See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/271348254

Interpretation of flow number test data for asphalt mixtures

Article  in  Transport · January 2013

DOI: 10.1680/tran.12.00070

CITATIONS READS

10 2,423

3 authors:

Veerararavan Amirthalingam Murali Krishnan

Indian Institute of Technology Madras Indian Institute of Technology Madras
80 PUBLICATIONS   710 CITATIONS    126 PUBLICATIONS   1,142 CITATIONS   

SEE PROFILE SEE PROFILE

Neethu Roy
Mar Baselios College of Engineering and Technology
7 PUBLICATIONS   45 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

HALF WARM ASPHALT MIXTURE USING POLYMER MODIFIED BITUMEN EMULSIONS View project

Investigation on the laboratory performance of bitumen modified with SBS, Elvaloy and Lotader. A joint venture of Total Oil India Private Limited, Bitumen Division and
I.I.T Madras. December 2014 to February 2017. View project

All content following this page was uploaded by Murali Krishnan on 01 October 2018.

The user has requested enhancement of the downloaded file.

Transport Proceedings of the Institution of Civil Engineers

Interpretation of flow number test data for http://dx.doi.org/10.1680/tran.12.00070

asphalt mixtures Paper 1200070
Roy, Veeraragavan and Krishnan Received 12/09/2012 Accepted 20/08/2013
Keywords: bitumen & tar/pavement design/roads & highways

ICE Publishing: All rights reserved

Interpretation of flow number

test data for asphalt mixtures
j
1 Neethu Roy PhD j
3 J. Murali Krishnan PhD
PhD Research Scholar, Department of Civil Engineering, Indian Associate Professor, Department of Civil Engineering, Indian Institute
Institute of Technology Madras, Chennai, India of Technology Madras, Chennai, India
j
2 Amirthalingam Veeraragavan PhD
Professor, Department of Civil Engineering, Indian Institute of
Technology Madras, Chennai, India

j
1 j
2 j
3

Flow number testing quantifies the permanent deformation characteristics of asphalt mixtures. For the flow number
to be determined correctly, it is necessary that the material exhibits the three-stage creep response when tested. The
current experimental protocol to be followed is not very clear for delineating such response. The post-processing
algorithm for data analysis also plays a critical role in the determination of flow number. In this study, experimental
investigations were carried out on an asphalt mix fabricated at air voids contents of 7% and 2%. The samples were
tested at two temperatures (408C and 558C) and three confinement conditions (unconfined, 100 kPa and 200 kPa) for
100 000 cycles. Post-processing of the data was carried out using three different methods. It was found that the onset
of the tertiary stage was strongly influenced by confinement conditions and air voids contents. The flow number
determined by the different post-processing algorithms was different for all the conditions of testing. It is thus
important for highways agencies to carefully select the testing conditions and post-processing algorithm before
using this method for checking the rutting susceptibility of asphalt mixtures.

Notation rutting or permanent deformation characteristics of asphalt con-

a, b, c, d regression constants crete mixtures. Out of these, two tests, which can be carried out
däi =dt strain rate at logged datum i in an asphalt mixture performance tester (AMPT), have shown
N number of loading cycles potential – the flow time (static creep) test and flow number
˜n sampling interval (dynamic creep) test (TRB, 2011; Witczak et al., 2002).
äiþ˜n strain at (i þ ˜n) samples
äi˜n strain at (i  ˜n) samples In the flow time test, an asphalt concrete mix specimen is
åp irrecoverable strain subjected to a static uniaxial or triaxial load along with an axial
contact load (approximately 5% of axial load) (Witczak et al.,
1. Introduction 2002). The accumulated deformation generally follows a three-
The Superpave mix design method was developed under the stage curve under certain conditions of loading duration, load
Strategic Highway Research Program to provide a more rational intensity, confinement pressure and temperature. However, a flow
method of asphalt concrete mixture design (Asphalt Institute, time test is less representative of the loading condition in the field
2001). The method consists of three phases, namely aggregate as the material is not allowed to recover. A flow number test is a
and binder selection, aggregate blend design and volumetric dynamic creep and recovery test in which the specimen is
analysis of specimens compacted using the Superpave gyratory subjected to repeated load cycles consisting of 0.1 s haversine
compactor. It was expected that materials selection and mix pulse load and 0.9 s rest time under confined or unconfined
design could control pavement distress such as permanent conditions (AASHTO, 2010a). Here also, it is hypothesised that
deformation, fatigue cracking and low-temperature cracking the accumulated deformation follows a three-stage (primary zone,
(Cominsky, 1994). However, there was a need for a practical secondary zone and tertiary zone) creep curve. The number of
performance test that could correlate pavement distresses over a load cycles corresponding to the onset of the tertiary zone is
range of traffic and climate conditions. Several laboratory tests defined as the flow number and has been recommended as a
have been developed as part of this effort to understand the significant rutting indicator of asphalt concrete mixtures (Witczak

