Você está na página 1de 9



Development and Evaluation of Dynamic

Flexural Beam Fatigue Test System
Although both mix variables and environmental variables are known to
affect the fatigue response of asphalt-aggregate mixes, other factors
including specimen fabrication procedure and test equipment and proceduresare equally important. The development of a dynamic flexural
beam fatigue test system is described, and the effects of specimen compaction method and equipment type on the precision of in situ fatigue
lives of asphalt-aggregate mixes predicted by using laboratory strainlife relationships are discussed. Results indicate a coefficient of variation of 41 percent in fatigue life for the new fatigue equipment compared
with one of 93 percent for an earlier electropneumatic version. The specimen compaction method was also found to influence significantly
the precision of the predicted fatigue life. A 33 percent difference in
coefficients of variation between the fatigue response of rolling wheel
compacted specimens and kneading-compacted specimens was observed.
Consequently, twice as many specimens are required to achieve a given
level of precision in in situ predicted fatigue life if kneading compaction
is used instead of rolling wheel compaction. Similarly, if a pneumatic
system and associated test procedure are used, approximately 12 times
as many specimens are required to achieve similar precision in predicted
fatigue life compared with the new servohydraulic fatigue test system.

The fatigue resistance of an asphalt-aggregate mix is its ability to

withstand repeated bending without fracture. Fatigue, one of the
common forms of distress in asphalt concrete pavements, manifests
itself in the form of cracking under repeated traffic loading.
Whereas simplified representations of fatigue response have been
developed (e.g., representative fatigue relationships, surrogate methods based on measurements of stiffness or dissipated energy, etc.), it
is generally accepted that accurate measurements require laboratory
testingfor example, using the third-point bending beam fatigue
test. In this test an imposed strain or stress is applied repetitively until
failure occurs, either by complete fracture or by a significant reduction in stiffness. The fatigue behavior of a mix is then characterized
by the relationship between strain or stress level and the number of
load repetitions to failure. Conventionally, in order to establish suitable relationships, repeated flexure tests may require as many as 36
specimens and up to 3 to 4 weeks of testing time for a mix.
Although mix variables (such as asphalt source, aggregate source,
and air void content) and environmental variables (such as temperature and moisture) are known to affect the fatigue response of
asphalt-aggregate mixes, other factors are equally important, including (a) specimen fabrication procedure and specimen size, and
(b) the test equipment itself. This paper describes some aspects of
the development of an improved dynamic controlled-strain flexural
beam fatigue (third-point loading) test system during Strategic
Highway Research Program (SHRP) Project A-003A. Discussed in
A. A. Tayebali, Department of Civil Engineering, North Carolina State University, Box 7908, Raleigh, N.C. 27695. J. A. Deacon, 3340 Hunter Road,
Lexington, Ky. 40502. C. L. Monismith, University of California at Berkeley, Richmond Field Station, 1301 South 46th Street, Building 452, Richmond, Calif. 94804.

this paper are the effects of specimen preparation method and equipment type on the fatigue response of asphalt-aggregate mixes. Physical improvements in the fatigue test module and the use of a servohydraulic mechanism significantly increased test reliability and
substantially reduced testing time. Consequently, a method has been
developed so that fatigue response of an asphalt-aggregate mix can
be reliably characterized in as little as 24 hr, thereby enabling a number of mixes to be evaluated in a relatively short period.


Beginning in the 1960s, results of flexural beam fatigue tests conducted at University of California at Berkeley (UCB) have been
reported in the literature (14). An early version of the third-point
flexural beam fatigue test equipment used for these tests, shown in
Figure 1, was the electropneumatic system. Load magnitudes were
regulated by air pressure, and loading time was controlled with
pneumatic logic control circuits. Note in Figure 1 the absence of any
load-measuring device within the equipment. The beam specimen
(38.1 3 38.1 3 381 mm) was held in place using four springloaded screw clamps; beam ends were constrained from moving
sideways by set screws. Deflections at the center of the beam specimen were measured using linear variable differential transducers
connected to a strip chart recorder. Testing was usually confined to
the controlled-loading mode, and the loading pulse was characterized by a square or haversine pattern with a frequency of 1 to 2 Hz
and a 0.1-sec loading time.
Later improvements to this same basic technology (Figure 2)
included the introduction of a load-measuring device to monitor the
load continuously and a computer-aided data acquisition and control system. Although the data acquisition system facilitated data
reduction and reduced the operator-related variance associated with
data reduction from strip chart recorders, overall testing time and
variability in test measurements remained relatively unchanged
because (a) specimen setup time did not change significantly,
(b) a frequency larger then 2 Hz could not be reliably achieved with
the pneumatic system, and (c) the magnitude of the load or deformation could not be controlled with a high degree of accuracy.

