You are on page 1of 16

This article was downloaded by: [Erciyes University]

On: 25 December 2014, At: 04:18


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Road Materials and Pavement Design


Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/trmp20

Determination of axle load spectra


based on percentage of overloaded
trucks for mechanistic-empirical
pavement design
a b a
Yanqing Zhao , Yiqiu Tan & Changhong Zhou
a
School of Transportation Engineering , Dalian University of
Technology , Dalian , 116024 , People's Republic of China
b
School of Transportation Science & Engineering , Harbin Institute
of Technology , Harbin , 150001 , People's Republic of China
Published online: 24 Oct 2012.

To cite this article: Yanqing Zhao , Yiqiu Tan & Changhong Zhou (2012) Determination of axle load
spectra based on percentage of overloaded trucks for mechanistic-empirical pavement design, Road
Materials and Pavement Design, 13:4, 850-863, DOI: 10.1080/14680629.2012.735796

To link to this article: http://dx.doi.org/10.1080/14680629.2012.735796

PLEASE SCROLL DOWN FOR ARTICLE

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
Downloaded by [Erciyes University] at 04:18 25 December 2014
Road Materials and Pavement Design
Vol. 13, No. 4, December 2012, 850–863

Determination of axle load spectra based on percentage of overloaded


trucks for mechanistic-empirical pavement design
Yanqing Zhaoa *, Yiqiu Tanb and Changhong Zhoua
a School of Transportation Engineering, Dalian University of Technology, Dalian 116024, People’s Republic

of China; b School of Transportation Science & Engineering, Harbin Institute of Technology, Harbin 150001,
People’s Republic of China

The mechanistic-empirical pavement design guide (MEPDG) developed under National Coop-
Downloaded by [Erciyes University] at 04:18 25 December 2014

erative Highway Research Program Project 1-37A requires the use of axle load spectra or load
distribution factors (LDFs) for pavement thickness designs. The MEPDG allows for various
levels of LDF inputs, varying from site-specific (level 1) to regional average (level 2) and
national average (level 3). There exists a concern that the MEPDG level 2 and level 3 inputs
do not take any site-specific information into consideration and thus may result in erroneous
thickness design. This study proposed a new approach to determine the axle LDF and num-
ber of axles per truck (NAPT) for level 2 and level 3 inputs. The proposed approach requires
that the design guide provide default LDFs and NAPTs for both normally loaded and over-
loaded trucks for various axle types and vehicle classes. Thus, design engineers can estimate
LDFs and NAPTs for a particular project site based on the percentages of overloaded trucks of
that site which are readily available from historical traffic data. The effectiveness of the pro-
posed approach was evaluated using weight-in-motion data collected from 26 sites in China.
The LDFs and NAPTs obtained from various approaches were used to predict truck factors
using the American Association of State Highway and Transportation Officials equation and
to predict pavement distresses using the MEPDG software. The prediction errors from the pro-
posed approach are substantially reduced when compared to those obtained using the MEPDG
approach, indicating the proposed approach to be a more accurate way for traffic loading charac-
terisation. The proposed level 2 inputs can further reduce the prediction errors when compared
to the proposed level 3 inputs. Among the three types of pavement distresses analysed using
the MEPDG software, namely rutting, bottom-up and top-down fatigue cracking, rutting is the
least sensitive to the variations in axle LDF and NAPT, while top-down cracking is the most
sensitive to the variations.
Keywords: traffic characteristics; load distribution factor; pavement design; trucks

Introduction
Background
Traffic data are one of the key inputs required for design and analysis of pavement structures.
Traditionally, traffic has been represented by equivalent single axle loads (ESALs), which convert
mixed traffic stream of different axle loads and axle configurations into one design traffic number.
This approach is currently used in the latest edition of the American Association of State High-
way and Transportation Officials (AASHTO, 1993) design guide. A new Mechanistic-empirical

*Corresponding author. Email: Yanqing_zhao888@126.com

ISSN 1468-0629 print/ISSN 2164-7402 online


© 2012 Taylor & Francis
http://dx.doi.org/10.1080/14680629.2012.735796
http://www.tandfonline.com
Road Materials and Pavement Design 851

pavement design guide (MEPDG) has been developed under National Cooperative Highway
Research Program (NCHRP) Project 1-37A (ARA, 2004). The MEPDG requires the use of axle
load distribution factors (LDFs), or axle load spectra, rather than the traditional ESAL inputs for
pavement thickness designs. The axle LDF represents the percentage of total axle load applica-
tions within each load interval for a specific axle type (single, tandem, tridem, and quad) and
vehicle class (VC). This is a more precise characterisation of traffic loads. Using axle LDFs, one
can calculate pavement responses due to each load interval, translate them into pavement damage
and arrive at the accumulated damage from all load intervals.
Accurate determination of axle LDFs plays a vital role in achieving a reliable pavement structure
design. Swan, Tardif, Hajek, and Hein (2008) reported that among various traffic inputs, the
axle LDF was the second most significant factor, after the number and type of trucks, affecting
the pavement performance. In the MEPGD, axle LDFs can be determined at three input levels
depending on the availability of weight-in-motion (WIM) data: level 1 – LDFs determined from
site- or segment-specific WIM data; level 2 – determined from statewide/regional WIM data;
Downloaded by [Erciyes University] at 04:18 25 December 2014