1
Transport Interpretation of flow number test data
for asphalt mixtures
Roy, Veeraragavan and Krishnan

et al., 2002). The strain rate is calculated and flow number is Furthermore, the current AMPT protocol for flow number sug-
taken as the point of minimum strain rate. gests a test duration of 20 000 cycles. Depending on the AVC,
one might need a greater number of cycles for the onset of the
The mechanical response of an asphalt concrete mixture to tertiary stage.
loading will depend on many factors, such as the state of
densification of the mixture, test temperature, loading frequency, Different post-processing algorithms can be used to estimate flow
load intensity and the confinement condition. At present, there is number from creep and recovery data, and these methods try to
little clarity regarding the test protocol to be followed for the flow capture the point corresponding to the minimum strain rate. The
number test. The recommended provisional test protocol as per minimum strain rate is expected to be the point corresponding to
NCHRP 9-19 for flow number test consists of testing the asphalt the onset of the tertiary stage. The different algorithms vary in
mix at one effective pavement temperature and one design stress the manner in which the raw data and the computed strain rate
level selected by the design engineer (Witczak et al., 2002). The values are smoothed before evaluating the cycle corresponding to
effective pavement temperature suggested is in the range 25– minimum strain rate. The main issue here is experimental noise
608C. The design stress level range is 69–207 kPa for unconfined in the data, which is inevitable, and the smoothing algorithm is
tests and 483–966 kPa for confined tests. The confinement level required to eliminate any such local minimum points. Depending
can be selected between 35 and 207 kPa and the specimens are on the different smoothing criteria, it is possible that the same
tested at a design air voids content (AVC) of 4%. Correlations experiment may lead to different flow numbers.
between dynamic modulus measured at 388C at a frequency of
0.1 Hz and flow number measured at 548C have recently reported In this investigation, specimens with AVC of 2% and 7% were
by Apeagyei (2011). used to capture the permanent deformation response. At a high
AVC of 7%, it is expected that the creep strain will be higher
The later NCHRP project 9-33 (NCHRP, 2011) suggested an axial than for 2% AVC specimen, as the material has the potential for
stress of 600 kPa for unconfined flow number tests on specimens further ‘densification and flow’. For a 2% AVC specimen one
with a target AVC of 7%; for a confined flow number test, expects creep to be predominantly due to ‘flow’ of material. The
NCHRP project 9-30A (TRB, 2009) recommended a confining confinement pressure plays a very important role in this. To
stress of 69 kPa, repeated deviatoric stress of 483 kPa and a target understand the influence of confinement on the creep behaviour,
AVC of 7% (Bonaquist, 2011). It is understood that a higher flow number tests were in the unconfined condition and at two
deviator stress is suggested so that the material attains the tertiary confinement levels (100 kPa and 200 kPa). The tests were carried
stage in a feasible time period (a few hours for instance) in the out at two temperatures (408C and 558C) to account for different
routine test procedure. However, such application of high stresses, temperature regimes causing rutting in pavements. A common
especially on specimens having a high AVC of 7%, may not be deviator stress of 200 kPa was used for both the unconfined and
effective in capturing the three stages of creep. Instead, the confined tests. The test was continued for 100 000 cycles and the
specimen may undergo an early tertiary stage without secondary test results were analysed using two procedures – the variable
creep. The use of such data for evaluating flow number and sampling rate approach (Bonaquist, 2008; Witczak et al., 2002)
eventually computing the rut-resistance characteristics of the and the Australian Road Research Board (ARRB, 1994) method.
mixture may not capture the precise mechanism of rutting. Since In addition, the data were fitted using the Francken model and
failures due to creep occur due to the build-up of deformation at flow number was determined using the Francken approach
a sufficiently small rate, it is necessary that laboratory experi- (Francken, 1977).
ments designed to mimic such failures also apply load levels that
are sufficiently small. This will result in well-defined primary,
2. Experimental investigation
secondary and tertiary regimes. In fact, to simulate the creep
rupture of plastics, it is suggested in ASTM D2990-09 (ASTM, 2.1 Materials and mixture
2009) that load levels leading to failures in less than 1000 h The mid-gradation of a bituminous concrete grade 1 (BC-grade 1)
should not be applied. wearing course with nominal maximum aggregate size of 19 mm,
as per the Indian specification (MoRTH, 2001), was chosen. An
Another major factor associated with the creep and recovery unmodified binder of viscosity grade VG30 provided by Chennai
response of asphalt concrete mixtures is the level of densification Petroleum Corporation Limited, India was used. A binder content
or AVC of the mixture. Immediately after construction, the of 5% was chosen for preparation of the specimens. The
material has an AVC in the range 5–8% (Asphalt Institute, 1997). gradation of BC-grade 1 is shown in Figure 1 and the pertinent
After many years of trafficking, the AVC reduces to the ‘refusal properties of the binder used are listed in Table 1.
air voids content’ of 2–3%. It is understood that mechanical
response to loading will be different at these stages of mix 2.2 Specimen preparation
volumetrics and hence to quantify the permanent deformation The asphalt mixture was prepared in accordance with AASHTO
characteristics of the asphalt mixture it might be necessary to practice to attain a Superpave gyratory compacted specimen of
delineate the three-stage creep curve at different air voids levels. height 170 mm (AASHTO, 2010b). The mixture was short-term