Improvements to Test Equipment

During the SHRP A-003A project, two major improvements were
made to the fatigue test equipment with the intent of not only improving the reliability of test results but also minimizing the setup and
testing time. These improvements included increasing the size of the
test specimen and designing and building a new fatigue beam module, one that could be used as a stand-alone device or adapted for use



FIGURE 2 Flexural beam fatigue electropneumatic test system

(computer-controlled data acquisition).

2. Linear and torsional bearings were improved to minimize any

extraneous stresses in the beam specimen and to maintain zero
moment at the beam ends.
3. Various components were redesigned or modified to accommodate the larger specimen and to adapt the module for accommodation in the SHRP Shear Test Equipment.
4. Temperature and test control, data acquisition, and data reduction were automated.


Flexural beam fatigue electropneumatic test

The new fatigue test equipment, operated by hydraulic pressure,

has better response and more precise control of induced stress and
strain in the specimen than did its electropneumatic predecessor. Sinusoidal, haversine, or step loads up to 20 Hz frequency, with or with-

within the SHRP Shear Test Equipment. Specific goals for improvements were targeted to increase the ease, simplicity, and reliability
of the fatigue test. Figures 3 and 4 show the new fatigue module and
an example of a stand-alone fatigue test system, respectively.

Specimen Size
The size of the test beam was increased from a 38.1- 3 38.1-mm
cross section to a rectangular cross section 50.8 mm high and
63.5 mm wide. This was the largest cross section that could be
accommodated given space restrictions within the SHRP Shear Test
Equipment. These restrictions also limited the beam length to
381 mm. However, the beam spanthe length between the reaction
pointswas increased from 305 to 356 mm to minimize shear
deformations in the beam.

Test Equipment
Specific changes in the test equipment included the following:
1. The test equipment was redesigned to simplify and reduce the
setup time. This was achieved by mechanizing specimen clamping
through the use of precision torque motors, reducing the original
setup time from 30 to 45 min to less than 5 min.


Dynamic flexural beam fatigue module.

Tayebali et al.


Specimen Compaction
Two types of compaction equipment were used for preparing
the flexural beam specimens. For the 2 3 2 and modified asphalt
mix test programs, specimens were sawed from 76.2- 3 76.2- 3
381-mm slabs compacted using a California kneading compactor.
For the 8 3 2 study, specimens were sawed from 610- 3 610- 3
76.2-mm slabs fabricated using a rolling wheel compactor.

Laboratory Test Programs

FIGURE 4 Stand-alone dynamic flexural beam fatigue test

system (courtesy of Cox & Sons, Colfax, California).

out rest periods, can be achieved at carefully controlled temperatures

ranging from 210 to 40C. The test itself, once the specimen is situated
in the loading frame (including temperature control), is completely
automated, including test control, data acquisition, and data reduction.

Specimen Fabrication and Laboratory Test Programs

Evaluation of the effect of compaction method and equipment type
in this paper is based on results of three extensive fatigue studies
conducted during the SHRP A-003A project. The first study, referred
to as the 2 3 2 study (because it involved two asphalts and two aggregates), was conducted during the pilot test program using the electropneumatic fatigue test equipment shown in Figure 2. This program
involved testing 16 mixes (128 specimens) under both controlledstress and controlled-strain modes of loading. The second and third
studies, called the 8 3 2 (because it involved eight asphalts and two
aggregates) and modified asphalt mix test programs, were carried out
using the new dynamic fatigue test module and involved testing 32
mixes (128 specimens) and 10 mixes (38 specimens), respectively.

The asphalts and aggregates were obtained from the SHRP Materials Reference Library (MRL). A total of eight asphalts and four
aggregates were used. Table 1 briefly identifies the binders and
aggregates used in the various studies, and Table 2 presents the target aggregate gradation for all mixes.

