson's ratio/~
= 0.35 the parenthetical term is equal to 1.158. That is to say, shear deformation
adds N 16 percent to the deflection due to bending, and the calculated modulus
of the beam would be in error by a like amount if shear deformation was not
considered. Because the modulus is used in calculating the strain (Eq 5), the
strain would be in error by 16 percent also, if shear deformation was neglected.
Corrections of this magnitude should not be neglected when analyzing test
results for beams having a 3-in. (76-ram) cross section.
Equations 6 and 7 have been obtained from solutions published by Roark
[15]. Equation 6 is identical to the Grashof stress equation used by Jimenez and
Gallaway [10], Jimenez [6], and Layman and Philippi [16]. However, Eq 7
differs from the Grashof deflection equation used by the above authors, in that
the support pressure, q, is not a factor in Eq 7. The function of the support
pressure is to maintain a zero mean stress condition in the specimen to protect it
from failure due to plastic flow. If the support pressure is too high, flow will
occur in one direction, whereas if it is too low, flow will occur in the other
direction. The magnitude of the plate deflection, Wo, is a function of the
amplitude of the cyclic load, P, provided the structural integrity of the specimen
is not disrupted due to excessive flow. The independence of deflection magni-
tude from support pressure was also verified experimentally in the laboratory.
The maximum stress in the circular plate specimens occurred in the extreme
fiber at the center of the plate. At that location the radial and tangential stresses
were equal. Using the generalized Hooke's law relationship between stress and
strain [17], Eq 8 was obtained, in which the Poisson effect of the support
pressure was neglected.
It can be seen in Eqs 6, 7, and 8 that a knowledge of plate specimen
deflection is not required to compute stress, although it is needed to be able to
compute the resilient modulus, and the modulus is needed to compute strain.
Because of difficulty experienced in measuring the plate deflection, results and
conclusions for the plate tests will be based solely on the stress fatigue behavior.
Regression analyses were performed to determine the slope and intercept
constants using the bilogarithmic models given in Eqs 1 and 2. A bivariate
multiple regression technique developed by Scrivner and Moore [18] was used,

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
IRWIN AND GALLAWAY ON ASPHALTIC CONCRETE 25

assuming the quality of each variable to be unity. The regression coefficients


determined in this manner correspond to the average of the coefficients
computed using the more common univariate regression technique by regressing
Nf on X and X on N[, where X is either the stress or the strain variable. Pell has
been reported by Epps and Monismith [1] to have used the averaging method.

Test Results
The results of the regression analyses, including the regression coefficients,
the correlation coefficients, and the standard errors of estimate, are given in
Table 4. The test data, the regression lines, and the standard error bands are
shown for each test series in Figs. 5 through 11.

TABLE 4-Stress and strain fatigue curve regression parametersfor East Texas
crushed stone with 7 percent asphalt, test temperature 70*F.

Number
Material and of Slope Correlation Standard
Test Method Specimens Intercept K n Coefficient Error
Stress Fatigue Curve
Laboratory beams
Full-wave sine loaded 22 1.74 X 1011 -4.15 -0.98 0.219
Laboratory beams
Half-wavesine loaded 14 5.89 X 1013 -5.32 -0.99 0.212
Field beams
Full-wave sine loaded 12 3.57 X 109 -3.75 -0.98 0.265
Laboratory plates
Full-wave sine loaded 12 5.75 X 1014 -5.30 -0.88 0.465

Strain Fatigue Curve


Laboratory beams
Full-wave sine loaded 22 5.37 X 101 o -2.76 -0.99 0.179
Laboratory beams
Haft-wavesine loaded 14 7.76 X 1011 -2.93 -0.98 0.238
Field beams
Full-wave sine loaded 12 5.62 X 1011 -2.98 -0.99 0.133

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
1000 : , u u u II| ! l I l I ,;i i T I p . I ,If| ; I I I w I l|| 9 I I , I vll
P~
STRESS FATIGUE CURVE M
m
EAST TEKAS CRUSHEOSTONE r
percent asphalt
f u l l wave stne loaded at 70~ G~
l a b o r a t o r y bars 0-fl
02_
M
loc c
I) z
0
r c
o)

t-
:u
m
Nf = 1.74 x 10I I G"4"15 o~
lo
o. r = -0.98

] .. |. i I I i ill I I i I J I * i| I i i I . a*'i I | i i * i l i | | i i i i iii


10 100 1000
NUNBEROF CYCLESTO FAILURE, Nf

FIG. 5 - Stress fatigue curve, East Texas crushed stone, 7 p e r c e n t asphalt, full wave sine
loaded at 70*F, laboratory bars.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
1OO0
STRESS FATIGUE CURVE
EAST TEXAS CRUSHED STONE
7 percent aspha]t
half wave sine loaded at 70~F
laboratory bars
O

loc
v

k
20

Z
z
o
G~
Tlf = 5 . 8 9 x 1013 0 - 5 . 3 2
u.J
_..1 r"
t-
r = -0.99
~ lO
-<
0
z

.-r
r
-I
1 i l JILl|lo0 I Z . i I I i i liO[O0 I i i , i a liJ]O00~ | J L I I 100000
i I lJ I i | [ i i lJ|000()00 (.)
10
0
z
NUMBER OF CYCLES TO FAILURE, Nf t~
m
-.t
m

FIG. 6 - S t r e s s fatigue curve, East Texas crushed stone, 7 percent asphalt, h a l f wave sine
loaded at 70"F, laboratory bars. .4

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
1000 ! , , . I ill| I i i l , gl,~ ! ~ , , i 1,1 l ! ! ! i . ii| i I ! . i ii
M
STRESS FATIGUE CURVE rm
EAST TEXAS CRUSHED STONE
z
7 percent a s p h a l t
full wave s~ne loaded a t 70~ 0
-11
field bars

100 c

z
0
c
ci)

x
-4
c
Nf = 3.57 x 109 r
m
r = -0.98
10
==

1 i I | i I i I I ~ i I I I | Ill| i l I I I l l i i i i i i * i l l [ l i l ~ I * *

10 100 1000 I{)~X) 100C.O0 l___m~___

NUMBER O F C Y C L E S TO FAILURE, Nf

FIG. 7-Stress fa•'gue curve, East Texas crushed stone, 7 p e r c e n t asphalt, full wave sine
loaded at 70~ field bars.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
1000 I ~ *,l,[ , r f , ,l~', I = ! i ! ! i 11| ; ! ; ! w l /

STRESS FATIGUE CURVE


EAST TEXAS CRUSHED STONE
7 percent asph41t
f u l l wave sine loaded at 70~
biaxlaI plates

100 I STD. E R R O R ~ -o ~ ~

z
z
E~

Nf = 5.75 x 1014 r
i-
r-
10 r = -0.88

,<
0
z
>
c/)
0
"..r

r-
,-I
I I * * JJl J l I I , I*l| L t i t ~ ~,=| a i , a * * ll| ! i l l l aJ

100 1000 10000 IO(X~O


O
NUNBEROF CY(~.LESTO FAILURE, Nf z
111
m
-.t

FIG. 8 - S t r e s s fatigue curve, East Texas crushed stone, 7 p e r c e n t asphalt, full v ~ v e sine
loaded at 70"F, biaxial plates.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
- ,=

, m ,ill I I 9 1 i ,,m,[ ) ~ i , ,'~,m I I I | | I I |1| "= ! | I I I I I 0


10000
STRAIN FATIGUE CURVE -4
m
EAST TEXAS CRUSHEDSTONE .-4
7 percent asphalt
6~
full wave sine loaded at 70~ 0
-n
laboratory bars
m_
-4
c
~ IOOC
z
o
lit c
i a:
x
c
~o
m

100 5.37 x lO 10 ~-?.76


Nf =
r = -0.99

I l I l i [[ i I | | l i I |] I I I ! i * | I| 9 I [ | I I Ill I I l I I I |]
10
10 100 1000
NUMBER OF CYCLES TO FAILURE, Nf

FIG. 9 - S t r a i n fatigue curve, East Texas crushed stone, 7 p e r c e n t asphalt, f u l l wave sine
loaded at 70~ laboratory bars.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
I0000 l I I I I I II I I ! I I I I I I I i , | | | 1~1 I .... ; w i | w lel I i i i i ii

STRAIN FATIGUE CURVE


EAST TEXAS CRUSHED STONE
7 percent asphalt
h a l f wave sine loaded at /O~
laboratory bars
+~

~, 1000
U
E

W
1 STD. ERROR
Z
~>
Z
0
0
Nf = 7.76 x 1011 dE-2"93
r"
r"
r = -0.98

-<
0
Z

r-
10 9 9 . i , | ,iS J , * , 1,,.I J j j j .,,,I ! I I i •'ill J ' ' i ''1" j_
0
10 loo 1000 10000 100000 "-"-'E~_=3
0
NUI'e~ER OF CYCLES TO FAILURE, ;if Z
:11
m
-.4
m
FIG. l O-Strain fatigue curve, East Texas crushed stone, 7 percent asphalt, half wave sine
loaded at 70~ laboratory bars. co

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
~0
i ! ! i iii~ ' - a ! ; [ i ii
10000 i ' , , ' ''"I ', , r ' ' '''I ' ' ' ' ' '''I '

STRAIN FATIGUECURVE
m
EAST TEXAS CRUSHEDSTONE
7 percent asphalt
f u l l wave sine loaded at 70~
0
f i e l d bars

M
lO00 C
z
0
c
o'}

X
-I
c
m
100 Nf = 5.62 x 1011 ~-2.98
.

r = -0.99

j i i I_I ~.li i l I. ! i | Ill 1 I j I i * i|i l I I I I ill| l t t i J 9


10
lO 100 lOgO Iggg~ 100000
NLII~ER OF C Y C L E S TO FAILURE, Nf

FIG. 11 -Strain faffgue curve, East Texas crushed stone, 7 percent asphalt, full wave sine
loaded at 70~ field bars.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
1000 ! ~ 1 $ Ill I I i i i r , i 1[ I i ! i i , ,I[ "'~ I I t r i l l [ I I 'I ~'1 ' l

STRESS FATIGUE CURVES


EAST TEXAS CRUSHED STONE

~ /-----Kallas and Puztnauskas 7 percent asphalt


" ~ j test temperature 70~

~ ~ ~ - - ~ j m - - b t a x i a l plates
v
IO0
b

Z
fullwavesine loaded L l y b a r s~~ / a b~o r a t o~r
Z
0
0
f u l l wave s i n e loaded / latoratory b a ~ f-
lO
half wavesine loaded
-<
0
Z

"o
-r

r-
i l i Ill| l * i * I ,ill i I i i i i.i| i i i P i Jill I i i i,,l l . i
1 O0 1000 10000 100000 1 @-=-=-=_-m
llO
C)
NUMBER OF CYCLES TO F A I L U R E , Nf 0
Z

m
-4
FIG. 12-Stress fatigue curves for asphaltic concrete. m

(.0
r

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
I I , . I II i I i I 1 I I II I I I , . I I, I] , I I I I I I11 I I I ] I I I
I0000
STRAIN FATIGUE CURVES M
Ill
EAST TEXAS CRUSHED STONE
laboratory bars 7 percent asphalt
~...~/ half w a v e sine loaded Z
test temperature 70~ 0
0
"11
=_
M
~ ~ allas and Puzinauskas t-
I000
U
Z
qlt
0
I ~ dwabvaers
sine loaded ~ ' ~ ' ~ ~ ~ ~ b ~ ~ a
x
e'-

/~1~ wavesine ~oaded m


r.~
~ 1o0

z i t .J,li i i i i i l i l i l i i i I i i i| l I i J i i I I| i ~ i * I i . i
I0
10 100 IO00 lo000 1OO0OO 101~,0
NI~BER OF CYCLES TO FAILURE, Nf

FIG. 13-Strain f a t i g u e c u r v e s f o r a s p h a l t i c c o n c r e t e .

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
IRWIN AND GALLAWAY ON ASPHALTIC CONCRETE 35

The four stress fatigue curves are shown together in Fig. 12, and the three
strain fatigue curves are shown in Fig. 13. Also shown in Figs. 12 and 13 are the
test results obtained by Kallas and Puzinauskas [5] for a material having similar
gradation and asphalt content. KaUas and Puzinauskas used a haversine loading
pattern with a 0.1 s loading time followed by a 0.4 s rest period. In the half wave
sine loading pattern used in the tests reported herein a 0.3 s loading time was
followed by an equal rest period. Reasonably good agreement was obtained
between the Kallas and Puzinauskas strain fatigue curve and the half wave sine
loading strain fatigue curve (Fig. 13), especially when it is recognized that the
Kallas and Puzinauskas curve should be corrected downward by 16 percent to
account for shear deformation in the test specimens.

Discussion

Full Nave VersusHalf WaveSine Loading


In the extreme fiber of the full wave sine loaded specimens there was
complete stress reversal during each load cycle, whereas in the half wave sine
loaded specimens there was no stress reversal. A comparison of the test results
obtained by the two methods for the laboratory-compacted beams permits an
assessment of the influence of stress reversal on fatigue behavior. Because the
rate of strain would be the same for equal strain amplitudes and loading
frequencies in beams subjected to the two loading patterns, a comparison
between half wave sine loading and full wave sine loading is preferable to a
comparison between block loading and full wave sine loading.
In general, both the stress and the strain fatigue curves for the laboratory-
compacted beams show that the specimens subjected to complete stress reversal
(full wave sine loading) had shorter fatigue lives than the specimens without
stress reversal (half wave sine loading) (see Figs. 12 and 13). Although they had
different slopes (Table 4), the stress fatigue curves crossed for the full wave and
half wave sine loaded beam test series (Fig. 12) and their error bands overlapped.
Similarly, the two strain fatigue curves were nearly parallel, and they fell in close
proximity to each other (Fig. 13).
In order to ascertain whether or not the fatigue curves for the two test
methods were statistically different, it was hypothesized that the slopes and
intercepts were equal, and the hypotheses were tested using the t-statistic [19].
It was determined that at the 0.05 significance level both the two stress fatigue
curves and the two strain fatigue curves were different.

Uniaxial VersusBiaxial Tests


A statistical comparison of the stress fatigue curves for the full wave sine
loaded biaxial plates and the laboratory-compacted, full wave sine loaded beam
test series was made. It was determined that at the 0.05 significance level the
curves were different.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
36 TESTING OF BITUMINOUS MIXTURES

From Eq 6 it can be seen that the specimen deflection did not enter into the
computation of stress in the circular plates. Thus, the precision of measurement
of plate deflection did not affect the precision of the computed stresses. It may
be noted from Table 4, however, that the regression for the biaxial plate tests
had the lowest correlation coefficient and the highest standard error among the
four stress fatigue curves. It is believed that the somewhat subjective method of
determination of failure used for the plate tests contributed to the dispersion of
the data on the stress fatigue curve.
The precision of the deflection data for the biaxial plate tests could have been
improved by using more sensitive techniques for the measurement of specimen
deflection. However since the biaxial specimens do not fail abruptly, as by
rupture, the problem of precise definition of failure for the biaxial test method
would remain. Although the state of stress in the biaxial plate test more closely
represents the state of stress in an actual pavement than does the uniaxial beam
test, the greater ability to define failure of the beam specimens would suggest
that the uniaxial beam test method should be preferred to the biaxial plate test
method.

Laboratory VersusField-CompactedSpecimens
Both the asphalt and aggregate, as well as the asphaltic mixture proportions,
were found to be nearly identical between the laboratory and field mixed
materials. The principal difference between the laboratory mixed and compacted
specimens and the field mixed and compacted specimens was in the density and
air void content. Other differences between the laboratory and field specimens
included using the as-molded surface versus a sawn surface on the field
specimens and loading perpendicular to the direction of compaction versus
loading in-line with compaction for the field specimens. While the full wave sine
loaded laboratory beams had an average density of 126.1 Ib/ft 3 (2021 kg/m 3)
and an air void content of 15.0 percent,the full wave sine loaded field beams had
an average density 120.5 lb/ft a (1932 kg/m 3) and an air void content of 18.8
percent. The higher air void content of the field beams resulted in a shorter
fatigue life for a given stress level (Fig. 12) and a longer fatigue life for a given
strain level (Fig. 13).
From the results of work published by Pell and Taylor [4] it would be
expected that where two asphaltic mixtures are otherwise the same except for
their air void content, the slopes of their fatigue curves would be the same but
the intercepts would differ. The fatigue curves shown for the full wave sine
loaded laboratory and field beams are very nearly parallel (Figs. 12 and 13).
Nevertheless, statistical hypotheses that the slopes and intercepts of the two
fatigue curves were equal were both rejected at the 0.05 significance level. It is
bdieved, however, that had the laboratory and field specimens been more nearly
equal in air void content, the hypotheses of equality of the fatigue curves could
have been accepted.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
IRWIN AND GALLAWAY ON ASPHALTIC CONCRETE 37

Number o/SpecimensRequiredfor Fatigue Characterization


Kallas and Puzinauskas [5] determined that eight or nine beam specimens
having 3 in. widths and depths gave correlation coefficients for the fatigue curves
comparable to those obtained using 19 or more specimens having 189in. widths
and depths. Their analysis led them to conclude that eight or nine 3-in.-
specimens were sufficient for a fatigue characterization.
In the study reported herein, 22 beam specimens having 3 in. widths and
depths were tested in full wave sine loading. In order to determine whether a
lesser number of specimens would yield an equivalent fatigue characterization,
the data from the 22 specimens were subdivided into eleven subsets. Three
groups, each having subsets of 9, 12, and 15 specimens were formed using a table
of random numbers. In addition, two subsets of eleven specimens were formed
by dividing the data by odd and even specimen numbers.
Regression analyses were conducted for each of the subsets, and the results
are given together with the regression data for the full set of 22 specimens in
Tables 5 and 6. Statistical comparisons with the slope and intercept of the full
set were made for each of the subsets. Those slopes or intercepts which were not
found to be equal to the full set at the 0.05 significance level are indicated by
footnote in Tables 5 and 6. Only two of the eleven subsets were found to be
equal to the full set stress fatigue curve (Table 5), and only three of the subsets
were found to be equal to the full set strain fatigue curve. Furthermore, only
one of the subsets was equal to the full set in terms of both stress and strain.
While most of the subsets were not found to be equal to the full set, the
question remains how different is "different". The allowable stress and strain for
one million load repetitions was calculated from the regression curves for the full
set and each subset, and the results are shown in the last column of Tables 5 and
6. The allowable stress varied from a low of 16.2 psi (112 kN/m 2) to ahigh of
20.7 psi (143 kN/m2). The allowable strain varied from a low of 45.0
microstrain to a high of 57.2 microstrain.
While it was anticipated that the largest subsets of 15 specimens would most
accurately predict the full set allowable stress or strain, this was not found to be
the case. The smaller subsets were closest to the full set allowable stress or strain
about as often as were the medium and larger subsets.
The interpretation of this analysis in terms of the minimum number of
specimens required for a fatigue characterization is not entirely clear. The results
of the analysis tend to suggest (but do not prove) that a small number of test
specimens (on the order of ten) would be adequate for a fatigue characterization
of asphaltic concrete when the laboratory methods used in this investigation are
employed. This recommendation is provisional, and a full scale determination of
repeatability and reproducibility should be made in order to properly assess the
minimum number of specimens required for fatigue characterization.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
m
o)

TABLE 5-Stress fatigue curve regression parameters. Sensitivity studies for laboratory beams, full wave sine loaded.

O
"11
Failure ttl
Number of Slope Correlation Standard Criteria for -d
Group Specimens Intercept K n Coefficient Error 106 Cycles, psi C
n
z
Full set 22 1.74 X 1011 -4.15 -0.98 0.219 18.31 0
c
o~
Random Group 1 9 0.445 a X 101 ] _3.84 b -0.98 0.231 16.24
m
12 0.457 a X 1011 _3.83 b -0.98 0.224 16.47 x
15 0.752 a X 1011 _3.95 b -0.98 0.217 17.16 -f
c
~o
Random Group 2 9 1.23 ~X 10 II -4.08 -0.98 0.164 17.68 m
12 1.75 X 1011 -4.15 -0.99 0.157 18.34
15 0.538a X I011 _3.83 b -0.99 0.022 17.19

Random Group 3 9 26.6a X I011 _4.92 b -0.97 0.206 20.22


12 22.1a X I011 _4.88b -0.97 0.211 19.96
15 21.4a X 1011 _4.82b -0.96 0.243 20.58

Odd numbered specimens 11 0.372 a X 1011 _3.76 b -0.99 0.177 16.43

Even numbered specimens 11 19.7 a X 1011 _4.78 b -0.98 0.248 20.74

a Statistically different from the full set intercept at the 0.05 level or lower.
b Statistically different from the full set slope at the 0.05 level or lower.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
TABLE 6-Strain fatigue curve regression parameters. Sensitivity studies for laboratory beams, full wave sine loaded.

Failure
Criteria for
Number of Slope Correlation Standard 10 ~ Cycles
Group Specimens Intercept K n Coefficient Error Microstrain

Full set 22 5.37 X I0 I0 -2.76 -0.99 0.179 51.73


z
Random Group 1 9 10.7 X I0 I~ -3.02 b -0.98 0.232 46.28
12 17.3a X I0 I~ -2.98/, -0.98 0.219 57.24 z
15 7.26 X I0 I~ -2.82 -0.98 0.207 52.93 o
o
Random Group 2 9 3.45 X 10 l~ -2.65 -0.99 0.125 51.57 f-
12 1.47 a X 101 o -2.50 b -0.99 0.114 46.44
15 I0.9 a X 1010 -2.89 b -0.99 0.193 55.34
.<
Random Group 3 9 2.51 a X 101~ -2.59 b -1.00 0.089 49.97 0
12 1.26 a X 101~ -2.48/' -0.99 0.127 45.02 z
15 1.21 a X 101~ -2.47 b -0.99 0.115 44.97
-o
-l-
Odd numbered specimens 11 9.06 X 101~ -2.87 -0.98 0.241 53.36
f-
Even numbered specimens 11 3.01 a X 1010 -2.64 b -1.00 0.108 49.71

0
z
a Statistically different from the full set intercept at the 0.05 level or lower. r
.11
b Statistically different from the full set slope at the 0.05 level or lower. m
.-I
m

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
40 TESTING OF BITUMINOUS MIXTURES

Conclusions and Recommendations


Based upon the work on fatigue characterization of asphaltic concrete
materials reported herein the following conclusions and recommendations
appear warranted.
1. From a comparison of half wave sine loading with full wave sine loading it
is concluded that test methods which incorporate full stress reversal in the
specimen lead to different fatigue characterizations than are obtained by test
methods in which stress reversal is not present.
2. An evaluation of fatigue characterizations obtained by using full stress
reversal in laboratory and field prepared beam specimens showed that the fatigue
curves were not statistically equal. However, differences in the air void contents
of the two sets of specimens may have adversely affected the comparison. It is
believed that had the laboratory and field specimens been more nearly equal in
air void content, their fatigue curves would have been more nearly the same.
3. It is recommended that the beam test method be preferred to the biaxial
test method due to thegreater ability to precisely define the number of cycles to
failure for controlled stress tests using the beam test method.
4. It is recommended that a minimum of ten beam specimens having 3 in.
widths and depths be used in the fatigue characterization of asphaltic concrete.
This recommendation is provisional and a full scale determination of repeatabi-
lity and reproducibility should be made in order to properly assess the minimum
number of specimens required for fatigue characterization.
5. Finally, it is concluded that factors involved in the laboratory test method
used in fatigue characterization of asphaltic concrete have an influence on the
resultant slope and intercept of the fatigue curve. These factors include the
degree of stress reversal and possibly also the type of stress field to which the
specimens are subjected. In this study only two stress fields were studied,
namely axisymmetric, center loaded biaxial, and third-point loaded uniaxial. It is
believed that other testing geometries such as cantilever and center loaded beams
would also have an influence on the slope and intercept of the fatigue curves.

APPENDIX I

Deflection of a Third-Point Loaded Beam, Considering Shear Deformation


To determine center deflection, a unit virtual (imaginary) force is applied at
the center of the beam. This will setup virtual moments, m, and virtual shear
forces, v (Fig. 14).
Next, the real forces axe applied, which for this case are P/2 at the third
points of the beam. The real forces will generate real moments, M, and real shear
forces, V (Fig. 15).

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
IRWIN AND GALLAWAY ON ASPHALTIC CONCRETE 41

LL.-x I ]
1,2,': .,2 II/2

v = +I/2 [
v : -I/2 1

f.~
_ x ~
~ 2 (x-~) :L_!

FIG. 14-Shear and moments due to virtual [orces.

~ P/2

[L_x I
I--,,3~.2
v : P/2 I v =o
v = P/2 I
PL
M : -g--
----~(x-~-~

FIG. 15-Shear and moments due to real [orces, third-point loaded beam.

Consider an element of the beam d x units long, which initially has virtual
moments, m, acting on it (Fig. 16a). Due to the real moments, M, plane sections
are rotated through an angle dO (Fig. 16b) which, when shear deformation is
neglected, is given by

dO - M d x
E1

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions aut
42 TESTING OF BITUMINOUS MIXTURES

and the internal work done by the virtual moments, m, due to the action of the
real forces is

du = redO - m M dx (9)
El

ml lm

d•

M
/
/, do

FIG. 16-(a) Virtual moments and (b ) real moments.

To include the work due to shearing we consider an element from the beam d x
long, d y wide, and dz tall (Fig. 17).
If the virtual shear stress on the element is r*, then the work done by z* due
to the shearing action of the real forces, r, (Fig. 18) is

du = r " dx d y X "y d z

force distance

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
IRWIN AND GALLAWAY ON ASPHALTIC CONCRETE 43

~X

Y
FIG. 17-Differential element of beam.

,T,A.

//_i
.-dx-~
FIG. 18-Element in shear.