level 3 – determined from national WIM data (Applied Research Associates [ARA], 2004). Level
1 is considered to be the most accurate because it uses actual axle weights measured at or near the
project site. However, a key issue in the context of mechanistic-empirical (M-E) design lies in
selecting the appropriate load spectra for a particular design project in the absence of site-specific
WIM data or for level 2 and level 3 inputs.
Hajek, Selezneva, Jiang, and Mladenovic (2002) and Hajek, Selezneva, Mladenovic, and Jiang
(2005) developed the Pavement loading guide (PLG) to estimate traffic load characteristics for
Long-Term Pavement Performance (LTPP) sections which do not have measured axle weight
data. The PLG is based on selecting surrogate axle load spectra from similar sites in the same
jurisdiction. Although reasonable traffic loading estimates can be obtained by judiciously selecting
surrogate data, the procedure is somewhat subjective and dependent upon designers’ experience
and judgment. As pointed out by Hajek et al. (2005), surrogate data can never replace site-
specific data.
Special efforts were made in the NCHRP 1-37A project using WIM data from the LTPP database
to relate load spectra to other highway or traffic characteristics, especially to the truck traffic
classification (TTC) groups. The results showed that load spectra for each axle type were as
variable within each TTC as between the classifications. No other traffic or highway parameter
was found within which to group the axle load data except for axle type and VC. As a result, all
WIM data were combined together and the average values were used as the default load spectra
for the MEPDG level 3 inputs (ARA, 2004). Other researchers have used regionally collected
WIM data to develop default axle LDFs for level 2 inputs (Elkins & Higgins, 2008; Lu & Harvey,
2006, Swan et al., 2008; Tran & Hall, 2007). The same approach as that adopted in the MEPDG
was used in these studies.
Researchers have shown that there are notable differences between the default axle LDFs pro-
vided in the MEPDG and regional axle LDFs (Swan et al., 2008; Tran & Hall, 2007). Kim,
Titus-Glover, Darter, & Kumapley (1998) analysed load spectra using data from 21 WIM sites
in the 12-state North Central LTPP Region. They found that the region could not be charac-
terised with a set of representative load spectra. Several studies have demonstrated that the
axle LDF is statistically different from site to site (Huang, Sung, & Lin, 2002; Kim et al.,
1998; Turochy, Baker, & Timm, 2005). A study of Texas load spectra recommended that site-
specific axle LDFs should be used for high-volume pavement designs (Prozzi & Hong, 2005).
Axle LDFs obtained using the MEPDG approach for level 2 and level 3 inputs only represent
the average load spectra and they do not contain any site-specific information. Therefore, the
MEPDG approach needs to be improved to better characterise traffic loading conditions of a
given project site.
852 Y. Zhao et al.

Objectives
Given the problems stated previously, the primary objective of this research is to develop a
methodology to more accurately determine axle LDFs for level 2 and level 3 inputs, and to
demonstrate the effectiveness of the new methodology by comparing the prediction errors of
various pavement distresses obtained using the proposed and MEPDG approaches.

Research methodology
An important piece of information usually collected and kept by local transportation agencies is
the percentage of overloaded trucks for various VCs. This information was used in the past for
traffic management purposes. It seems that the percentage of overloaded trucks for a given project
is the only type of historical data available which relates to site-specific axle or truck weights and
consequently can be used in the determination of LDFs.
For level 1 LDF inputs, the approach proposed in this study is the same as the one adopted in
Downloaded by [Erciyes University] at 04:18 25 December 2014

the MEPDG. This study proposes a new approach for level 2 and level 3 inputs. The new approach
utilises the information on the percentage of overloaded trucks which is readily available from
local transportation agencies. Following the new approach, the M-E design guide should provide
the following default traffic parameters for various axle types and VCs as defined by the Federal
Highway Administrations (FHWA, 2001) vehicle classification system:

• Default axle LDFs for overloaded trucks;


• Default axle LDFs for normally loaded trucks;
• Default numbers of axles per truck (NAPTs) for overloaded trucks;
• Default NAPTs for normally loaded trucks;

A normally loaded truck is defined herein as a truck whose total gross weight is less than the
standard weight limit, as opposite to the overloaded truck. It is obvious that the sum of the
percentages of overloaded and normally loaded trucks is 100%. In performing a pavement structure
design, a design engineer needs to obtain site-specific percentages of overloaded trucks for various
VCs from historical traffic data and then use the following equation to determine the axle LDFs
for use in the MEPDG software:
LDFNijk × NAPTNij × PNi + LDFOijk × NAPTOij × POi
LDFijk = , (1)
NAPTNij × PNi + NAPTOij × POi

where LDFN and LDFO denote the default LDFs for normally loaded and overloaded trucks,
respectively, NAPTN and NAPTO the default NAPTs for normally loaded and overloaded
trucks, respectively, PN and PO the site-specific percentages of normally loaded and overloaded
trucks, respectively and subscripts i, j and k the ith VC of FHWA class 4–13 (FHWA, 2001), jth
axle type (single, tandem, tridem and quad) and kth load interval. Similarly, the NAPTs for use
in the MEPDG software can be determined according to the following equation:

NAPTij = NAPTNij × PNi + NAPTOij × POi . (2)

For the proposed level 2 inputs, regional default LDFs and NAPTs are used in Equations (1) and
(2), while for level 3 inputs, national default LDFs and NAPTs are used. The proposed approach
incorporates site-specific information (percentage of overloaded trucks) in the determination of
axle LDFs, rather than using nationally or regionally averaged LDFs, and therefore more accurate
results would be expected.
Road Materials and Pavement Design 853

Table 1. WIM site information.