Salient features of the three test programs are presented in Tables 3

and 4. For the 2 3 2 program, the electropneumatic test system was
used. Testing was done in both the controlled-stress and controlledstrain modes of loading. Step loading was used for the controlledstress mode and haversine loading for the controlled-strain mode;
the loading time and frequency were 0.1 sec and 1.67 Hz, respectively. For the 8 3 2 and modified asphalt mix test programs,
the new dynamic fatigue test module with servohydraulic loading
was used (Figure 3). The sinusoidal loading was applied in the
controlled-strain mode at a frequency of 10 Hz. Response variables
for each of the test programs included the initial stiffness, fatigue
life, and total (cumulative) dissipated energy to failure. Fatigue life
and total dissipated energy to failure were defined to correspond to
a stiffness reduction of 50 percent for the controlled-strain testing,
and to fracture for the controlled-stress testing. Details of each of the
test programs can be obtained from SHRP Report SHRP-A-404 (5).


Because of the differing nature of the 2 3 2 and 8 3 2 test programs,
a direct one-to-one comparison of the repeatability of the test results
was not possible. Accordingly, results from an analysis of variance
(ANOVA) from a general linear model (GLM) were used to assess
the repeatability in the test results. The variance estimates for the
two test programs include the variability in results due to equipment
differences, compaction differences, and mix differences. The coefficient of variation was calculated using the mean square error from
ANOVA and the following relationship (5):
CV 5 100(e MSE 2 1)0.5


CV 5 coefficient of variation,
MSE 5 mean square error resulting from ANOVA (natural log
transformed data), and
e 5 base of natural logarithm.
Table 5 gives results of the general linear modeling for stiffness,
fatigue life, and total dissipated energy. These results show a
marked improvement in the 8 3 2 test data: coefficient of variation
of 41 percent compared with 93 percent for the 2 3 2 data for both
fatigue life and total dissipated energy. This improvement in test
reliability is attributed to the combined effects of improved control
of induced deformation, use of the larger test specimens, and use of
rolling wheel compaction, which minimizes aggregate fracture
when compared with kneading compaction.




Asphalt Binders and Aggregates Used in Fatigue Studies


The effect of the compaction method on the repeatability of test
results was evaluated using fatigue data from the 8 3 2 and modified asphalt mix test programs. As indicated earlier, specimens for
the 8 3 2 study were compacted using a rolling wheel compactor,
whereas specimens for the modified asphalt mix study were compacted using a kneading compactor. Except for the differences in
mix type and specimen preparation, all other aspects of testing
were the same for both studies, including specimen size, test


Aggregate Gradation

equipment, test procedure, strain levels, loading pattern, frequency, and temperature.
Because specimens were tested with full replication at each strain
level for both test programs, it was possible to evaluate repeatability in test results directly from sample variance estimates obtained
by pooling variances between replicate tests using the following

2 [ln( Nrep1 /Nrep2 )]

s =
( # Obs)


Tayebali et al.



Features of 2 3 2 and 8 3 2 Expanded Fatigue Test Program

s 5 sample variance of ln cycles to failure associated with
fatigue testing,
Nrep1 5 fatigue life of first replicate,
Nrep2 5 fatigue life of second replicate, and
# Obs 5 number of replicate pairs.

Repeatability in test results was evaluated by computing coefficients of variation using Equation 1 with the MSE replaced by the
sample variance as calculated using Equation 2. Table 6 gives coefficients of variation for the stiffness, fatigue life, and total dissipated
energy for both test programs. The coefficient of variation for the
kneading compacted specimens is approximately 54 percent for
fatigue life, compared with 41 percent for rolling wheelcompacted
specimens, a difference of almost 33 percent. Because of a higher
incidence of aggregate fracture, more variance is expected in the
fatigue life of kneading-compacted specimens than in the fatigue life
of rolling wheelcompacted specimens. However, it should be noted
that the differences in coefficients of variation could also be affected
by the mix differences (modified versus unmodified binders) in
addition to compaction methods.


Because of the smaller coefficients of variation and, hence, more
reliable measurements, the new fatigue system enables mix charac-

terization within a relatively short time and with fewer specimens.

