Professional Documents
Culture Documents
net/publication/267752142
CITATIONS READS
2 99
1 author:
Litao Liu
Texas A&M University
13 PUBLICATIONS 41 CITATIONS
SEE PROFILE
All content following this page was uploaded by Litao Liu on 05 November 2014.
To cite this article: Litao Liu (2014): Characterisation of variability in highway pavement materials and construction,
International Journal of Pavement Engineering, DOI: 10.1080/10298436.2014.943219
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
International Journal of Pavement Engineering, 2014
http://dx.doi.org/10.1080/10298436.2014.943219
Pavement materials and construction (M&C) are inherently variable. Unavoidable inhomogeneous materials, inconsistent
construction methods and equipment, and changing weather conditions during construction, among other factors, result in
M&C variability. Motivated by the need to provide realistic inputs to pavement quality assurance programs, design
procedures and reliability analysis, this paper presents the results of an investigation of variability in key acceptance quality
characteristics (AQCs) for both hot-mix asphalt (HMA) and Portland cement concrete (PCC) pavements using empirical
data (field and laboratory test results from recently completed construction projects). Variability is measured at three levels:
within-lot, within-project and within-state. The variability values found in this study are compared with the variability
values reported about 17 years ago. The results showed that major progress had been made over the past two decades in
reducing variability in pavement M&C.
Keywords: variability; PWL; standard deviation; COV; specifications
Downloaded by [Litao Liu] at 11:15 23 October 2014
*Email: liulitao11@tamu.edu
q 2014 Taylor & Francis
2 L. Liu
uncertainty in M&C, as well as evaluate its impact on the concrete pavements. The coefficients of variability offered
performance of pavements (for example, see Madanat in their study were 15% for compressive strength, split
1993; Swei et al. 2013). The variability measures (e.g. SD) tensile strength and modulus of elasticity, and 12% for
that are used in pavement specifications and design flexural strength (Tayabji and Wu 2002).
procedures represent the total variability that results from Gharaibeh et al. (2001) used the PaveSpec software to
the above sources. Little research has been made to evaluate the effect of the mean and variability of key AQCs
determine the portion of the total variability that can be on pay factor (which is an indicator of quality). PaveSpec
attributed to each one of these sources. Hughes (1996) uses Monte Carlo simulation for estimating the effects of
synthesised the levels of variability in key AQCs of M&C variability on pavement performance. Thus, the
highway construction and materials based on data gathered Monte Carlo simulation procedure yields theoretical, but
from several state DOTs. One of his conclusions was that logical, effects of M&C variability on pavement
the variability of some mix properties (e.g. gradation and performance. The Gharaibeh et al. (2001) study showed
asphalt content) had decreased over time while others that, according to the PaveSpec simulation approach, the
appear to be unchanged. Hughes attributed the decrease in pay factor always decreases or remains the same (for
variability to contractor quality control, specifications conservative designs) as the variability increases. Another
requiring a measurement of variability, improved industrial theoretical relationship between variability and perform-
technology and improved test methods (Hughes 1996). ance (the PAVAN model), however, does not always result
While there is a general agreement in the pavement in decreased performance with increased variability
community that M&C variability adversely affects (Gharaibeh et al. 2001).
pavement performance, how variability affect perform- A recent National Highway Cooperative Research
ance is still largely unknown. There has been no major Program (NCHRP) research effort developed a perform-
attempt to correlate variability in the initial as-constructed ance-related specification for hot-mix asphalt pavement
quality characteristics and actual pavement performance (Fugro Consultants, Inc. and Arizona State University
data. While quantifying these effects using actual/ (ASU) 2011). This research used historical variance of
measured data may be ideal, it is very challenging because variables in its simulation of actual pavement perform-
pavement construction data and performance data are ance. The variance values have been dated early back to
usually stored in disparate databases and thus are difficult 1975. Lack of practical variability information in key
to match. Additionally, existing quality assurance data AQCs for both HMA and PCC pavements makes this
normally indicate total variability (including testing study necessary to look into a number of pavement
variability) and not just M&C variability. A recent construction projects and investigate the variability in
research effort was made to propose a methodology for some key AQCs using empirical data.