Province Number of sites Site no. Year

Jiangsu 13 A1–A13 2009


Shandong 6 B1–B6 2009
Shanxi 7 C1–C7 2008

Determination of axle LDFs and NAPTs using the proposed approach


Traffic data
In order to use Equations (1) and (2) to determine the axle LDFs and NAPTs for a particular
project site for use in the MEPDG software, default axle LDFs and NAPTs for normally loaded
and overloaded trucks have to be developed for the design guide at both regional and national
levels. This requires extensive efforts of collecting WIM data regionally and nationally.
Downloaded by [Erciyes University] at 04:18 25 December 2014

In this research, WIM data were collected from freeway networks in three provinces (Shanxi,
Jiangsu and Shandong) in China. Since 2003, China has gradually adopted a new highway toll
policy that toll rates are determined based on both VC and gross weight (China Department of
Transportation, 2003). To this end, WIM devices, tire detectors and other relevant equipment were
installed at toll plazas. Traffic data collected at toll plazas are of high quality since axle weights are
measured when vehicles stop completely or travel at very low speeds (usually less than 10 km/h),
which minimises the dynamic loading effects frequently taking place when vehicles travel at
operating speeds. Moreover, WIM devices are well maintained and calibrated to ensure their
proper functioning and avoid any dispute between drivers and toll collectors.
A preliminary check of the WIM data showed that 26 stations provided data which were
continuously collected for more than 1 year. To avoid any possible effects of traffic seasonal
variations, 1 whole year of data from each of the 26 stations were used in this study. Information
on WIM data collection is summarised in Table 1. It is noted that for safety reasons, multi-trailers
(FHWA class 11–13) are not allowed to travel on freeways in China, and consequently no data for
multi-trailers are available for this study. As a matter of fact, the number of multi-trailers dropped
dramatically in China in the last decade. A survey conducted in 2009 on roadways of various
functional classes showed that out of 28,000 trucks surveyed, only 22 trucks were multi-trailers
(Zhao, Wang, & Wang, 2010).
It is worth noting that buses (FHWA class 4) are usually not overloaded. The weight of a bus
is directly related to the number of seats installed on the bus. To make the proposed approach
consistent, buses are also separated into two groups, namely type III (20–39 seats) and type IV (40
seats or more) groups. Thus, Equations (1) and (2) can also be applied to buses with the subscripts
for normally loaded and overloaded trucks referring to type III and type IV buses, respectively.
Because the new toll policy states that toll rates for buses are determined based solely on the
bus type (China Department of Transportation, 2003), axle weights of buses are not necessary
for toll collection purpose and they are not saved at most toll stations. Out of the 26 stations
used, only 5 stations from Jiangsu province provided axle weight data of buses. The bus axle
LDFs and NAPTs determined from these five stations were also used for other provinces and for
national-level analysis.

Default LDFs and NAPTs for normally loaded and overloaded trucks
Each record of the WIM data collected was examined to identify if the truck is overloaded. The
weight limits for various truck types, as specified in the Chinese specifications (China Department
of Transportation, 2004), are listed in Table 2. The corresponding FHWA VC is also presented in
854 Y. Zhao et al.

Table 2. Truck weight limits.

Truck description Weight limit (ton) Corresponding FHWA VC

Two-axle, six-tire trucks 17 5


Three-axle trucks 25 6
Four-axle trucks 35 7,8
Five-axle trucks 43 9
Six or more axle trucks 49 10

16

14 Jiangsu

12 Shandong
Shanxi
Axle LDF, %
Downloaded by [Erciyes University] at 04:18 25 December 2014

10
National
8

0
0 100 200 300 400
Axle Weight, KN

Figure 1. Default axle LDFs for normally loaded trucks (tandem axle, VC 9).

12

10 Jiangsu
Shandong

8 Shanxi
Axle LDF, %

National
6

0
0 100 200 300 400
Axle weight, KN

Figure 2. Default axle LDFs for overloaded trucks (tandem axle, VC 9).

the table. All WIM data were separated into normally loaded and overloaded truck groups. Axle
LDFs and NAPTs of the two truck groups were determined for each axle type and VC and for
each WIM site. Regional default LDFs and NAPTs were determined by taking the average of
the LDFs and NAPTs of all WIM sites in that region. National default LDFs and NAPTs were
obtained by taking the average of all WIM sites from the three provinces. Figures 1 and 2 present
Road Materials and Pavement Design 855

the regional and national default LDFs for normally loaded and overloaded trucks, respectively.
For clarity, only LDFs for tandem axle of FHWA VC 9 are shown in these figures.
Regional and national default NAPTs for normally loaded and overloaded trucks are listed in
Table 3. It can be observed from the table that there is no distinct discrepancy between the default
NAPTs for normally loaded and overloaded trucks. By assuming that the default NAPTs for the
two truck groups are same, Equation (1) can be reduced to the following simple form:

LDFijk = LDFNijk × PNi + LDFOijk × POi . (3)

However, since the default NAPTs for normally loaded and overloaded trucks can be easily
determined, once trucks are divided into two groups, Equation (1) is used in this study to achieve
better LDF predictions.
Because WIM data were collected from only three provinces in China, the results could hardly
be representative of national defaults. Nevertheless, the primary objective of this study is to
Downloaded by [Erciyes University] at 04:18 25 December 2014

propose a new approach of determining axle LDFs. Representative default inputs can be obtained
using the proposed approach once more WIM data across China become available.