2
Transport Interpretation of flow number test data
for asphalt mixtures
Roy, Veeraragavan and Krishnan

100 Upper limit

Temperature: 8C 40, 55
Lower limit
Mid-gradation (adopted in study) Maximum deviator stress: kPa 200
80
Confinement stress: kPa 0, 100, 200
Percentage passing

Load and recovery time per cycle 0.1 s haversine load,

60 0.9 s recovery period
Number of cycles 100 000
40 Table 2. Flow number test matrix

20
10 000 Unconfined

Accumulated residual microstrain

100 kPa confinement
0
0·01 0·1 1 10 100 200 kPa confinement
Sieve size: mm

Figure 1. Gradation of mix chosen for the study

1000

Absolute viscosity at 608C, Pa s 270.28

Kinematic viscosity at 1358C, cm2 /s 5.306
Penetration at 258C, 100 g, 5 s: 0.1 mm 58
Softening point (ring and ball): 8C 54.5
100 1000 10 000 100 000
Ductility at 258C after thin film oven test: cm .100 Number of cycles
Viscosity ratio at 608C after thin film oven test 3.36
Figure 2. Flow number test on 2% AVC specimens at 408C under
Table 1. Properties of binder used in the study
different confinement conditions

aged at the mixing temperature for 4 h  5 min before casting 100 000 Unconfined
Accumulated residual microstrain