considering the variability in initial as-constructed quality
characteristics in predicting long-term pavement perform-
ance, and its application in predicting pavement roughness Data sets
especially after thin overlay treatment was demonstrated The data sets used in this study were obtained from Florida
to be encouraging (Liu and Gharaibeh in press). DOT (FDOT) and Kansas DOT (KDOT) for construction
International Journal of Pavement Engineering 3
projects completed between 2005 and 2009. The KDOT 3. characterise the variability values at the levels of
database included both HMA and PCC data. The databases within-lot, within-project and within-state, respect-
included project information, contractor information, mix ively, and compare the variability values to those
design, date of sampling, lot size, lot and sublot index, and offered in Hughes (1996) to assess change in
sample test results. The FDOT HMA database consisted of variability over the past 14 years.
five AQCs, which are:
A preliminary analysis of the data sets found that the
. asphalt content AQC test data within a particular lot generally follows
. percent aggregate passing #8 (2.38 mm) sieve normal distribution, with the mean close to lot target AQC
. percent aggregate passing #200 (0.075 mm) sieve value. Figure 2 shows the AQC distributions from several
. air voids randomly selected HMA projects by FDOT, and it can be
. mix density seen that the AQC test values at project level have a rough
shape of normal distribution.
The KDOT HMA database had two AQCs, air voids and
density, and the PCC database had also two AQCs, 28-day
compressive strength and thickness. From the databases Characterisation of HMA variability
large projects that contained 20 or more lots were selected HMA within-lot variability
for analysis. In this study, 12 HMA projects were selected The SDs and COVs for each single lot in the selected
from the FDOT database, and 12 HMA projects and 5 PCC HMA projects are calculated. Histograms are drawn to
projects were selected from the KDOT database. show the SD and COV distributions for each AQC in the
Downloaded by [Litao Liu] at 11:15 23 October 2014
These sets of data were analysed as follows: Florida and Kansas HMA data sets.
Figure 3. Histograms for SDs of five HMA AQCs in Florida (sample size ¼ 348).
asphalt content has the smallest range (Figure 3(b)). The (Figure 5(a)), which is noticeably different from the air
SD distributions appear fairly to follow chi-square voids distribution in Florida HMA (Figure 3(d)). The SD
distribution, except for that of percent passing #200 of density is normal distributed with some sknewness to
sieve (P#200), which appears rather uniformly distributed. the left-hand side.
For P#8, air voids and density, the distributions are skewed The air voids data also have a very big range of COV in
to the left-hand side. Kansas HMA, which is about 27% as shown in Figure 6(a),
The distributions of COV for each HMA AQC in comparing to only 1.8% for density in Figure 6(b). The
Florida are shown in Figure 4. Air voids has the largest distributions poorly conform to normal with a large
range of COV (as high as 35%), followed by P#200, P#8, skewness to the left-hand side.
and asphalt content. The range of COV for density is about
2.1%, dramatically lower than the other four. All the five
AQCs conform roughly to normal distribution.
HMA within-project variability
Pooled SDs and COVs of each AQC were calculated at
Kansas HMA project level for the HMA and PCC databases. The
Figure 5 shows the SD distributions for HMA AQCs in formula used to calculate pooled SDs is as shown in
Kansas. The distribution for air voids is fairly uniform Equation (1):
International Journal of Pavement Engineering 5
Downloaded by [Litao Liu] at 11:15 23 October 2014
Figure 4. Histograms for COV of five HMA AQCs in Florida (sample size ¼ 348).
Figure 5. Histograms for SDs of two AQCs in Kansas HMA (sample size ¼ 407).
6 L. Liu
Figure 6. Histograms for COV of two AQCs in Kansas HMA (sample size ¼ 407).
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uPk Figure 7 graphically describes the difference in the
u
i¼1 ðni 2 1Þsi
2
sp ¼ t Pk ; ð1Þ within-project COV of the five AQCs. The COVs spread
i¼1 ðni 2 1Þ out well from less than 1% to about 17% across the five
AQCs with some overlapping between asphalt content and
Downloaded by [Litao Liu] at 11:15 23 October 2014
where sp is the pooled SD in a project, ni is the sample size P#8. It can be seen that, among the studied AQCs, density
in the ith lot, si is the SD in the ith lot and k is the number has the lowest and air voids has the highest within-project
of lots in that project. COV.
