You are on page 1of 14

C H A P T E R 7 : PA P E R 1 1 0

53 3

PA P E R 1 1 0

Statistical Analysis of LTPP SPS-3


Experiment on Preventive Maintenance of
Flexible Pavements
Hamid Shirazi
Applied Research Associates, Elkridge, Maryland, United States
Regis L. Carvalho
Applied Research Associates, Elkridge, Maryland, United States
Manuel Ayres Jr
Applied Research Associates, Elkridge, Maryland, United States
Olga Selezneva
Applied Research Associates, Elkridge, Maryland, United States

ABSTRACT
This paper describes the evaluation of preventive treatments in mitigating the rate of distress propagation in
flexible pavements. The analysis was based on data from preventive maintenance treatments data collected in
the Long Term Pavement Performance (LTPP) program. Data were obtained from 81 sites across the United
States and Canada which were part of the specific pavement experiments (SPS-3). SPS-3 was designed to
monitor the performance of four treatments: thin overlay, chip seal, crack seal and slurry seal under different
design conditions. Design conditions considered were precipitation, temperature, traffic, subgrade materials
and pavement condition prior to applying preventive treatment. Fatigue cracking, rutting and longitudinal
roughness data collected during the LTPP program were used to compare the overall performance of different treatments. A weighted average index was defined to represent the overall performance of the sections over
the years. Statistical techniques were used to compare the effectiveness of each treatment in relation to others and the control section, which did not receive any treatment. Conclusions from the analyses indicated that
thin overlay and chip seal are effective treatment options for most design conditions with respect to fatigue
cracking. Thin overlay outperforms other treatments in most design conditions with respect to rutting and
in some cases with respect to roughness. The difference between the performance of crack seal, slurry seal and
control section was not found to be statistically significant with respect to any distress type and design factor.

534

Compendium of Papers from the First International Conference on Pavement Preservation

B A C KG R O U N D
Rehabilitation and pavement preservation represent the majority of pavement maintenance activity in the
U.S. The importance of pavement network preservation cannot be overemphasized. Preventive maintenance
includes treatments that are applied to pavements primarily to prevent development of distress or to reduce
the rate of damage development. Preventive maintenance activities are focused primarily on improving pavement functional performance and prolonging pavement life through the following:
Waterproofing the pavement by sealing fine cracks or fissures.
Restoring proper characteristics of the pavement surface (e.g., to counteract weathering, aging, hardening, or raveling of asphalt).
Preventing further deterioration of cracks from contamination with incompressible materials that lead to
crack enlargement, and from hardening and raveling of the exposed crack surfaces.
Preventive maintenance activities are not intended to have a direct improvement of pavement structural
capacity. Selection of the appropriate type and timing of treatment application forms the basis for a sound
preventive maintenance practice. The selection of the preventive treatment is often based on the experience
and local practice specific to a region or district within a public highway agency.
To help highway agencies to make the most informative and cost effective decision regarding preventive
maintenance options, the Strategic Highway Research Program (SHRP) initiated project H-101, Pavement
Maintenance Effectiveness in late 1980s. The purpose of project H-101 was to develop a database that would
permit increased understanding of selected maintenance treatments in extending pavement service life or reducing the development of pavement distress, including an evaluation of the cost effectiveness of pavement
maintenance treatments.
Flexible pavement sites selected for SHRP H-101 continued to be monitored under FHWA LTPP program as Special Pavement Studies-3 (SPS-3) experiment. The LTPP SPS-3 experiment was designed in 1990
to evaluate the effectiveness of maintenance alternatives and to determine the optimum timing for applying
treatments for flexible pavements. In addition, local highway agencies were encouraged to integrate supplemental experimental sections into collection and analysis associated with the SPS-3 experiment. The purpose
of supplemental sites was to compare differences in performance of treatments due to different material characteristics and construction techniques employed locally by the agencies.
This research focused on evaluation of the results of the LTPP SPS-3 experiment. The objective of this
study was to assess the effect of the above mentioned maintenance treatments on pavement performance and
to develop quantitative information on performance and optimum timing of preventative maintenance treatments. The assessment of treatments included identification of important pavement performance indicators, analysis of the impact of each treatment on pavement performance, and analysis of the impact of
treatment timing on pavement performance.
Over the years, numerous SHRP, FHWA and NCHRP sponsored studies were conducted to assess SPS
experiments status, key data element availability (e.g. traffic, subgrade, materials, monitoring), as well as conduct preliminary analyses of the collected data. Smith, Freeman and Pendleton (1993) implemented damage
modeling approach with an index varying from 0 to 1. The index is dependent on accumulated traffic/age,
expected traffic/age to failure and the shape of the performance trend. This model became the base for
remaining life analysis used in developing the AASHTO 1993 design guide.
Morian, Epps and Gibson (1997) concluded that thin asphalt concrete overlay treatments performed best
after 5 years as compared to other treatments. Chip seal performance was best in the southern region with predominantly wet-no freeze environment. They also concluded that crack seal treatment performed very well