Determination of LDFs and NAPTs


Once the default LDFs and NAPTs for normally loaded and overloaded trucks were available,
LDFs and NAPTs for each WIM site were determined according to Equations (1) and (2). For
comparison purposes, all WIM data were also analysed following the MEPDG approach in which
no separation of trucks into normally loaded and overloaded groups was performed. Thus, a total
of five different sets of inputs, designated as M1–M5, were produced for axle LDFs and NAPTs
and they are listed as follows:

• M1: MEPDG and proposed level 1 inputs; each WIM site has its own LDFs and NAPTs.
• M2: MEPDG level 2 inputs; all WIM sites in a given province use the same default LDFs
and NAPTs.
• M3: MEPDG level 3 inputs; all WIM sites from various provinces use the same default
LDFs and NAPTs.
• M4: proposed level 2 inputs; each WIM site has its own LDFs and NAPTs.
• M5: proposed level 3 inputs; each WIM site has its own LDFs and NAPTs.

Figure 3 presents an example of axle LDFs obtained using various approaches (M1–M5) for site
A1. For site A1, 54.5% of VC 9 trucks are overloaded. It is not uncommon to find that a significant
amount of trucks are overloaded in China highway network. Timm, Robbins, Huber, and Yang
(2011) conducted a full-scale pavement experiment in Shandong province, China, and they found
that 42% of single axles and 70% of tandem axles travelling on the test sections were overloaded
by US standards.

Evaluation of the effectiveness of the proposed approach


It is difficult to evaluate the effectiveness of the proposed approach by directly examining the
similarities between predicted and actual axle LDFs, because each LDF represents a curve of
distribution frequency versus load interval and the distribution frequency of heavy axle is much
more critical than that of light axle in terms of its damage effects on pavement structures. In this
study, the effectiveness of the proposed approach was evaluated in two ways. The first way was
to convert the axle LDFs and NAPTs obtained from various approaches (M1–M5) to traditional
856
Y. Zhao et al.
Downloaded by [Erciyes University] at 04:18 25 December 2014

Table 3. Default NAPTs for normally loaded and overloaded trucks.

Axle Normally loaded trucks Overloaded trucks


Province Type VC4 VC5 VC6 VC7 VC8 VC9 VC10 VC4 VC5 VC6 VC7 VC8 VC9 VC10

Jiangsu Single 1.99 2.00 2.13 1.99 2.13 1.94 1.88 1.97 2.00 1.98 2.00 2.07 1.97 2.00
Tandem 0.01 0.00 0.44 0.93 0.87 0.13 0.56 0.03 0.00 0.51 0.93 0.93 0.06 0.50
Tridem 0.00 0.00 0.00 0.07 0.00 0.93 1.00 0.00 0.00 0.00 0.07 0.00 0.97 1.00
Shandong Single 1.99 2.00 1.51 1.70 2.09 1.84 1.38 1.97 2.00 1.49 1.99 2.04 1.89 1.46
Tandem 0.01 0.00 0.75 0.81 0.91 0.34 0.81 0.03 0.00 0.75 0.89 0.96 0.23 0.77
Tridem 0.00 0.00 0.00 0.05 0.00 0.83 1.00 0.00 0.00 0.00 0.11 0.00 0.88 1.00
Shanxi Single 1.99 2.00 2.30 2.00 2.06 1.92 1.94 1.97 2.00 2.50 2.00 2.01 1.98 2.17
Tandem 0.01 0.00 0.35 0.97 0.94 0.24 0.53 0.03 0.00 0.25 0.99 0.99 0.06 0.42
Tridem 0.00 0.00 0.00 0.03 0.00 0.87 1.00 0.00 0.00 0.00 0.01 0.00 0.97 1.00
National Single 1.99 2.00 2.01 1.92 2.10 1.91 1.77 1.97 2.00 1.99 2.00 2.05 1.95 1.91
Tandem 0.01 0.00 0.49 0.91 0.90 0.21 0.62 0.03 0.00 0.51 0.93 0.95 0.10 0.55
Tridem 0.00 0.00 0.00 0.06 0.00 0.89 1.00 0.00 0.00 0.00 0.07 0.00 0.95 1.00
Road Materials and Pavement Design 857

8
M1
7 M2
6 M3

Axle LDF, %
M4
5
M5
4

0
0 100 200 300 400
Downloaded by [Erciyes University] at 04:18 25 December 2014

Axle weight, KN

Figure 3. Axle LDFs for tandem axle of VC 9 for site A1.

truck factors for various VCs and compare the results. For the M1, M2 and M3 approaches, the
truck factor was calculated using the following equation:
 