Within-project arithmetic mean values for each AQC
were computed. The within-project COV values were
determined by dividing the pooled SD by the arithmetic Kansas HMA
mean value. For KDOT’s HMA AQCs, the within-project SD of air
voids ranges from 0.2% to 1%, relatively smaller than the
range of density, which is from 0.5% to 1.3% (Table 2).
Florida HMA Similar to the COV trend found in FDOT’s data set,
Table 1 presents the within-project SDs and COVs for density has lower within-project COV than air voids
HMA AQCs in Florida. The range of within-project SD is (Figure 8).
about 0.1 – 0.2% for asphalt content, 0.2 –0.5% for P#200,
0.4 –0.7% for air voids, 0.4 – 0.9% for density, and 1.3 –
2.3% for P#8. The range of within-project COVs is 0 –1% HMA within-state variability
for density, 2– 5% for asphalt content and P#8, 4 –9% for The arithmetic average, SD and sample size in each
P#200, and 10– 15% for air voids. project are used to compute the pooled within-state SDs
50%
Density
Asphalt Content
40%
P#8
P#200
% of Projects
20%
10%
0%
0.5 0.6 0.7 0.8 0.9 1 2 2.5 3 3.5 4 4.5 5 6 7 8 9 11 12 13 14 15 16
Coefficient of Variation, %
and COVs for FDOT’s and KDOT’s data sets. The results Kansas HMA
Downloaded by [Litao Liu] at 11:15 23 October 2014
are presented in Table 3, which also includes the typical The within-state SDs of air voids and density are 0.49%
SDs and COVs from the NCHRP Synthesis 232 (Hughes and 0.82%, respectively, which are fairly close to the
1996). Florida values. The SD for air voids is below the lower end
of the reported range in NCHRP Synthesis 232.
Florida HMA The COVs for air voids and density are 11.7% and
0.9%, respectively, also very close to the Florida values.
As shown in Table 3, the within-state SDs are comparable
to the variability values obtained about 14 years ago. The
SDs of P#8 and P#200 are smaller than the lower ends of the
Characterisation of PCC variability
ranges for the two AQCs in the NCHRP Synthesis 232.
The within-state SDs of asphalt content and air voids are PCC within-lot variability
just on the brink of the reported ranges 14 years ago. Also, The within-lot SD of 28-day compressive strength in
the SD of P#8 is favourably comparable to the range Kansas PCC has a range of about 30 –1400 psi, with 95%
provided in a study by Villiers et al. (2003), which was of the data ranging between 100 and 800 (Figure 9(a)). The
from 0.7% to 2.32%. within-lot SD of thickness has a range of 0– 1.2 in., with
Regarding COV, air voids has the largest within-state 95% of the data ranging between 0.05 and 0.55 (Figure 9
value at 14%, significantly higher than the only 0.8% of (b)). The distribution of the thickness within-lot SD
density which is the lowest in the five AQCs. conforms well to a normal curve.
70%
60% Density
% of Projects 40%
30%
20%
10%
0%
0.5 0.6 0.7 0.8 0.9 1.4 11 12 13 14 23
Coefficient of Variation, %
has a range of about 0 –28%, comparing to a range of As Table 5 shows, the pooled within-state SD is 498 psi for
about 0 –10% for thickness. The two distributions are 28-day compressive strength, and 0.29 in. for thickness in
shown in Figure 10(a),(b). Kansas PCC. Comparing the 1996 variability values
(Hughes 1996), the SD of 28-day compressive strength is
PCC within-project variability within the reported range but fairly close the lower end; the
The pooled within-project SD ranges between 400 and SD of thickness is smaller than the lower end of the range
630 psi for 28-day compressive strength and ranges for thickness SD reported by Hughes (1996).
between 0.2 and 0.4 in. for thickness in Kansas PCC, as The within-state COV of 28-day compressive strength
shown in Table 4. The COV of 28-day compressive is 8.9%, which is much higher than the 2.5% of thickness.
strength ranges between 7% and 12%, much larger than The COV of 28-day compressive strength is just slightly
the COV of thickness that ranges between 1.8% and 3.5%. larger than the lower end of the range of 8.5– 21.8%
Figure 9. Histograms for SDs of two AQCs in Kansas PCC (sample size ¼ 235).