C H A P T E R 7 : PA P E R 1 1 0

53 5

in wet-freeze environments where the wide shallow sealant reservoir was routed. Crack seal performance in
the other two regions was not as successful. The study indicated that the question of timing could not be totally resolved from the visual observation of the SPS-3 sites, but indications were that earlier applications of
the preventative maintenance treatments provided greater benefits than later application.
A year later in another study, Morian, Gibson and Epps (1998) concluded that structural adequacy did not
have a significant effect on the performance of SPS-3 treatments. Thin overlay had significant effect in rutting and roughness reduction while other treatment options were either slightly effective or non-effective.
They developed multiple regression models for cracking, rutting, ride quality, and friction predictions and an
index called Pavement Rating Score (PRS).
Eltahan and Von Quintus (2001) conducted survival analysis of SPS-3 sites in the Southern LTPP region
in 1999 to obtain life expectancy of each treatment, effect of timing and the benefit of treatment to the life span.
The study revealed that, after six years of service, sections which received maintenance when in poor condition had a probability of failure twice as much as sections initially in fair or good conditions. Also, sections
in fair and good condition had about the same probability of failure. The overall median survival times for
thin overlay, slurry seal, and crack seal were 7, 5.5, and 5.1 years, respectively. A median survival time for chip
seal could not be determined because fewer than 50 percent of these sections had failed at the time of the
analysis. The study concludes that chip seals outperformed thin overlay, slurry seal, and crack seal treatments
with respect to controlling the reappearance of distress.
Hall, Correa and Simpson (2002) showed that more than 40 percent of SPS-3 sites had problems in the
application of maintenance treatments, mostly chip seal. Analysis showed that the most effective maintenance
treatments was thin overlay treatment, followed by chip seal treatment and slurry seal in terms of roughness,
rutting, and fatigue cracking. Thin overlay treatment was the only one of the four SPS-3 maintenance treatments to produce an initial small reduction in roughness, and the only one of the four to have a significant
effect on long-term roughness, relative to the control section. For rougher pavements, however, there was
some evidence that chip seals and slurry seals also had some effect on long-term roughness, rutting, and cracking, relative to the control section. Crack seals did not have any significance on long-term roughness, rutting,
or fatigue cracking.

DESIGN EXPERIMENT
SPS-3 experiment was initiated in the United States and Canada in 1990 and 1991. An experimental design
was developed to help determine the impact of important factors on flexible pavements. Design factors selected
include the following:
1. Moisture (2 categories: wet, dry)
2. Temperature (2 categories: freeze, no-freeze)
3. Subgrade type (2 categories: fine grained, coarse grained)
4. Traffic loading (2 categories: low, high)
5. Pavement condition (3 categories: good, fair, and poor)
Therefore, the design experiment contained 11 categories. This created a total of 48 combinations (classes)
in the design experiment. 33 States and Canadian provinces participated in the experiment. Some States assigned few sites to the experiment bringing the total number of experiment sites to 81. In practice however,
there were some deviations from design. The number of sites in each design category is presented in Figure 1.
As shown in the figure, the distribution of the sites is not uniformly based on actual experiment design. For
example, there are more sites in freeze zone as compared to no-freeze zone; so are pavements in initial poor
conditions compared to fair and good conditions.

536

Compendium of Papers from the First International Conference on Pavement Preservation

Figure 1. SPS-3 Experiment Breakdown Based on Major Design Factors

Table 1 illustrates the SPS-3 sites based on actual design factors. As an example, site labeled KE-A in Table
1 is site A in Kentucky which is located in wet-freeze zone with fine subgrade material and low traffic. The site
was initially in good condition. Although based on the experiment design, at least one site should have been
assigned to each of the 48 classes, many classes were left without a designated site in the actual experiment as
shown in Table 1.