With this shortened procedure, fatigue tests can be performed at a
given temperature and at relatively high frequency (10 Hz) over a
range of strain levels (strain levels that will give a fatigue life varying between 5,000 and 500,000 cycles). Results can be interpreted
conventionally (i.e., log strainversuslog cycles to failure relationship) or by using dissipated energy concepts.
For typical mixes containing conventional asphalt binders, testing can often be completed within an interval as short as 24 hr to
ascertain whether the mix is suitable for the intended pavement
application. A minimum of four beam specimens are tested in the
controlled-strain mode of loading. For atypical mixes containing
modified binders, full fatigue testing with more than four specimens
is recommended.
The short procedure is described as follows:
1. Conduct a test at a strain level expected to yield a fatigue life
between 5,000 and 10,000 cycles. Typically, an appropriate strain
will be in the range of 800 to 1,000 microstrain. If the fatigue life at
this strain level is more than 10,000 cycles, increase the strain level
for the second specimen. If it is lower, reduce the strain level. These
tests are expected to require a total of about 2 hr.
2. From the first two tests, a crude estimate of the slope of the
log strainversuslog cycles relation can be determined. This relationship, tempered by experience, can then be used to estimate the
strain level that will provide a fatigue life of approximately 100,000
cycles. This test will require approximately 4 hr.
3. Reestablish the log strainversuslog cycles relationship from
Step 2 and estimate the strain required for the specimen to fail between



TABLE 4 Features of Modified Asphalt Mix Study


400,000 to 500,000 cycles. This test will take 13 to 15 hr.This long test
can be scheduled to begin at the end of the workday (or earlier if the
other three tests have been completed) and to continue overnight. The
specimen will have reached its fatigue life the next morning.
Figure 5 shows the results of fatigue tests for two typical densegraded mixes evaluated using the methodology just described. Testing for both mixes was done at strain levels of 700, 400, 300, and

200 microstrain. As indicated in Table 7, coefficients of determination (R2) for the log strainversuslog life relationships for the two
mixes are .99 and .97, respectively. Using the regression line, an
extrapolation is often required to estimate the fatigue life at a smaller
strain level, one expected under real traffic loading in situ. The variability associated in predicting the life at the smaller strain level and
the effect of testing replicate specimens are briefly explored in the
following section.

TABLE 5 Effect of Test Equipment and Procedure on Fatigue Test Results

Tayebali et al.


TABLE 6 Effect of Compaction Method on Repeatability



Fatigue testing is usually conducted at relatively large strain levels
in order to minimize the required laboratory testing time. Assuming
linearity in the log strainversuslog life relationship such as that
shown in Figure 5, a regression line can be fitted to the data. An estimate (prediction) of fatigue life at the design strain level, that is, the
in situ tensile strain under the standard traffic load, is then made by
extrapolation from the regression line. Layered elastic analysis provides the in situ strain under simulated field conditions.
The precision (variability) of the fatigue life estimate at a specific
design strain level depends on the laboratory fatigue test equipment,
test procedure, specimen compaction method, total number of test
specimens, number of replicate specimens at each strain level, and
number of strain levels. The variance expected at the in situ strain
level can be computed using the following relationship (5,6 ):

( X x )2
Var{Y } = s 2 1 + +
n q( x p x )2



Var{Y } 5 variance of predicted ln(fatigue life),

s2 5 sample variance of ln(fatigue life) in fatigue testing,
n 5 number of test specimens,
X 5 ln(in situ strain) at which predicted ln(fatigue life) is
_ estimated,
x 5 average ln(test strain),
q 5 number of replicate specimens at each test strain level,
p 5 number of strain levels, and
xp 5 ln(strain) at pth test strain level.
The sample variance (s2) is the best estimate of the variance associated with fatigue tests on replicate beam specimens, that is, beamto-beam variance. This best estimate of sample variance for fatigue
life for the A-003A fatigue testing equipment and procedure was
0.152 for specimens compacted using rolling wheel compaction
and 0.252 for specimens compacted using kneading compaction, as
indicated in Table 6.
With these estimates of sample variance from laboratory fatigue
testing, variance estimates for predicted in situ fatigue life were
made for asphalt mix A (Figure 5) to determine the influence of
replicate testing. Table 8 presents the variance of the predicted

Strain-life relationship for two mixes developed using short procedure.