 
TFi = NAPTij (EALFijk × LDFijk ) , (4)
j k

where TF is the truck factor, EALF the equivalent axle load factor, and subscripts i, j and k
the ith VC, jth axle type and kth load interval. The EALFwas computed using the AASHTO
1993 equation for flexible pavements (AASHTO, 1993). A structure number of 5 and a terminal
serviceability of 2.5 were used in the analyses. For the M4 and M5 approaches, the truck factor
was determined using the following equation:

TFi = TFNi × PNi + TFOi × POi , (5)

where TFN and TFO are the truck factors for normally loaded and overloaded trucks, respectively.
The TFN and TFO were calculated using Equation (4) and the default LDFs and NAPTs for
normally loaded and overloaded trucks.
The second way of evaluating the proposed approach was to input the axle LDFs and NAPTs
obtained from various approaches (M1–M5) into the MEPDG software and compare the predicted
pavement distresses. Although various inputs required in predicting pavement distresses are site-
dependent, one purpose of this study is to investigate the effectiveness of the proposed approach,
and therefore all design inputs were held constant among software runs except for axle LDFs
and NAPTs. All analyses were conducted using a 15-year analysis period. A typical three-layer
asphalt pavement, comprising hot-mix asphalt over an unbound granular base on top of subgrade
soil, was used in the analyses. Important information related to the analyses is summarised in
Table 4. Three predicted distresses of interest are total rutting, top-down and bottom-up fatigue
cracking.
The predicted distress or truck factor from M1 inputs was used as benchmarks for comparisons.
The prediction error, represented as the difference between the distresses or truck factors predicted
858 Y. Zhao et al.

Table 4. Inputs for pavement distress predictions.

Description Input

Reliability 50%
Two-way AADTT 3000
Percent of trucks in design lane 95%
Monthly distribution factors 1.0 for all months for all VCs
VC distribution factors National default
Axle LDFs Variable
NAPT Variable
Asphalt layer
Thickness 20.3 cm
Mix 12.5 mm and PG 70–22
Modulus Level 3
Granular base
Thickness 35.6 cm
Downloaded by [Erciyes University] at 04:18 25 December 2014

Modulus 207 MPa


Subgrade
Classification A-7-6
Modulus 55 MPa

Table 5. Predicted truck factors and normalised errors for VC 9.

Truck factora Normalised error (%)


Site no. PO (%) M1 M4 M5 M2 M3 M4 M5

A1 54.5 3.45 3.31 3.21 10.4 6.5 4.2 7.2


A2 43.7 2.86 2.9 2.78 8.1 12.9 1.3 2.7
A3 46.2 3 2.99 2.88 3.2 7.8 0.1 3.9
A4 74 3.97 4.04 3.97 22.1 18.6 1.9 0
A5 34 2.27 2.53 2.41 36.2 42.3 11.6 6
A6 46.1 3.11 2.99 2.88 0.4 4 3.7 7.3
A7 46.1 2.96 2.99 2.88 4.6 9.3 1.2 2.6
A8 44.5 3.08 2.93 2.81 0.4 4.8 5 8.7
A9 41.2 2.62 2.8 2.69 18.2 23.4 7.2 2.6
A10 45.5 2.79 2.97 2.85 10.9 15.8 6.4 2.4
A11 51.6 3.07 3.2 3.09 0.7 5.2 4.1 0.7
A12 40.2 2.54 2.77 2.65 21.6 27.1 8.8 4.1
A13 76.7 4.49 4.15 4.07 31.1 28.1 7.6 9.3
B1 46.7 2.2 2.76 2.9 29.4 46.9 25.3 31.9
B2 49.9 3.13 2.88 3.03 9 3.3 8 3.2
B3 47.9 2.54 2.8 2.95 11.8 27 10.1 15.9
B4 48.6 2.88 2.83 2.98 1.3 12.1 1.9 3.2
B5 44.8 2.79 2.69 2.83 1.9 15.7 3.8 1.2
B6 57.6 3.52 3.15 3.33 19.2 8.3 10.4 5.5
C1 41.3 2.17 2.58 2.69 75.9 48.9 19.1 24.1
C2 85.3 4.32 4.5 4.41 11.7 25.2 4.2 2
C3 68.8 3.76 3.78 3.76 1.4 14.2 0.5 0.1
C4 90.9 4.38 4.74 4.62 13 26.3 8.2 5.5
C5 41.7 2.19 2.6 2.7 73.9 47.2 18.4 23.2
C6 71.5 4.44 3.9 3.87 14 27.2 12.1 12.8
C7 74.2 5.44 4.01 3.97 29.9 40.7 26.3 27
Average 17.7 21.1 8.1 8.2
a Truck factors are 3.09, 2.84 and 3.81 for M2 inputs for Jiangsu, Shandong and Shanxi,
respectively. It is 3.23 for M3 inputs.
Road Materials and Pavement Design 859

80.0

70.0

60.0 M2
Normalised error,%. M3
50.0
M4

40.0 M5

30.0

20.0

10.0

0.0
4 5 6 7 8 9 10
Downloaded by [Erciyes University] at 04:18 25 December 2014

Vehicle class

Figure 4. Normalised errors of predicted truck factors.

Table 6. Predicted rutting and normalised errors.