M A I N T E N A N C E T R E AT M E N T S
Four preventive maintenance treatments were considered in the experiment. These include thin HMA overlay, slurry seal, crack seal and chip seal. The effectiveness of combinations of treatments was not considered.
Test sites were constructed with five consecutive sections each receiving one of the treatments and a control
section with no treatment. All sections of a site were subject to the same traffic loading and environmental conditions. In many cases these sites were linked to GPS sections serving as control sections. The following is a
brief description of the treatments used in the LTPP SPS-3 study:
Thin Hot-Mix Asphalt Overlay The primary purpose of thin HMA overlays was to improve pavement
surface condition, protect the pavement structure, reduce the rate of pavement deterioration, correct surface deficiencies, reduce permeability and improve the ride quality of the pavement, particularly when accompanied by surface milling. For SPS-3 experiment, thin HMA overlay thicknesses ranged from 0.5 in
to 1.6 in.
Slurry Seal The primary purpose of slurry sealing was to seal low-severity surface cracks, waterproof the
pavement surface, and improve skid resistance at speed 30 mph. Slurry seal is a mixture of well-graded aggregate (fine sand and mineral filler) and asphalt emulsion that is spread over the entire pavement surface with a squeegee or a spreader box attached to the back of a truck. Thickness is generally less than 0.4
in.

C H A P T E R 7 : PA P E R 1 1 0

53 7

Table 1. Actual SPS-3 Categories

Crack Seal The primary purpose of crack sealing was to prevent the intrusion of moisture through
existing cracks. Crack sealing is applied to working cracks ( i.e., those that open and close with changes
in temperature). Sealants are typically thermo-plastic (bituminous) materials that soften upon heating and
harden upon cooling.
Chip Seal The primary purpose of chip sealing was to seal pavement surface and improve friction. An
asphalt emulsion is directly applied to the pavement surface (0.35 to 0.50 gal/yd2) followed by the application of aggregate chips (15 to 50 lb/yd2), which are then immediately rolled to imbed chips.

538

Compendium of Papers from the First International Conference on Pavement Preservation

A S S E S S M E N T O F S P S - 3 D ATA AVA I L A B I L I T Y
An important aspect of this study is in-depth knowledge of the contents of the LTPP database. The information in the LTPP database is divided into the following modules:
Inventory: section location, characteristics, and material characteristics.
Materials Testing: material properties and characteristics.
Climatic: temperature, humidity, precipitation, cloud cover, and wind statistics.
Maintenance: activities performed since inclusion in the LTPP program.
Rehabilitation: major improvements since inclusion in the LTPP program.
Traffic: annual traffic summary statistics since it was opened to traffic.
Each of these modules contains tables that provide information on the various design features and performance measurements of a particular pavement section. Figure 2 provides a summary of SPS-3 data availability. Figure 2a presents the histogram of the number of times the sites were surveyed. As shown in the
Figure 2a, the sites were surveyed between 2 and 9 times during the experiment. Most sites were surveyed 6
times. Figure 2b presents the number of years the sites were monitored. As shown, 88% of the sites were monitored for at least 4 years and about 22% of the sites were surveyed for 10 years or more. Figure 2c illustrates
the number of times the treatments were applied to the sites. In 74% of the sites the treatment was applied
only once as expected, however, 22% of the sites have received the treatment twice and 4% for 3 times. Treatment application more than 1 time might be due to the long time some sites were monitored. By March 1 2006,
all SPS-3 sites were de-assigned from the experiment.

P E R F O R M A N C E I N D I C AT O R
To evaluate the effectiveness of preventive maintenance treatments regarding the performance of pavement
sections, the alternatives at each site were compared among each other and the control section. Previous studies have used many different performance indicators. The most common were: most recent survey measures
of distress (Hall et al, 2002), expert Task Group field reviews of distress (Morian et al, 1997), evaluation of distress trends (Morian et al, 1998; Von Quintus, Simpson, and Eltahan, 2001; Ambroz and Darter, 2001), distress
regression models (Morian et al, 1998), area under/above condition indicators (Chatti et al, 2004), distress level
immediately after rehab/treatment (Ambroz et al, 2001), and average distress over survey period (Ambroz et
al, 2001; Von Quintus, 2001)
Different alternatives for performance indicators were evaluated in this study. The objective was to find
an indicator that could represent the pavement performance over the monitoring period and that provided a
simple, stable and comparable parameter that could minimize the effect of survey measurement errors. The
parameter selected was the weighted average of the distress normalized over the survey period, as calculated
using the following equation:

Where
WD is the weighted distress value (e.g. area of fatigue cracking) over the total survey period
i is the survey number (i=0 is the initial distress level immediately after the treatment)
Di is the distress value measured at the ith survey
Pi+1 is the period (in years) between survey i and survey i+1
n is the total number of surveys for the section

C H A P T E R 7 : PA P E R 1 1 0

Figure 2. Assessment of SPS-3 data availability

53 9

540

Compendium of Papers from the First International Conference on Pavement Preservation

The weighted average, in reality, represents the total area under the distress versus time curve normalized
by the total time period between first and last surveys. As such it is a measure of pavement performance
relative to the specific distress over the entire survey period. Moreover, the normalized value allows for comparisons of pavement sections at different survey periods. The advantages of using the weighted distress value
are the following:
The measured variable is related to the pavement performance over the whole analysis period;
The analysis is repeatable and not subjective;
The concept is similar to the relationship between performance and serviceability;
The effect of variability from measurements by different surveyors is reduced;
It offers a parameter that can be used to compare sections with different survey periods.
The effectiveness of treatments in prolonging the pavement life was evaluated using a couple of load-associated distresses and ride quality surveyed in the LTPP program for the SPS-3 experiment. Performance was
evaluated as the deterioration measured by fatigue cracking, rutting and roughness (using the international
roughness index, IRI).
Figures 3 to 5 illustrate Box-Whisker plot of weighted distress index for fatigue cracking, rutting and IRI
of all SPS-3 sites. The boundaries of the box present lower and upper quartiles and the middle line is the median. The whisker marks the minimum and maximum limits of the distress. Only the surveys that contained
distress measurements for all treatments were used to draw the graphs, and some outliers were excluded.
As shown in Figure 3, thin overlay and chip seal exhibited lower fatigue cracking quantities as compared
to slurry seal, crack seal and control section. The median of chip seal was slightly lower than thin overlay. Chip
seal also presented the smallest min-max range. Sections that were treated with crack seal had lower weighted
average fatigue cracking as compared to slurry seal and control section. Slurry seal seems to be less effective
in mitigating the progression of fatigue cracking over the monitored period.
Figure 4 illustrates a significant difference between sections that were treated with thin overlay over other
sections. Thin overlay was effective to reduce rutting immediately after the treatment and, as consequence,
presented a lower level of rutting over the analysis period. Although chip seal marginally outperformed the
rest of treatments, the difference has little advantage over the remaining treatments and it is not statistically
meaningful.
Similar conclusion is found in terms of IRI performance, as show in Figure 5. Thin overlay was the most
effective maintenance option in mitigating IRI over the years. Possibly, this type of treatment was the most
effective to reduce initial roughness immediately after the treatment, rather than providing a significant structural improvement of the pavement section However, inferring more conclusion from other treatments require
more rigorous statistical analysis which is presented in the next section.

C H A P T E R 7 : PA P E R 1 1 0

Figure 3. Weighted average fatigue cracking; median min-max box chart

Figure 4. Weighted average rutting; median min-max chart

54 1

542

Compendium of Papers from the First International Conference on Pavement Preservation

Figure 5. Weighted average IRI; median min-max chart

S TAT I S T I C A L A N A LY S I S A P P R O A C H
The statistical test selected for the analysis, was the Friedman Test. This is a non-parametric test (distributionfree) used to compare paired observations on a subject. It is also called a non-parametric randomized block
analysis of variance. Unlike the parametric repeated measures ANOVA or paired t-test, this test makes no
assumptions about the distribution of the data (e.g., normality). In addition, unlike the t-test, Friedman test
can be used for multiple comparisons, as is the case for the SPS-3 experiment with 4 different types of treatment, in addition to the control section. Friedman test, like many non-parametric tests, uses the ranks of the
data rather than their raw values. The test statistic for the Friedmans test is a Chi-square with n-1 degrees of
freedom, where n is the number of repeated measures.
The performance of the pavement sections with the preventive maintenance treatment was compared
with the performance of similar pavements without the treatment in the control sections, as well as between
the different treatment types. Friedman test was applied for all 11 categories of the design factors for each distress type. The values used were the weighted average distresses normalized for the analysis period. The results
indicate whether a statistically significant difference exists between any pair of treatments.
As an example, Table 2 presents the results for wet climate. A total of 41 sites were located in wet climatic
region with fatigue cracking measurements available for all treatments. Based on test results, thin overlay and
chip seal had the lowest sum of ranks, and slurry seal and control section presented the highest. A lower sum
of ranks indicates better performance.