Noting that by Hooke's law

r=TG
we obtain

r*Xr
du - - - clx X dy X dz (10)
G

The external work done by the virtual force due to the application of the real
forces is

IXwo
where Wo is the center deflection of the beam due to the real forces. The
external work done by the virtual force must equal the internal work done by
the virtual moments and virtual shear stress, thus

1 X Wo =
fo L mM
E1
dx +
fofoh b L
G
dxX dyXdz (11)

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
44 TESTING OF BITUMINOUS MIXTURES

where

G - E (12)
2(1+ u)
In a rectangular beam, shear stress is distributed parabolically according to

r = (v) ((h)5 _ zs)


2I 2

From the shear diagrams, over the distance x = 0 to x = L[3 we obtain

?. = 14i ((-2-h)5- z 5) (13)

r = 4--I-
e ( (2-h)5 _ z 5) (14)

and over the distance x = L]3 to x = L[2

~'=0

Also, over the distance x = 0 to x = L/3

m - x (15)
2

M= Px (16)
2

and over the distance x = L[3 to x = L/2

X
m ~ m

M = PL (17)
6

Noting that the loading is symmetric with respect to the center of the beam and
that the shear stress is distributed symmetrically with respect to the top and
b o t t o m of the beam, we can perform the integrations, substituting Eqs 13-17 in
Eq 11

Wo = ~ f L/3 x X .Px X dx
. +. 2 f.. L / 2 x • PL X dx
EI J 2 2 EI J 2 6
o L/3

+ ~bf L/3 dx • 2 ] Fh/214t -- X e


--4I ( ( _~
)2 _ z2)~clz
0 0

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
IRWIN AND GALLAWAY ON ASPHALTIC CONCRETE 45

from which we obtain

23 PL 3 PL
WO --
+
1296 E1 5 GA (18)

where A = bh. The first term in Eq 18 is the customary strength of materials


solution for third-point loaded beams where shear deformation is neglected. The
second term in Eq 18 reflects the influence of shear deformation.
Substituting Eq 12 in Eq 18 and collecting terms we obtain

23 PL 3 216 h 2 ( 1 + #)
Wo - (1 § )
1296 E1 115 L 2 (19)

It can be seen from Eq 19 that when the length of the beam, L, is large with
respect to its depth, h, the term in brackets approaches unity, yielding the
customary strength of materials solution.
For the case where L = 12 in., h = 3 in., and/~ = 0.35, Eq 19 becomes

Wo _ 23 PL 3 (1.158)
1296 E1

or, in a beam of this size which is subjected to third-point loading, shear


deformation adds ~ 16 percent to the deflection due to bendin~

References
[1] Epps, J.A. and Monismith, C.L. in Proceedings, Association of Asphalt Paving
Technologists, Vol. 38, I969, pp. 423-464.
[2] Epps, J.A. and Monismith, C. L. in Fatigue of Compacted Bituminous Aggregate
Mixtures, ASTM STP 508, American Society for Testing and Materials, 1972, pp.
19-45.
[3] PeU, P. S. in Proceedings, Second International Conference on the Structural Design
of Asphalt Pavements, 1967, pp. 577-593.
[4] PeU, P. S. and Taylor, I. F. in Proceedings, Association of Asphalt Paving Technolo-
gists, Vol. 38, I969, pp. 371-422.
[5] Kallas, B. F. and Puzinauskas, V. P. in Fatigue of Compacted Bituminous Aggregate
Mixtures, ASTM STP 508, American Society for Testing and Materials, 1972, pp.
47-65.
[6] Jimenez, R. A. in Fatigue of Compacted Bituminous Aggregate Mixtures, ASTM STP
508, American Society for Testing and Materials, 1972, pp. 3-17.
[7] Deacon, J.A., "Fatigue of Asphalt Concrete," Graduate Report, The Institute for
Transportation and Traffic Engineering, University of California, Berkeley, 1965.
[8] Bazin, P. and Saunier, J. B. in Proceedings, Second International Conference on the
Structural Design of Asphalt Pavements, 1967, pp. 553-569.
[9] Ramsamooj, D.V., Majidzadeh, K., and Kauffman, E.M. in Proceedings, Third
International Conference on the Structural Design of Asphalt Pavements, 1972, pp.
692-704.
[10] Jimenez, R.A. and Gallaway, B.M. in Proceedings, Association of Asphalt Paving
Technologists, Vol. 31, 1962, pp. 477-506.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
46 TESTING OF BITUMINOUS MIXTURES

[11] Annual Book of ASTM Standards, Part 11, American Society for Testing and
Materials, April 1972.
[12] MaterialsManual, Vol. 1, State of California, Department of Public Works, Division
of Highways, Sacramento, Calif., 1963.
[13] Manualof TestingProcedures, Vol. 1, Texas Highway Dept., Austin, Tex., 1962.
[14] Popov, E. P., Mechanics of Materials, Prentice-Hall, Edgewood Cliffs, N.J., 1952, pp.
391-414.
[15] Roaxk, R. J., Formulas for Stressand Strain, 4th ed., McGraw-Hill, New York, 1965,
pp. 217-218.
[16] Layman, A.H. and Philippi, S.W. in Proceedings, Association of Asphalt Paving
Technologists, Vol. 38, 1969, pp. 684-705.
[17] Timoshenko, S. and Goodier, J. N., Theory of Elasticity, 2nd ed., McGraw-Hill, New
York, 1951, p. 7.
[18] Scrivnex, F.H. and Moore, W. M., "Evaluation of the Stiffness of Individual Layers
on a Specially Designed Pavement Facility from Surface Deflections," Research
Report 32-8, Texas Transportation Institute, Texas A&M University, College Station,
Tex., June 1966, pp. 55-72.
[19] Miller, I. and Freund, J. E., Probability and Statistics for Engineers, Prentice-Hall,
Edgewood Cliffs, N.J., 1965, pp. 231-235.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
R. L. Terrel, I L S. A w a d , 1 and L. R. Foss ~

Techniques for Characterizing


Bituminous Materials Using a
Versatile Triaxial Testing System

REFERENCE: Terrel, R.L., Awad, I. S., and Foss, L. R., "Techniques for
Characterizing Bituminous Materials Using a Versatile Triaxlal Testing Sys-
tem," Fatigueand Dynamic Testing of Bituminous Mixtures, ASTM STP 561,
American Society for Testing and Materials, 1974, pp. 47-66.
ABSTRACT: This paper is part of an experimental investigation of the resili-
ent properties and the stress-strain relationship of asphalt treated base mater-
ial. It describes the triaxial testing system which was designed for this purpose
and the different tests employed in the material stress-strain characterization.
KEY WORDS: fatigue tests, triaxial tests, bituminous, viscoelasticity, resili-
ence, dynamic modulus of elasticity, Poisson's ratio, flexible pavements

In the search for a "rational" design method of asphalt pavements, much of


the work reported in recent years emphasized the importance of laboratory
testing. Research into the response of asphalt pavement materials is now
extensive as well as intricate.
As newer theoretical techniques have become available, it has become
increasingly evident that the theoretical aspects of pavement design have
progressed further than the conventional test ability to provide suitable
parameters to complete the design. Parameters such as resilient modulus, creep
compliance, complex modulus, etc. are needed to describe the elastic as well as
the viscoelastic properties of the asphalt mixture.
A number of experimental methods have been adopted to obtain a
relationship between stress and strain for asphalt mixes. These methods include
the stiffness modulus method computed from beam or slab bending tests, and
the dynamic modulus and resilient modulus methods computed from axial or
triaxial testing. Due to the nonlinear strain distribution across the beam or slab
pavement section, stiffness is only a measure of the overall flexural properties of
that section. The dynamic and the resilient modulii, on the other hand, ade-
quately describe the properties at or in the neighborhood of a point within the
pavement section.
The dynamic modulus test is obtained by applying an axial stress that is

1Associate professor of civil engineering and lecturer in civil engineering, respectively,


University of Washington, Seattle, Wash. 98195.
2Engineering aid, Research Engineering Division, Boeing Company, Seattle, Wash.
98124.
47
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Copyright9 1974 bybyASTM International
Downloaded/printed www.astm.org
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
48 TESTING OF BITUMINOUS MIXTURES

oscillating sinusoidally with time and measuring the corresponding oscillating


strain. This test furnishes valuable information on the viscoelastic properties of
the specimen tested.
Although the concept of the resilient modulus was introduced for subgrade
materials, it was adopted to characterize asphalt mixes as well. In this test, the
resilient modulus is def'med as the ratio of the applied axial stress to the
corresponding recoverable axial strain. Since asphalt mixtures show some degree
of viscoelastic behavior, it should be noted that the material response and the
recovered strains are time dependent. The computed value of the modulus,
therefore, becomes dependent upon the rest period between individual stress
pulses. This dependency is quite insignificant at low temperatures and for short
stress durations. At high temperatures, however, this dependency becomes
considerably more apparent and, thus, must be included as a parameter in the
material characterization.
Asphalt treated base (ATB) material, as used by the state of Washington, is a
dense-graded, crushed reek or natural gravel mixed with paving grade asphalt.
Generally, the mixtures contain relatively low asphalt percentages (2.5 to 4.5),
while the required Hveem stabilometer value of 25 approaches those of conven-
tional mixtures.
The objective of this paper is to describe a triaxial test system and to report
the techniques utilized in the characterization of ATB.

Triaxial Test System


Figure 1 illustrates the main features of the triaxial chamber designed to
carry out the proposed investigation. Unlike conventional triaxial cells, the
designed cell provides continuous circulation of silicone oil at a constant
temperature and under the required pressure. This can be accomplished by
having the entire temperature control system under the same pressure. For this
purpose, therefore, the heating tank and the chamber are equipped with
bdloframs connected to two pressure reservoirs which provide two types of
pressure: static and pulsating.
A special type bellofram is provided on the bottom of the chamber to act as a
seal and a separation between the triaxial chamber and a subchamber. This
subchamber can be either vented or connected to the same pressure as the
chamber by a shut-off valve. By doing this, it is possible to separate or isolate the
axial and radial stresses.

Load Application System


As traffic moves on a highway pavement, the axial and the radial stresses (ol
and o3) change for a period of time such that each wheel pass can be considered
as a stress pulse. The magnitude, shape, and duration of these pulses vary with
the wheel load, its speed, and the depth in the pavement at which the stress is
considered.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
TERREL ET AL ON T R I A X I A L TESTING SYSTEM 49
12
10
"14 1t

t
19
(1) 4 x 8 Inch Sample (12) Chamber Pressure (,,~)
(2) Strain Gages (13) Plastic Cylinder Lifting Device
(3) Controlling Load Cell (14) Aluminum Frame With Three Openings
(4) Porous Stone (15) Plastic Cylinder
(5) Rubber Membrane (16) Drainage Outlet
(6) O - Rings (17) Pressure (,r~) Bottom Inlet
(7) Temperature Sensors (18) Rectilinear Ball Bearings
(8) Pressure Cell (19) Axial Loads (MTS)
(9) Oil Circulation Outlet (20) Special Bellotram
(10) Outlet of Electric Wiring (21) 2" Aluminum Platform
(11) Oil Circulation Inlet

FIG. 1-Triaxial chamber.

Little information is available concerning the actual shape of the axial stress
pulse and its variation with depth. Most researchers used the data obtained from
full-scale test tracks (Fig. 2). In order to select a pulse shape, art experimental
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
01
0

Direction Of Traffic - ~
Stockton Test Track m
80
Seed et al (1955)
Seed & Feed (1959) 60
Pagen & Jagannath ffl
O3
(1968) '~ / \ 0
p.s.i.
Le Tirant & Sarda W
(1965) (Sine Wave) 20 / N
.-t
C Larew & Leonards (A) c
0 ~J
(1962) Time
z
Yandell (1965) 0
Depth = 27 In.
U~
Stockton Test Track (After BIBL 32)
X
(/) .-I
80 c
m

60 m
o
,m \ o- z
40 /\
> p.s.i.
X 20
0 --'" (B)
Time
Load -- 4550 Lb.
d~ Ii I
(c) Speed -- 11.3 M.P.H.
Time Depth -, 25.3 In.
Whiffi._.._~n(After BIBL 13)

FIG. 2-Vertical stress f u n c t i o n s used b y various investigators [after shackel).

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
TERREL ET AL ON T R I A X I A L TESTING SYSTEM 51

investigation was carried out which included the full sine, the square, and the
triangular stresses. In this uniaxial stress experiment, a stress al of 30 psi was
applied in the sinusoidal and triangular modes for 0.1, 1.0, and 10 s. The
specimen was then subjected to combinations of percentages of both the stress
duration and magnitude in the square stress mode. The result of this investiga-
tion is shown in Fig. 3, and the following are the conclusions:
1. There is no significant difference in the magnitude of the total or the
resilient strains between the triangular or the sinusoidal stress pulses. 3
2. An equivalent square pulse can be obtained by (a) applying the same
stress, but, for a duration of 33 percent of the equivalent sinusoidal, or (b)
applying 66 percent of the stress with the same duration as the equivalent
sinusoidal.
It was also concluded that a square vertical stress pulse is a reasonable

1000

A
c
"~ 100

2
~' Coarse Gradation
~ lo - - ~" ~ Temperature = 70*
< Asphalt Content ~ - 2 . 5 %
Square
-- ~ - - Sine (30 psi max.)
~ Triangular (30 psi max.)

1 . 1 I 1 II 1 t ! 11 1 [ i i 1 ~ I [ I
0.01 0.1 1.0 10 100

Stress Duration (sec.)

FIG. 3-Axial strains resulting from different stress pulse shapes and durations.

approximation of the actual conditions within a pavement layer. This approxi-


marion simplifies the model fitting and the subsequent analysis of the data)

aTerrel, R.L. and Awad, I.S., "Resilient Behavior of Asphalt Treated Base Course
Materials," Washington State Highway Department research Program,Report No. 6.1, Aug.
1972.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
52 TESTING OF BITUMINOUS MIXTURES

The axial load was controlled by a load cell which was fixed to the upper side
of the test specimen inside the triaxial cell. By doing this, problems resulting
from friction were essentially eliminated. Moreover, in designing the load cell, it
was important to isolate the effects of the chamber pressure and the axial stress;
in other words, to have a load cell that is insensitive to chamber pressure varia-
tions. Figure 4 shows the load cell used in this experimental study.
Figure 5 shows the system designed to pulse the chamber pressure. In this
system, the pressure was kept constant in two tanks (a3 andAa3). The pressure
o3 was equivalent to the in situ radial pressure and alternatively applied with

StrainGageWires

TriaxialFrame

i~ ' StrainGages

k 3

q
= i

I t:StrainGages
1111C Specimen
Strain
GageWires

Section-A-A
I[ Drainage 1I

FIG. 4 - L o a d cell used inside the triaxial chamber.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
TERREL ET AL ON TRIAXIAL TESTING SYSTEM 53
Ao3 which corresponds to the total radial pre~ure during the load pulse. A
special electronic circuit (Fig. 6) was designed to syc~onize the axial and
chamber pressure pulses (ol and Ao3).
Temperature ControlSystem
There has been an increasing awareness that the test temperature plays the

Main PressureSupply
~ L (~ Iofram
Triaxial
Cell
oad D%
ell

~3Control Unit
Control Unit (MTS)
t_

FIG.5-System used in pulsing the chamber press~e,

~o I~ I e~O ! ]
ce,, I
Triaxial [
I~'~ e'"l Sequence
Timer
1-22sec.

i TimeDelay ~ Generator
Fu.ction1
[
I
Air
Valve Hydraulic1
Servo
[ Or;er ] Controller

I Air
I Pra.ure I I "xla' I
Load

l
FIG.6-Sychronization of axial and chamber pressures.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
54 TESTING OF BITUMINOUS MIXTURES

major role in the behavior of asphalt treated materials. It is, therefore, highly
important that the testing system should have the capability of controlling and
maintaining a constant temperature throughout the experiment.
Figure 7 diagrammatically shows the main features of the closed-loop
temperature control system. In this system, the refrigerator and the heating unit
work opposite to each other to obtain a constant temperature in the tank. The
silicone oil at this temperature was then circulated through a centrifugal pump
to the triaxial chamber where the temperature was measured by three sensors
located at different elevations. In order to obtain the required temperature, the
heating coil or the refrigerator were activated by a thermostat to compensate for
any heat losses or gains.

Measurement and Monitoring Systems


Strains were measured by two axial and two circumferential strain gages.
SR-4, A-1 wire gages 0.75 in. in length and MM EA-13-19 CDK foil gages 2.0 in.
long produced satisfactory preliminary results. The 2.0.in.-long gages were used
since the maximum aggregate size was 3/4 in. AE-10 adhesive was used to obtain
a smooth base for the strain gage. The M-Bond 200 was used in the strain gage
installation.
The axial load was controlled by a 4-in.-diameter load cell inside the triaxial
chamber and continuously checked against a special load cell in the loading
frame.
The triaxial chamber pressure was measured by a pressure transducer located
at the bottom of the cell. The pressure at this point was slightly higher than that
at the center of the specimen due to the hydrostatic pressure of the oil.
During each test, continuous recordings of data were made using an ultra-
violet recorder. The desired information was then read from this recording using
a magnifying lense and an accurate scale.

Material
A total of 27 specimens representing nine different mixtures were tested
(three asphalt content, 2.5, 3.5, and 4.5 percent and three aggregate gradations).
These specimens were fabricated utilizing the Triaxial Institute kneading com-
pactor. The compaction procedure was essentially the same as the Hveem meth-
od recommended by The Asphalt Institute. The 4-in.-diameter and 8-in.-high
specimens were compacted in three lifts and were stored in sealed plastic bags at
40~ until tested. 4
Natural pit-run aggregate was used in the form of three gradations that
envelop the Washington State Highway specifications for ATB. The asphalt
cement used was 85 to 100 penetration and was supplied by the Chevron As-
phalt Company.

4Awad, I. S., Rutz, F. R., and Terrel, R. L., "Characterization of the Stress-StrainRela-
tionship for Asphalt Treated Base Materials," Proceedings, Association of Asphalt Paving
Technologists, Vol. 42, 1973.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
TERREL ET AL ON TRIAXIAL TESTING SYSTEM 55

L___J L Triaxial
Cell

Temperature
Sensors
./
Plexiglass
Cylinder

Silic~
Oil Feedback
Th~rmo~ounl~_

Circulation
Valve

Agitator

Polyurethane ---.
Foam
Insulation

_C)-"
v

Circulation
Pump

Heating Coil

FIG. 7-Temperature control system.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
56 TESTING OF BITUMINOUS MIXTURES

Tests - Procedure and Results


Conventional tests included the standard density, air voids, and the Hveem
stability and cohesion. Stability and cohesion, however, were performed at
temperatures that correspond to the temperatures under which the triaxial tests
were performed, rather than the conventional 140~
The following tests were carried out in the triaxial test system:
1. Repeated load (both axial and radial).
2. Creep (constant axial compressive stress).
3. Dynamic (sinusoidal axial).

Repeated Load Tests


One basic question in the characterization of engineering materials is the
material nonlinearity. In investigating this phenomenon, the specimen should be
subjected to stress conditions simulating those of the field. Most of the tests
which utilize the tdaxial cell are performed under constant chamber pressure.

140

120 9 o~-" 0 psi


O a0=5
A L~ ~o - - 10

c: 100 o~=15
i_.
[] 0~
o
'-
0
80

E
C
L_
go~ / 70 ~

<x 40

20 Coarse Gradation
Asphalt Content - - 2.5%
Stress Duration - - 0.1 sec.

10 20 30 40 50
Axial Stress (psi)
FIG. 8-Stress-strain states under sustained confining pressures.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
TERREL ET AL ON TRIAXIAL TESTING SYSTEM 57

Under these conditions, the response of the material is dependent on the confin-
ing pressure. For this reason, two separate series of tests were conducted: (a)
utilizing the constant confining pressure, and (b) subjecting the same specimen
to simultaneous axial and confining stresses.

140

9 ~3 - - 0 psi
0 ~3=5
120
A ,~3-" 10
* ~3~----15
A
U)
r 100 90 ~
,m
m
!,-

o
'- 80
, uo

E
70*
.u
m
Im
60
U)
.~
x 4O

20 Coarse Gradation
Asphalt Content = 2.5%
Stress Duration = 0.1 sec.

10 20 30 40 50

Axial Stress (psi)

FIG. 9-Stress-strain under cyclic axial and radial stresses.

In linear elasticity, the radial and axial strains, er and ez can be written in the
form

er = BI or + B 2 0 z
(1)
ez = B3 Or + B4 oz

where B1, B2, B3, and B4 are constants which can be obtained from linearly
fitting the experimental data.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
58 TESTING OF BITUMINOUS MIXTURES

Since most of the available elastic solutions utilize the isotropic conditions,
an approximate value for the modulus and Poisson's ratio are required. These
values can be obtained by

2 2 (B2 + B3)
E " - - + --
BI +B4 3 [(B 1 + B4) (B2 + B3) ] (BI + B4)
3

~, 2 (B2 + B3)
v 3 (B1 + B4) (B2 + B3) (2)

ff the strains utilized in Eqs 1 and 2 are the recoverable strains, then E is
usually referred to as the resilient modulus. The following empirical relationships
were obtained using the least square curve fitting technique for 0.1 and 1.0 axial
stress durations

log MR (0.1 s) = 6.8203 - 0.0000029944 (a[c) 2 (T) 2 - 0.00011927 (7)


- 1.414 (% air)/(/) (3)

log MR (1.0 s) = 6.98798 - 0.001765 (a/c) (7') - 0.01165 (7)


- 0.0022964 (% air)2 - 0.973085 (% air) 2/(7) 2 (4)

where
a/c = asphalt content, percent,
O
T = temperature, F, and
% air = air void, percent.

The following correlations were also obtained between the resilient modulus
and the Hveern stability determined at the same temperature as that of the
resilient modulus rather than 140~

log M R (0.1 s) = (5196240 + 1.697 S a) X 10-6 (Fig. 10) (5)

log M R (1.0 s) = (4457240 + 214.13 S 2) • 10-6 (6)


Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
TERREL ET AL ON TRIAXIAL TESTING SYSTEM 59

5x10 6
O Fine Gradation ~////
A Medium Gradation Oj[~
[] Coarse Gradation ///~

10+ ~o ~
ul

[]
10 5
Stress Ouralion = 0.1 sec.
LOG]oMD = [5196240 q- 1.697S3] x 10-6

I I I I I I I I I I I I I I I
20 30 40 50 60 70 80 90 1O0

Stability Value

FIG. l O-Correlation between the stability value and the resilient modulus.

where S = Hveem stability determined at same temperature as that of the


resilient modulus.
Two conclusions related to the resilient modulus were drawn:
1. Within practical ranges of temperature and stresses, the resilient strains
are linear indicating no material nonlinearity with respect to axial stresses.
2. The value of the resilient modulus when computed under constant
confining pressures is found dependent on the magnitude of the confining
pressure 03 (slope of ol - el diagrams in Fig. 8). Under repeated radial stresses,
however, no such dependency exists since the resilient modulus is computed by
Eq 2 based on equal values for B1, B2, B3, and B4 at different confining
pressures (Fig. 9).

Dynamic Modulus Tests


In addition to their fundamental use in f'mding the stress-strain relationship of
viscoelastic material, dynamic modulus tests provide a good basis for evaluating
and comparing various materials in terms of their rheological properties.
The complex modulus E * may prove very helpful in the mix design criteria
and in evaluating the performance of the material in the road.
The dynamic modulus tests were conducted under zero confining pressures
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
60 TESTING OF BITUMINOUS MIXTURES

and at the same temperatures and axial stress levels as those of the resilient
modulus tests (10, 30, 50 psi). The sinusoidal axial load was provided by the
MTS so that the stress was always in the compression mode. The stress
frequency was varied from 0.01 to 20 Hz. Stresses, strains, and phase angles
were obtained from the continuous recordings as described before.
The following observations were drawn concerning the complex modulus E*,
complex Poisson's ratio v*, and the phase lag angles ~1 and ~3 .a
1. The value of the complex modulus drops considerably as the temperature
increases and the frequency decreases. At low temperatures, the value of this
modulus is in close agreement with the resilient modulus. Furthermore, this
value is also equivalent to the stiffness of the free elastic parameter in the four
element model computed from the creep tests.
2. The value of the complex modulus is influenced by the asphalt content.
As the asphalt content increases the effect of temperature becomes more
significant and the value of the complex modulus becomes less and less affected
by the frequency. This result is expected, since unbound material (a/c = 0) is not
at all influenced by temperature and possibly very little by frequency. Pure
asphalt (a/c = I00 percent), on the other hand, is quite viscous, and its response
is dependent on both temperature and frequency.
3. Since different gradations have different asphalt film thicknesses for the
same asphalt content, the coarse gradation which has the thickest asphalt films
has the highest temperature or frequency susceptibility or both.
4. The value of the complex Poisson's ratio has very low values at low
temperatures and at high frequencies. This value, however, tends to approach
the limiting value of 0.5 at high temperature at at low frequency. This was more
prominent in mixtures with high asphalt contents and coarser gradations.
5. In general, the axial phase lag angle is smaller than that in the radial or
circumferential directions. The difference is more noticeable at high tempera-
tures and,under low frequencies.
6. The phase lag, ~1 or r increases with temperatures and decreases with
frequency. The phase lag also increases with thickness of the asphalt fdm coating
the aggregate. This was noticed particularly in the coarse gradation at 4.5
percent asphalt.

Time-Temperature Superposition
It is very difficult to analyze the temperature dependency by seeking an
analytical form for the complex parameters at a constant frequency. Instead, the
method of reduced variables or viscoelastic corresponding states was utilized.
This method affords a valuable simplification in separating the two principal
variables of time and temperature on which the viscoelastic properties depend,
and expressing the properties in terms of a single function of each. In principle,
this method is based on an empirical assumption which essentially means that
time or frequency or both and temperature are interchangeable for linear
viscoelastic materials.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
TERREL ET A L ON T R I A X I A L TESTING SYSTEM 61

Figure 11 illustrates the shift function ~r for different mixtures with 70~ as
reference temperature. For ideally elastic materials with no temperature
susceptibility, time shifts equal zero, and the material response is unchanged
regardless of the test temperature. Untreated aggregate, for all practical
purposes, falls within this category. As the asphalt content increases, the
temperature susceptibility and the shift factor increase, indicating more viscous
behavior.