(AASHTO, 2002). To achieve test specimens with target AVC of 100 kPa confinement
7  0.5% and 2  0.5%, trial specimens were prepared by 200 kPa confinement
varying the number of gyrations. A specimen of 150 mm
diameter and 170 mm height was then cored to obtain a test
specimen of 100 mm diameter. Approximately 10 mm was sawn 10 000
from each end of the cored out portion using a twin-saw
arrangement so that the end flatness was within the tolerance
limit of <0.5 mm and end perpendicularity was <1.0 mm
(AASHTO, 2010c). The final dimensions of the AMPT specimens
were 100 mm diameter, with a standard deviation of diameter
<0.5 mm and 150  2.5 mm height. 1000
10 100 1000 10 000 100 000
Number of cycles
2.3 Test matrix
Flow number testing was carried out using the AMPT equipment
Figure 3. Flow number test on 2% AVC specimens at 558C under
at different temperatures and confinement levels as listed in the
different confinement conditions
test matrix (Table 2). A beta version of the flow number test
software (UTS014) supplied by M/s IPC Global, which can run
the experiments up to 100 000 load repetitions, was used for this
investigation. The experiments were carried out twice for all test respectively. Similarly, Figures 4 and 5 show the response of the
conditions to check the repeatability of results. 7% AVC specimens at these temperatures. The coefficient of
variation of the experimental observation for identical testing
3. Experimental results conditions was found to be within an acceptable tolerance.
The deformation response collected was converted to axial strain
and plotted against the number of cycles for each confinement The effect of temperature on the mechanical response of the
condition considered in the study. Figures 2 and 3 show the material was as expected. At higher temperatures, the accumulated
response of the 2% AVC specimens tested at 408C and 558C strain was considerably higher than at lower temperatures. The

3
Transport Interpretation of flow number test data
for asphalt mixtures
Roy, Veeraragavan and Krishnan

Unconfined different. However, a considerable difference in creep behaviour

100 000 100 kPa confinement
Accumulated residual microstrain

was observed when the 7% AVC specimen was tested at 558C

200 kPa confinement
under 100 kPa and 200 kPa confinement levels.

4. Data analysis
10 000
4.1 Variable sampling rate approach (Bonaquist, 2008;
Witczak et al., 2002)
Witczak et al. (2002) and Bonaquist (2008) recommended a
numerical differentiation approach to compute the flow value. In
1000 this method, the operator has the facility of choosing the data
sampling interval from 1 to 20 cycles for the flow number test. In
10 100 1000 10 000 100 000 this study, a sampling interval of 1 cycle was chosen so that the
Number of cycles
response at all cycles of loading could be captured. The
numerical differentiation algorithm for evaluating flow number is
Figure 4. Flow number test on 7% AVC specimens at 408C under
different confinement conditions
däi äiþ˜n  äi˜n
1:
¼
dt 2˜n
100 000 Unconfined
100 kPa confinement
Accumulated residual microstrain

1200 kPa confinement

where däi =dt is the strain rate at logged datum i, äiþ˜n is the
strain at i þ ˜n samples, äi˜n is the strain at i  ˜n samples
and ˜n is the sampling interval. The derivatives produced are
then smoothed by a five-point moving average method. The flow
10 000 number is taken as the point at which the minimum value of the
smoothed strain rate occurs. In the case that more than one point
has the minimum strain rate, the first minimum point is taken as
the flow number (Bonaquist, 2008; Witczak et al., 2002).

4.2 ARRB (1994) method

1000 Due to the sensitivity associated with determining the flow
0·1 1 10 100 1000 10 000 100 000
Number of cycles
number, the ARRB (1994) procedure suggests smoothing of the
experimental data using different filters. The experimental data
Figure 5. Flow number test on 7% AVC specimens at 558C under are first filtered using a seven-point running median method.
different confinement conditions After carrying out the zero offset adjustment, the data series is
further filtered by a running median (base 7) filter. To ensure that
such double filtering and zero offset adjustment does not change
accumulated strains were found to be higher for the 7% AVC the nature of the experimental data, a check is made with the
specimens than for the 2% AVC specimens. The 7% AVC original data; in the case of difference, a linear interpolation
specimens exhibited a three-stage creep curve within the test between the two nearest adjacent points that have identical values
period at 408C and 558C under unconfined conditions, as shown to their corresponding uncorrected datum values is carried out.
later. However, for the 2% AVC specimen, the three-stage creep The slope is then calculated for each datum of data series and is
curve was noted only in the unconfined test at 558C. At 408C for processed by a running average (base 25) filter. The minimum
the 2% AVC specimen, the material was only in the primary or slope is taken as the lowest value occurring in final data set. Due
secondary stage of creep at the end of 100 000 test cycles. The to the requirement of filling the running median and running
confinement conditions have a significant influence on the creep average filters, filtered slope values cannot be generated within
response of the mixtures. As the confining pressure increased, 17 entries from the first and last data elements of the data set.
strains were found to reduce drastically when compared with
testing under unconfined conditions. Flow numbers for all the test conditions were evaluated using the
variable sampling approach and the ARRB method, as shown in
The strain at cycle number 100 000 was used to study any Table 3. It can be seen from the results that the two method-
noticeable difference in performance of the mixtures when tested ologies give widely varying flow number values. This could be a
under 100 kPa and 200 kPa confinement levels. It was observed serious issue when one attempts to use the flow number from the
that under 200 kPa confinement, the creep strain was less than test data to estimate the rutting characteristics of an asphalt
that in the 100 kPa confinement condition, but not significantly concrete mixture. Consider the case of the 2% AVC sample tested