C H A P T E R 7 : PA P E R 1 1 0

54 3

Table 2. Friedman Results of Fatigue Cracking for Treatments in Wet Climate

Analysis results indicated that the difference in performance of thin overlay and chip seal compared to slurry
seal and the control sections was significant. Also, chip seal performed significantly better than crack seal.
The difference between other treatments was not significant.
The results for all design conditions are summarized in Table 3. To compare the treatments, a design category from the left side of the table and a distress from the top of the table are chosen. One of the treatments
from the left and another one from the top are selected. If the intersecting cell is empty, there is no significant difference identified; otherwise, the cell is filled with treatment performing better. In the table, TH, SL,
CR, CH and CO are acronyms for thin overlay, slurry seal, crack seal, chip seal and control section respectively.
As shown in the table, chip seal and thin overlay performed better with respect to fatigue cracking. Thin
overlay was the preferred treatment with respect to rutting. No treatment was found prevalent in treating IRI;
however, thin overlay outperformed other treatments in some cases.
Although conclusions such as these are valuable, the main contention of the SPS-3 is to identify the
effects of the major design factors on treatments performance. In other words, the experiment was intended
to identify if different climate conditions, subgrade material, traffic level or initial pavement condition affects
the choice of a preferred treatment. Therefore, the results from different categories of design factors were compared for each distress, as summarized in the following sections.

544

Compendium of Papers from the First International Conference on Pavement Preservation

Fatigue Cracking

Temperature makes a significant difference when comparing thin overlay with slurry seal and the control section. Thin overlay outperforms both treatments in freeze zone. Temperature does not affect the performance
of other treatments since chip seal outperforms slurry seal, crack seal and control section in both freeze and
no freeze zones. The same conclusion is valid for precipitation. Thin overlay outperforms slurry seal and control section only in wet regions. There are no other significant differences in the performance of other treatments with respect to temperature.
It was found that the subgrade type affects the performance of the thin overlays. Thin overlay generally
performs better than slurry seal if the subgrade is coarse material. Traffic affects the performance of the thin
overlay as compared to slurry seal and control section. Under high traffic condition, the performance of thin
overlays prevails over both the control and slurry seal sections.
The performance of the maintenance treatments with respect to initial pavement condition is regarded
as application timing. Timing of treatment application influences its performance. If the pavement is in poor
conditions, thin overlay performs better than slurry seal, crack seal and the control section. Also, chip seal outperforms crack seal with a significant statistical difference under such pavement conditions. It is important
to note that the comparisons were based on fatigue cracking surveys and chip seals, in general, may only mask
the cracks, rather than correcting the distress.
Rutting

In most cases thin overlay performs better for rutting when compared to the other treatments and the control section. Temperature plays a significant role when comparing chip seal and slurry seal performance.
Under freezing condition, chip seal outperforms slurry seal. Only precipitation was statistically influential in
rutting when comparing chip seal with thin overlay and slurry seal. In wet conditions, thin overlay outperforms
chip seal, and in dry condition chip seal performs better than slurry seal.
Under higher traffic roads, thin overlay performs better than the control section while in low traffic there
is not a significant difference. Also surprisingly, slurry seal performs worse than control section under low traffic condition with respect to rutting. No other design factor affects the performance of the pavement in a statistically significant way with respect to rutting
International Roughness Index

Temperature is an important design factor to define maintenance activities in regard to surface smoothness.
Thin overlay performed significantly better than slurry seal, crack seal and chip seal in freezing condition. In
no-freeze condition, there were not a significant differences among treatments with respect to IRI. Precipitation does not affect the performance of the treatments. There is no significant difference among treatments
in dry and wet conditions.
In pavements with coarse subgrade, thin overlay was superior to chip seal. In pavements with fine subgrade, thin overlay outperforms control section. Subgrade material does not influence the performance of the
treatments in any other combination.
Traffic also affects the performance of maintenance treatments when roughness is considered. Thin overlay outperforms crack seal, control section and chip seal in roads under higher traffic. In low traffic roads there
were no significant differences.
Pavement initial condition affects the performance of the treatments. Thin overlay performs better than
crack seal, control section and chip seal if the condition of the pavement is poor. There is no significance difference in other combinations.