102

10

1.0

-~ 10-I 9 / /

10-2
iF.
0 0 @ A/C ~-- 2.5%
ZkA~i A/C = 3.5%
1114
rll i1 A/C = 4 . 5 %

r
I: =
10-4
40 50 60 70 80 90

Temperature (~

FIG. 11-Shift factor versus temperature (reference temperature = 70~

Figures 12-14 show the master curves of the complex modulus versus reduced
frequencies for the asphalt mixtures tested. At low asphalt content, little
difference was observed between the medium and the coarse gradations. The fine
gradation, however, assumed considerably lower values even at high frequencies.
This may be due to the high air voids associated with this gradation. As the
asphalt content increases, no appreciable difference between the three gradations
was noticed at high frequencies. At low frequencies, however, the coarse
gradation which has the thickest asphalt films, resulted in low values of the
complex modulus.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
62 TESTING OF BITUMINOUS MIXTURES

u I~ u-

~ N

n o
T-

~m o ~ . II

"~ , = - o o < l o
el
1.1= I~ a, IJ.
&

m
I1.

U
"0

Ilg

? r~

('~

\
:llllJ I J 1 L lllll [ I J = iJll I I I I I

o
~ ,r- T-

(! sd) . 3 snlnpolAI xeldtuoo

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
10 7

E M
_~ J. ll.
C
IF

10 6
LO

-= m
"o
O
m
x r-
m
~ , 10 s
0 - F r-
0
C.~ ~' Z
Asphalt Content = 3.5% -I

X
3>
10 4 I I IIIIII I I I J Ill| I I I IIIII I I I I IIII I i Illllll I I IIIIII I I Illllll I I iiiiii r-
10-~ 10 -z 10 -t 10 o 10' -t
10 z 101 10 4 10 m
r~
Reduced Frequency (cps) -I

-<
"-I
m

FIG. 13-Master c o m p l e x m o d u l u s IE*I at r e f e r e n c e t e m p e r a t u r e = 70*F. <~

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
64 TESTING OF BITUMINOUS MIXTURES

~UU.
o

o II

c-
o
U
a
r
_-_,-
o
,<

Q.

o 0
=.- e-
~P m

O"

IL
___o
o
"0

"0

CC
7o

u. C)
III I I I Ill I I I I I III I ! I I L

(!sd) . 3 snlnp01N x01duJoo

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
TERREL ET AL ON T R I A X I A L TESTING SYSTEM 65

Concluding Remarks

The most important factor in material characterization is simulating the field


conditions in terms of temperature and stress. In order to obtain meaningful
relationships between stresses and strains that can be used in multilayer pave-
ment analysis, both axial stress and chamber pressure must be of a repetitive
nature.
As a result of this investigation, material nonlinearity for ATB may be
ignored within the practical range of temperatures. It is recommended that no
further tests should be investigated in this regard.

Acknowledgment
The authors are appreciative of the Washington Highway Department and the
Federal Highway Administration for their sponsorship of the project and the
Department of Civil Engineering at the University of Washington for providing
facilities. Jim Wilson, formerly at the University, was responsible for the original
design concepts of the test system, and the civil engeineering machine shop,
under the direction of Jim Ritchie, fabricated most of the triaxial equipment.
Also, Fred Rutz was most helpful in the specimen preparation and testing
conducted during his graduate student tenure.

Bibliography
[1 ] Barksdale, R. D., "Compressive Stress Pulse Times in Flexible Pavements for Use in
Dynamic Testing," The 50th Annual Meeting of the Highway Research Board, Wash-
ington, D.C. 1971.
[2] Deacon, J.A., "Materials Characterization-ExperimentalBehavior," in HRB Special
Report 126, Structural Design of Asphalt Concrete Pavement Systems, Proceedings
of a Workshop, Austin, Tex. 1970.
[3] Deacon, J. A., "Fatigue of Asphalt Concrete," Ph.D. dissertation, University of Cali-
fornia, Berkeley, 1965.
[4] Dehlen, G.L., "The Effect of Non-Linear Material Response on the Behavior of
Pavements Subjected to Traffic Loads," Ph.D. dissertation, University of California,
Berkeley, 1969.
[5] Dixon,W. J., Ed., "Biomedical Computer Programs," University of California Press,
Berkeley and Los Angeles, 1967.
[6] Griggs, G.E., "Permanent Deformation Subsystems - Quasi-Elastic and Visoelastic
Approaches."
[7] Heukelom,W. and Klomp, A. J. G., "Road Design and Dynamic Loading," Proceed-
ings of the Association of Asphalt PavingTechnologists, Dallas, Tex., 1964.
[8] KaUas,B. F., "Dynamic Modulus of Asphalt Concrete in Tension and Tension-Com-
pression," Association of Asphalt PavingTechnologists, 1970.
[9] Kallas, B. F. and Riley, J. C., "Mechanical Properties of Asphalt Pavement Materials,"
Proceedings, Second international Conference on the Structural Design of Asphalt
Pavements, 1967.
[10] Kasianchuk, D. A., "Fatigue Considerations in the Design of Asphalt Concrete Pave-
ments," Ph.D dissertation, University of California, Berkeley, Calif.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
66 TESTING OF BITUMINOUS MIXTURES

[11] Material Research and Development, Inc., "Translating AASHO Road Test Findings
Basic Properties of Pavement Components," Project Nos. 1-10 and 1-10/1, Oakland,
1970.
[12] Mitchell, J.K. and Monismith, C. L., "Behavior of Stabilized Soils Under Repeated
Loading," Report No. 2 for U. S. Army Material Command, University of California,
Berkeley, 1966.
[13] Moavenzadeh, F. and Elliot, J.F., "Moving Load on Visoelastic Layered System,
Phase II," Research Report R69-64 of Civil Engineering, Massachusetts Institute of
Teehnology, Cambridge, Mass.
[14] Monismith, C. L., "Design Considerations for Asphalt Pavements," Seminar, Univer-
sity of Washington, Seattle, 1970. [
[15] Monismith, C. L., Alexander, R. L., and Secor, K. E., Rheologic Behavior of Asphalt
Concrete," Proceedings, Association of Asphalt Paving Technologists, Vol. 35, 1966.
[16] Monismith, C. L., and Secor, K. E., "Visoelastic Behavior of Asphalt Concrete Pave-
ments," International Conference on the Structural Design of Asphalt Pavements,
Michigan, 1962.
[17] Papazian, H.S., "The Response of Linear Visoelastic Materials in the Frequency
Domain with Emphasis on Asphaltic Concrete," International Conference on the
Structural Design of Asphalt Pavements, Michigan, 1962.
[18] Remain, J. E., "Rut Depth Prediction in Asphalt Pavements," Research Report No.
150/JER/1969, Centre De Recherehes Routieres, Brnxelies, France.
[19] Rutz, F. R., "Correlation of Resilient Properties with Conventional Test Values for
Asphalt Treated Materials," M.Sc. thesis, University of Washington, Seattle.
[20] Secor, K. E., and Monismith, C. L., "Analysis of Triaxial Test Data on Asphalt Con-
crete Using Viscoelastic Principles," Proceedings, Highway Research Board, Vol. 40,
1961.
[21 ] Seed, H. B., Chan, C. K., and Lee, C. E., "Resilience Characteristics of Subgrade Soil
and Thin Relation to Fatigue Failure in Asphalt Pavements," Proceedings, First Inter-
national Conference on the Structural Design of Asphalt Pavements, Ann Arbor,
1962.
[22] Seed, H. B., Mitry, F. G., Monismith, C. L., and Chan, C. K., "Prediction of Pavement
Deflection from Laboratory Repeated Load Tests," Report No. Te 65/6, University
of California, Berkeley, 1965.
[23] Shook, J. F., and KaUas, B.F., "Factors Influencing Dynamic Modulus of Asphalt
Concrete," Proceedings, Association of Asphalt Paving Technologists, Vol. 38, 1969,
pp. 162-164.
[24] Terrel, R. L., "Factors Influencing the Resilient Characteristics of Asphalt Treated
A~gregates," Ph.D. thesis, University of California, Berkeley, 1967.
[25] Van der Peel, C., "A General System Describing the Visoelastlc Properties of Bitu-
rains and its Relation to Routine Test Data," Journal of Applied Chemistry, May
1954.
[26] Washington State Highway Commission, Department of Highways, "Standard Specifi-
cations for Road and Bridge Construction," Washington State Highway Commission,
Olympia, Wash., 1969, Section 9-03.6.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
M. W. lCitczak 1 and R. E. R o o t I

Summary of Complex Modulus


Laboratory Test Procedures
and Results

REFERENCE: Witczak, M. W. and Root, R. E., "Summary of Complex Modu-


lus Laboratory Test Procedures and Results," Fatigueand Dynamic Testing of
Bituminous Mixtures, ASTM STP 561, American Society for Testing and
Material, 1974, pp. 67-94.
ABSTRACT: The fundamental properties of an asphalt mixture are an impor-
tant input to the new design procedures being developed. The complex modu-
lus test, which measures both elastic and viscous material properties, is one of
the several tests available for use with these procedures. In order to use these
test results for design, it is necessary to assume that the mixture exhibits both
linear and isotropic behavior. A review of current literature is presented which
shows both assumptions valid for the elastic properties of an asphalt mixture
under certain limiting conditions. The literature review is also used to provide
background information for the development of a standaxd test method by
deeming the effect of test variables on complex modulus results.
The results of a study of complex modulus data from both cored speci-
mens and laboratory molded specimens are used to determine testing variabi-
lity associated with various testing strategies. The testing error for the elastic
part of the complex modulus, as defined by the coefficient of variation is
equal to 4 percent for cored specimens and 8 percent for laboratory molded
specimens. The variation between specimens was 8 and 17 percent, respective-
ly, for laboratory molded and field cored specimens. Based on these results, a
testing strategy consisting of three specimens tested twice at each test condi-
tion for laboratory molded specimens, and six specimens tested once at each
test condition for cored specimens is recommended.
The final section of the paper develops a method of predicting the elastic
component of complex modulus. An equation is presented which can be used
to predict this value from the results of the flexural fatigue tests.
KEY WORDS: fatigue tests, flexible pavements, dynamic modulus of elasti-
city, phase lag, elastic properties, viscoelasticity, isotropy, linearity, stiffness,
stresses, strains, design, test equipment, coefficient of variation

Purpose
The c o m p l e x m o d u l u s test is one o f several laboratory procedures that have
been developed to evaluate the fundamental stress strain response o f an asphaltic
paving mixture. During the past decade, The Asphalt Institute, along w i t h other
research agencies, has devoted considerable time and effort toward the develop-

ZResearch engineer and assistant research engineer, respectively, The Asphalt Institute,
College Park, Md. 20740.

67
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Copyright* 1974 bybyASTM International
Downloaded/printed www.astm.org
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
68 TESTING OF BITUMINOUS MIXTURES

ment of this test procedure as well as the direct application of the results into a
new fundamental asphalt pavement design system.
The purpose of this paper is to summarize the results and experiences gained
to date with the laboratory test method. Included in this summary is a review of
the effect of the salient test variables upon the test results, the investigation of
the variability components involved in the test procedure, and a study dealing
with the prediction of the complex modulus from flexural stiffness (modulus)
results obtained from flexural fatigue testing of asphalt concrete specimens.

Definition
The complex modulus is a complex number which defines the relationship
between stress and strain for a linear viscoelastic material subjected to a
sinusoidal loading. The real part of the complex modulus is a measure of the
materials elasticity and the imaginary part is a measure of its viscosity. When a
linear viscoelastic material is subjected to a loading stress of the form a = Oo sin
(cot), the resulting strain response is of the form e = eo sin ( w t - ok) which lags
the stress by the phase angle, r With this loading form, the complex modulus,
E*, is defined by Eqs 1 and 2 [1] .2

E * = E' + (1)

where

E* = complex modulus,
E' = (ooleo) cos r
E" = (ao/eo) sin r
/ = imaginary number,
do = amplitude of the repeated axial loading stress, psi,
eo = recoverable axial strain, in./in., and
= phase angle determined from test results, deg.

E* = IE*le/r
(2)

where

IE*I = Oo/eo, dynamic modulus, psi

Based on this def'mition, the absolute value of complex modulus, IE*I (the
real part), is a measure of the materials elasticity while the phase lag, or angle, r

2The italic numbers in brackets refer to the list of references appended to this paper.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
WITCZAK AND ROOT ON MODULUS LABORATORY TEST 69

(the imaginary part), is a measure of the viscous response. The absolute value of
complex modulus, IE* I, is commonly referred to by some researchers as the
dynamic modulus [2].

Applications
The majority of the work on applying complex modulus test results to design
has been done by Ohio State University and The Asphalt Institute. At Ohio
State, the results of the complex modulus test have been compared to the results
of creep tests (another means of linear viscoelastic materials characterization)
quite successfully [3]. Both test results have been used to predict pavement
performance [4].
At The Asphalt Institute, complex modulus tests have been conducted on
both laboratory molded specimens and field cores from many different projects.
The values of dynamic modulus, IE*I, from these tests have been used to predict
pavement deflection, fatigue cracking, and required overlay thickness [5-7].
They have also been used in the development of the fundamental thickness
design procedure used in The Asphalt Institute Manual Series (MS-11),
Full-DepthAsphalt Pavementsfor Air CarrierAirports, Jan. 1973 [8].

Equipment
The test equipment required for the complex modulus test consists of a load
application system, temperature control system, and stress and strain measure-
ment system. The load application system must be capable of applying repeated
sinusoidal loads to the test specimen at several different load frequencies. Cur-
rently, most researchers employ an electrohydraulic testing machine capable of
applying the stress repetitions over a wide range of loads and frequencies [5].
A temperature control system is necessary for complex modulus testing
because of the influence of temperature on the material properties. Temperature
control can be accomplished by either bringing the specimens to a constant
temperature in a controlled temperature cabinet and then quickly testing after
removal from the cabinet [5], or by enclosing the specimen and part of the
loading equipment with a small temperature control cabinet.
The continuous measurement of the applied loading and the resulting axial
strain is obtained through use of an electronic system consisting of a load cell,
strain gages or linear variable differential transformers (LVDT), amplifier, and
recorder. This equipment provides the necessary accuracy for the measurements
and a permanent record of the test results.

Specimen Type
The test specimens used for complex modulus testing are either laboratory or
field cored specimens. Both the gyratory and kneading compactors have been
used to compact specimens [9,10]. The specimens normally have a length to
diameter ratio of two. A 4-in.-diameter by 8-in..high specimen has been used for

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
70 TESTING OF BITUMINOUS MIXTURES

the majority of complex modulus testing [2,4,5,11].

Effect of Test Variables


General
The complex modulus test, like any other test, has a large number of variables
which affect the final test results. This section of the paper summarizes seven
controlled test variables and their effect upon test results. The results of this
section are summarized in the recommendations and conclusions for use as
background information for the future development of a standard test method
on complex modulus testing.

Temperature
The major effect of temperature on the modular response of asphalt mixtures
has been demonstrated by many researchers. Kallas and Riley have shown a
range of dynamic modulus, [E*I, from 170 000 to 2 000 000 psi for an asphalt
concrete surface mixture as the testing temperature decreased from 100 to 40~
[5]. Coffman has shown that both IE*i and phase lag, r are a function of
temperature [4]. It is, therefore, necessary to define the complex modulus over
the range of temperatures expected in the field if the pavement design is to be
based either upon some form of cumulative damage theory similar to the one
used in Ref 6 or if only a "critical" design temperature will be used. Witczak has
shown, using an elastic pavement model and cumulative damage theory, that it is
possible to define a unique temperature for a particular environment in the
design of thick asphalt pavements [8].

Frequency
The variation of complex modulus with load frequency, like temperature, is a
basic property of an asphalt concrete mixture. The def'mition of complex
modulus states that for a linear material at a given temperature, the complex
modulus is a function o f frequency. Work by several researchers has shown the
dependence of IE*I and ~ on frequency [1,4,5]. Kallas and Riley, for an asphalt
concrete surface mixture, have shown ranges in IE*I from 1 600 000 to
2 200 000 psi and 100 000 to 260 000 psi for frequency changes from 1 to 14
Hz at temperatures of 40 and 100~ respectively [5]. Therefore, it is necessary
to def'me the material properties over a range of frequencies expected in the
field.
To develop this frequency range, it is necessary to relate vehicle speed to
loading frequency. Based on a study of pavement deflections at various vehicle
velocities, Coffman concluded that the vehicle acts as a cyclic load with a
wavelength of 6 ft [4]. Barksdale, in an analysis based on linear elastic theory,
calculated theoretical loading times for various depths and vehicle speeds [12].
A comparison of the two methods indicates that Coffman's is somewhat
conservative for depths in the pavement between 0 and 12 in. and approximately
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
WITCZAK AND ROOT ON MODULUS LABORATORY TEST 71

equal to Barksdale from 12 to 28 in. Either one of these methods can be used to
convert vehicle speed to loading frequency.

Testing Mode
The majority of all complex modulus testing to date has been done by
applying a compressive loading to the specimens. It is possible, however, to
determine the complex modulus from tension and tension-compression loadings.
The value of complex modulus will be independent of testing mode if the
asphalt mixture behaves as an isotropic material. Isotropy is assumed in both
elastic and viscoelastic analysis to allow the calculation of tensile and compres-
sive strains from material properties determined by one or the other loading
modes.
Based on previous research, it is possible to compare the complex modulus
values from various testing modes to substantiate the assumption of isotropy.
Work by Kallas has shown similar values of IE*I from tests using compression,
tension-compression, and tension modes of testing for several loading frequen-
cies and test temperatures. However, considerable differences were shown in the
values of r determined from the tests with the three testing modes. On the
average, the values of ~ determined from tension and tension-compression tests
were, respectively, 50 and 25 percent greater than the values determined from
compression testing [2].
A similar study has been made by Coffman et al on cores from an asphalt
concrete pavement. The first part consisted of testing cores cut in three
orthogonal directions from the pavement. A comparison of IE*I values showed
good agreement for the range of temperatures and frequencies studied. The
second part of the study compared the values of IE*I determined from
compressiontesting with values obtained from a bending test. The levels of
loading stress were controlled so that the material response was approximately
independent of loading stress for both tests. The results of this part of the study
showed similar values of IE*I for the two different testing modes. Based on this
work, Coffman et al concluded that an asphalt concrete mixture behaves as an
isotropic material [13].
The results of the two studies confirm that the elastic component of an
asphalt concrete mixture behaves isotropically and is, therefore, independent of
testing mode. Kallas' work has shown that the viscous response of an asphalt
concrete mixture does depend on the testing mode used. Therefore, when a
viscoelastic design method is used, a testing mode consistent with the expected
inplace loading of the specimen should be used. This will probably be a
tension-compression test in most cases.

Load Stress Level


A basic assumption in the development of the theory behind the complex
modulus test is that the asphalt mixture behaves as a linear viscoelastic material.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
72 TESTING OF BITUMINOUS MIXTURES

By definition, a linear material is one whose stress-strain ratio (dynamic


modulus) is independent of the level of stress applied. It is commonly known
that an asphalt concrete mixture is not a linear material. However, it is possible
to characterize an asphalt mixture at low stress levels where its behavior
approximates that of a linear material.
Several research studies have been conducted to study the extent of the linear
range for asphalt concrete mixtures. Kallas and Riley have shown linear results
from compression tests with loading stresses up to 70 psi at temperatures of 40,
70, and 100~ [5].
Coffman et al have shown similar results for load levels up to 80 psi at 40 and
77~ and 35 psi at l l 0 ~ [11]. Cragg and Pell have conducted a series of
tension-compression tests on asphalt concrete mixtures varying the axial stress
level from +5 to +50 psi (+ = compression and - = tension). The results indicated
that the IE* I becomes stress dependent at a stress level between -+25 to +30 psi
for tests conducted at 68 and 86~ [14]. Kallas has shown that the maximum
loading stress for a linear response is a function of temperature, frequency, and
testing mode. For tests at 70~ and 1 Hz, the maximum loading stress used was
17.5 psi, +35 psi, and 70 psi for tension, tension-compression, and compression
modes of testing respectively [2]. These results have shown that an asphalt
concrete mixture does behave linearly over a range of stresses consistent with
those experienced within a highway pavement.
When developing a testing program, a low stress level should not be selected
to guarantee that the results will be in the linear range. A brief study of complex
modulus test results indicated a higher standard deviation for tests at 17.5 psi
when compared to tests at 35 psi over a range of temperatures and frequency for
the compression test mode.
Little work to date has been done on the effect of axial load level on r Cragg
and Pell showed that values of ~ are approximately constant over the linear
range of IE*t response [14]. A brief review of the unpublished data from the
Kallas and Riley paper shows no consistent variation of ~ with loading.
Therefore, ~b can be assumed to be independent of the loading stress over the
same range of load levels as IE* I.

Confining Pressure
An element of an asphalt concrete pavement under load is subjected to a state
of triaxial stress. In materials characterization, it is possible to simulate this
loading by applying both axial and confining stresses to the test specimen. The
literature search did not yield any information on the effect of conffming
pressure on complex modulus.
Some work has been done in this area using the resilient modulus test, which
is another method of measuring the elastic properties of a material. Tests with
confining pressures from 0 to 20 psi at 70 and 90~ have indicated that
confming pressure has little or no effect on the elastic response of asphalt
concrete when tested in the linear range of stress-strain response [15]. These
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
WITCZAK AND ROOT ON MODULUS LABORATORY TEST 73

results should also apply to values of dynamic modulus determined from the
complex modulus test.
With regard to r further research should be done to establish the influence of
conf'ming pressure on test results. It is very likely that the viscous properties of
asphalt concrete will be influenced by toni'ruing pressure.

Test Specimens
One of the primary problems in laboratory testing is to obtain the same
results from laboratory molded and field cored specimens. Several studies have
shown that it is possible to get comparable values of complex modulus if a
special laboratory compaction method is used [4,10]. The laboratory specimens
were compacted using the California Kneading Compactor. The number of
tamps, foot pressure, and static load are adjusted for various mixtures to give a
similar voids level to that obtained from field compaction. An example of this
method with values of the compaction variables used has been outlined by Kallas
[101.

Test Measurements
It is necessary to continually monitor the values of load and the resulting
strain experienced by the specimen during the test. The load measurement is
easily accomplished by using a commercially available load cell with appropriate
amplification and recording equipment. However, strain measurements are not as
easy and, subsequently, require further consideration. Some factors to be con-
sidered are: type of measurement device, length of strain gage (if used), and
procedure for determining phase lag, (~, from the recording.
The strain measurements are also done electronically and, thus require an
amplifier and recorder. The majority of strain measurements have been made by
bonding two wire foil strain gages to opposite sides of the specimen [5]. Recent-
ly, some work has been done using LVDT's is that strain gages are fairly expen-
sive and they can only be used on one specimen. Studies at Ohio State Univer-
sity have shown good agreement between the two measurement techniques with
regard to IE*I [4,16]. However, little work has been done to compare the values
of ~ from the different systems.
The effect of strain gage length is of some concern when testing base courses
composed of large aggregate. The most commonly used strain gage has a gage
length of 13/16 in. Coffman has compared measurements from these gages with
those from a pair of 3-in. gages on large aggregate base course specimens. Care
was taken so that the short gages were attached over areas between the aggregate
faces. The averages of the test results from the two different gage lengths were
essentially the same for both [E*l and ~. However, there was considerably less
scatter between the individual test results with the longer gages [4]. Coffman's
results indicate that for the worst case (large aggregate, short gage length)
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
74 TESTING OF BITUMINOUS MIXTURES

complex modulus values are not affected by gage length as long as a sufficient
number of specimens are tested to overcome scatter. Based on these Findings, the
assumption that complex modulus is independent of gage length can be made.
The values of r are determined from measurements made on the recording of
the complex modulus test. As defined previously, r is a measure of the time
delay between the loading of the specimen and its response. It is possible to
measure r on both the loading and unloading portions of the response curves
and according to linear viscoelastic theory, these two values should be the same.
To check the effect of measurement on r the variations with regard to
measurement location were analyzed. The values of r from the loading and
unloading portions of the curve were compared for four separate sets of data
representing a wide range of materials. A summary of these data are given in
Table 1 for test temperatures of 40, 70, and 100~ and the two measurement
locations. Each value shown in the table represents the average ~ from tests on
several different specimens at loading frequencies of 1,4, and Hz. At the higher
temperature, there is a considerable degree of difference between the two
measurement locations. Based on these results, The Asphalt Institute modified
its procedure with regard to the determination of r in 1968. The current
procedure determines r from the average of the loading and unloading value.
If the complex modulus data are to be used in a viscoelastic design procedure,
further work is required on the methods of determining r from the test results.
Pagen has compared values of IE*I from the complex modulus test and the
creep test and has shown good agreement [3,9]. It is felt that this type of
comparison could be extended to the r data to help establish measurement
procedures for the complex modulus test.

TABLE 1- Variation of phase lag with measurement location.

Temperature
40*F 70*F 100*F
Samples eLa ~b ~ eu eL eu
Colorado test road 18.3 15.1 30.8 26.4 34.3 35.9
Brampton test road 18.3 16.0 38.4 35.8 29.4 42.3
San Diego test road 15.3 12.4 27.0 25.9 25.8 32.4
Washington State University
test track 13.0 12.6 29.7 31.2 25.2 37.9

Average 16.2 14.0 31.5 29.8 28.7 37.1

a ~bL = measurement from loading portion of curve.


o r = measurement from unloading portion of curve.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
WlTCZAK AND ROOT ON MODULUS LABORATORY TEST 75

V~riability of the Dynamic Modulus Test

General
During the past several years, The Asphalt Institute has conducted numerous
complex modulus tests on various asphalt concrete mixtures. In order to
quantitatively assess the variance components and magnitude attributable to this
test method, an analysis of variance (ANOVA) was conducted on these past
results. In the testing program, the specimens were categorized into one of two
major groups. The first category, termed field, represents tests conducted on
cores taken directly from an in situ (existing) pavement structure. The other
category of tests, identified as lab, were those projects dealing with the fabrica-
tion within the laboratory, of specimens that simulated, as nearly as possible, the
field conditions of a proposed or existing pavement structure.
Two test replicates were conducted on each of three specimens for each
project studied, regardless of whether field or lab specimens were utilized. Each
test was conducted at a factorial combination of three levels of test temperature
(T = 40, 70, and 100~ and three load frequencies (f = 1, 4, and 16 Hz). Both
the dynamic modulus, IE* I, as well as the phase angle, ~, were measured for each
replicate. The ANOVA study was conducted on four field projects and nine lab
projects. This resulted in a total of 216 test measurements analyzed for the field
IE*I values, 216 for the field ~ values, 486 for the lab IE*I values, and 486 for
the lab r values.