Ruttinga (mm) Normalised error (%)


Site no. M1 M4 M5 M2 M3 M4 M5
A1 20.3 19.8 19.6 3.9 3.4 2.1 3.1
A2 19.3 19.3 19.1 0.7 1.2 0.1 1.1
A3 19.3 19.3 19.1 0.3 0.8 0.3 1.2
A4 20.6 20.6 20.3 5.6 5.1 0.4 0.7
A5 18.5 18.8 18.8 4.2 4.8 1.4 0.4
A6 19.6 19.3 19.3 0.4 0.1 0.7 1.7
A7 19.8 19.6 19.3 1.3 0.8 1.2 2.2
A8 19.1 19.1 18.8 2.4 2.9 0.4 0.4
A9 18.3 18.8 18.8 6 6.5 3 2.1
A10 19.1 19.1 18.8 2.7 3.2 0 0.8
A11 18.8 19.1 19.1 3.1 3.6 1.6 0.8
A12 18 18.8 18.5 7.3 7.9 3.8 2.9
A13 21.3 20.8 20.6 8.5 8 1.6 2.8
B1 19.3 19.3 19.3 0.4 1.6 0.1 0
B2 18.5 19.3 19.3 3.8 5.1 3.8 3.7
B3 20.1 20.3 20.3 3.3 2.2 1.5 1.1
B4 21.1 19.8 19.8 7.9 6.8 5.2 5.6
B5 19.3 19.3 19.3 0.4 0.8 0 0.3
B6 20.6 20.3 20.3 5.8 4.7 0.9 1.4
C1 18.3 18.8 19.3 7.9 6.5 2.8 5.8
C2 20.8 20.6 21.3 4.5 5.8 0.9 2.6
C3 20.8 20.6 21.1 5.2 6.4 2.1 1.1
C4 20.6 20.6 21.3 3.8 5.1 0.1 3.1
C5 17.3 18.3 18.5 14.2 12.8 5.6 6.7
C6 19.6 19.6 20.1 1 0.3 0.3 2.7
C7 20.1 20.1 20.6 1.6 2.9 0.6 2.5
Average 4.1 4.2 1.6 2.2
a Rut depths are 19.6, 19.3 and 19.8 mm for M2 inputs for Jiangsu, Shandong
and Shanxi, respectively. It is 19.6 mm for M3 inputs.
860 Y. Zhao et al.

from M1 inputs and other input types, was normalised using the following equation:

|DMx − DM 1 |
NE = × 100%, (6)
DM 1

where NE is the normalised error, DM 1 the predicted distress or truck factor from M1 inputs and
DMx the predicted distress or truck factor from other input types (M2–M5).

Results and discussion


Truck factor analysis
The predicted truck factors for VC 9 and their normalised errors are listed in Table 5 for various
sites. It can be seen that for most of the sites, the normalised errors of truck factors predicted
Downloaded by [Erciyes University] at 04:18 25 December 2014

using the proposed approach (M4 and M5) are substantially reduced when compared to those
obtained using the MEPDG approach (M2 and M3). The table also shows that predicted truck
factors from M1 inputs vary significantly from site to site, indicating that the actual traffic loading
characteristics are distinctly different from site to site. The fact that the current MEPDG uses

Table 7. Predicted bottom-up fatigue cracking and normalised errors.

Bottom-up crackinga (%) Normalised error (%)


Site no. M1 M4 M5 M2 M3 M4 M5
A1 13 12 11.5 15.4 12.3 7.7 11.5
A2 9.44 10.5 10.1 16.5 20.8 11.2 7
A3 9.78 10.6 10.2 12.5 16.6 8.4 4.3
A4 14 14.2 13.7 21.4 18.6 1.4 2.1
A5 8.31 9.75 9.36 32.4 37.2 17.3 12.6
A6 10.8 10.8 10.4 1.9 5.6 0 3.7
A7 10.7 11 10.5 2.8 6.5 2.8 1.9
A8 11.2 10.3 9.9 1.8 1.8 8 11.6
A9 9.11 9.82 9.46 20.7 25.1 7.8 3.8
A10 9.9 9.96 9.6 11.1 15.2 0.6 3
A11 10.2 10.3 10 7.8 11.8 1 2
A12 8.89 9.82 9.48 23.7 28.2 10.5 6.6
A13 16.9 14.6 14 34.9 32.5 13.6 17.2
B1 6.99 8.61 9.93 32.6 63.1 23.2 42.1
B2 10.4 8.98 10.3 10.9 9.6 13.7 1
B3 8.97 10.9 12.3 3.3 27.1 21.5 37.1
B4 11.3 10.4 11.7 18 0.9 8 3.5
B5 10.3 9.37 10.7 10 10.7 9 3.9
B6 13.2 11.9 13.3 29.8 13.6 9.8 0.8
C1 9.67 11.4 10.9 52 17.9 17.9 12.7
C2 17.1 17 16.4 14 33.3 0.6 4.1
C3 20.2 17.5 17 27.2 43.6 13.4 15.8
C4 17.2 17.1 16.5 14.5 33.7 0.6 4.1
C5 8.81 10.6 9.71 66.9 29.4 20.3 10.2
C6 13.9 13.6 12.9 5.8 18 2.2 7.2
C7 12.9 14.8 14.2 14 11.6 14.7 10.1
Average 19.3 20.9 9.4 9.2
a Bottom-up cracks are 11.0, 9.3 and 14.7% for M2 inputs for Jiangsu, Shandong
and Shanxi, respectively. It is 11.4% for M3 inputs.
Road Materials and Pavement Design 861

Table 8. Predicted top-down fatigue cracking and normalised errors.