4
Transport Interpretation of flow number test data
for asphalt mixtures
Roy, Veeraragavan and Krishnan

AVC: % Test conditions: Flow number

temperature; confinement
Variable sampling ARRB
approach method

2 558C; unconfined 4284 8015

558C; 100 kPa 47 348 94 548
558C; 200 kPa 58 809 58 801
408C; unconfined 47 348 94 548
408C; 100 kPa 29 491 94 251
408C; 200 kPa 19 136 69 108
7 558C; unconfined 224 578
558C; 100 kPa 83 277 75 290
558C; 200 kPa 81 252 91 798
408C; unconfined 1931 2748
408C; 100 kPa 50 021 93 366
408C; 200 kPa 61 806 95 625

Table 3. Flow number computed using variable sampling

approach and ARRB method

at 408C and 200 kPa confinement pressure. The raw experimental mum value of strain rate at cycle 69 108 is reported as the flow
data indicate that the material is in a stable secondary stage, but number (Figure 7). Clearly, even the algorithm for numerical
the variable sampling approach and the ARRB method estimated differentiation for the determination of flow number needs con-
flow numbers of 19 136 and 69 108 respectively. Figure 6 shows a siderable refinement.
plot of the strain rate after five-point averaging in the variable
sampling rate approach. It can be seen that there are certain local It is proposed here to improve the ARRB method by further
minimum values in the strain rate data, which differ only from filtering the data to avoid the inclusion of such local minimum
the twelfth decimal place. Of these four points (as can be seen in data points. The strain rates evaluated are rounded off to six
Figure 6), cycle number 19 136 has the minimum strain rate value decimal points and the data further smoothed with a 50-point
and hence this point was reported as the flow number. It is running average method. These two additional steps were in-
required that the protocol should be able to discard such local cluded on the basis of the quality of data being collected and the
minimum values. Similarly in the ARRB method, a local mini- noise level in the data. The filtered data were then used to find
the minimum slope as corresponding to the flow number. The
2 ARRB modified method results are listed in Table 4.
Strain rate (after five-point averaging)

0·4

1 0·3

0·2
0
Strain rate

0·1

0
⫺1
⫺0·1

⫺2 ⫺0·2

0 2 4 6 8 10 ⫺0·3
Number of cycles ⫻104 0 2 4 6 8 10
Number of cycles ⫻104
Figure 6. Variable sampling rate approach: strain rate versus
number of cycles for 2% AVC specimen, test at 408C and 200 kPa Figure 7. ARRB method: strain rate versus number of cycles for
confinement 2% AVC specimen, test at 408C and 200 kPa confinement

5
Transport Interpretation of flow number test data
for asphalt mixtures
Roy, Veeraragavan and Krishnan

AVC: % Test conditions: Flow number

temperature; confinement
ARRB modified Francken model
method approach

2 558C; unconfined 8015 9730

558C; 100 kPa 92 668 .100 000
558C; 200 kPa 99 969 .100 000
408C; unconfined 99 388 .100 000
408C; 100 kPa 99 890 .100 000
408C; 200 kPa 99 963 .100 000
7 558C; unconfined 578 205
558C; 100 kPa 99 982 .100 000
558C; 200 kPa 96 236 .100 000
408C; unconfined 2750 2823
408C; 100 kPa 99 875 .100 000
408C; 200 kPa 99 963 .100 000

Table 4. Flow number computed using ARRB modified method

and Francken model method

@ 2 åp
4.3 Francken model ¼ ab(b  1)N (b2) þ cd 2 edN
3: @N 2
Empirical models were originally used to represent the creep
behaviour of asphalt concrete mixtures and the most common one
was a power law model correlating strain and time data (Moni- The cycle number at which the tertiary stage is reached is given
smith et al., 1975). Francken (1977) proposed a three-stage creep by the point where the rate of change of slope changes sign. This
curve model on the basis of repeated triaxial load tests under indicates the inflection point in the curve of accumulated strain
various stress levels and temperatures (Equation 2) versus number of cycles at which the tertiary stage begins.