Replicate Tests on Same Specimen


In the testing procedure used by The Asphalt Institute, replicate test
measurements of both IE*I and r are performed on the same specimen. In
general, each of three specimens is tested once at a particular temperature and
load frequency. Upon completion of the initial measurements for all specimens,
they are retested for the replicate result. Thus, a time interval varying from ~1
to 4 h may occur between replicate tests on the same specimen. It is, therefore,
important to determine whether any bias in test results occurs due to the
influence of the initial test upon the second replicate value because of the
assumed random nature of the results in the ensuing ANOVA study.
When considering either IE*I or ~ results, if a bias is introduced by the first
measurement, a convenient and powerful statistical test to identify this effect is
a hypothesis test concerning the difference between paired measurements.
Analytically, the difference of the ith paired difference is d i = q~li - ~2 ior IE* I1i
- IE* 12i with the specific test hypothesis being H : d'= 0.
The test statistic for the analysis is

(e - ao)vCff
t' = sa (3)

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
76 TESTING OF BITUMINOUS MIXTURES

where

d = mean difference between paired observations,


do = expected mean difference (d o = 0 for hypothesis),
n = number of paired difference measurements, and
Sa = standard deviation of the distribution.

The appropriate tabular value o f t is tet/2; v - ~ for a two tailed significance


test. Acceptance of the hypothesis occurs i f - t < ~ t ' <<,+t. In the analysis
conducted, both laboratory and field specimens were grouped together and the
statistical test accomplished for 39 paired measurements at each temperature
(40, 70, and 100~ and frequency (1, 4, and 16 Hz) combination. For n = 39
and cx = 5 percent, to.o2s;3s = 2.334.

TABLE 2-Replicate test hypothesis summary.

Dynamic Modulus Phase Angle


Test Load Test Test Test Test
Temperature, *F Frequency, Hz Statistic, t ' Result Statistic, t ' Result

40 1 0.965 NSa -1.220 NS


4 -0.907 NS 0.351 NS
16 -2.259 NS -1.599 NS
70 1 -2.125 NS 1.874 NS
4 -1.319 NS -0.150 NS
16 -0.475 NS 0.150 NS
100 1 -0.887 NS -0.709 NS
4 -0.756 NS 1.241 NS
16 -1.879 NS 2.101 NS

to.o2s;38 = 2.334

a NS -- nonsignificant.

Table 2 summarizes the results o f the test for both IE*I and ~bdeterminations.
The results for both parameters at all temperature-frequency combinations
indicate that the hypothesis was accepted. Since no consistent difference or bias
between replicates was observed, it can be concluded that the replicate measure-
ments on the same specimen are random in nature.

A N O VA Model
The model used for the ANOVA was a one way nested classification, Model II
(random effect). Each ANOVA was computed for a fixed combination of core

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
WITCZAK AND ROOT ON MODULUS LABORATORY TEST 77

group type (field or lab), test parameter (IE*I or r and specific temperature-
frequency level combination. Thus, within any ANOVA computation the
components of variance analyzed were the project op 2 , core (specimen) Oe2 , and
test or replicate ot 2 variability. Analytically, for the type model used and
conditions noted

Xi]k = l~ + Pi + Ci] + ~qk (4)

for constant (T, f)(lE*l or ~) (field or lab)

where xilk is the measured value of the it~ project (i = 1 . . a ) , j th specimen (j =


1 .. b), and k th replicate (k = 1 .. n). As previously stated, for both the field
and lab groups, n = 2 and b = 3, while a = 4 for the field study and a = 9 for the
lab study.
The parameter, /~, represents the true mean value, while Pi, Cq, and ~ijk
represent the random effects due to the project effect Pi, specimen effect Cq,
and test effect ~i/k. All of these random variables are assumed normally
distributed and independent with Pi = N (0, op2), Cq = N ( 0 , Oc2), and ~ilk = N
(0, O't2).

TABLE 3-GeneraI ANOVA table.

Mean
Sum
Source of Degrees of of Parameters
Variation Freedom Sum of Squares Squares Estimated
Projects a- 1 ~'X2i . . /bn - ( X . . . ) 2 / a b n MSp ot 2 + nOc2 + bnOp2

Specimens a (b - 1) ~,Xi j .2In _ ~,Xi " . 2 /bn MSc Ot2 +nOe2


in projects
Replicates ab (n - 1) ~, Xi]k2 _ ~, Xij. 2 /n MSt Ot2
on specimens

The general format of the ANOVA utilized is shown in Table 3. As can be


seen, the mean square (MS) value for a particular variation source estimates a
combination of the true variance components desired from the study. Since
these values are only estimates, hypothesis tests were determined to ascertain
whether either oc 2 or op 2 is zero. This is accomplished by the following
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
78 TESTING OF BITUMINOUS MIXTURES

H " Oc2 = O ; F = MSc estimates


MSt
(5)
tTt2 + nOe2 , f = Vc, Pt
Ot 2

where vc and vt are the degrees of freedom associated with the specimen and
replicate (test) variation source, respectively. Similarly,

H : op 2 = 0 ; F = MSp estimates
MSt
(6)
~ + nOc2 + bntrP z , f = Vp, vc
Ot 2 + nOc 2

If the test is significant at the significance level, a, selected, estimates of at 2 ,


oc 2 , and op 2 may be obtained by

ot: = MSt (7)

(MSc - MSt )
o~ 2 - (8)
n

(MSp -- M S c ~
op 2 - (9)
bn

If the hypothesis test is not significant, the ANOVA results are pooled for a
single subclassification analysis and the pertinent variance components com-
puted.

Results
Significance Tests-Table 4 summarizes the significance test results of all the
ANOVA studies conducted to determine whether either ac 2 = 0 or ap 2 = 0. The
results are shown for two significance levels, t~ = 5 and 1%. The implication of a
nonsignificant result (NS) implies that the hypothesis of zero variance (ac 2 or
op 2) is accepted at the level of risk assumed.
~*1 R e s u l t s - F r o m the ANOVA study of the IE*I, it can be seen that for the
field group, both hypotheses (H : %2 = 0 and H : op 2 = 0) were significant for
each temperature-frequency combination. This implies, as would be anticipated,
that when cores are taken from an existing pavement structure for testing the
IE*I varies both from specimen to specimen and from project to project. The
study concerned with specimens fabricated in the laboratory showed similar,

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
WlTCZAK AND ROOT ON MODULUS LABORATORY TEST 79

TABLE 4-Summary of significant test results ftom ANOVA.

Results
Test Load I~"1 Analysis ~bAnalysis
Test Temperature, Frequency,
Hypothesis Group *F Hz ot =5% ot=1% o~=5% 0t =1%

H : o c 2 =0 field 40 1 Sa S NSb NS
4 S NS NS NS
16 S S NS NS
70 1 S S NS NS
4 S S NS NS
16 S S NS NS
100 1 S S NS NS
4 S S S S
16 S S NS NS
lab 40 1 S S NS NS
4 NS NS NS NS
16 S S NS NS
70 1 NS NS NS NS
4 S NS NS NS
16 NS NS NS NS
100 1 S S NS NS
4 S S NS NS
16 S S NS NS
H:ap2=O field 40 1 S S S S
4 S S NS NS
16 S S S S
70 I S S S S
4 S NS S S
16 S S S S
100 1 S S NS NS
4 S S S NS
16 S NS S S
lab 40 1 S S S S
4 S S NS NS
16 S S S S
70 1 S S S S
4 S S S S
16 S S S S
100 1 S S S S
4 S S S S
16 S S S NS

a S = significant at (1 - or).
b NS = not significant at (1 - O0.

though n o t as strong, results as the field group.


It can be noted that for the test o f H : oc2 = 0, three o f the
temperature-frequency combinations resulted in nonsignificant results at the 5
percent level. This is not entirely u n e x p e c t e d because indirectly this result is
illustrating that the ability to fabricate identical specimens in the laboratory for
a given project is quite good. However, because no distinct trend in these
nonsignificant results were observed and because the majority o f the
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authoriz
80 TESTING OF BITUMINOUS MIXTURES

temperature-frequency levels were found to be significant, it was subsequently


assumed that for the lab group, the core to core variability was not equal to zero
for all combinations of test conditions.
All of the results for the lab group were found to be significant at both tx
levels for the between project, at, 2 , variance. In summary, it can be concluded
that for IE*[ results, the variability consists of test to test, ot 2, specimen to
specimen oe 2, and project to project op 2 components regardless of whether
specimens are cored from the field or fabricated in the laboratory.
qbResults-The analysis of the test data concerned with the measurement of
the"~ revealed some important results. From Table 4, it can be seen that for the
test 6f H : oe 2 = 0, the results are strongly nonsignificant for both the field and
lab groups. Significance test results of both groups indicated, however, that the
project to-project variability did exist. This implies that when measuring ~ on
specimens obtained from either field coring or laboratory fabrication, the
measurements vary only from test to test (or 2) and project to project (op2).
From a statistical viewpoint, the specimen to specimen variability within a given
project is zero (oe 2 = 0).
Variability Analysis-From the results of the significance tests, it was
demonstrated that the components of variance for IE*I measurements are or, 2,
oe 2, and Or2; while for ~ determinations, op 2 and ot 2 are the primary
components. Using the procedures previously described, the various variance
components for each test parameter, specimen type and temperature-frequency
combination were calculated from the ANOVA results. In analyzing these values,
it was observed that each component variance tends to increase with an increase
in the mean value of the measurement. This implies that the coefficient of
variability (CV = variance divided by mean) is a better indicator of the variability
than the variance.
Figure 1 illustrates the CV values by variance components as a function of the
mean IE*I value irrespective of the temperature-frequency test condition for
both lab and field groups. In general, the CV values tend to be relatively con-
stant over the range of values considered. The only exception where this may
not be observed is for conditions of very low mean modulus values (high temper-
ature-low frequency combination). This observation also appears to be more
noticeable for the between project component, CVp.
However, since the two most important variability components for any given
design or project evaluation are gc 2 and ot 2 and they do appear to be relatively
constant, it would be logical to state that the use of the CV parameter is valid to
describe the variability of the IE*I procedure. Table 5 is a summary of the CV
values for the results of this study.
[E*I Results-From the analysis of the IE*I variability, it can be seen that the
average CVp (project to project) value for both field and lab groupings is about
equal to 30 percent. The CVt values for specimens cored from existing pavement
structures is relatively small at 4 percent. This implies that relatively good
reproducibility of test measurements can be obtained in the same specimen.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
WITCZAK AND ROOT ON MODULUS LABORATORY TEST 81

40 .... i .... i .... i .... i .... i .... i .... i ' '


x
x x LAB CORES

CVp ( PROJECT ) = 3 0 . 6 %
30 X x
x x x
x

Z I0 -o ~ 9 CVc ~ CVt (CORE 8t T E S T ) = 7 . 8 % o 9 .


O N rl 9
I~ o o :
E"

. . . . i . . . . i . . . . I . . . . i . . . . I . . . . i . . . . i , ,
0
9 ~,.25~'~ 5o .... ~,5 .... io,.o, . . ,2,.5, , , ,5o .... ,~,.5..,9.o
(~ 4 0 (

ktJ x FIELD CORES

C'Vp ( PROJECT ) = 2 9 . 5 % j
uJ 3 0 ....... -x- . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . }<. . . . . . . . . 3<. . . . . . . . ~'~
8
n X X j
20 x
9 C-'Vc(CORE) = 16.6 %
9 9

ID

o ~,(TEsTI.,3% ~ :.
q 0 O;
0 E
f o o 0 o
. . . . i . . . . i . . . . i , i i , I . . . . I [ i , , i i i , [ [ i

0 2 5xlO 5 5.0 7.5 I00 12 5 15.0 17..5 190

I E*I- DYNAMIC MODULUS ( psi )

FIG. 1-Relationship between coefficient of variation by component to mean dynamic


modulus value.

However, the same variability component is near 8 percent for specimens


fabricated in laboratory molds. This demonstrates that better reproducibility is
obtained on replicate tests of field cores than laboratory prepared specimens.
Although the specific reason for this occurrence is not known, one possible
explanation may be due to the fact that in coring, a smoother surface texture is
obtained on the cut face of the specimen, thus making strain gage mounting less
difficult and sensitive relative to the rougher and more discontinuous face ob-
tained on a molded laboratory compacted specimen.
As one would expect, the core to core variability, CVc, for field specimens is
more than twice the variability associated with laboratory prepared specimens
(16.6 versus 7.8 percent). For the field specimens, the C V c value is about four
times as large as the test variability, CVt; while for laboratory specimens, the
specimen to specimen variability is about equal to the test variability (CV c =
cv,).
@ R e s u l t s - A significant observation made from the analysis of @test results is
the relatively large error associated with the test reproducibility on the same

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
82 TESTING OF BITUMINOUS MIXTURES

TABLE 5-Summary of component coefficient of variation.

Coefficient of Variation Value for Parameter Indicated


Test Load
Test Temperature, Frequency, IE*[ Test ~bTest
Group *F Hz CVp (%) CVc (%) CVt (%) CVp (%) CVt (%)

Field 40 1 29.5 12.5 4.1 23.6 12.4


4 28.9 9.2 7.7 9.1 14.2
16 27.6 16.0 3.5 14.1 22.9
70 1 25.6 16.7 6.2 20.5 10.7
4 26.2 19.0 2.9 21.9 11.8
16 20.1 11.3 3.6 11.8 13.3
100 1 42.8 22.5 4.7 12.3 14.6
4 36.0 22.1 2.2 14.9 8.7
16 28.9 20.5 3.4 17.8 11.0
Average 29.5 16.6 4.3 16.2 13.3

Lab 40 1 26.9 8.2 6.9 12.0 17.0


4 27.2 4.2 10.1 10.1 14.9
16 25.1 10.6 6.7 19.9 26.5
70 1 31.2 3.8 8.1 16.9 9.9
4 30.5 6.3 6.9 18.0 9.1
16 27.0 4.1 8.2 17.5 13.7
100 1 36.7 13.9 10.1 10.4 11.6
4 37.1 10.2 8.7 13.1 12.0
16 36.5 9.1 4.7 9.6 14.3
Average 30.9 7.8 7.8 14.2 14.3

specimen. It can be seen that for b o t h field cores and lab specimens, the CVt
value is " 14 percent. Another factor is the comparison between the project to
project variability, CVp, to the CV t value. This shows that the variability
associated with measuring q~ in the same specimen is about as large as the
variability associated between various projects.

Limit o f Accuracy Curves


Development-Knowing the component variability magnitude determined in
the previous section, it is possible to develop precision curves illustrating how
many specimens, either field cored or laboratory prepared, and number o f test
replicates per core will be required to achieve a certain specified limit o f
accuracy for a particular confidence level. Since this type o f information is
primarily required when dealing with a particular project, the two major
variabilities o f importance are the test variability, at 2, and the within project
(specimen to specimen) variability, oc 2 .
F o r any given project, the specimen mean variance, 002 , for b = 1 core and n
= 1 replicate per core is

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
WITCZAK AND ROOT ON MODULUS LABORATORY TEST 83

002 = ot 2 + o~ 2 (10)

For b cores and n replicates, this specimen mean variance is

ot 2 ffc2
Oo~- = - - + (11)
bn b

If the test parameter (~E*[ or ~O)measured is denoted by 'Is, the characteristic


of this distribution, assuming normality, is

,~ = N (u, bo ~) (12)
From two tailed confidence interval theory, the upper, U, and lower, L, limits
of the mean value can be shown to be equal to
m A

U = ~ + ooKa/2 (13)
- - A

L = kO - aoKotl2 (14)

where Ka/2 is the normal deviate for the selected ot value of the confidence
interval. The limit of accuracy value, +R, for this condition is simply
^

R = Oo Kotl2 (15)

Since it has been shown that the C V value is a more relevant variability
indicator

at = (CVt) qr (16)
I

Oe = (CVe) tI, (17)

Substituting these expressions into the specimen mean variance equation and
then further substitution into the limit of accuracy equation, yields the
following equation
[CVt2 + nCVe 2 1/2
(R_) = Ka/~ ] (18)
bn

This is the basic expression derived to develop the limit of accuracy diagrams
for IE*I and ~bmeasurements. It yields the accuracy of the test value, expressed
in the form of a ratio to the mean measurement value as a function of the b and
n combinations selected. The input used for this equation is summarized in
Table 6 for all test combinations. The analysis was made for an r = 5 percent. It
should also be recalled that for the ~ analysis, it was found that oc 2 = 0 for all

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
84 TESTING OF BITUMINOUS MIXTURES

temperature-frequency conditions. Thus, as far as the analysis is concerned, the


number of specimens is statistically equivalent to the number of test replicates,
or b --- n for this case only. Figures 2 to 4 show the curves developed for IE*I and
r testing for both field and lab conditions.

TABLE 6-Basic input summary for limit o f accuracy equation.

Confidence Component Coefficient of Variation


Test Parameter Test Group Level
fig r C lr t CVc

[E* I field 5% 0.04 0.16

IE*I lib 5% 0.08 0.08

field and lab 5% 0.14 0

24 I I~t ' I ' I ' I ' I ' I ' I ' ~z I ~2

20
t)
i,i
\\ N IN
o.
,-~ 16 \\ L u
J
u.
\\ ~176
Iv ~ IE*I-FIELDPROJECTS
w XN~..in=l TESTREPLICATES
o. 12
ir
o

o 8
w
=li
3
z

I I I , I , I , I , I , I
. ,06 .OB .iO .12 .i4 .16 .~8 ,20 .22

( R/I [*l ) VALUE

FIG. 2-Limit o f accuracy curves [or IE*l evaluation on cores from field projects (+-R;
et = O.05).

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
WlTCZAK AND ROOT ON MODULUS LABORATORY TEST 85

24
' I ' I ' I ' I ' I ' I ' I
~=' ~ z I z'
Crcotes

a: 20
O
o

, \ I.-.§
\ ~ L IE'I u
\ \ a :5~

>-
I1:
0 12
n=2
n=4
J
b 8 n=8
{1:
w
{IQ

z~ 4

0 I 1 I I , I , I , I , I ,
.04 .06 08 .10 .12 .14 .16 .18 .20 .22

(R/I~*I) VALUE

F I G . 3-Limit of accuracy curves for [E*I evaluation on laboratory fabricated core


specimens (+ R; 0t = 0.05}.

z4 . I' ~i ' \i 9 i v I ' ~ ' =~

2o

"7 n=2
J2
n=4

n=8

I , I I , I , I , I I ,
.02 .04 06 .08 .,0 .12 .14 .,6 A8

F I G . 4-Limit o f accuracy curves for ~)evaluation on either field or laboratory fabricated


cores (+-R; Ot = 0.05).
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
86 TESTING OF BITUMINOUS MIXTURES

Interpretation-The choice of the number o f core specimens, b, as well as the


number of tests per specimen, n, for any level of accuracy should ideally be
predicated upon optimization procedures using the economics of the cost per
specimen and the cost per replicate test. The selection o f a suitable limit value,R,
satisfactory to the engineer is also a very difficult decision. For example, what is

the true significance between an (R_R___) value of 0.15 relative to 0.13 from a
'~*'tEI
design and performance viewpoint? What is the difference in cost for the testing
program between these two values? These questions are in reality the real prob-
lem that faces the engineer. Nonetheless, the curves shown should provide an
invaluable tool to the engineer to at least quantitatively assess each o f the factors
considered to help in the development of a more fundamentally based testing
program.
Discussion-Using the curves shown in Figs. 2 to 4, limit of accuracy values R ,
were calculated for several combinations of b and n for both IE*I and
measurements developed from field and lab specimens to help illustrate some
basic results of the study. These results are summarized in Table 7.

TABLE 7-Limit of accuracy valuesfor selected test conditions.

Number of +R for t~ 30 deg


Number of Tests per +R for IE*I = 100000 psi
Cores, (b) Core, (n) Field Cores, psi Lab Cores, psi Field Cores, deg Lab Cores, deg

1 1 32300 22200 8.2 8.2


1 6 31500 16900 3.5 3.5
1 12 31400 16300 2.4 2.4
3 1 18700 12800 5.0 5.0
3a 2a 18500 10900 3.5 3.5
3 4 18200 I0000 2.4 2.4
6 1 13200 9100 3.5 3.5
12 1 9400 6400 2.4 2.4

a Prior testing procedure utilized by The Asphalt Institute.

IE*I Analysis-fleld specimens - A s can be seen from both Table 7 and Fig. 2,
there is a rather negligible effect upon the accuracy obtained by increasing n.
This is directly attributable to the low test variance and high within project
variance for cores obtained from existing pavement structures.
For example, by increasing the number of test replicates from n = 1 to n =
12, for b = 1, the confidence interval only decreases from + 32000 psi (E =
100000 psi) to + 31000 psi. However, for the same total number of test results
(bn = 12), the interval is decreased to + 9400 when b = 12 and n = I. The
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
WlTCZAK AND ROOT ON MODULUS LABORATORY TEST 87

previous procedure used by The Asphalt Institute was b = 3 and n = 2. This


resulted in a limit ratio of 0.185. Based upon the results shown, it would appear
that if six specimens were taken (b = 6) and only one test measurement per
specimen used (n = 1), a reduction of almost 30 percent in the confidence
interval could be obtained with a limit ratio value of 0.132.
IE*I Analysis-lab specimens-From Table 7 and Fig. 3 it can be observed that
for identical b and n combinations, the R value for lab group specimens is less
than that for field. This result, as one would normally expect, is directly due to
the larger test variance and reduced core to core variance relative to the field
group. It should also be noted that the value of n exerts slightly more influence
upon the confidence interval because of the aforementioned reasons.
If the number of tests per core were increased fromn = 1 to n = 12 for a b =
I condition, a decrease in the limit value from + 22200 to + 16300 psi occurs.
However, letting b = 12 and n = 1, the limit value decreases to _+6400 psi. The
interval ratio obtained with the previous Asphalt Institute procedure (b = 3, n =
2) is 0.109 for lab conditions.
Analysis-field and lab specimens-The largest influence in reducing the
confidence interval by an increase in n, relative to the latter two parameters
discussed, occurs for testing of the asphalt concrete mix ~b. Since for both lab
and field groups the effects of b and n are equivalent, any combination of the
values that yield the same product (bn) will give the same confidence interval.
This result is observed in Table 7. The procedure used by the Institute results
in a confidence interval of -+ 3.5 deg for a mean phase angle value of 30 deg. An
increase in either b or n that results in a bn = 12 combination (twice the number
of measurements for b = 3, n = 2) only results in a confidence interval reduction
of 1 deg.
One final observation that should be made concerns the fact that since IE*[
and q~ are measured for each test, it is more logical to select a testing procedure
for the more critical of these, two parameters and use this for both measure-
ments. At the present time, the IE*I value is used directly in several multilayered
elastic design approaches. Because of this, the need for as accurate a measure-
ment of this factor is much more critical than the phase angle.

Relationship Between Dynamic Modulus and Flexural Stiffness Tests


Another direct and rational test method used to measure the elastic modulus
of asphaltic mixtures, besides the dynamic modulus IE*l procedure discussed in
this report, is the flexural stiffness, Es, obtained from constant stress flexural
fatigue tests. The details of specimen preparation, test procedure, and
measurement have been previously reported by Kallas and Puzinauskas [17]. In
order to define what relationship, if any, exists between [E*[ and Es, an analysis
was conducted on the laboratory results of 17 different asphalt concrete mix-
temperature combinations that had been tested by both modulus procedures.
Kallas and Riley have previously noted that the measured E s value, for a

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
88 TESTING OF BITUMINOUS MIXTURES

particular mix and test temperature, is a function of the flexural stress, o, used
in the repetitive flexural test [5]. Regression studies to determine the relation-
ship between E s and a for each mix and temperature combination were conduct-
ed. The model used was

Es = E o A a o (19)

where

Es = flexural stiffness, psi,


Eo = flexural stiffness for a = 0 condition, psi,
a = flexural stress, psi, and
Az = regression constant dependent upon particular mix-test temperature.

The Eo value is the intercept of the regression line for the condition of zero
flexural stress. A typical plot of one set of data is shown in Fig. 5.
Using the results of this study, a multiple regression analysis was then
conducted to determine the relationship between the measured IE*I value at
various test frequencies, f (Hz) and Eo value at identical test temperatures. From
this study, it was found that the IE*I, measured at a given load frequency, was
related to the Eo value by

IE*I = A2 .f= Eo [c (20)

,o~ ! I I I I I
8

r
(n
uJ
z

F-
~o
2
.J

x
..J
IJ_
!
'!I A._c.
~ SURFACEMtX~
T~ 70-F

I I t I I
-

I
4/
0 50 I00 150 200 250 300 350
o ' - FLEXURAL STRESS (psi)

FIG. 5-Relationship between stiffness modulus and flexural stress level.


Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
WITCZAK AND ROOT ON MODULUS LABORATORY TEST 89

where

IE*I = dynamic test modulus, psi, and


f = test frequency, Hz.
A 2, a, b, c and d = regression constants with
A2 =0.18089,
a = 2.1456,
b = 14.6918,
c = 0.01, and
d = 13.5739.