TABLE 7 Calibration of Strain-Life Relationship

fatigue life as a function of the type of compaction method and the

number of replicate test specimens at each of the four strain levels
used. The following observations can be made from the variance
data given in this table:
1. As anticipated, the precision of the predicted life increases
(variance decreases) with increasing numbers of replicates tested at
each of the four strain levels.
2. The precision of the predicted value decreases with increasing
magnitude of the predicted lifethat is, variability increases as the
extent of required extrapolation increases.
3. Precision based on laboratory fatigue testing using rolling
wheelcompacted specimens is greater than that using kneadingcompacted specimens.
4. To achieve similar precision in the predicted fatigue life,
the required number of kneading-compacted specimens is approximately twice the number of rolling wheelcompacted specimens. Consequently, to achieve the same level of precision, required
testing time will double if kneading-compacted specimens are used.
If the best estimate of the sample variance in Equation 3 is that
associated with the electropneumatic testing device and test
procedure (s2 5 0.62, RMSE 5 0.787; Table 5), for a predicted
fatigue life of 20 million cycles, approximately 48 specimens
will be required (Table 9) to achieve precision similar to that
obtained with a servohydraulic test device using four rolling
wheelcompacted specimens. Use of the variance values for a
reliability-based asphalt-aggregate mix design and analysis system has been explored in some detail by Deacon et al. (5) and
Tayebali et al. (6 ).

This paper presents some aspects of the development of an
improved dynamic flexural beam fatigue test system during the
SHRP A-003A research program. The new fatigue equipment operates under hydraulic pressure control, resulting in better response
and more precise control of induced load and deformation. Results
of this research effort indicate that precision of the test results is
greatly influenced by the type of test equipment as well as the
method of specimen preparation. A servohydraulic-controlled
dynamic flexural fatigue test module exhibits superior precision and
repeatability in test results compared with its predecessor (an electropneumatic system), as indicated by a coefficient of variation of
41 percent in fatigue life as compared with 93 percent. Due to lower
variability, fewer test specimens are required to determine the
fatigue life relationship without sacrificing precision. Consequently,
a short fatigue test procedure has been developed that facilitates
determination of the fatigue relationship for a given mix and temperature in as few as 24 hr.
Results of this study also indicate that the specimen compaction
method influences the precision of the predicted fatigue life. A 33
percent difference was observed in coefficient of variation between
rolling wheelcompacted specimens and kneading-compacted
specimens. Consequently, twice as many specimens are required to
achieve a given level of precision in in situ predicted fatigue life if
kneading compaction is used instead of rolling wheel compaction.
Similarly, if a pneumatic system and associated test procedure are
used, approximately 12 times as many specimens would be
required to achieve similar precision in predicted fatigue life compared with the new servohydraulic fatigue test system.

TABLE 8 Effect of Compaction Method and Replicate Specimens on Variance of Predicted ln(Fatigue Life),
Hydraulic Test System

Tayebali et al.


TABLE 9 Effect of Replicate Specimens on Variance of Predicted ln(Fatigue Life), Pneumatic Test System

The research reported herein was conducted as part of the SHRP
project Performance-Related Testing and Measuring of AsphaltAggregate Interactions and Mixtures, which was conducted principally by the Institute of Transportation Studies of UCB. SHRP was
a unit of the National Research Council authorized by Section 128 of
the Surface Transportation and Uniform Relocation Assistance Act
of 1987.

1. Deacon, J. A. Fatigue of Asphalt Concrete. Graduate report. Institute
of Transportation and Traffic Engineering, University of California,
Berkeley, 1965.
2. Epps, J. A. Influence of Mixture Variables on the Flexural Fatigue and
Tensile Properties of Asphalt Concrete. Doctor of Engineering thesis.
University of California, Berkeley, 1969.

3. Monismith, C. L., and J. A. Deacon. Fatigue of Asphalt Paving Mixtures.

Journal of Transportation Engineering, ASCE, Vol. 95, No. 1969,
pp. 317346.
4. Monismith, C. L. Fatigue Characteristics of Asphalt Paving Mixtures and
Their Use in Pavement Design. Proc. 18th Paving Conference,
University of New Mexico, Albuquerque, 1981.
5. Asphalt Research Program, Institute of Transportation Studies, University of California, Berkeley. Fatigue Response of Asphalt-Aggregate
Mixes. Report SHRP-A-404. Strategic Highway Research Program,
National Research Council, Washington, D.C., 1994.
6. Tayebali, A. A., J. A. Deacon, and C. L. Monismith. Development and
Evaluation of Surrogate Fatigue Models for SHRP A-003A Abridged
Mix Design Procedure. Journal of the Association of Asphalt Paving
Technologists, Vol. 64, 1995, pp. 340366.
This paper represents the views of the authors only and is not necessarily
reflective of the views of the National Research Council, SHRP, or SHRPs
sponsor. The results reported here are not necessarily in agreement with the
results of other SHRP research activities; they are reported to stimulate
review and discussion within the research community.
Publication of this paper sponsored by Committee on Characteristics of
Bituminous Paving Mixtures to Meet Structural Requirements.