C H A P T E R 7 : PA P E R 1 1 0

54 5

Table 3. Summary of Friedman tests results

*TH, SL, CR, CO and CH are the acronyms for thin overlay, slurry seal, crack seal, control section and chip seal, respectively.

546

Compendium of Papers from the First International Conference on Pavement Preservation

CONCLUSIONS
Based on the analysis of SPS-3 sites of LTPP program, the following could be concluded:
Fatigue Cracking: Thin overlays and chips seal were more effective than slurry seal and crack seal treatments in mitigating fatigue cracking;
Rutting: Thin overlay was the prevalent option with respect to rutting. Chip seal was more effective than
slurry seal in freeze zones and in wet regions. There were no significant differences between slurry seal,
crack seal and the do nothing scenario with respect to rutting.
If the pavement is in freeze zone or with high traffic or initially in poor condition, thin overlay performs
better than when the pavement is in non-freeze zone, with low traffic and in fair or good condition with
respect to fatigue cracking and IRI;
Thin overlay was also more effective in wet regions only with respect to fatigue cracking;
Chip seal performance was not affected by any of the design factors with respect to fatigue cracking;
Design factors have very little or no influence on treatments with respect to rutting. Chip seal is only
marginally more effective in freeze zone and in dry climate.

ACKNOWLEDGMENT
The authors would like to acknowledge FHWA and LTPP program for sponsoring this study conducted under
a LTPP research project.

REFERENCES
Ambroz, J. A. and Darter, M. I., (2005). Rehabilitation of Jointed Portland Cement Concrete Pavements: SPS-6 Initial Evaluation
and Analysis. Federal Highway Administration, Office of Infrastructure Research and Development, Report No. FHWA-RD-01-169.
October 2005.
Chatti, K., Buch, N., Haider, S. W., Pulipaka, A., Lyles, R. W., and Gilliland D. (2004). LTPP Data Analysis: Influence of Design and Construction Features on the Response and Performance of New Flexible and Rigid Pavements, Draft Final Report. National Cooperative
Highway Research Program, NCHRP Project 20-50(10/16), Transportation Research Board, National Research Council. 2004.
Eltahan, A. A., Von Quintus, H. L., (2001). LTPP Maintenance and Rehabilitation Data Review. Federal Highway Administration,
Office of Infrastructure Research and Development, Washington, D.C., Report Number FHWA-RD-01-019. 2001.
Hall, K. T., Correa C. E., Simpson, A.L., (2002). LTPP Data Analysis: Effectiveness of Maintenance and Rehabilitation Options, National
Cooperative Highway Research Program, NCHRP Web Document 47 (Project 20-50[3/4]): Contractors Final Report. June 2002.
Morian, D. A., Epps, J. A., and Gibson, S. D., (1996). Pavement Treatment Effectiveness, 1995 SPS-3 and SPS-4 Site Evaluations.
Federal Highway Administration, Tech Brief, Report No. FHWA-RD-96-208. 1996.
Morian, D. A., Gibson, S. D., and Epps, J. A., (1998). Maintaining Flexible Pavements The Long-Term Pavement Performance
Experiment SPS-3 5-Year Data Analysis. Federal Highway Administration, Office of Infrastructure Research and Development,
Washington, D.C., Report No. FHWA-RD-97-102. 1998.
Smith, R. E., Freeman, T. J., and Pendleton, O., (1993). Pavement Maintenance Effectiveness. Strategic Highway Research Program,
Report SHRP-H-358. 1993.
Von Quintus, H. L., Simpson, A. L., and Eltahan, A. A., (2001). Rehabilitation of Asphalt Concrete Pavements Initial Evaluation of
the SPS-5 Experiment. Federal Highway Administration, Office of Engineering Research and Development, Report No. FHWA-RD01-168. 2001.

You might also like