Figure 6 compares the measured IE*I and the regression lines for a given
frequency to the corresponding Eo value determined from the individual rela-
tionships between the Es and a. Figure 7 illustrates the predicted results of the
multiple regression equation previously noted. The equation has an R 2 value of
0.9117 and a standard error of 0.0694. Also shown is the relationship between
the line of equality, (Eo = IE*I) and the predicted IE*I line from the regression
study evaluated for a test frequency o f f = 2 Hz.

Z
0
-r
)-
0 ixi06
Z
Ld 8

W 6
L~
g~

..J
f =]i H
o
0
IE
f= H / ' 0,,/ EGEND:
j f= I Hz (cps) - - - - - - o
1
Q
ixa05
8
f=l H z / / f=4Hz(cps) ~ - ~ $
i
f=16 Hz (cps) ~z~
6

4xlO 4 . . . . I . . . . . . . . . . . I . . . .
4xlO 4 6 8 ixlO5 2 4 6 8 ixlO6 2 4

Eo-FLEXURAL STIFFNESS FOR o'=0 STRESS (psi)

FIG. 6-Comparison o f measured dynamic modulus to computed flexural stiffness value


at zero stress condition.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
90 TESTING OF BITUMINOUS MIXTURES

, , , ,i I , , , i , , i I ~ J '1 i ~

z 2
0
-I-

D• s
Z
8 /,.///"
0
,,, 6 / ,,.C/
Ix.
tw 4
,///'/
r
,-.6., / .//'"
- >" / , / / E E.o:
r~
0
2 //,,," ...... _u_eF
f =4 H Z / r PREDICTED BY: b ~,
~/" =A2f~ ~ -d)
ICI

>-
Ixlo ~ / - / " "r=ZHz f=l.,(~,) . . . .
e~ 8
I
g 6

4xlO
d i i tl i , t i I i , , i ,ll ~,,,
4 x l O4 ~ 8 ixlO5 2 4 6 81• 6.v
E o - FLEXURAL STIFFNESS FOR o" =0 STRESS (psi)

FIG. 7-Dynamic modulus valuespredicted from multiple regression equation.

For all practical purposes these two relationships are the same. This fact does
have technical justification as the loading procedure used in the flexural fatigue
test is applied to repeated loads in the form of one-half sine wave with a load
duration of 0.1 s and a 0.4 s rest period between loads. This load condition,
from a frequency viewpoint, is equivalent to the 2 Hz value used in the dynamic
modulus test for the line of equality even though it is applied in a continuous
load wave fashion.
Table 8 summarizes the asphalt concrete mixes used in the analysis along with
the input data used for both prediction models. The predictive equation is
important in that it allows the IE*I value, at any desired frequency, to be
computed only from the results of flexural stiffness measurements. Even though
this is a significant step toward the development of reduced laboratory testing
time and expense, it is hoped that future research will allow the prediction of
the actual A1 value from the particular mix properties. This, of course, would
only necessitate the laboratory measurement of the dynamic modulus to predict
the flexural stiffness and its stress dependency for each asphalt mix tested.

Recommendations and Conclusions


The objective of this study was to present a summarization of research
findings concerning the effect of testing variables, material, and test variability,
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
TABLE 8-Summary o f flexural stiffness and dynamic modulus results.

Test E s (flexural stiffness) [E*I (dynamic modulus)


Temperature,
Mix Description "F Eo, psi A t , constant f = 1 Hz, psi [ = 4 Hz, psi f = 16 Hz, psi
-4
Colorado surface 70 5.77 x l 0 s 0.9932 4.73 x l 0 s 7.12 x l 0 s 9.44 x l 0 s N
Colorado base 70 5.89 0.9938 5.11 7.48 10.60
Ontario surface 70 3.79 0.9955 2.30 3.99 6.47 >
California base 70 3.23 0.9961 2.36 3.49 5.01 Z
WSUa surface 70 4.73 0.9940 3.73 5.65 7.81 t~
WSU base (AT) 70 4.83 0.9958 5.15 8.13 11.20
O
TAI b lab " A " 55 7.93 0.9978 6.30 8.40 11.40 O
TAI lab "A" 70 4.92 0.9950 3.30 4.70 6.40 -4
WESe item 2B 80 1.65 0.9951 1.30 2.25 3.60 O
Z
BFIAdA-3 surface 70 4.87 0.9992 3.50 5.70 7.50
BFIA A-5 base 55 7.92 0.9978 7.30 9.80 11.00 O
BFIA A-6 surface 55 13.49 0.9970 9.80 13.00 15.50 C
BFIA A-6 surface 70 5.55 0.9982 5.80 8.20 10.00 f-
C
BFIA A-6 surface 85 3.69 0.9940 3.50 5.00 6.30 (/1
BFIA A-6 base 55 10.38 0.9973 9.80 13.00 15.50 I--
BFIA A-6 base 70 5.56 0.9956 5.80 8.20 10.00
BFIA A-6 base 85 5.00 0.9929 3.50 5.00 6.30 O
~D

-I
0
a WSU = Washington State University. .<
b TAI = The Asphalt Institute. -I
c WES = Waterways Experiment Station, Vicksburg. m

d BFIA - Baltimore Friendship International Airport. -I

,--b

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
92 TESTING OF BITUMINOUS MIXTURES

and predictive techniques for the complex modulus testing on asphalt mixtures.
The f'mdings for each of these factors are presented in the following paragraphs
in a format consistent with The Asphalt Institute's recommended procedure for
conducting the complex modulus test.

Test Variables
1. The complex modulus is comprised of an elastic component (dynamic
modulus, IE*l)and a viscous component (phase lag or angle, ~). Because the
effect of test variables may act differently upon each parameter, IE*l and ~, it is
imperative that a set of test conditions be selected in accordance with the
anticipated use of the results.
2. The two most important variables affecting the complex modulus value,
for a particular asphalt mixture, are the test temperature and load frequency.
Accordingly, appropriate levels of temperature and frequency should be selected
to cover the anticipated range expected to occur under the field conditions. Test
temperatures of 40, 70, and 100~ along with factorial levels of load frequencies
of 1, 4, and 16 Hz have been found to be suitable for most anticipated
temperature and load frequency conditions.
3. The selection of an appropriate testing mode should be made on the basis
of the proposed use of the results. If the results are to be used exclusively in a
linear elastic layered model, a compression mode of test can be utilized. If a
viscoelastic utilization of the data is to be made, it is felt that the
tension-compression mode would be better suited for the test procedure because
of the differences found in the phase angle between the two testing modes.
4. The elastic component of the complex modulus, IE*l, is linear up to an
axial load stress of at least 70 psi when tested in a compression test mode at
temperatures of up to 100~ on a dense graded asphalt concrete mixture. For
the tension-compression test mode, nonlinearity at 86~ has been found at
about _+25 psi. Little or no variation in the viscous component, ~, was observed
within the stress conditions noted for each test mode. Accordingly, an axial
stress level recommended for compression mode testing is 35 psi. Caution should
be exercised to ensure that the specimen will not fail under this stress level when
tested at conditions of 100~ and 1 Hz load frequency. If this occurs, the axial
stress should be reduced to 17.5 psi.
5. Based upon limited observations the dynamic modulus of asphalt mixtures
is independent of confining pressure. Further work is needed to ascertain what
influence, if any, confining pressure has on the viscous portion of the complex
modulus test.
6. Comparable mean test values can be obtained from laboratory molded
specimens and field cored specimens.
7. The dynamic modulus can be measured equally well with strain gages or
LVDT's. No information is available to show that similar results in the phase
angle will be obtained through both measuring devices.
8. Both dynamic modulus and phase angle test results using strain gages both
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized
WlTCZAK AND ROOT ON MODULUS LABORATORY TEST 93

13/16 and 3 in. long resulted in essentially identical mean values. The data
scatter with the longer gage was less than that observed with the smaller gage.
9. Significant differences in the phase angle determination on a given
specimen were observed depending upon whether the measurement was made on
the loading or unloading portion of the load curve. Therefore, until future
research proves different, it is recommended that the phase lag be computed as
the average of the phase lags determined from the loading and unloading curve
portions.

Variability
1. No test result bias was found to exist, in either dynamic modulus or phase
angle measurements, between replicate tests conducted on the same specimen
for temperature and load frequency combinations of 40, 70, and 100~ and 1,
4, and 16 Hz, respectively. It can be concluded that the replicate test procedure
employed resulted in random measurements between replicates.
2. Based upon a statistical study, the primary component variances in
measuring the dynamic modulus on either laboratory fabricated specimens or
field cored specimens are the specimen to specimen variability, %2, and the
testing variability, at 2 .
3. When measuring the phase angle, the primary variance component is only
the test variability, at 2 , on either lab or field specimens from a given project.
4. Regardless of the temperature and load frequency test conditions, each
variance component increased linearly with an increase in the mean measure-
ment value for both dynamic modulus and phase angle determinations. This
indicates that a better overall indicator of variability is the coefficient of
variation, CV.
5. For dynamic modulus measurements, the average C V value for specimen
to specimen variability (CVc) was found to be 16.6 percent for field cored
specimens and 7.8 percent for laboratory specimens. The test variability, CVt,
was found to be 4.3 percent for field cored specimens and 7.8 percent for
laboratory prepared specimens.
6. The test variability associated with the phase angle determination was 13.3
percent (CVt) for field cored specimens and 14.3 percent for laboratory
specimens.
7. It is recommended that when testing cores obtained from existing
pavement structures, a total of six cores be used with one set of test
measurements per core. This results in a confidence interval range of + 13.2
percent of the mean dynamic modulus value, regardless of test temperature and
frequency conditions. The subsequent limit of accuracy for the phase angle
determination using this sampling procedure would be near -+ 11.5.
8. When laboratory specimens are prepared for testing, it is recommended
that three specimens and two sets of measurements per core be used. This results
in a confidence interval range of -+ 10.9 percent of the mean dynamic modulus
value and + 11.5 percent for the phase angle measurements.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
94 TESTING OF BITUMINOUS MIXTURES

Prediction of Dynamic Modulus


1. The elastic part of the complex modulus may be predicted at any load
frequency from the results o f flexural stiffness values obtained by flexural
fatigue beam tests.
2. The flexural stiffness, Es, of any mixture is a function o f the applied
flexural stress level, o. The relationship between Es and o is of the form Es =
EoA1 o.
3. The Eo value is the extrapolated Es value when the flexural stress value is
zero.
4. Once the Eo value for a particular mixture and test temperature has been
determined from a regression analysis of the flexural stiffness results, the
dynamic modulus at the same temperature can be predicted for any desired load
frequency by the predictive equation noted in the report.

References
[1 ] Papazian, H. S., Proceedings, International Conference on the Structural Design of
Asphalt Pavements, Aug. 1962, p. 454.
[2] KaUas, B.F., Proceedings, Association of Asphalt Paving Technologists, Vol. 39,
1970, p. 1.
[3] Pagan, C. A., Highway Research Record, No. 67, April 1965, p. 1.
[4] Coffman, B.S., Proceedings, Second International Conference on the Structural
Design of Asphalt Pavements, Aug. 1967, p. 819.
[5] Kallas, B.F. and Riley, J. C., Proceedings, Second International Conference on the
Structural Design of Asphalt Pavements, Aug. 1967, p. 931.
[6] Kingham, R. L and Kallas, B. F., Proceedings, Third International Conference on the
Structural Design of Asphalt Pavements, Sept. 1972, p. 849.
[7] Witczak, M.W., "Design Analysis-Asphalt Concrete Overlay Requirements for
Runway 18-36, Washington National Airport," Research Report 72-4, The Asphalt
Institute, College Park, Md., 1972.
[8] Witczak, M. W., Proceedings, Third International Conference on the Structural Design
of Asphalt Pavements, Sept. 1972, p. 550.
[9] Pagan, C. A.,Proceedings, Third International Conference on the Structural Design of
Asphalt Pavements, Sept. 1972, p. 290.
[10] Shook, J.F. and Kallas, B. F., Proceedings, Association of Asphalt Paving Techno-
logists, Vol. 38, 1969, p. 140.
[11] Coffman, B. S., Kraft, D.C., and Tamayo, J., Proceedings, Association of Asphalt
Paving Technologists, Vol. 33, 1964, p. 54.
[12] Barksdale, R. D., Highway Research Record, No. 345, 1971, p. 32.
[13] Coffman, B. S., lives, G., and Edwards, W. F., Highway Research Record, No. 329,
1970, p. 65.
[14] Cragg, R. and Pelt, P. S., Proceedings, Association of Asphalt Paving Technologists,
Vol. 40, 1971, p. 126.
[15] Awed, I.S., Rutz, F.R., and Terrel, R.L., Proceedings, Association of Asphalt
Paving Technologists, Vol. 42, 1973.
[16] Majidzadeh, K., private communication, 29 March 1973.
[17] Kallas, B.F., Proceedings, Association of Asphalt Paving Technologists, Vol. 39,
Mixtures, ASTM STP 508, American Society for Testing and Materials, 1972, p. 47.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
C. L. S a r a f I a n d K a m r a n M a ] i d z a d e h I

Dynamic Responseand Fatigue


Characteristics of Asphaltic Mixtures

REFERENCE: Saraf, C. L. and Majidzadeh, Kamran, "Dynamic Response and


Fatigue Characteristics of Asphaltic Mixtures," Fatigue and Dynamic Testing
of Bituminous Mixtures, ASTM STP 561, American Society for Testing and
Materials, 1974, pp. 95-114.
ABSTRACT: This investigation deals with the influence of several mix
variables in respect to asphaltic concrete on its fatigue response.
The investigation was based on a simple model representation comprising a
beam supported on elastic foundation subjected to sinusoidal repeated load-
ing, thus providing a two-dimensional simulation of actual pavement condi-
tions under traffic loads. The dimensions of the beam were designed such that
plan strain conditions prevailed during fatigue loading.
Fracture mechanics concepts using linear elastic theory were employed to
analyze test data, including stress and deformation near the crack tip. It was
also hypothesized that the parameter, A, in the rate-of-crack-growth equation,
dc/dN = A K 4 would adequately reflect the changes induced in the fatigue
response of the mix by the selected mix variables.
The effect of mixture variables such as the amount of binder, density,
aging, and viscosity on the parameter A is investigated,
KEY WORDS: fatigue tests, fracturing, dynamic modulus of elasticity,
mixtures, asphalts, viscosity, density

The structural deterioration of pavements due to repeated load applications


and the interaction with adverse environmental conditions appear in two distinct
forms: cracking and rutting. The problem o f damage accumulation and resulting
pavement deterioration due to cracking has been extensively studied over the
past three decades. In these studies, the damage process has been associated with
the repeated application of stresses inducing fluctuating stresses and strains in
the asphaltic-bound layer. In one method o f analysis, often designated as the
phenomenological approach, the pavement life expectancy is expressed as the
nonlinear-inverse relation o f induced stresses and strains in the pavement struc-
ture [1-7] .~
The mechanistic model o f pavement damage, however, utilizes the fracture

1Research associate and professor, respectively, Department of Civil Engineering, Ohio


State University, Columbus, Ohio 43210.
2The italic numbers in brackets refer to the list of references appended to this paper.
95
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Copyright9 1974 byby
Downloaded/printed ASTM International www.astm.org
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
96 TESTING OF BITUMINOUS MIXTURES

mechanics principles to describe the interrelation of the pavement life exhaus-


tion process and the loaded material and geometrical variables involved.
In this mechanistic model, fatigue is considered as a damage process, when
under a given state of stress, damage grows according to a crack propagation
model from an initial stage to a critical level. The damage model utilized in this
method of analysis is a power law relating the rate of damage accumulation and
the parameters associated with the material and geometric characteristics. The
damage model is given by

ac l,~V = A K "

where A and n are material constants and K is stress intensity factor. The stress
intensity factor K, in general is written as

on = r -~ (K, ~i' (0) + K, , h / ~ (0) + K , , ,hi ~' ')

where r and 0 are polar coordinates introduced at the crack tips and symbols 1,
11, and 111 connotate different modes of fracture. The Mode I or bending mode
is probably the most common type of fraction in pavement systems. The inter-
relation between stresses and stress intensity for Mode I fracture is given as
follows

"1 1 - s i n 0__ sin 30 -]


~xx [ KI 2 2
0 1+ sin 0 sin 30
OyyJ ~l~r I

- - COS - -

2 2 2

Oxy sin O sin 30


2 2

The laboratory analysis of asphaltic mixtures has indicated that bending


mode fracture is predominant in mixtures having small size aggregate as in
asphaltic surface course mixtures and sand asphalts. However, in mixtures having
large size aggregates, the crack propagation does not remain perpendicular to the
direction of the principle stress responsible for crack opening. For such mix-
tures, the crack surface follows a zigzag pattern. For an irregular crack pattern,
the combined effects of Modes I and II must be incorporated in the analysis.
Previous research [8-11] has indicated that for fine-grained asphaltic mix-
tures, such as sand asphalt and asphaltic concrete surface course, the exponent n
in the damage equation is ~_4.0. However, for mixtures with large size aggregate,
due to irregular crack pattern, the experimental results have indicated that n is

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
SARAF AND MAJIDZADEH ON ASPHALTIC MIXTURES 97

smaller than 4.0. The parameter A in the damage equation is known to be a


material constant and related to mixture characteristics..It has been previously
shown [11] that fatigue crack growth parameter A is independent of mode of
loading and geometric characteristics of the specimen.
The experimental results on fatigue and fracture studies of asphaltic mixtures
have indicated that the parameter A in the proposed damage model is the same
for both load and stroke mode of loading. Similarly, a fatigue damage model
obtained for beam on elastic foundation test configuration is also applicable for
pavement systems as slabs resting on elastic foundation. Therefore, the para-
meters A and n of the damage model, when determined for a given paving
material, can be utilized to predict the performance of pavement system under
actual field conditions. In this paper, the effect of various mixture properties on
the fatigue crack growth parameter A is presented.

Materials and Testing Procedures


Materials
In this study an asphaltic mixture conforming with the Ohio Department of
Transportation (DOT) specifications was used. All laboratory specimens for
surface course were prepared using limestone aggregates meeting the Ohio DOT
specification for Mix 404 and different grades of asphalt cements. Table 1 shows
the composition of this mix. Table 2 shows a typical grading of aggregates used
for laboratory mixtures.

TABLE 1-Item ,t04 asphalt cona, ete specifications (Ohio Department of Highways).

U.S. SieveSize %Passing, by weight


1/2 in. 100
3/8 in. 90 to I00
No. 4 45 to 75
No. 6 36 to 65
No. 50 3 to 22
No. 200 0 to 8
Design asphalt cement,
% by weight of total mix 4.5 to 9.5

TABLE 2-Composition of aggregate mix used in laboratory specimens.

U.S. SieveSize % Passing, by weight


1/2 in. 100
3/8 in. 100
No. 4 66.1
No. 6 58.5
No. 50 14.0
No. 200 3.0
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
98 TESTING OF BITUMINOUS MIXTURES

Three different penetration grades (60/70, 85/100, and 140/150) of asphalt


cement (AC) were used in this study. 851100 AC was aged in a thin-fdm oven
test at 4 2 5 ~ for 2, 4 and 6 h to obtain three additional viscosity grades of the
binders used. Measured viscosities of all six AC types used are shown in Table 3.

TABLE 3-Measured viscosities o f asphalts used in mixes.

Asphalts Viscosity, (dynes-s/cm2 X 106

60/70 2.10
85/100 1.65
140/150 0.45
2 h aged 3.30
4 h aged 4.80
6 h aged 5.30

TABLE 4-Mix specifications.

Binder
Mix Penetration Grade % Useda Density, lb/ft3

Laboratory Specimens:

AID4 - 6.5 60/70 6.5 132


AID2 - 6.5 60[70 6.5 144
AID1 - 6.5 60/70 6.5 138
AID1 - 6.0 60/70 6.0 138
AID1 - 7.0 60/70 7.0 138

AIID1 - 6.0 85/100 6.0 138


AIID1 6.5- 85/100 6.5 138

AIIID1 - 6.5 140/150 6.5 138

2 h aged 140/150 6.5 138


4 h aged 140/150 6.5 138
6 h aged 140/150 6.5 138

Plant Mix Specimens:

AIID1 - 6.0 85/100 6.0 138


AIID1 - 6.5 85[100 6.5 138

a Percentage by weight of aggregate for all laboratory specimens. Percentage by weight of


total mix for plant mix specimens.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
SARAF AND MAJIDZADEH ON ASPHALTIC MIXTURES 99

Asphaltic mixture was prepared in the laboratory by taking 4 kg of preheated


aggregate and mixing the same with the required amount of asphalt (calculated
on the basis of percent AC of total aggregate weight measured for each speci-
men).
A mold of size 2 by 2 by 24 in. was preheated to about 375~ filled with the
predetermined weight of the asphaltic mixture, and then compacted under static
load to the desired size (2 by 2 by 24 in. exact).
For notched specimens, a notch plate of required size was placed in the mold
before idling it with the mixture.

Testing Procedures
Fatigue Tests-Fatigue tests on beams were performed by resting them on an
elastic foundation (Fig. I). A dynamic load function of haversine shape as
represented in Fig. 2 was used to apply fatigue load (repeated loading) on the
specimen using an MTS electrohydraulic testing setup. To ensure complete re-
covery of the specimen before next load cycle, a rest period of 0.8 s was allowed
between each load application. The duration of load application in all tests was
kept constant at 0.2 s. However, the amplitude of load was varied.

10 psi
f'TTTD
72"
BEAM
g- C :'ack

/ 7
Smooth Interface
FOUNDATION 4"

24"
BASE
'1

FIG. 1-Beam on elasticfoundation (2-D system).

The specimen was periodically observed for any crack development during
fatigue testing. Ink-staining technique was used to locate a crack in the speci-
men. A log of crack length and the corresponding fatigue cycle at which it was
observed was maintained at regular intervals of the test duration. Since the
length of crack on the two sides of a beam are not necessarily identical, crack
growth on both sides was recorded during testing. In Fig. 3, the fatigue crack
propagation of an asphaltic concrete beam is shown.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
100 TESTING OF BITUMINOUS MIXTURES

LOAD

1 see. Scale = 1 lb./Dlv.

A
l i f l l l f l i l l l l l l
Dead Load

I I I I I I I I I I I I I

DE FLECTION
Scale = 2 x 10 -4 in./Div.

L
L ,.I I I, I I I I I I I I I j l * I s ~ ~ * I ~ I m j j

FIG. 2-Loading function used in fatigue tests.

All fatigue tests were controlled-stress tests, and the load was kept constant
and continuously monitored on the Brush recorder for any slight changes in
amplitude, which were corrected immediately upon detection.

Modulus o f Elasticity Tests-The same equipment used for fatigue tests was
employed in measuring the complex modulus, E*, of the specimen. In addition
to regular MTS equipment and Brush recorder, a separate linear variable
differential transformer (LVDT) was used for these tests. A specially designed
support enabled placing of the LVDT exactly beneath the loading point of the
beam. This testing arrangement is shown in Fig. 4.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
SARAF AND MAJIDZADEH ON ASPHALTIC MIXTURES 101

FIG. 3-Fatigue crack propagation o f asphaltic beam on elastic foundation.

FIG. 4-Modulus of elasticity (402 beam}.


Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
102 TESTING OF BITUMINOUS MIXTURES

In these experiments, 12-in.-long beams were simply supported on rollers


attached to a rigid plate and were tested for 10-in. span and 7 Hz frequency.
From the load deflection data, the dynamic modulus of elasticity, E*, was
determined using the simply supported beam formula making proper adjust-
ments for beam curvature.
The dynamic modulus of asphaltic concrete specimens was also determined
using cylindrical specimens of 5.6 in. in height and 2.8 in. in diameter. A half
sine compression load function was used for loading at a rate of 8 Hz along with
a small load ranging from 10 to 20 lb. The deflection was measured with a
LVDT placed underneath the aluminum plate as shown in Fig. 5. To ensure
reliability of results, deflection was measured at four perpendicular positions,
and an average value of measured deflections was used in the calculation of
dynamic modulus.

FIG. 5-Modulus of elasticity (301 cores}.

For specimens having nonuniform end condition, it was found necessary to


cap these specimens before testing.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
SARAF AND MAJIDZADEH ON ASPHALTIC MIXTURES 103

Analysis of Results
Introduction
Analysis of fatigue test data is a two-step process; namely, (1) stress intensity
factors, K, are calculated for the given geometry and boundary conditions and
(2) these K values are combined with the fatigue crack growth rate, dc/dN, to
obtain the material constants A and n, using the following formula

dr = AK n
dN

The finite element method is used to calculate the stress intensity factors. A
typical finite element mesh used for a given case is shown in Fig. 6.
Fatigue data are recorded in the form of c-N data, and the value of A and n
are calculated from this information and using dc[dN and K~c relations. The
experimental results on fine-grained asphaltic concrete surface course have
indicated that fatigue crack growth parameter n = 4.0. For asphaltic concrete
mixtures having large size aggregate, however, the results indicate that n is
smaller than 4.0 and ranges between 2.5 and 4.0 depending on mixture com-
position.
Therefore, considering n = 4 for f'me-grained asphaltic concrete mixtures, the
fatigue life is inversely proportional to the crack growth parameter, A, and is
given by

, 1
dc
Nr- ae" K/e4
Co

where

co = initial flaw or crack size at the commencement of fatigue loading,


cf = final crack size at the end of Nf fatigue cycles,
N/ = Final count of cycles when crack length is equal to cf, and
P = total applied load.

The parameters Co and cf are geometrical variables associated with flaws and
discontinuities in the mixture and the specimen thickness.
The parameter A is, however, influences by material characteristics of the
specimen. The dynamic modulus E* of the asphaltic mixture, in particular, has a
significant influence on the fatigue crack growth parameter A. Of course, the
dynamic modulus of the compacted asphaltic mixture is, in turn, affected by
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
Total Number of Nodal Points = 358; Total Number of E l e n ~ e n t s = 334 i11

J z
J
0
"11
J
-d
V C
2"
E Z
0
C
J

x
L c
.02" Bars
l J
--g-
0
m

T
J 0
L h
,. 12"
_!
F- r i

FIG. 6-Finite element mesh used for analyzing 2-in.-deep beams resting on 1-in.-deep
elastic foundation.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
SARAF AND MAJIDZADEH ON ASPHALTIC MIXTURES 105

other mixture variables as asphalt type, viscosity, percent binder, etc. The effect
of these material variables on the crack growth parameter, A is discussed in the
following sections.