The Francken model parameters estimated for the accumulated

2: åp ¼ aN b þ c(edN  1) strain data for 2% and 7% AVC specimens at different test
conditions are listed in Table 5. The second derivatives of
accumulated strains were computed using Equation 3 for the
unconfined test data of the 7% AVC specimen at 558C and 408C
in which åp is the irrecoverable deformation or irrecoverable and the 2% AVC specimen at 558C, where a tertiary stage was
strain, N is the number of loading cycles and a, b, c and d are observed (Table 4). This approach found that the tertiary stage was
regression constants. Francken (1977) proposed that ‘bituminous reached in the unconfined 558C test of the 2% AVC specimen at
road mixes’ can exhibit two types of creep behaviour. The first is 9730 cycles. For the 7% AVC sample at 558C unconfined test, the
where c ¼ 0 and the response is parabolic and the second is flow number was 205 cycles. The 7% AVC sample also experi-
where the curve has a parabolic and exponential shape repre- enced a three-stage creep when tested at 408C in the unconfined
sented by the first and second terms of Equation 2 respectively. condition; the flow number obtained for this case was 2823. In all
The exponential term indicates a rapid increase in irrecoverable other test conditions, the specimen attained only a primary or
strain, leading to failure of a pavement. Hence, in the three-stage primary and secondary stage of creep, which can be explained
creep curve, the regression constant c can be considered an using the power law function (the first portion of Equation 1). Here,
indicator of whether tertiary flow has occurred or not. the total number of load cycles in the test is recorded as the flow
number, which indicates that the flow number is at a cycle greater
In this investigation, the Francken model was fitted to accumu- than the test duration. The fitting parameters a and b are derived
lated strain as a function of the number of cycles. Typical curve from the linear or secondary portion of the accumulated strain
fitting of the data using the Francken model is shown in Figure 8. versus number of cycles relationship. It can be seen that b is a good
It is possible to identify the point at which the tertiary stage indicator of irrecoverable deformation behaviour within the sec-
begins by considering the rate of change of slope of irrecoverable ondary stage of creep behaviour. As the test condition varies from
deformation (Biligiri et al., 2007; Dongré et al., 2009). The unconfined to confined, b decreases, indicating a reduction in the
second derivative of the Francken model is given by slope of the creep curve. However, no significant variation was

6
Transport Interpretation of flow number test data
for asphalt mixtures
Roy, Veeraragavan and Krishnan

0·005 0·12
Accumulated residual strain

Accumulated residual strain

0·004

0·08
0·003

0·002
0·04

0·001 Experimental data

Francken model
0 0
0 20 000 40 000 60 000 80 000 100 000 0 4000 8000 12 000 16 000 20 000
Number of cycles Number of cycles
(a) (b)

0·10 0·10
Accumulated residual strain
Accumulated residual strain

0·08 0·08

0·06 0·06

0·04 0·04

0·02 0·02

0 0
0 2000 4000 6000 8000 10 000 12 000 0 1 2 3 4 5 6 7
Number of cycles Number of cycles
(c) (d)

Figure 8. Fitting of Francken model with experimental data:

(a) 2% AVC, 408C, unconfined test; (b) 2% AVC, 558C,
unconfined test; (c) 7% AVC, 408C, unconfined test; (d) 7%
AVC, 558C, unconfined test

observed in the slope of the secondary stage of creep between test improved flow number estimation, with the Franken model and
conditions of 100 kPa and 200 kPa confinement pressure. This modified ARRB method producing compatible results (Table 4).
indicates that, even though pressure is a significant parameter in
the creep response of asphalt concrete mixtures, the effect of The major issue associated with the use of an empirical model to
100 kPa and 200 kPa confinements for the 7% and 2% AVC describe the data is that there are numerous combinations of
samples considered in the study is very similar. However, a parameter values for the Francken model that can best fit the
variation in the intercept parameter a between the 100 kPa and data. The model assumes an additive decomposition of a
200 kPa confinement tests indicates a higher deformation level in parabolic part and an exponential part. As per Francken (1977),
the primary stage for less confined conditions. the presence of parameter c will indicate the occurrence of a third
stage in the creep curve. However, it was observed that a value
Table 3 shows that the flow numbers estimated using the first two closer to zero (c ¼ 3:268 3 107 for the 2% AVC specimen at
methods discussed in this paper are entirely different. The 558C unconfined test) is also possible in a three-stage creep
variable sampling approach clearly fails to identify the onset of behaviour. A more careful interpretation of the physical signifi-
the tertiary stage, which may be due to noise in the data. The cance of the parameters is thus needed before following a curve-
ARRB method is a more robust smoothing algorithm and gives fitting algorithm such as the Francken model. This becomes far
reasonable outputs, as can be observed from the raw data. A more important when one deals with experiments carried out in
further modification by smoothing the strain rates gave an the confined state.

7
Transport Interpretation of flow number test data
for asphalt mixtures
Roy, Veeraragavan and Krishnan

AVC Test conditions: a b c d R2

temperature; confinement

2% 558C; unconfined 0.001001 0.2873 3.268 3 107 0.000578 0.9690

558C; 100 kPa 0.000591 0.2546 0a 0 0.9988
558C ; 200 kPa 0.000584 0.2545 0 0 0.9471
408C; unconfined 0.000864 0.1507 0 0 0.9972
408C; 100 kPa 0.000591 0.1184 0 0 0.9086
408C ; 200 kPa 0.000318 0.1182 0 0 0.9739
7% 558C; unconfined 0.001210 0.5175 0.0003536 0.00932 0.9984
558C; 100 kPa 0.001090 0.2598 0 0 0.9748
558C ; 200 kPa 0.000592 0.2560 0 0 0.9874
408C; unconfined 0.000880 0.3474 0.0009662 0.000376 0.9983
408C; 100 kPa 0.000865 0.1688 0 0 0.9811
408C; 200 kPa 0.000553 0.1672 0 0 0.9969
a
Values less than 1 3 1015 approximated to zero

Table 5. Parameters estimated using Francken model

5. Conclusion AASHTO (2010a) TP-79-10: Determining the dynamic modulus

j The current flow number protocols do not explicitly consider and flow number for hot mix asphalt (HMA) using the
the influence of air voids content (AVC) and confinement asphalt mixture performance tester (AMPT). AASHTO,
pressures when mapping the complete three-stage creep Washington, DC, USA.
behaviour. AASHTO (2010b) PP60-09: Preparation of cylindrical
j To precisely delineate such creep behaviour, it is required that performance test specimens using the Superpave gyratory
the material be subjected to considerably longer loading compactor (SGC). AASHTO, Washington, DC, USA.
durations. To this end, all samples in this work were AASHTO (2010c) T-312: Preparing and determining density of
subjected to a deviatoric stress of 200 kPa for 100 000 cycles. hot mix asphalt (HMA) specimens by means of the
The current practice of determining the flow number, Superpave gyratory compactor. AASHTO, Washington, DC,
however, involves application of 20 000 cycles with a USA.
deviatoric stress of 600 kPa and hence it is possible that the Apeagyei AK (2011) Correlating rutting with dynamic modulus
material may not exhibit the secondary stage completely. of asphalt concrete. Proceedings of the Institution of Civil
j The flow number was estimated using the variable sampling Engineers – Transport 164(4): 241–249.
rate approach, the ARRB method and the Francken model, ARRB (Australian Road Research Board) (1994) Data Processing
and a wide variation in results was observed. of Dynamic Creep Data to Determine Creep Test Parameters.
j The predictions of the ARRB method and the Francken ARRB, Vermont South, VIC, Australia, Austroads Pavement
model were reasonably close for testing conditions in which Reference Group, Report 94/07(MA).
the material exhibited a three-stage creep curve. Asphalt Institute (1997) MS-02: Mix design methods for asphalt
j Flow number determination depends on the numerical concrete and other hot-mix types, 6th edn. Asphalt Institute,
differentiation algorithm and this requires standardisation. Lexington, KY, USA.
The required rigour in quantifying such behaviour will Asphalt Institute (2001) SP-02: Superpave mix design. Asphalt
involve the development and validation of a proper non-linear Institute, Lexington, KY, USA.
viscoelastic constitutive model for the material. ASTM (2009) D2990-09: Standard test methods for tensile,
compressive and flexural creep and creep rupture of
Acknowledgements plastics. ASTM International, West Conshohocken, PA,
The authors acknowledge M/s IPC Global, Australia, for sharing USA.
the beta version of UTS014 software. Biligiri KP, Kaloush KE, Mamlouk MS and Witczak MW (2007)
Rational modeling of tertiary flow for asphalt mixtures.
REFERENCES Journal of the Transportation Research Board 2001:
AASHTO (American Association of State Highway and 63–72.
Transportation Officials) (2002) R-30: Standard practice for Bonaquist R (2008) Refining the Simple Performance Tester for
mixture conditioning of hot mix asphalt. AASHTO, Use in Routine Practice. National Academies Press,
Washington, DC, USA. Washington, DC, USA, NCHRP Report 614.