Top-down crackinga (m/km) Normalised error (%)


Site no. M1 M4 M5 M2 M3 M4 M5
A1 14.2 12.7 11.8 19.5 17 10.4 16.4
A2 10.1 10.5 9.8 12.9 16.5 4.1 2.4
A3 10.6 10.8 10.2 7.3 10.7 2 4.1
A4 17.3 17.7 16.3 34.1 32 2.4 5.7
A5 7.9 9.2 8.7 44 48.6 16 9.6
A6 12.4 11.4 10.7 8.2 5.3 8.1 14
A7 13.8 11.8 11.0 17.5 14.9 14.4 20.3
A8 9.2 9.6 9.1 23.9 27.8 4.1 1.6
A9 7.6 9.4 8.8 50.5 55.3 23.8 16.8
A10 10.6 9.5 9.1 7.7 11.1 10.2 14.5
A11 8.7 10.5 9.9 31.2 35.3 20.5 14.2
Downloaded by [Erciyes University] at 04:18 25 December 2014

A12 6.4 8.8 8.3 77.6 83.2 36.6 29.5


A13 22.5 20.5 18.7 49.4 47.8 9.2 17.2
B1 16.0 12.6 11.9 23.6 26.7 21.4 25.7
B2 7.8 12.1 11.4 56.7 50.4 55 46
B3 17.1 16.0 14.8 28.4 31.3 6.6 13.8
B4 19.1 13.9 12.8 35.9 38.5 27.5 33
B5 11.8 12.0 11.3 3.9 0.3 1.9 4.2
B6 17.7 16.2 14.9 30.9 33.7 8.9 16
C1 7.6 9.2 11.7 54.3 54.1 20.1 52.9
C2 17.2 15.7 21.2 31.6 31.7 9 23.2
C3 13.1 12.9 17.2 10.4 10.5 1.9 31
C4 15.2 15.2 20.3 22.3 22.4 0.4 33.8
C5 4.5 6.6 7.3 160.3 159.8 46.9 61.1
C6 13.0 12.3 15.9 9.1 9.2 5.4 22.7
C7 15.9 13.5 17.9 25.9 26 15.1 12.8
Average 33.7 34.6 14.7 20.9
a Top-down cracks are 11.4, 12.3 and 11.8 m/km for M2 inputs for Jiangsu, Shandong
and Shanxi, respectively. It is 11.8 m/km for M3 inputs.

average LDFs and NAPTs for level 2 and level 3 inputs makes it impossible to capture these
site-specific characteristics and consequently leads to large prediction errors.
Figure 4 presents the averages of normalised errors of predicted truck factors for various VCs.
It can be seen from Table 5 and Figure 4 that for most of the sites and VCs, the proposed approach
can reduce the normalised error to half or even less of that obtained using the MEPDG approach.
The proposed level 2 inputs (M4) can further reduce the normalised error when compared to the
proposed level 3 inputs (M5). The results indicate that the proposed approach is a better way of
traffic loading characterisation for design methods using the traditional ESAL concept.

Pavement distress analysis


For each of the 26 WIM sites, three sets of axle LDFs and NAPTs were produced by methods
M1, M4 and M5. Methods M3 and M2 provided one set and three sets of inputs, respectively.
Thus, a total of 82 software runs were performed. The predicted rutting, bottom-up cracking and
top-down cracking after 15 years and corresponding normalised errors for various sites are listed
in Tables 6–8, respectively. The tables show that for most of the cases analysed, the proposed
approach (M4 and M5) can reduce the normalised error to half or even less of that obtained using
the MEPDG approach (M2 and M3), indicating that the proposed approach is a more accurate
way for traffic loading characterisation.
862 Y. Zhao et al.

The tables show that among the three types of distresses predicted rutting exhibits much smaller
normalised errors than bottom-up and top-down cracking, indicating rutting is the least sensitive
to the variations in axle LDF and NAPT, while top-down cracking is the most sensitive to the
variations. An examination of the results for M4 and M5 shows that the proposed level 2 inputs can
further reduce the prediction error when compared to the proposed level 3 inputs. The averages of
normalised errors are about the same for the MEPDG level 2 and level 3 inputs (M2 and M3). As
discussed previously, the default axle LDFs and NAPTs for level 3 inputs were determined using
WIM data collected from only three provinces. The difference between the prediction results from
MEPDG level 2 and level 3 inputs could become more distinct once the level 3 default inputs are
developed using WIM data from more provinces.