Effect of Mixture Variables on Dynamic Modulus and Fatigue Parameters


Effect of Binder Aging on the Dynamic Modulus (E*J-85/lO0 penetration
asphalt was aged for 2, 4, and 6 h in an oven heated to a constant temperature of
425~ 6.5 percent of this grade of AC was used to compact beams of 2 by 2 by
12 in. size of 138 lb/ft 3 using 404 Ohio DOT mix.
Data from eight observations on each treatment (0,2,4, and 6-h aging) were
analyzed. The statistical analysis produced a highly significant value of the test
statistic indicating that the aging of binder has statistically significant affect on
the dynamic modulus of the compacted mix. This fact is illustrated in Fig. 7
which shows that the increased aging of asphalt produces higher dynamic
modulus of the compacted mix.

5
77 ~
t~

to

4 -

2 , I , [ i I I i I , I
0 2 4 6

Aging of Bh]der, hrs.

FIG. 7-Effect of binder aging on E* of the compacted mix.

Effect of Binder Viscosity on the Dynamic Modulus-Analysis of variance


performed on all six types of asphalt consistencies produced highly significant
value of the test statistic indicating that the effect of viscosity on the dynamic
modulus of the compacted mix is significant. Since the relationship between
viscosity of binder and E* of compacted mix is nonlinear on the arithmetical
scale, the data are plotted on a double logarithmic scale (see Fig. 8) showing that
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
10 6

O
"11

C
i

Z
O
C

X
"-4
C
m
f~

i0 ~

5 8 10 6 2 5 10 7

Viscosity ol A p h a l ' , dynes-second/era 2

FIG. 8-Mix dynamic modulus (E*) versus viscosity of asphalt (log-log plot).

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
SARAF AND MAJIDZADEH ON ASPHALTIC MIXTURES 107

the following relationship holds between E* and viscosity

E* = 3.12 x 104 (viscosity of binder) ~176

Effect of Asphalt Content on Dynamic Modulus (E*) of the Compacted


Mix-Three different asphalt contents were used in this study. Using 60/70
penetration grade of asphalt, 2 by 2 by 12 in. beams were compacted to 138
lb/ft a using 404 mix specified by the Ohio DOT. Asphalt contents of 6.0, 6.5,
and 7.0 percent were used in these beams.
The beams were tested as a simply supported beam (three-point load) at 8 Hz
with a haversine dynamic load of 10 lb amplitude. The results of these tests are
shown in Fig. 9.

4 . 0 x 10 5
77~

3.~

2.0 J , I J I
6.0 6.5 7.0

Asphalt Content, %

FIG. 9-E*: asphalt content.

It can be seen from this figure that 6.5 percent asphalt yields maximum E*
values. However, the difference for beams of different asphalt contents is not
very significant.

Effect of Compacted Mix Density on E* of the Mix-Specimens for this test


were compacted in the same manner as described earlier, except that in this case
6.5 percent AC was used for all specimens compacted to known densities of 132,
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
108 T E S T I N G OF B I T U M I N O U S M I X T U R E S

138, and 144 l b / f t 3 .


Test results for these specimens (tested as simply supported beams) are shown
in Fig. 10. A rising trend in theE* value with increase in compacted mix density
is noticeable. A significant increase in E* value was observed when the
compacted mix density was increased from 132 to 138 lb/ft 3 . Further increase
in density of the compacted mix by another 6 lb/ft 3 did not increase the E*
value in the same proportion as did the density range of 132 to 138 lb/ft 3 .

4.0 x 105 77~

"=3.0

,
132 138 144

Density of Compacted Mix, pcf

FIG. 10-E*: compactedmix density.

Effect of Mix Dynamic Modulus s


(E*) on the Stress Intensity Factor of
Cracked Beams-Stress intensity factors for beams with crack size of 0.2, 0.4,
0.6, 0.8, and 1.0 in were calculated by the finite element method using a
two-dimensional system as shown in Fig. 11.
Stress intensity factor, K, normalized with the total load,P, on unit width of
beam, for example, KIP values for different moduli of the beam, are shown in
Fig. 1 I. These values were calculated for the same foundation having modulus of
elasticity of 70 psi. It is evident from this figure that the KIP ratio increases with
increase in beam modulus, the rate of increase being greater for increasing crack
lengths. A typical K/P=c curve is shown in Fig. 12.
Effect of Binder Grade On Fatigue Parameter A-Values of fatigue
parameter, A, obtained for three different grades of asphalt were subjected to
analysis of variance. The analysis gave a test statistic value of 10.958 (F2,,~)

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorize
SARAF AND MAJIDZADEH ON ASPHALTIC MIXTURES 109

C = i"
11

de~
9
C = .8"
0

7 C = .6"

e~

C = .4"
5

L
C = .2"

250 300 350 400 450


Dynamic Modulus (E,b) of Beam, 103 psi

FIG. 11-Effect of beam modulus (E'b) on the K/P ratio (2-in.-deep beam resting on
4-in.-deep foundation ofEf = 70 psi.

which is highly significant, indicating that the grade of binder affects the fatigue
parameter, A. However, it was found that there was no significant difference
between the binder grades 60/70 and 85/100. Thus, to examine any variation in
parameter A, a large difference in the binder grades, as is the case with 85/100
and 140/150, must exist. A general trend as observed from this analysis is that
the increase in binder penetration grade increases the parameter A value. Since
the fatigue life, iV/, is inversely proportional to the fatigue parameter A, softer
grades of asphalt will show less endurance to fatigue loads than harder grades.
This conclusion is very similar to earlier results obtained by Monismith [51 and
Pell [1,2] under controlled-stress testing.

Effect o f Binder Aging on Fatigue Parameter A-Beams using 85/100


penetration grade asphalt, aged for 2,4, and 6 h at 425~ were tested under
fatigue loading. The values of fatigue parameter A, as obtained from these tests,
are plotted in Fig. 13. The effect of binder aging on fatigue parameter A is
significant as indicated by the test statistic value of F3,17 = 26.405. The
continuous line in Fig. 3 shows the average trend as obtained from these tests. It
is evident that parameterA decreases at a faster rate during initial aging than at a
subsequent stage of aging. This indirect trend indicates that slight aging of
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
110 TESTING OF BITUMINOUS MIXTURES

O
10

="

;4
2

0 , I , [ l l l l ~ [ , l l I J l , l , I
0 .2 .4 .6 .8 1.0

Crack Length, C, inches

FIG. 1 2 - K / P - C curve.

77OF

o Individual Observations
I
8
X
O O Average Values
<

o I , I , I , ,
0 2 4 6

Aging, hours

FIG. 1 3 - E f f e c t o f binder aging on fatigue parameter A.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
SARAF AND MAJIDZADEH ON ASPHALTIC MIXTURES 111

asphalt in the initial stage helps in improving its fatigue properties, but the
asymptotic trend toward increasing aging is definitely an indication of the limit
to this property. Although these results are in conformity with those obtained
by other investigators, the possibility of making the mix brittle and, hence, less
flexible cannot be ruled out completely if the aging is continued further.

Effect of Asphalt Content on Fatigue Parameter A-Figure 14 shows the


effect of binder content on fatigue parameter A. The continuous line in the
figure is drawn through the average values of the parameter A for three different
asphalt contents. A continuously decreasing trend with increase in AC indicates
that increase in AC from 6.0 to 7.0 percent is beneficial in improving the fatigue
life of the mix. However, the rate of decrease of A is much slower between 6.5
and 7.0 percent than between 6.0 and 6.5 percent AC in mix.

77~
1"0f C

,.~ .8

<

i6
.4

.2 I J I f I i I r I
6.0 6.6 7.0
BinderContent,

FIG. 14-Effect o f binder content on A.

Analysis of variance on these data gave a test statistic o f F 2 , 7 = 3.455, which


is not significant at 95 percent confidence interval. However, the difference
between 6.0 and 7.0 percent AC was significant at the same level of confidence.
This indicates that asphalt contents varying by 0.5 percent are not significant
from fatigue life point of view. This is an important conclusion for practical
applications of these results as specifications for mix design may require stricter
control over bitumen content in plant mix which may not be so important after
all.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
112 TESTING OF BITUMINOUS MIXTURES

Effect o f Dynamic Modulus {E*} o f Compacted Mix on its Fatigue Parameter


A-Results of all size mixes described earlier are combined to study the effect of
the dynamic modulus, E*, of the compacted mix on its fatigue parameter, A.
Figure 15 is the graphical plot of the data, where a circle represents the average
values and the lines extending on both sides represent the range of the observed
values. Statistical analysis of the data gave the test statistic, Fs ,22 = 22.19, which
is significant, indicating that the effect of E* on its fatigue parameter, A, is
statistically significant at 95 percent confidence interval. A dotted line drawn
through the mean values indicates that the value of parameter A decreases with

5!m
Test Series B, 77~

!,
10-12

\
10-13

\
I i I , I I\l I J
2 3 4 5 6 x IQ 5
D y n a m i c Modulus (E*) of the Mix, psi

FIG. 15-Effect of dynamic modulus of the compacted mix on its fatigue parameter A.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
SARAF AND MAJIDZADEH ON ASPHALTIC MIXTURES 113

increase in mix modulus. Since parameter A is inversely p~oportional to fatigue


life, N/,, it shows that mixes with higher modulus are likely to be stronger under
fatigue loads.
It is important to note here that Fig. 15 can be used to estimate the value of
fatigue parameter A if the dynamic modulus of the material is known. Since the
value of stress intensity factor can be determined from the considerations of
geometry and boundary conditions of the problem, it is possible to estimate the
fatigue life, N[, of the mix using Eq 1 for any given load, P.

Summary and Conclusion


Performance of any asphaltic mixture used in a pavement structure is
generally evaluated from several considerations, among which fatigue response is
perhaps an important one. Since the properties of the mixture are very much
dependent on the ingredients used, it is logical to study the problem by
investigating the effects of each ingredient on the overall performance of the
mixture.
The fracture mechanics concept using linear elastic theory has been employed
to analyze the stress and deformation near the crack tip. In this analysis stress
intensity factor K provides a single parameter characterization of the system
including the effects of specimen configuration, boundary conditions, and load.
The material constant A which is known to remain unaffected by the mode of
loading, can be determined from the crack growth rate by using the relationship

ac/aN=A(K)"
where the exponent n has been taken to be 4.
It has been hypothesized that the material constant A represents the material
characteristics when undergoing fatigue loads. Therefore, if the mixture
variables, asphalt grade, AC, and density of compacted mix are varied, these
variations should be reflected on the parameter A. With this assumption in mind,
studies were carded out in the laboratory using asphaltic mixtures containing
Ohio DOT Mix 404 and different penetration grades of asphalts.
The analysis is based on linear elastic theory, and it is assumed that the
amount of creep in the beam was negli#ble compared to live load deformations.
Following are the conclusions that emerge from this investigation:
1. The dynamic modulus of the compacted mix increases with increase in the
binder viscosity.
2. There is an optimum asphalt content for maximum dynamic modulus of
the mix, for a given grade of asphalt.
3. The stress intensity factor K for beams representing different mixes
increases with increase in the dynamic modulus of the mix. The rate of increase
is very small for beams of small crack length but becomes more pronounced as
the crack length grows bigger.
4. The fatigue parameter A decreases as the dynamic modulus of the mix
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
114 TESTING OF BITUMINOUS MIXTURES

increases, provided the mix has been compacted to the same density and tested
on the same foundation.
5. Asphaltic beams tested on an elastic foundation with different dynamic
moduli show a unique relationship between their fatigue life, Nf, and fatigue
parameter, A.
6. Aging of asphalt, to a certain extent, increases the viscosity and dynamic
modulus and, subsequently, improves the fatigue life o f asphaltic mixtures.

References
[1 ] Pell, P. S., "Fatigue Chaxacteristics of Bitumen and Bituminous Mixes," Proceedings,
First International Conference on the Structural Design of Asphalt Pavements,
University of Michigan, Ann Arbor, 1962.
[2] PeU, P. S., "Fatigue of Asphalt Pavement Mixes," Proceedings, Second International
Conference on the Structural Design of Asphalt Pavements, University of Michigan,
Ann Arbor, 1967.
[3] Monismith, C. L., "Flexibility Characteristics of Asphaltic Paving Mixtures,"
Proceedings, Association of Asphalt Paving Technologists, VoL 27, 1958.
[4] Monismith, C. L., Secor, K. E., and Blackmer, E. W., "Asphalt Mixture Behavior in
Repeated Flexure," Proceedings, Association of Asphalt Paving Technologists, Vol.
30, 1961.
[5] Mortismith, C. L., "Asphalt Mixture Behavior in Repeated Flexure," Report No. TE
67-4, University of California, Berkeley, 1967.
[6] Monismith, C. L., Epps, J. A., and Kasianchuk, D. A., "Asphalt Mixture Behavior in
Repeated Flexure," Report No. TE 68-8, University of California, Berkeley, 1968.
[71 Epps, J. A. and Monismith, C. L., "Influence of Mixture Variables on the Flexural
Fatigue Properties of Asphalt Concrete" in Proceedings, Association of Asphalt
Paving Technologists, VoL 38, 1969.
[8] Majidzadeh, K., Chan, A. T., and Ramsamooj, D. C., "Fatigue and Fracture of
Bituminous Paving Mixtures," Annual Meeting of the Highway Research Board, Jan.
1970.
[91 Majidzadeh, K. and Kanffman, E. M., "Analysis of Fatigue and Fracture of
Bituminous Paving Mixtures-Phase II," The Ohio State University Research Founda-
tion Final Report, Project RF 2845, Sept. 1971.
[101 Majidzadeh, K. Kauffman, E. M., and Saxaf, C. L. in Fatigue of Compacted
Bituminous Aggregate Mixtures, ASTM STP 508, American Society for Testing and
Materials, 1972, pp. 67-84.
[11] Ramsamooj, D. V., "'Analysis and Design of the Flexibility of Pavements," Ph.D.
thesis, The Ohio State University, Columbus, 1970.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
J. Morris I and R. C. G. Haas I

Dynamic Testing of Bituminous


Mixtures for Permanent
Deformation Response

REFERENCE: Morris, J. and Haas, R. C. G., "Dynamic Testing of Bituminous


Mixtures for Permanent Deformation Response," Fatigue and Dynamic Test-
ing of Bituminous Mixtures, ASTM STP 561, American Society for Testing
and Materials, 1974, pp. 115-131.
ABSTRACT: Permanent deformation, fatigue, and shrinkage fracture are the
three major distress mechanisms found in flexible pavements. To date, how-
ever, little quantitative information is available on the fundamental material
characteristics related to load-associated permanent deformation. The response
and characterization of paving materials have proven to be complex problems.
This situation has retarded the development of a reliable deformation predic-
tive technique and, consequently, a systematic pavement design process.
This paper briefly discusses available permanent deformation predictive
approaches. It is indicated that current techniques are inadequate in their
present form. A new approach based upon laboratory tests under simulated
field conditions is advanced, and the essential operating characteristics of asso-
ciated materials characterization test equipment are outlined. The apparatus,
which includes a temperature-controlled triaxial test chamber, electrohyd-
raulic systems for independent control of axial and lateral dynamic stresses,
and ancilliary measuring instruments is described.
The equipment can be used to evaluate a wide variety of conditions and
paving materials. The test results have been used as input data to the develop-
ment of statistically based models for predicting the permanent deformation
of full-depth sections at the Brampton Test Road, Ontario. Excellent agree-
ment between measured and predicted values have been obtained.
KEY WORDS: fatigue tests, flexible pavements, design, deformation, bitu-
mens, material characterization, dynamic tests, electrohydraulic equipment,
triaxial tests

Background
In the past decade, highway and airfield authorities have expended consider-
able effort to develop "rational" methods of pavement design. A vast amount of
information has been accumulated; however, the knowledge gained has generally
not been coordinated to the best degree possible. The problem has been signifi-

t Research assistant and professor associate chairman, respectively, Department of Civil


Engineering, University of Waterloo, Waterloo, Ont., Canada. Mr. Morris is now airport
engineer, Department of National Defense, Ottawa, Ont., Canada.

115
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Copyright* 1974 by
Downloaded/printed byASTM International www.astm.org
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
116 TESTING OF BITUMINOUS MIXTURES

cantly advanced in recent years by employing the principles of system engineer-


ing. Today, the systems approach is accepted by all major pavement authorities
in North America as the best means of structuring and organizing pavement
design and management technology [1-5] .2
Pavement studies indicate that three principal structural distress mechanisms
exist [6, 7]. These are:
1. Load-associated fracture (namely, fatigue).
2. Load-associated permanent deformation.
3. Nonload-associated shrinkage fracture.
The fatigue and shrinkage fracture subsystems have been developed to a
working state [8,9]. Permanent deformation, on the other hand, has received
comparatively little attention. Existing technology attempts to preclude exces-
sive deformation by indirect empirical means but cannot predict its magnitude
under a given set of conditions. Permanent deformation is a major world wide
distress mechanism today [6,10]. The trend towards heavier axle loads, tire
pressures, and volumes will further increase the importance of permanent
deformation and augment the need for a predictive technique.

Scope and Ob/ectives of the Paper


The general purpose of this paper is to describe the procedures and test
methods developed at the University of Waterloo to characterize bituminous
concrete for the prediction of permanent deformation. More specifically, the
paper discusses:
1. Design approaches for predicting permanent deformation.
2. Operational characteristics of test equipment required to characterize
bituminous mixtures for permanent deformation predictions.
3. Test methods and apparatus developed to satisfy the operational require-
ments outlined in No. 2.

Design Approaches for Prediction of Permanent Deformation

The problem of predicting permanent deformation is largely one of materials


characterization. Ideally, if one could formulate constitutive equations for
paving materials and mathematically describe the input variables, then the
solution of permanent deformation would be a relatively simple matter.
The properties of paving materials, however, are dependent upon many
variables including the rate of loading, the stress levels, temperature, etc. To
date, it has not been possible to develop constitutive equations for even the
simplest of paving materials.
Existing pavement design procedures, therefore, are largely empirical and
based upon extensive experience. Material characterization, likewise, is generally
based upon an arbitrary test procedure.
Permanent deformation is indirectly controlled by specifying minimum layer

2The italic numbers in brackets refer to the list of references appended to this paper.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
MORRIS AND HAAS ON PERMANENT DEFORMATION 117

thicknesses and minimum material strength or stability indexes; its magnitude,


however, cannot be predicted.
Klomp and Dorman [11] have developed a semiempirical means of control-
ling permanent deformation in pavement subgrades by limiting the maximum
vertical stress on the subgrade surface. Again, no attempt is made to predict its
magnitude.
More recently, two direct permanent deformation predictive approaches have
been proposed [12]. These are:
1. The quasi-elastic method [11-15].
2. The linear viscoelastic method [16,17].
The first approach hypothesizes the existence of material deformation laws
but is nonoperational since the laws, if they exist, have not yet been determined.
The linear viscoelastic approach has a sounder theoretical base than the
quasi-elastic method, but the assumption that the materials are linear has been
shown to be an over-simplification [12,14,18]. A recent study [19] aimed at
assessing the capabilities of the method indicates that the theory appreciably
underestimates the actual permanent deformation. The research employed a
layered viscoelastic program based on a Voigt model.
The behavior of paving materials has been shown to be quite complicated
[20,21,28]. It would be extremely difficult and presumptuous, at this time, to
postulate mathematical relationships or laws which attempt to describe the
distortional characteristics. The approach adopted at the University of Waterloo,
therefore, is based on the assumption that the material's characteristics are best
obtained by conducting tests under simulated field conditions. Because of the
probable interactions between the variables and the need for estimates of error,
models are developed on a statistical basis. It should be noted that this approach
does not presume that a fundamental deformation law is nonexistent. In fact, it
should provide a useful guide if any such law does apply.
The Waterloo investigation is initially concerned with developing a permanent
deformation model for the bituminous concrete employed at the Brampton Test
Road, Ontario, Canada. Permanent deformation measurements of full-depth
asphalt sections at this site indicate that rutting is due almost entirely to defor-
mation in the bituminous concrete [22]. A means of comparing observed and
predicted behavior is, therefore, available.

Equipment Operational Requirements


The prime factors that must be considered when simulating field conditions
in the laboratory testing of a given bituminous material are temperature, stress,
and the number of stress repetitions. Reliable analytical methods, which predict
temperature and stress distributions in the bituminous layer of a pavement, are
available [18,23,24].
Figure 1 schematically shows typical temperature and stress variations in the
upper layer of a full-depth pavement at a point below the center of a statically
loaded area. If a moving wheel is considered, then the elements are subjected to
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
118 TESTING OF BITUMINOUS MIXTURES

CIRCULAR UNIFORM LOAD

NEU

I ASPHALT CO'CRETE I

C -- COMPRESSION
I - TENSION

(O) ELEMENTAL (b) DISTRIBUTION (c) DISTRIBUTION (d ) TEMPERATURE ,T ,


STRESSES OF oil OF 0"3 DISTRIBUTION

FIG. 1-Typical pavement stress and temperature distributions in center o f wheel paths
under static loading.

stress variations as indicated in Fig. 2. The presence of shear stress implies the
rotation of the principal stress trajectories as described in Refs 25 and 26.

STRESS ~ TIME
0-' I I" PULSE
TIME "l~ REST PERIOD L
c -- COMPRESSION

c t- TENSION

/NEUTRAL AXIS

(b) HORIZONTAL
STRESS TIME

o"3 t , ./
BELOW
~. ,,
-
NEUTRAL AXIS,.,.~-"

~TRANSVERSELY L ~
(c) SHR
STEE
ASR
SES TIME

LONGT
IUON
I AL
FIG. 2-Stress variations in typical asphalt concrete element due to moving loads.

It is not possible to completely reproduce these conditions in the laboratory


at the present time. The triaxial test, despite a number of shortcomings, is
considered by many experts to be the most reasonable device for characterizing
paving materials [6,25]. Barksdale [26] points up that for permanent deforma-
tion studies the application of vertical stress pulses rather than principal stresses
are more appropriate. The shape and duration of compressive stress pulses, that
should be employed in the dynamic testing of paving materials, are also reported

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
MORRIS AND HAAS ON PERMANENT DEFORMATION 119

by Barksdale. It is shown that in the case of full-depth pavements, sinusoidal


pulses are appropriate for the upper and middle elements of the asphalt, but
below these depths a triangular pulse is a better approximation.
Raithby and Sterling [27] reporting on fatigue studies in bituminous mix-
tures indicate that pulse shape is not as significant as rest period. Considering the
deficiencies associated with laboratory testing and using the philosophy that
permanent deformation characterization, at this stage, must concentrate on what
seem to be the most important factors, the application of sinuosidal pulses for
all elements seems justifiable.
Pulse times vary with the depth of an element and the speed of the moving
load. For full-depth construction employing 5 to 12 in. of bituminous concrete
and vehicle speeds of 50 to 60 mph pulse times of 0.03 to 0.05 a seem to be
appropriate [26].
Rest periods are a most significant factor in the fatigue life of bituminous
mixtures [27]. The effect of this variable on permanent deformation, however,
has not been established. McLean [28] using only some 103 load repetitions
reports that rest period is an important factor; Snaith [29], on the other hand,
employing many more applications than McLean, concludes that rest period is
not a significant parameter. Preliminary studies at Waterloo indicate that rest
periods are relatively unimportant in compression testing but significant when
tension is applied. In keeping with the philosophy stated earlier, rest periods,
should be treated as a factor of secondary importance at this stage. However, the
longest possible rest period, commensurate with the test time available, should
be employed.
A major hypothesis employed in the Waterloo investigation is that the
behavior of bituminous concrete near the surface, where o3 is compressive, is
substantially different from that near the bottom, where o3 is tensile. The
experimental program developed, therefore, provides for the derivation of two
independent models.
Temperature is a major factor in any work concemed with the behavior of
bituminous materials. In this study, temperature control to within +_.O.5~ is
considered desirable.
The system developed at Waterloo, therefore, is designed primarily as follows:
1. To operate on specimens 8 in. high by 4 in. in diameter under triaxial
conditions.
2. To apply stresses, ol and o3, dynamically, in-phase, sinusoidal in shape,
and with pulse time equal to 0.04. Stress ol, must be capable of application in
either a compressive or tensile mode. A range of rest period is also required.
3. To operate under controlled temperature conditions within +0.5~
accuracy.

Description of Apparatus and Test Methods


The basic testing approach employs an electrohydraulic closed-loop system
[30]. These systems have revolutionized materials testing in the past decade
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
120 TESTING OF BITUMINOUS MIXTURES

because of their versatility and ability to accurately control pressures, frequen-


cies, and pulse shapes.

LVOT
LVDT. . . . ~ SERVOVALVE
LIMIT
SWITCHES- - OIL
AXIAL LOAD
I~YSTEM
THERMO
TEMPERATURE
CONTROL SYSTEM
I COUPLE
AXIAL
LOAO
CONTROL~R
] COOLING ~
CONTROL --
PRESSU~[
TRANSDUCER LATERAL
PRESSURE
REATER~
Jr0
TRI A XIAI,,,
~ LOAO CELL SERVO CONTROL
SYSTEMS

LATERAL DEFORMATION
MEASUREMENT SYSTEM
r CELL
LATERAL
PRESSURE
SYSTEM

HYORAULIC
CONTROL MAIN HYDRAUI,,IC
SYSTEM

FIG. 3-General schematic of equipment.

A general schematic and photograph showing the arrangement of the


equipment are given in Figs. 3 and 4, respectively. The main components of the
system are:
1. Triaxial cell.
2. Axial and lateral (or confining pressure) systems.
3. Control system.
4. Temperature control system.
5. Measurement and data acquisition systems.
These elements are detailed in the following sections.