8
 Transport Interpretation of flow number test data
for asphalt mixtures
Roy, Veeraragavan and Krishnan

Bonaquist R (2011) Precision of the Dynamic Modulus and Flow repeated loading. Journal of the Transportation Research
Number Tests Conducted with the Asphalt Mixture Board 537: 1–17.
Performance Tester. National Academies Press, Washington, MoRTH (Ministry of Shipping, Road Transport and Highways)
DC, USA, NCHRP Report 702. (2001) Specifications for Roads and Bridge Works. MoRTH,
Cominsky RJ (1994) The Superpave Mix Design Manual for New New Delhi, India.
Construction and Overlays. National Research Council, NCHRP (National Cooperative Highway Research Program)
Washington, DC, USA, Strategic Highway Research Program (2011) A Mix Design Manual for Hot-Mix Asphalt. TRB,
Report A-407. Washington, DC, USA, NCHRP Project 9-33.
Dongré R, D’Angelo J and Copeland A (2009) Refinement of TRB (Transportation Research Board) (2009) Calibration of
flow number as determined by asphalt mixture performance Rutting Models for Structural and Mix Design. TRB,
tester: use in routine quality control–quality assurance Washington, DC, USA, National Cooperative Highway
practice. Journal of the Transportation Research Board 2127: Research Program Report 719.
127–136. TRB (2011) A Manual for Design of Hot Mix Asphalt with
Francken L (1977) Pavement deformation law of bituminous road Commentary. TRB, Washington, DC, USA, National
mixes in repeated load triaxial compression. Proceedings of Cooperative Highway Research Program Report 673.
the 4th International Conference on the Structural Design of Witczak M, Kaloush K, Pellinen T, El-Basyouny M and Von
Asphalt Pavements, Ann Arbor, MI, USA, pp. 483–496. Quintus HL (2002) Simple Performance Test for Superpave
Monismith CL, Ogawa N and Freeme CR (1975) Permanent Mix Design. National Academies Press, Washington, DC,
deformation characterization of subgrade soils due to USA, NCHRP Report 465.

WHAT DO YOU THINK?

To discuss this paper, please email up to 500 words to the
editor at journals@ice.org.uk. Your contribution will be
forwarded to the author(s) for a reply and, if considered
appropriate by the editorial panel, will be published as a
discussion in a future issue of the journal.
Proceedings journals rely entirely on contributions sent in
by civil engineering professionals, academics and students.
Papers should be 2000–5000 words long (briefing papers
should be 1000–2000 words long), with adequate illustra-
tions and references. You can submit your paper online via
www.icevirtuallibrary.com/content/journals, where you
will also find detailed author guidelines.

9
View publication stats