International Journal of Pavement Engineering 9
Figure 10. Histograms of COV of two AQCs in Kansas PCC (sample size ¼ 235).
a b a
1 45 627 5314.5 11.8 0.41 11.804 3.5b
2 245 573 5704.9 10.0 0.31 10.745 2.9
3 207 460 5616.1 8.2 0.23 12.863 1.8
4 134 402 5585.6 7.2 0.27 12.898 2.1
5 127 439 5546.6 7.9 0.30 10.631 2.8
a
The largest within-project SD for the AQC in column.
b
The largest within-project COV for the AQC in column.
reported by Hughes (1996). The COV of thickness is lower technologies, more rigorous quality control procedures
than the 1996 reported range. and other related factors.
The empirical results are of practical value in that they
Summary and conclusions provide insights concerning variability values that can be
used in establishing agency’s pavement quality acceptance
The variability, in terms of SD and COV, in contractor-
criteria, pavement design procedures and reliability
performed test results was evaluated and characterised at
analysis. But as these variability values (e.g. SDs and
the within-lot, within-project and within-state levels for
COVs) are derived from Florida and Kansas data only, they
both HMA and PCC pavements. Within-project and
should be used cautiously for pavements in other locations.
within-state COVs are computed using the pooled SD and
arithmetic mean of each AQC in a project.
The derived SDs and COVs are compared across Acknowledgements
different AQCs within the same pavement type and also The author is grateful for the FHWA for sponsoring the research
compared to past results reported in the 1996 NCHRP and also for the Florida and Kansas Departments of
Synthesis 232 (Hughes 1996). Considering the within- Transportation for providing the data used in this research.
state variability, major progress has been made over the
past 17 years in reducing SDs and COVs in AQCs for both References
HMA and PCC pavements. This progress can possibly be Amirkhanian, S.N., Burati, J.L., and Mirchandani, H.C., 1994.
contributed to advancements in pavement construction Effect of testing variability on contractor payment for asphalt
10 L. Liu
pavements. Journal of Construction Engineering and Newton, L.A. and Christian, J., 2006. Impact of quality on
Management, 120 (3), 579– 592. building costs. Journal of Infrastructure Systems, 12 (4),
Fugro Consultants, Inc. and Arizona State University (ASU), 199– 206.
2011. A performance-related specification for hot-mix Patel, A. and Thompson, M., 1998. Consideration and
asphalt. Washington, DC: Transportation Research Board
characterization of pavement construction variability.
of the National Academies NCHRP Report 704.
Gharaibeh, N.G., et al., 2001. Effect of variability and central Transportation Research Record: Journal of the Transpor-
tendency in performance-related specifications for concrete tation Research Board, 1632, 40 – 50.
pavement. In: Proceedings, 7th international conference on Stubstad, R.N., 2002. LTPP data analysis: variations in
concrete pavements. Vol. 1, Orlando, FL, 513–524. pavement design inputs. Washington, DC. National Coop-
Hughes, C.S., 1996. Variability in highway pavement construc- erative Highway Research Program (NCHRP) Project No.
tion. National Cooperative Highway Research Program 20-50(5) Transportation Research Board.
(NCHRP) Synthesis 232. Washington, DC: Transportation
Swei, O., Gregory, J., and Kirchain, R., 2013. Probabilistic
Research Board, National Research Council.
Jiang, Y., et al., 2003. Estimation of pavement layer thickness characterization of uncertain inputs in the life-cycle cost
variability for reliability-based design. Transportation analysis of pavements. Transportation Research Record:
Research Record: Journal of the Transportation Research Journal of the Transportation Research Board, 2366, 71 – 77.
Board, 1849, 156– 165. Tayabji, S.D. and Wu, C.L., 2002. Variability of concrete
Liu, L. and Gharaibeh, N., in press. A Bayesian model for materials data in long-term pavement performance program.
predicting the performance of pavements treated with thin Transportation Research Record: Journal of Transportation
HMA overlays. Transportation Research Record: Journal of Research Board, 1813, 181– 190.
the Transportation Research Board.
Madanat, S., 1993. Optimal infrastructure management decisions Villiers, C., et al., 2003. Evaluation of percent-within-limits
Downloaded by [Litao Liu] at 11:15 23 October 2014
under uncertainty. Transportation Research Part C: Emer- construction specification parameters. International Journal
ging Technologies, 1 (1), 77 – 88. of Pavement Engineering, 4 (4), 221– 228.