Triaxial Cell
The triaxial cell is standard in concept but highly "over-designed" in a struc-
tural sense in order to minimize the compliance of the system. The ends are
constructed in aluminum for lightweight and ease of machining. An 8-in.-inter-
nal-diameter steel tube which houses two pyrex glass viewing windows forms the
cylinder. Compliance is an important factor in systems where high pressures and
frequencies are employed. The cell and the filling operations, therefore, are
designed to minimize the possibility of air being trapped in the cell. Likewise,

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
MORRIS AND HAAS ON PERMANENT DEFORMATION 121

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
122 TESTING OF BITUMINOUS MIXTURES

high pressure fittings and stainless steel tubing are employed and kept to a
minimum.
An oil-storage vessel is located to provide a minimum head of approximately
12 in. above the cell. This enables the cell to be f'dled by gravity and also
minimizes the possibility of developing air bubbles during the filling operation.
The cell is emptied by applying a vacuum to the oil-storage vessel. Filling and
emptying operations are quick and trouble-free.
A leak-proof, durable membrane is essential in a closed-loop control system.
Various combinations of liquid, membrane type, and sealing technique have
been employed. At Waterloo, the most efficient and economic solution was
found to be a clear, mineral, additive-free oil with silicone rubber membranes
and O-ring seals. Silicone rubber is applied to O-ring grooves in the caps before
placing the membrane as an added precaution.

Axial and Lateral Pressure Systems


The axial and lateral pressure systems consist of hydraulic actuators and
servovalves fed by a 70 gal/min, 3,000 psi hydraulic power supply. A hydraulic
control is installed between the main supply and the systems to guard against
surge forces. Limit switches located on the axial ram and error detectors ensure
that the hydraulic power supply will shut down in the event of specimen failure
or a hydraulic fault, respectively.
The axial loading system comprises a l~-in.-diameter actuator capable of
exerting a maximum compressive force of approximately 3 kips. The lateral
pressure actuator is designed as a pressure-reducer. An increase in bore from l S
to 4 in. reduces the 3000 psi input pressure to a maximum of approximately 420
psi in the cell. This is well within the safe capacity of the chamber. This
arrangement also ensures that the main hydraulic supply will not be contamin-
ated in the event of membrane or specimen failure.

Control System
The actuator systems just described are regulated independently by means of
two servocontrollers. A block diagram showing one typical control loop is given
in Fig. 5. The desired load or stress on the specimen is set by adjusting the span
and set point controls on an input module. This operation produces the com-
mand signal. At the same time, a feedback signal indicating the actual load or
stress on the specimen is obtained by an appropriate transducer. The serve-
controller compares the command and feedback signals. If they are not equal,
then the servocontroUer produces an error signal proportional to the difference
which causes the actuator to stroke in the direction that reduces the error.
The feedback signals for the axial and lateral pressure systems are obtained by
means of a load cell and a pressure transducer, respectively. The load cell is
situated in the triaxial chamber below the test specimen to ensure that friction
in the axial ram is not included in the measurement of the load.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
MORRISAND HAASON PERMANENTDEFORMATION 123
FEEDBACK
AMPLIFIER

TRANSDUCER
SIGNAL
SERVO - -

CONTROLLER

COMMAND SIGNAL @
9~ SET SERVO
<~ POINT VALVE
STATIC < ~
SIGNAL _ ~

SIGN~._~SPAN

FIG. 5-7~pical servocontrol loop.

Although the span control produces command signals in direct proportion to


its setting, the amplitude of the controlled variable is not always directly
proportional to the setting. Factors such as size of servovalve, size of actuator,
specimen stiffness, etc., determine the dynamic performance capability of the
system. Proper span settings are obtained by monitoring the feedback signals on
an oscilloscope.
The desired pulse characteristics are controlled by combining function and
pulse generators as described in the following manner. The function generator is
set to provide a haversine pulse shape at a frequency of 25 Hz (that is, a pulse
time 0.04) in the gate position. In this mode, the function generator only pro-
duces pulses when triggered by an external positive-going pulse. If a pulse genera-
tor emits square waves as shown in Fig. 6 a, then the function generator will be
triggered at times corresponding to the points shownA 1, A2, etc. It can be seen
that frequency of the square wave controls the time between pulses on the
specimen and, hence, the rest period.
The function generator frequency will generally be much greater than that set
on the pulse generator, and, therefore, a number of pulses may be generated in
the time that the square wave is positive-going, namely, A 1, A a. This difficulty is
overcome by placing a capacitor in the circuit as shown in Fig. 6. The capacitor
differentiates the square wave into the form shown in Fig. 6 c and, therefore,
triggers the function generator once only at times shown A4, A s, etc.
If the two hydraulic systems are triggered at the same time, then the axial and
lateral stresses on the specimen will be found to be out of phase. The axial ram
bears directly on the specimen, but the response of the lateral system is delayed
by the inertia of the system. The two stresses can be brought into phase by
delaying the axial pulse. This is accomplished by connecting the differentiated

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
124 TESTING OF BITUMINOUS MIXTURES

CYCLE
(c) DIFFERENTIATED
TIME
POSITIVE PULSE
PULSE
TIME ~ A4 AS
At A3 A2

PULSE "

Ib)~ _+
(----O TO LATERAL

I~'---O TO AXIAL
SYSTEM
FUNCTION
GENERATOR

BI B2

r L

(e) DIFFERENTIATED
NEGATIVE PULSE

FIG. 6-Method of cycling and phasing pulses.

negative pulse to the axial function generator. The generated negative square
wave and its derivative are shown in Fig. 6 b and d, respectively. It can be seen
that single axial pulses will be generated at times shown B1, B2, etc. Hence the
phase difference between the systems can be controlled by adjusting the fre-
quency of the pulse generator square wave.

FIG. 7-Typical test stress pulses.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
MORRIS AND HAAS ON PERMANENT DEFORMATION 125

An oscillograph showing typical al and 03, pulses obtained in this study is


given in Fig. 7. It can be seen that good pulse shape and control are obtained.
The method employed to transmit axial tensile forces to the specimen is
illustrated in Fig. 8. Steel plates are epoxied to the ends of the specimen and
threaded into aluminum caps. The specimen ends are carefully sawn to ensure
good bonding and alignment.

IAL RAM

,UMINUM CAP

EEL PLATE

FUMINOUS
ECIMEN

EEL PLATE

UMINUM BASE

~.D CELL

FIG. 8-Details o f specimen capping and load cell.

The same approach is employed when applying compressive axial forces. In


the upper part of a pavement, the lateral stress o3 exceeds the axial stress trl.
The corresponding laboratory test, therefore, must be capable of applying a
negative deviator stress or a tensile force in the axial ram.

Temperature ControlSystem
Temperature control is a major requirement in the characterization of
bituminous materials. In this research, the specimen temperature during testing
is controlled by maintaining the temperature of the surrounding off to within a
tolerance of -+0.5~ The system employed has been shown diagrammatically in
Fig. 3. It consists of the following components:
1. Chromel heating dement.
2. Iron-constantin thermocouple.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
126 TESTING OF BITUMINOUS MIXTURES

3. Indicating thermocouple controller.


The arrangement operates as a closed-loop system similar to that outlined for
the servocontrol. The desired temperature (or command) is set on the controller.
If the oil is not at the required temperature as indicated by the immersed
thermocouple (or feedback) then the controller activates the heater until the set
temperature is achieved.
To prevent overshoot and cycling at the set point, the controller provides
proportional power output over an adjustable span. The power output initially is
equal to approximately 95 percent of the load rating, but, as the oil temperature
enters the proportional band, the power is reduced smoothly. When the oil
temperature reaches the set point a precise balance between power input and
losses is maintained by adjusting the reset control.
Under normal circumstances, the desired temperature is fixed on a set-point
scale. Resetting by this method can lead to an error up to +--1.5~ which is well
above the tolerance desired in this work. The experimental program adopted for
this investigation may be described as a two-level factorial design. By incorporat-
ing a two-way switch set at the experimental upper and lower temperature levels,
the repeatability of temperature control was diminished to + 0.5~
The controller behaves most efficiently when the minimum operating output
is approximately 20 percent of maximum load. Therefore, the cell includes
%-in.-diameter copper tubing wound inside the upper perimeter of the cylinder
for the passage of cold tap water. Oil temperature stability is generally achieved
within 40 m of heating but may involve some 2 h when cooling.
The heating element is helically wound, formed into a circular configuration,
and attached to the cell base to aid heat circulation. The thermocouple tip is
placed at the rnidheight of the specimen so that the effects of temperature
gradients within the specimen are minimized. Specimen temperatures were not
investigated in this work. Awad [31], however, has shown that stability can be
achieved within 5 h. If a specimen is set up at night and tested the following
morning, then temperature stability may be assumed.

Measurement and Data Acquisition Systems


In the course of a test, the axial and lateral stresses are monitored while the
longitudinal and radial deformations are measured at various times (or number
of repetitions). Details are given in the following discussion.

Axial and Lateral Stresses-As previously mentioned, the axial and lateral
stresses acting on a specimen are measured by a load cell and a pressure
transducer, respectively. The load cell is located below the specimen as shown in
the lower portion of Fig. 8. It consists of a thin-walled aluminum pedestal with
pairs of axial and circumferential strain gages mounted on opposite sides. The
full-bridge so obtained is completely insensitive to any loading other than that in
the axial direction [32]. The cell is designed for a maximum working load of

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized
MORRIS AND HAAS ON PE RMANENT DE FORMATION 127

2000 lb or approximately 160 psi on the (specimen) and behaves linearly over
this range.

Longitudinal Deformations-The longitudinal deformation is measured by


means of a linear variable differential transformer (LVDT) connected to the
axial ram. The output voltage, 5 V/in. of movement, is monitored on an oscillo-
scope. More sophisticated methods of measuring longitudinal deformations, em-
ploying LVDT's between sprung collars, have been developed [28,33]. These
were not seriously considered in this investigation because cell space was limited
and interference with lateral deformation gages could not be tolerated. Sub-
sequent test results indicate that lateral measurements are more important than
those in the longitudinal direction.

Radial Deformations-Specimen radial deformations are measured by deflec-


tion transducers mounted at three pairs of diametrically opposite points as
shown in Fig. 9. Details of the transducer and the method of attachment can be
seen in Fig. 10. The transducer consists of three main dements:

FIG. 9-Lateral deformation transducers on specimen and heater.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
128 TESTING OF BITUMINOUS MIXTURES

FIG. l O-Lateral deforraation transducer.

1. A beryllium copper beam supporting four active strain gages.


2. Aluminum arms which support the beam and forks.
3. Brass forks which house ~mbles.
The beams are designed to measure permanent changes in diameter up to
+-0.75 in. Since specimen bulging or necking induces twisting moments in the
transducer, gimbles are incorporated to permit rotations of the supporting pins.
The gages are fixed by Allen screws to pins protruding from the specimen.
The pins are screwed into thin disks which are epoxied to the specimen. A jig is
employed to ensure correct placement and alignment of the disks. During
assembly the pins are passed through the membrane, and, therefore, the sleeve
must be carefully resealed. A nut and washer on the pin are tightened so that the
membrane is squashed against the disk. If silicone rubber is now applied to the
general area, then a highly successful seal is obtained. The major requirements of
the gage measuring system are high stability and sensitivity. Stability is obtained
by employing a d-c power supply and d-c amplifier as shown in Fig. 11. A d-c
amplifier and differential comparator enable diameter changes in the order of
2.5 microstrain to be measured.
The transducers are calibrated using the apparatus shown in Fig. 12.
Accurately machined circular disks are placed on a central location pin, and the
corresponding transducer output is measured. An excitation of approximately
12 V provides a linear output of 1 V per 0.10 in. change in diameter.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
MORRIS AND HAAS ON PERMANENT DEFORMATION 129

I DC POWER
SUPPLY

TRANSDUCERS DIFFERENTIAL
COMPARITOR

FIG. 11 -Lateral deformation transducer measuring system.

FIG. 12-Lateral deformation transducer calibration device.

Conclusions
An electrohydraulic servocontrolled testing facility has been designed,
assembled, and employed to characterize bituminous materials for the prediction
of permanent deformation.
The equipment is capable of producing vertical and lateral pressure pulses
with rest periods in a temperature-controlled triaxial environment. A technique
for measuring lateral deformations of test specimens has also been devised.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
130 TESTINGOF BITUMINOUS MIXTURES

The test results have been employed to develop models for the prediction of
permanent deformations in full-depth sections at the Brarnpton Test Road,
Ontario, Canada. Excellent agreement between measured and predicted values
have been obtained [34].

Acknowledgments
The equipment described in this paper was designed for a research project
sponsored by Gulf Oil Canada Ltd. Their support, assistance, and cooperation,
particularly that offered by Ted I-Iignell, is gratefully acknowledged. Apprecia-
tion is also extended to the Ministry o f Transportation and Communications of
Ontario, particularly Fred Field and Bill Phang, for their cooperation in
supplying specimens for the test program.

References
[1] Haas, R. C. G. and Hutchinson, B. G., "A Management System for Highway
Pavements" in Proceedings, Australian Road Research Board, 1970.
[2] Hudson, W. R., Finn, F. N., McCullough, B. F., Nail K., and Valerga, B. A.,
"Systems Approach to Pavement Design," Final Report, NCHRP Project 1-10, Mater-
ials Research and Development Inc., March 1968.
[3] Hutchinson, B. G. and Haas, R. C. G., "A Systems Analysis of the Highway Pavement
Design Process," Highway Research Board Research Record No. 239, 1968.
[4] Hudson, W. R., McCullough, B. F., Scrivner, F. H., and Brown, J. L., "A Systems
Approach to Pavement Design and Research," Research Report 123-1, Texas
Highway Dept., Texas A & M University and University of Texas at Austin, March
1970.
[5] Wilkins, E. B., "Outline of a Proposed Management System for the Canadian Good
Roads Association," Pavement Design and Evaluation Committee, Proceedings,
Canadian Good Roads Association, 1968.
[6] Highway Research Board, "Structural Design of Asphalt Concrete Pavement
Systems," Special Report 126, 1970.
[7] Kasianchuk, D. A., Terrel, R. L., and Haas, R. C. G., "A Design System for
Minimizing Fatigue, Permanent Deformation and Shrinkage Fracture Distress of
Asphalt Pavements" in Proceedings, Third International Conference on Structural
Design of Asphalt Pavements, London, 1972.
[8] Structural Design of Asphalt Concrete Pavements to Prevent Fatigue Failure,
Highway Research Board Annual Meeting, Washington, 1973.
[9] Haas, R. C. G., "A Working Method for Design Asphalt Pavements to Minimize
Low-Temperature Shrinkage Cracking," Asphalt Institute Research Report No. 72-6,
July 1972.
[10] Proceedings, Third International Conference on the Structural Design of Asphalt
Pavements, University of Michigan, Ann Arbor, 1972.
[11] Klomp, A. G. J. and Dorman, G. M., "Stress Distribution and Dynamic Testing in
Relation to Road Design" in Proceedings, Australian Road Research Board, 1964.
[12] Griggs, G.E., "Permanent Deformation in Asphalt Pavements: Qnasi-Elastic and
Viscoelastic Approaches," report prepared for the Asphalt Institute, April 1972.
[13] Heukelom, W. and Klomp, A. J. G., "Consideration of Calculated Strains at Various
Depths in Connection with the Stability of Asphalt Pavements" in Proceedings, Sec-
ond International Conference on the Structural Design of Asphalt Pavements, Univer-
sity of Michigan, Ann Arbor, 1967.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
MORRIS AND HAAS ON PERMANENT DEFORMATION 131

[14] Remain, J. E., "Rut Depth Prediction in Asphalt Pavements," Centre de Recherehes
Routieres, Brussels, Research Report No. 150/JER/1969, Dec. 1969.
[15] Remain, J.E., "Rut Depth Prediction in Asphalt Pavements" in Proceedings, Third
International Conference on Structural Design of Asphalt Pavements, London, 1972.
[16] Perioff, W.H. and Moavenzadeh, F., "Deflection of Viseo-elastie Medium Due to a
Moving Load" in Proceedings, Second International Conference on the Structural
Design of Asphalt Pavements, University of Michigan, Ann Arbor, 1967.
[17] Moavenzadeh, F. and Elliot, J.F., "Moving Load on Visco-elastic Layered System
Phase II," Research Report R69-74, Department of Civil Engineering, Massachusetts
Institute of Technology, Cambridge, Mass., 1969.
[18] Morris, J., "The Prediction of Traffic-Induced Permanent Deformations in Flexible
Pavements," unpublished comprehensive report, Department of Civil Engineering,
University of Waterloo, Ontario, Canada, March 1971.
[19] The Asphalt Institute, Research and Development Progress, Sept. 1972.
[20] Busehing, H.W. Goetz, W.H., and Hart, M.E., "Stress-Deformation Behavior of
Anisotropic Bituminous Mixtures," Joint Highway Research Project, Purdue Univer-
sity, Lafayette, Ind., 1967.
[21] Barksdale, R.D. and Leonards, G.A., "Predicting Performance of Bituminous
Surfaced Pavements" in Proceedings, Second International Conference on Structural
Design of Asphalt Pavements, Michigan, 1967.
[22] Haas, R. C. G., Kamel, N. I., and Morris J., "Brampton Test Road: Analysis of Per-
formance by Elastic Layer Theory and Its Application to Pavement Design and
Management in Ontario," Ontario Joint Transportation and Communication Re-
search Program, Report RR 182, Ontario, Canada, Nov. 1972.
[23] Barbarber, E. S., "Calculation of Maximum Pavement Temperatures from Weather
Reports," Highway Research Board Bulletin No. 169, 1957.
[24] Christison, J. T. and Anderson, K. O., "The Response of Asphalt Pavements to Low
Temperature Climatic Environments" in Proceedings, Third International Conference
on Structural Design of Asphalt Pavements, London, 1972.
[25] PeU, P.S. and Brown, S.F., '`The Characteristics of Materials for the Design of
Flexible Pavement Structures" in Proceedings, Third International Conference on
Structural Design of Asphalt Pavements, London, 1972.
[26] Barksdale, R.D., "Compressive Stress Pulse Times in Flexible Pavements for Use in
Dynamic Testing," Highway Research Board Annual Meeting, Washington, D.C.,
1971.
[27] Raithby, K. D. and Sterling, A. B., "Some Effects of Loading History on the Fatigue
Performance of Rolled Asphalt," Transportation and Road Research Laboratory,
TRRL Report LR 496, 1972.
[28] McLean, D. B., private communication, 1972.
[291 Snaith, M. S., private communication, 1972.
[30] "An Introduction to M.T.S. Closed-Loop Testing Systems," MTS Systems Corp.
[31] Awad, I. S., "Characterization of the Stress-Strain Relationships of Asphalts Treated
Base Material," Graduate Report of the University of Washington, Seattle, Aug.
1972.
[32] Cook, N. H. and Rabinowicz, E., Physical Measurements and Analysis, Addison-
Wesley, Reading, Mass., 1963.
[33] Dehlen, G.L., "The Effect of Non-Linear Material Response on the Behaviour of
Pavements Subjected to Traffic Loads," Ph.D. thesis, University of California,
Berkeley, 1969.
[34] Morris, J., Haas, R. C. G., Reilly, P. M., and I-Iignell,E. T., "Permanent Deformation
in Asphalt Pavements Can Be Predicted," paper presented to Annual Meeting, Asso-
ciation of American Paving Technologists, Williamsburg, Va., Feb. 1974.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
W. J. Kenis I

Material Characterizations for


Rational Pavement Design

REFERENCE: Kenis, W. J., "Material Characterizations for Rational Pave-


ment Design," Fatigue and Dynamic Testing of Bituminous Mixtures, ASTM
STP 561, American Society for Testing and Materials, 1974, pp. 132-152.
ABSTRACT: This paper describes a basis framework which is composed of an
assemblage of mathematical models to form a structural subsystem intended
to be used in the structural analysis of flexible pavements. Emphasis is placed
upon describing the newer type laboratory tests which provide the necessary
inputs for the operation of the structural subsystem computer program pack-
age referred to in this paper as VESYS II. A method is presented whereby
variations of the material properties which might be expected to occur in the
field cart be determined from the results of the laboratory tests. Several exam-
ples of computer output are given as well as those significant limitations which
should be imposed on its use.
KEY WORDS: fatigue tests, material characterization, viscoelasticity, struc-
ture subsystems, computer programs

The ultimate objective of the Federal Highway Administration's (FHWA)


federally coordinated research project "New Methodology for Flexible Pavement
Design" is the development of a rationally based modular framework for the
analysis and design of bituminous pavements. The modular capability within
such a framework will allow the designer a choice of different mathematical
model formulations and will permit modification or replacement of existing
formulations to improve the prediction capability of any of the modular
subsystems. A total design framework developed at The Massachusetts Institute
of Technology (MIT) under FHWA Contract 7776 [1-3] 2 is shown diagrama-
tically in Fig. 1. It can be viewed as consisting of modular sets of subsystems
which when integrated together form a predictive pavement design system.
These methodology are presently being further developed and calibrated under
current FHWA contract and Highway Planning and Research (HPR) studies. It
should be noted that the solid lines connecting the subsystems result in an
optimized design which reflects cost and maintenance activities. Existing

1Highway research engineer, Federal Highway Administration, Depaxtment of Transpor-


tation, Washington, D.C. 20591.
2The italic numbers in brackets refer to the fist of references appended to this paper.

132
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Copyright* 1974 bybyASTM International
Downloaded/printed www.astm.org
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
KENIS ON RATIONAL PAVEMENT DESIGN 133
methods which follow such a scheme are those being developed under HPR
studies at the University of Texas and through the National Cooperative
Highway Research Program (NCHRP) research at the Texas Transportation
Institute [4,5]. The immediate objective of FHWA is the development of a
structural subsystem which is based upon fundamental principles. Before such a
subsystem can be incorporated in the overall framework, it must be verified that
the underlying principles can be utilized within reason to predict real pavement
behavior and distress. Thus the dotted lines infer an approach whereby a pro-
posed thickness design is analyzed for specific distress mechanisms and then
modified in accordance with limiting design design criteria established by the
state highway departments.

t Analysis 1
Pavements

Structural Analysis Procedure


SubsystemW/Modular ........
Components A
I
, t II
M;d;:e~antem~

[ I L=re]
PresentD~ign|

Highway Cost
Subsystem
1

FIG. 1-Pavements analysis and design.

A structural subsystem in the form of a computer program package and


known as VESYS II has been developed to provide the user with an analysis of
the structural integrity of a given pavement system. It is the intent of this paper
to present methods for characterizing the material properties basic for the
operation of the VESYS II program. Influences on these properties as time of
loading, temperature, roll of construction variability, and type of laboratory test
are also discussed in detail. A method is presented whereby the variability of a
material property can be determined in the laboratory. A short discussion of the
limitations and usage of the computer program is given in order to demonstrate
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
134 TESTING OF BITUMINOUS MIXTURES
the versatility of the program and to stimulate discussion. It is hoped that these
discussions will evolve as an invaluable tool for guiding the continued develop-
ment and refinement of the modular components.

Structural Subsystem(VESYSII) Interactions


The composition of the structural subsystem (VESYS II Computer Program)
consists of three basic models as shown in Fig. 2. Specified output from one
model is used as input to another. The internal construction and formulations of
the models can be found elsewhere [1-3, 6-10]. In order to familiarize the reader
with the primary operational techniques of the VESYS II program, a brief
description of each model and their interactions follows.

,n,~ ]/ J,
Material Properties
Time-Temperature [
Primary
J ResponseModules :
[ inputs
OistributiomOf
Traffic, Temperature
Shift Factor -[

Fatigue l J Distress
Characteristics] - [ IndicatorModules

Pllffofmsnoa
Indicator Modules

FIG. 2-Structural subsystem.

Primary Response Model


The primary response model presently consists of a three layer linear
viscoelastic boundary value problem which incorporates a probabilistic solution
to account for the stochastic nature of specified input variables. Output from
this model consists of distributions in terms of the mean value and variance of
the resilient (elastic) and accumulative (time-dependent) stresses, strains, and
deformations at any point in the pavement due to a stationary load applied at
the pavement's surface.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
KENIS ON RATIONAL PAVEMENT DESIGN 135

Distress Indicator Models


The distress indicators are computed assuming three independent load
associated failure mechanisms: (1) fatigue failure, (2) accumulative deformation
in the wheel paths, and (3) longitudinal roughness. The fatigue failure model
utilizes Miner's law for the prediction of the initiation of cracking. Development
of this law for use in pavement design and analysis is attributed to the work
accomplished at the University of California [8-10]. It has been modified to
account for variations of the fatigue characteristics and of the resilient strains on
the underside of the pavement's surface layer by incorporating closed form
probabilisfic solutions [1,2]. Inputs to the model are accordingly the radial
resilient strains occurring on the underside of the bituminous layer as computed
through the primary response model, the fatigue characteristics determined from
laboratory fatigue tests, and the magnitude and number of repeated loadings
applied to the surface. Output is presented as the mean value and variance of
square yards of cracks per 1000 square yards of pavement surface occurring at
any time prescribed by the user.
The accumulative deformation model utilizes a probabilistic form of Boltz-
man's superposition principle to predict wheelpath rut depth occurring under
repeatedly applied traffic. Inputs are the time dependent vertical deformations
computed from the primary response model and the magnitude, velocity, and
number of repeated wheel load applications applied at the pavement's surface.
Output is given in inches of wheelpath rut depth in terms of a mean value and its
variance after any number ofloadings as prescribed by the user.
The roughness model "utilizes the distributions of rut depth as predicted by
the accumulative deformation model and a spatial correlation coefficient which
represents the longitudinal variation of material properties along the roadway.
This coefficient is determined by the user to reflect the statistical correlation of
two points along the pavement at a given distance of each other. Roughness is
computed in terms of the American Association of State Highway and Transpor-
tation Officials (AASHTO) definition of slope variance.