Conclusions and recommendations


Downloaded by [Erciyes University] at 04:18 25 December 2014

This study proposed a new approach to determine axle LDFs and NAPTs for level 2 and level 3
inputs for the use in M-E pavement design. The proposed approach requires that the design guide
provide default LDFs and NAPTs for both normally loaded and overloaded trucks for various
axle types and VCs. Thus, design engineers can estimate LDFs and NAPTs for a particular project
site based on the percentages of overloaded trucks of that site. Based on the analyses and results
presented, the following observations and conclusions can be made:

(1) The proposed approach incorporates site-specific information (percentage of overloaded


trucks) in the determination of axle LDFs and NAPTs, rather than using nationally or
regionally averaged values, and therefore can achieve much better prediction results. For
most of the cases analysed, the proposed approach can reduce the prediction error to half
or even less of that obtained using the current MEPDG approach.
(2) The fact that the percentage of overloaded trucks is readily available from local
transportation agencies makes the proposed approach easy to implement in design
projects.
(3) The proposed approach is a more accurate way for traffic loading characterisation for
design methods using the traditional ESAL concept.
(4) Among the three types of pavement distresses analysed, namely rutting, bottom-up fatigue
cracking and top-down fatigue cracking, rutting is the least sensitive to the variations in
axle LDF and NAPT, while top-down cracking is the most sensitive to the variations.
(5) The proposed level 2 inputs can further reduce the prediction error when compared to the
proposed level 3 inputs.

With respect to the work presented herein, the following recommendations are made:

(1) Further studies are needed to evaluate the effectiveness of the proposed approach when
applied to rigid pavements.
(2) Data collected from 26 WIM stations in three provinces in China were used in this study.
Studies using data from more provinces and other countries are needed to further validate
the proposed approach.

Acknowledgements
This research was sponsored by the China Western Transportation Research Program (#200831800099),
Specialised Research Fund for the Doctoral Program of Higher Education (#20100041120005), and China
National Science Foundation (#50778057). The supports are gratefully acknowledged.
Road Materials and Pavement Design 863

References
American Association of State Highway and Transportation Officials. (1993). AASHTO guide for design of
pavement structures. Washington, DC: Author.
Applied Research Associates. (2004). Guide for mechanistic-empirical design of new and rehabilitated
pavement structures, National Cooperative Highway Research Program (NCHRP) Project 1-37A (Final
Report). Albuquerque, NM: Author.
China Department of Transportation. (2003). Guidelines for operation and management of toll roads. Beijing:
Author.
China Department of Transportation. (2004). Standards for truck sizes, axle load and weight limits (GB1589).
Beijing: Author.
Elkins, L., & Higgins, C. (2008). Development of truck axle spectra from Oregon weight-in-motion data for
use in pavement design and analysis (FHWA-OR-RD-08-06). Corvallis: Oregon Statute University.
Federal Highway Administrations. (2001). Traffic monitoring guide (FHWA-PL-01-021). Washington, DC:
Author.
Hajek, J.J., Selezneva, O., Jiang, J.Y., & Mladenovic, G. (2002). Improving the reliability of pavement
loading estimates using the pavement loading guide (Transportation Research Record, No. 1809)
Downloaded by [Erciyes University] at 04:18 25 December 2014

(pp. 93–104). Washington, DC: Transportation Research Board.


Hajek, J.J., Selezneva, O.I., Mladenovic, G., & Jiang, Y.J. (2005). Estimating cumulative traffic loads. II:
Traffic data assessment and axle load projection for the sites with acceptable axle weight data, final
report for phase 2 (FHWA-RD-03-094). Washington, DC: Federal Highway Administration.
Huang, W., Sung, Y., & Lin, J. (2002). Development of axle load distribution for heavy vehicles. Presented
at 81st Annual Meeting of the Transportation Research Board, Washington, DC.
Kim, J.R., Titus-Glover, L., Darter, M.I., & Kumapley, R.K. (1998). Axle load distribution characterization
for mechanistic pavement design (Transportation Research Record, No. 1629) (pp. 13–17). Washington,
DC: Transportation Research Board.
Lu, Q., & Harvey, J.T. (2006). Characterization of truck traffic in California for mechanistic-empirical design
(Transportation Research Record, No. 1945) (pp. 61–72). Washington, DC: Transportation Research
Board.
Prozzi, J.A., & Hong, F. (2005). Hierarchical axle load data for mechanistic-empirical design. Presented at
the 84th Annual Meeting of the Transportation Research Board, Washington, DC.
Swan, D.J., Tardif, R., Hajek, J., & Hein, D.K. (2008). Development of regional traffic data for the
mechanistic-empirical pavement design guide (Transportation Research Record, No. 2049) (pp. 54–62).
Washington, DC: Transportation Research Board.
Timm, D.H., Robbins, M.M., Huber, G.A., & Yang, Y. (2011). Analysis of perpetual pavement experi-
ment sections in China (Transportation Research Record, No. 2226) (pp. 94–103). Washington, DC:
Transportation Research Board.
Tran, N.H., & Hall, K.D. (2007). Development and influence of statewide axle load spectra on flexible
pavement performance (Transportation Research Record, No. 2037) (pp. 106–114). Washington, DC:
Transportation Research Board.
Turochy, R.E., Baker, S.M., & Timm, D.H. (2005). Spatial and temporal variations in axle load spectra and
impacts on pavement design, ASCE. Journal of Transportation Engineering, 131(10), 802–808.
Zhao, Y.Q., Wang, G.Z., & Wang, Z.C. (2010). Traffic inputs for the development of a new flexible pavement
design method (Western Transportation Research Program (No. 20083180009), Final Report). Beijing:
China Department of Transportation.

You might also like