Performance Model
The AASHTO definition for serviceability index is used as the model to
reflect this measure of pavement performance. More realistic models could even-
tually be used; however, there exists little information at present to make such
assumptions. The model is included as a first attempt to relate pavement distress,
computed using rational principles, to the pavement's serviceability life history.
Imputs include the distress computed by the damage indicator models after
prescribed periods of time. Inasmuch as the inputs are in terms of means and
variances, the output is given as the mean and variance of the serviceability index
at those specified times prescribed by the user.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
136 TESTING OF BITUMINOUS MIXTURES

Variability of Material Properties


A basic premise of the VESYS II structural subsystem is that of providing
each model with the capability to account for variations in material properties,
traffic, and environment. The distress indicators as discussed previously are
computed as probabilistic distributions. The reliability of these predicted
estimates depends in part on the designer's estimates of expected variations in
the input parameters, for example, traffic, environment, and material properties.
Because of past experiences the highway designer has had little difficulty in
estimating traffic or temperature distributions although some may question the
reliability of these estimates. On the other hand expected variations in construc-
tion practices, moisture contents, material quality, and type are in general quite
difficult to estimate. All of these parameters influence the behavior of the mater-
iat properties of the pavement layers, and it is this variability with which we are
concerned. The material property inputs to the VESYS II structural subsystem
include the expected in-service variability of both the fatigue characteristics and
the elastic viscoelastic properties of each layer of the pavement in the form of
probabilistic distributions. In keeping with the theme of this symposium the
following discussions are limited to describing certain behavioral characteristics
of bituminous concrete; however, the concepts may be applied in general to
other layer materials.

Variability of Creep Compliance


It is acknowledged that the total response behavior of any complex material
can only be obtained when its response under load excitation as well as its
rebound response behavior are known. As an example, a load pulse is applied to
a material specimen as shown in Fig. 3. The response is seen to consist of four
material responses. Under load the material responds instantaneously. It is not
known, however, whether this instantaneous behavior is entirely recoverable or
if some "plastic" nonrecoverable component is included. The material then
deforms with time. The time dependent response may consist of a recoverable
viscoelastic deformation and a nonrecoverable viscous deformation. The true
nature of the materials behavior is, however, revealed upon unloading. In this
example the permanent deformation is seen to consist of a viscous time
dependent plus plastic component. Such behavior exists for most pavement layer
materials, however, to greater or lesser degrees. In addition, as the load pulse is
repeated the response components will be different. For large numbers of appli-
cations of load, the plastic component may increase, however, at a decreasing
rate. Such nonlinearities as discussed here are being presently developed for
incorporation into the VESYS II modular structural subsystem. The ability to
continually upgrade such a modular system is inherent in the use of rational
principles and should not inhibit their use. It is practical usage which will in
general dictate those changes necessary for the evolvement of more reliable
models. The current version of the VESYS II program considers only the behav-

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
Excitation

Load Curve
9 Response / Unload Curve

Load
Amplitude
I' Dr

Time
'"-V-_.':::="- \ l'"'"'"" Ill
Z
_1
"1 e,r"
Load Duration o
Z
Time - T
~>
.-i
6
Z
9v = Viscous Strain ~>
f-
eVE = V i s c o e l a s t i c S t r a i n "~o
~>
9s = Elastic <
m
ep = "Plastic" Strain
rn
Z
-4
0
Ill
t.n
FIG. 3-Material response.
Z

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
138 TESTING OF BITUMINOUS MIXTURES

ior of materials under loading and, hence, the ensuing discussions will be con-
cemed only with this aspect of the material's behavior.
It is generally accepted that the behavior of asphalt concrete is time depend-
ent and that under certain limiting strain conditions it earl be characterized as a
linear and isotropic viscoelastic material [11,12]. It has also been demonstrated
that temperature has a major influence on the behavior of bituminous materials
and that a time.temperature superposition principle can be applied to predict
such behavior [12]. The influence of state of stress on material behavior is not
as well defined, although it has been indicated that confining stresses have more
influence on the material behavior at higher temperatures. One must be caution-
ed, however, to use the best engineering judgment in determining the conditions
for the conduct of the laboratory tests. For instance one cannot duplicate in the
laboratory the actual regimens of temperature distributions occurring in the
field. Nor can one duplicate the actual state of stress which exists in the pave-
ment. We do not know, for instance, the effects of compaction and its tendency
to generate "locked-in" stress conditions. These stresses may be a great deal
different than what we have normally assumed to exist in the pavement.
In order to look at some of these effects, we will det~me the viscoelastic
modular creep compliance function, D (t), determined from uniaxial tension or
compression tests with confinement by

ezz (t)
o (t) = (1)
Ozz - 2 ~t (t) or,

where

Ozz = axially applied load,


orr = confinement pressure,
ezz = axial strain, and
/a (t) = Poission's ratio.

The influence of temperature and type of test on D (t) can be seen from the
curves in Fig. 4. These curves were assembled from work conducted at Materials
Research and Development Corp. [11]. In general, it can be observed that the
influences on D (t) of conFmement are much less significant than are tempera-
ture or type of test. The most significant factor contributing to the variation of
this material property function is noted by observing the differences between
the results of tension and compression tests at high temperatures. The rate of
deformation is considerably greater under tension than it is under compression
at 100~
It was also demonstrated in the same research effort that asphalt concrete
closdy resembles a thermorheologic simple material (application of the time
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
60O

T .
/
/ C : Compression Test
w~7 ~ T = Tension Test
f wo/C = Without Confinement
o
f w / C = W i t h Confinement
~4oc Temperature = 100~ F
..... Temperature = 70~ F
a . . . . Temperature = 40~ F 7~
Note: Curves Obtained From Tests m
.! A t Approximately 4.5% Air Voids 2~

(J 0

(J
/ Z
"-n

-4
m200
| Spread in Test Results . Z
~>
~. -~ ~ C wo/C I-
"o
~>
100 <
,~ / ~Cw/C m
a:
~ ' ~ l ~ " ~ - --a~- ' - ' " C wo/C m
~ ~ ~-- T wo/Cj -" "" "" "~ . . ~. .~ . m~ ~ C w/C Z
A l l Type Tests
-4

1 10 100 1,000 m
Time After Loading (Seconds)
Z
FIG. 4-Variation of modular creep compliance with temperature and type test (data
obtained from Materials Research and Development, Inc. Contract 7319}.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
-4
m
lo 3 , j F

s
f O
"11
r m

z
O
c
o9

_8 101
c
~0
m

8 Uniaxial Compression
. . . . . . Confined Compression ,
.~ 1@ ~ _ _ ~ . ~
S Uniaxial Tension
I
Note: Curves Represent Average Values A t
0
Approximately 4.5% Air Voids

10-110-3 10- 2 10-1 100 101 102 103 104


1
105
Time After Loading, (Seconds)

FIG. 5-Curve fitted master/modular creep compliance, D (t) at reference T = 70*F for
all stress states tested in contract FH-J 1-7319 [after Materials Research and Development,
Inc. 7785 quarterly report).

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
m

11111.

| 0o=lO, 5=22 ~" l -~- ,L,. J


B

:~ 10.
s
s
Q. 7~
E Ill
0 Z

0
Z
m :33

0 -t

Z
r-
"o

Note: Curves Obtained Using Heukelom & Klomp's <


m
Nomograph And Test Results On Mixes A t
Approximately 4.5% Air Voids m
Z
-1
1 [ _k.____ 0
10-3 10-2 10-1 100 101 102 103 104 m
Time (Seconds) (I)

FIG. 6 - U n c o n f i n e d compressive master modular creep compliance D (t) at reference T =


70~F.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
142 TESTING OF BITUMINOUS MIXTURES

temperature superposition principle) and can be characterized as such provided


certain limitations are placed on the method of test used to formulate the
associated shift functions. Master creep compliance curves for uniaxial tension
and compression and for confined compression data are shown plotted in Fig. 5.
Because of the vertical logarithmic scale, the results of tension tests do not
appear to be a great deal different from compression test data. These differences
were demonstrated to be quite significant when fed into the VESYS II fatigue
failure module where cracking initiation was seen to occur five years sooner
using tension data [23].
So far we have only discussed those influences on a material property func-
tion which result from variations of temperature, type of test, and stress state. In
addition, material properties may vary considerably because of variations in
density, asphalt content, material inhomogeneities, methods of specimen fabrica-
tion, air void content, etc.
In order to emphasize the significance of the influence of air void content
variation on creep compliance, master creep compliance curves for air void
contents of 3 and 6 percent obtained from unconfined compressive tests are
plotted in Fig. 6. These curves were extracted from the research results
presented under contract 7785 with the Materials Research and Development
Corp. [141. The eye bar curves superimposed on the compliance curves at 0.01,
10, and 1000 s represent the spread in compliance associated with type of test as
evidenced in Fig. 5. These comparisons show that variation of air void content
has an equal or even greater influence on compliance variability as does type of
test.

Variability of Fatigue Characteristics


The fatigue failure model which is utilized in the VESYS II program is that
developed through the research work at the University of California [8-10]. The
fatigue characteristics are represented by the foUowing equation

AT/ X Kl = (l/Ae) K2

where

= number of cycles to failure,


Ae = tensile strain amplitude, and
K1,K2 = coefficients related to material characteristics.

In order to utilize this concept, sets of fatigue curves are obtained in the
laboratory similar to those shown in Fig. 7. The values of KI and K2 in the
preceding equation represent a measure of the vertical intercept and slope,
respectively. It is seen that KI and K2 differ considerably for changes in asphalt
concrete modulus which can also be related directly to temperature changes. In
the VESYS II program fatigue damage is computed using K's which relate to the
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
KENIS ON RATIONAL PAVEMENT DESIGN 143

u. t',,,
o
I-
~

~ o 6 ~ ~
o o~=~ ~ ~ ~==

('Ul/'Ul 1/) "u!eJtS 8u!pu*8 le!;!Ul

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
144 TESTING OF BITUMINOUS MIXTURES

daily or monthly temperature over the analysis period. The analysis also provides
for the variation of KI and K2 due to other causes. Some information has
recently been obtained at the University of California which indicates the fatigue
characteristics vary significantly for different air void contents. More research is
needed, however, to determine the precise orientation of these curves and to
what extent air void content changes have on these relationships. Later in this
paper a method is presented whereby the variation of compliance can be obtain-
ed in the laboratory from a knowledge of the air voids contents expected in the
field.

Other Layer Materials


In the previous sections, the basic viscoelastic material property functions of
asphalt concrete required for input to the VESYS II computer program were
defined. In addition, certain influences affecting these properties were discussed.
The material properties of other layers may also be defined in a similar manner.
Engineering judgment at this point, however, must be utilized as the behavior of
other layer materials may differ considerably. If it is known that certain layers
behave as elastic materials, then a viscoelastic characterization should not be
necessary. In these cases it is suggested that the resilient modulus be determined
for the stress levels and environment anticipated in the field [15,19]. A creep
test, however, will serve as an indicator of the material's elastic behavior, for
example, if a creep test is conducted on such materials then little or no time
dependency would be observed. Ongoing research has indicated that for these
materials the value of the creep compliance at 0.1 s is reasonably close to the
resilient modulus. In any event, the material property will vary for different
moisture contents, densities, stress states, and testing and fabrication techniques.
If the variations of the modulus which are expected to occur in a real pavement
system are known, then they may be utilized as input to the VESYS II program.
For other materials where it is anticipated that time dependent accumulative
deformations will take place, a viscoelastic characterization is appropriate [20].
Such materials as clays and silty soils would fall into this category. Changes in
moisture content are probably the most significant factors affecting the variabi-
lity of the behavior of these materials.

Suggested Laboratory Method for Estimating Compliance Variation


In the previous discussions the importance of recognizing those material
property variations which might be expected to take place in an in-service
pavement were emphasized. In order to account for these variations, one must
first determine the extent of those conditions occurring in the field which
significantly influence the materials behavior. If such temperatures, moisture,
densities, air void contents, and stress states are known, then laboratory speci-
mens can be tested under these conditions. Since a unique feature of the VESYS
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
KENIS ON RATIONAL PAVEMENT DESIGN 145

II computer program is that of accounting for the statistical variation of the


material properties it would be advantageous to attempt to explain how one
might obtain this information from laboratory tests.
It will be assumed that both air void content and creep compliance are
distributed normally and, therefore, can be described by the mean and variance
of the density distribution curves.3 Construction control tests might also be used
to establish a frequency distribution curve of air void content. Let us assume
that for a particular project, it is estimated that a mean air void content of 4.5
percent with a standard deviation of 1.5 percent is specified. If specimens are
fabricated to plus and minus one standard deviation of the mean air void content
and tested for determination of the creep compliance function, then the curves
shown in Fig. 6 might be obtained. The spread in compliance at any time can
then be used as a crude estimate of the standard deviation of compliance. At
1000 s, the compliances at plus and minus one air void standard deviation are
100 and 58 X 10 -6 psi -x . One may then estimate that the spread in compliance
over this range fall within plus or minus one standard deviation of the mean
compliance. One standard deviation ofD (t) may be given as

100 58
21X 10-6 psi-1
-

OD - -
2

and the estimated coefficient of variation C o i s

ao 21
CD - ~ - 79 = 0.27

Estimated values at three different times are

o o • 10-6 CD
Standard Deviation Coefficient of D X 10 -6
Time (s) o f D (t) Variation ofD (t) Mean ofD (t)
0.01 0.30 0.28 11
10 10.00 0.43 22
1000 21.00 0.27 79

The preceding method is not recommended to be used as part of an overall


analysis procedure; however, it does give some estimate of the material property
variability under the imposed normal distribution assumption. The tabulation
also demonstrates the large changes in oD with time relative to the smaller
changes in Co. The VESYS II computer program assumes Co to be constant
with time and that the standard deviation varies as a constant multiple of the
mean. An average value of CD equal to 0.33 would have been selected to
represent the variation in this set of curves.

3Indications are that compliancesor voids might be better representedby a chi square or
log-normal distribution.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
146 TESTING OF BITUMINOUS MIXTURES

Limitations and Use of the VESYS II Program


The VESYS II structural subsystem computer program package is currently
being assembled for use by highway agencies on a research basis. For instance,
research at the Penn State University will attempt to utilize the program to
predict cracking, rutting, and roughness of the accelerated test facility which is a
part of an HPR study with the Pennsylvania Department of Highways. Already a
great deal of work has been accomplished in cooperation with the Asphalt
Institute and National Crushed Stone Association to obtain the necessary mater-
ial characterizations. The material properties were obtained for anticipated varia-
bilities in the test track using procedures similar to those described in this paper.
Results of these tests should be available through the Pennsylvania Department
of Transportation (PennDOT) shortly. It is hoped that this study will provide
information which will allow the adjustment and calibration of the damage
models for more accurate predictions of distress.
Several of the damage models of the subsystems have been used with some
success to predict pavement behavior. HPR work with the California Department
of Highways is currently utilizing the fatigue failure model to predict pavement
cracking. Professor Monismith at the University of California has also used the
fatigue model to predict fatigue life, and it is indicated that within reasonable-
ness the model can be used for this purpose; however, local conditions will
dictate certain improvements and modifications to the model.
Staff research at FHWA [17,18,21] utilized a very early version of the rutting
subsystem (VESYS I) to predict accumulative deformation occurring in two
prototype flexible pavements under both static and repeated loads. The
predictive capability of the model is seen to be excellent for the static loading
case, Fig. 8. For the dynamic loading case, Fig. 9, predicted accumulative
deformation after 1000 load repetitions were approximately 10 percent of the
measured values. In each case, the creep compliance at 70~ was determined and
used for the asphalt concrete, and the resilient modulus was used for the
granular base and subgrade. Two factors would appear to have contributed to
the differences evidenced between measured and predicted values. Since the
granular base and subgrade are characterized as being elastic, accumulative
deformations do not occur in the analysis for these layers. It is not known if
these layers actually behaved elastically or if some permanent deformation did
take place. Probably the most significant factor which contributed to this
discrepancy involves the characterization of the asphalt concrete. As it was
demonstrated previously, asphalt concrete may behave differently under load
than it does upon rebound. In such cases the unload curve must also be used in
the analysis of the pavement. As the loading is repeated, residual deformations
would increase, however, at a decreasing rate. Based upon these findings the
current version of the VESYS II program is being modified through FHWA
contract with the University of Utah, to accept more generalized forms of
material characterization. These modifications will include use of the rebound
curves and will attempt to account for plastic deformations through the use of a
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
30

STATIC TESTS
27

24

-i-

~, 21

~ 18

Z
oI - 15
w m
,.1 z
u. 12
Q
O
z
< 9 ::o
I,,- ~>
0 .-I
I,,- !
Z
TESTS: I-
ASSUMED INSTANTANEOUS mm m am VESYS Z: "o
~>
- ELASTIC DEFLECTION <
i rn

ol l J I ,I I I m
Z
,01 .10 1.00 10.00 100.00 1000.00 10000.00
0
m
TIME, SECONDS
Z

FIG. 8-Deflection for 40 and 80 psi static Ioadh~g center location.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
10
DYNAMIC TESTS
9 Ill

B
0
"11
"r"
r
Z 7
E

Z
6 0
i c
Z
o_ 5:-
I--
r x
LU
.-I I c
t.I. 3o
ILl 4; 111
oo
..I
<
t- 3~m
o

MEASURED
2- RESILIENT
DEFLECTION
1--

0 i i l i I
.01 .10 1.00 10.00 100.00 1000,00 1000

REPETITIONS

F I G . 9-Deflection f o r 40 psi dynamic loading center location.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
KENIS ON RATIONAL PAVEMENT DESIGN 149

decay factor similar to those generated by Barksdale at the Georgia Institute of


Technology [22].
In addition to the previously mentioned usages of the models, sensitivity
analysis on a limited scale have also been performed [23]. One such analysis
indicated the following differences in fatigue life using the compliance curves of
Fig. 5 for uniaxial compression and tension

Type Test Fatigue Life, years


Uniaxiai compression
load duration = 0.05 s 7.7
Uniaxial tension
load duration = 0.05 s 1.1

Other limitations which should be mentioned are those pertaining to the deter-
mination o f the viscoelastic functions. Table 1 shows limiting strain levels which
have been recommended for obtaining consistent results [11]. This means that
the creep compliance functions, determined by limiting strain levels to the values
shown in the table, should enable one to construct well defined master compli-

TABLE 1-Limiting Strain Levels for Consistent Results (for determination of viscoelastic
functions, from MR&D final report, Contract FH-11. 7319).

Temperature = 40"F
Results are similar for stress levels and stress states utilized in this project.

Temperature = 70*F
Type of Test Limiting Strain Levels
Unconfined axial tensile strains = 200/.t in./in.
axial compressive strains = 1200/2 in./in.

Confined and unconfined axial compressive strains = 300 bt in./in.


radial tensile strains = 25/2 in./in.
Temperature = 100"F
Type of Test Limiting Strain Levels
Unconf'med axial tensile strains = -250/2 in./in.
axial compressive strains = 600/a in./in.
Confined and unconfined axial compressive strains = 270/.t in./in.
radial tensile strains = - 3 0 / ~ in./in.
Conclusion
If the axial compressive strains are kept below 270 p in./in., the results are independent of
t h e test method.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
150 TESTINGOF BITUMINOUS MIXTURES

ance curves over a rather wide range of temperatures. These recommendations


would appear within reason when compared with those actually measured in the
FHWA test pavements under repeated loadings at 70~ For the 6% and 289
bituminous pavements subjected to 80 psi load intensities, the accumulative
tensile strains on the underside of the surface layer did not exceed 180/a in./in.
[21].

Summary
A structural subsystem (VESYS II) developed for FHWA was described in
order to familiarize the reader with some of the significant parameters associated
with it. The program offers the design engineer a unique procedure whereby
selected pavement sections can be analyzed for their structural integrity. The
development of pavement distress is stochastic in nature; therefore, the predic-
tion process must also be stochastic. The VESYS II program incorporates a
probabilistic analysis so that the designer can account for those significant vari-
abilities which effect the structural integrity of the pavement. Ranges of com-
puter output in terms of distributions of pavement behavior and distress are
calculated provided realistic estimates of the variations of the input variable are
provided by the user. A method was presented whereby the designer could
estimate the variations expected in the viscoelastic modular creep compliance
function which are due to variations in air void content.
Laboratory tests conducted for determination of creep compliance on asphalt
concrete specimens should be limited to the strain levels listed in Table 1. It is
recommended for the present that the test method also be limited to confined or
unconfined compression tests in order that uniform results will be obtained. Use
of these data in the program, together with feedback information of pavement
performance, may indicate that other test procedures such as tension tests or the
use of higher than normally assumed confinement pressures would give more
realistic values of distress.
Inasmuch as the structural analysis system is being developed to be modular,
it will allow continual upgrading of analysis techniques and testing procedure.
This in turn will offer the design engineer to select alternate techniques for
analysis purposes when local conditions so dictate.
Subsequent to the writing of this paper major modifications have been made
to the structural subsystem models, and henceforth the computer programs shall
be referred to through changes in the Roman numeral prefLx to the VESYS
designation.

Acknowledgment
Work was conducted under the supervision and guidance of Dr. T.F. Mc-
Mahon. Credit is due Mrs. Sally DeCuzzi for typing the report and to the efforts
of HPR and contract research.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
KENIS ON RATIONAL PAVEMENT DESIGN 151

References
[1] Moavenzadeh, F., Soussou, J.E., and Findakly, H.K., "Synthesis for Rational
Design," (Vol. I) Final Report for FHWA Contract 7776, Federal Highway
Administration, Jan. 1974.
[2] Moavenzadeh, F., Soussou, J.E., and Findakly, H.K., "Synthesis for Rational
Design," (Vol. II) Final Report for FHWA Contract 7776, Federal Highway
Administration, Jan. 1974.
[3] Moavenzadeh, F., Soussou, J.E., and Findakly, H.K., "Synthesis for Rational
Design," (Operating Instructions and Documentation). Final Report for FHWA
Contract 7776, Federal Highway Administration, Jan. 1974.
[4] Lytton, R.L. and McFarland, W.F., "Systems Approach to Pavement Design-
Implementation Phase," Final Report for NCHRP Project 1-10A, National Coopera-
tive Highway Research Program, Sept. 1973.
[5] Hudson, W.R., McCullough, B. F., Scrivner, F. H., and Brown, J.L., "A Systems
Approach Applied to Pavement Design and Research," Research Report 123-1,
published jointly by Texas Highway Dept., Texas Transportation Institute, Texas
A&M University, and Center for Highway Research, University of Texas at Austin,
April 1970.
[6] Kenis, W. J. "Mathematical Models to Predict Pavement Response," FHWA Report
No. RD-72-9, Federal Highway Administration, April 1972.
[7] Kenis, W. J. and McMahon, T. F., "A Flexible Pavement Analysis Subsystem," paper
presented at the 53rd Annual Meeting of the Highway Research Board, Jan. 1974.
[8] Monismith, C.L., "Asphalt Mixture Behavior in Repeated Flexure," Report No.
TE68-8, Institute for Transportation and Traffic Engineering, University of Cali-
fornia, Berkeley, Dec. 1968.
[9] Kashianchuk, D.A., Monismith, C.L., and Garrison, W.A., "Asphalt Concrete
Pavement Design, A Subsystem to Consider the Fatigue Mode of Distress," HPR
N29I, Highway Research Board, 1969.
[10] Monismith, C. L. et al, "Asphalt Mixture Behavior of Repeated Flexure," Report No.
TE 70-5, Institute for Transportation and Traffic Engineering, University of
California, Berkeley, Dec. 1970.
[11] Nail K., Smith, W., and Chang, C-Y, "Materials Research and Development
Characterization of Asphalt Concrete and Cement-Treated Granular Base Course,"
Final Report for FHWA Contract 7319, Feb. 1972.
[12] Pagen, C.A., "An Analysis of the Thermorheological Response of Bituminous
Concrete," Ph.D. thesis, Ohio State University, Columbus, March 1963.
[13] Pagen, C. A., "Rheological and Compressive Strength Characteristics of Laboratory
and Field Compacted Asphalt Concrete Mixtures," Report No. EES 258-2, Ohio
State University, Columbus, Sept. 1968.
[14] Materials Research and Development, Inc., "Characterization of Untreated Granular
Base Course and Asphalt Treated Base Course," Final Report for FHWA Contract
7785, Federal Highway Administration, Oct. 1973.
[15] Nail K., and Chang, C-Y, "Translating AASHO Road Test Findings Basic Properties
of Pavement Components," Final Report prepared for the Highway Research Board
by Materials Research and Development, Inc., 1970.
[16] Hicks, R.G., "Factors Influencing the Resilient Properties of Granular Materials,"
Ph.D. thesis, University of California, Berkeley, 1970.
[17] Kenis, W.J., "Response Behavior of Flexible Pavements," paper presented at the
1973 annual meeting of Association of Asphalt Paving Technologists.
[18] Kenis, W.J., "Comparisons Between Measured and Predicted Flexible Pavement
Responses," FHWA Report No. RD-72-10, Federal Highway Administration, Jan.
1973.
[19] Monismith, C.L. and McLean, D.B., "Design Considerations for Asphalt Pave-
ments," Report No. TE 71-8, Institute for Transportation and Traffic Engineering,
Dec. 1971.
[20] Pagen, C.A. and Jagannath, B.N., "Theological Properties of Hybla Valley Clay,
Engineering Experiment Station," Report No. EES-248-7, The Ohio State University,
Columbus, 1969.
Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.
152 TESTING OF BITUMINOUS MIXTURES

[21 ] Kenis, W. J., "Responses of Prototype Flexible Pavements Subjected to Repetitive


and Static Loads," FHWA Report No. RD-72-8, Federal Highway Administration,
/an. 1972.
[22] Barksdale,R. D., "Repeated Load Test Evaluation of Base Course Materials,t' Georgia
Highway Department, Final Report No. 7002, Georgia Institute of Technology,
Atlanta, May 1972.
[23] Materials Research and Development unpublished Quarterly Report, Nov.-Feb. 1972,
FHWA Contract 7785, Federal Highway Administration.

Copyright by ASTM Int'l (all rights reserved); Thu Dec 22 07:48:30 EST 2011
Downloaded/printed by
Indian Institute of Technology Kharagpur pursuant to License Agreement. No further reproductions authorized.

You might also like