Documente Academic
Documente Profesional
Documente Cultură
Abstract: Truck traffic is a crucial factor that results in pavement damage. The increasing demand of freight transport contributes to rising
truck volumes. In some countries there is a trend toward using larger vehicles, and road authorities are concerned about their effect on
pavement because of the lack of pavement maintenance funding. This paper investigates the efficiency of freight transport by comparing
the effect of six-axle articulated trucks (semi-trailers) and nine-axle articulated trucks (B-doubles) with regard to pavement performance when
carrying various payloads based on weigh-in-motion data. Mathematical models were developed to help decision makers consider how to
distribute road freight, to minimize pavement damage induced by the different types of trucks. It was found that there is reduced pavement
damage when one or more of the following conditions are met: The payload of the trucks gets closer to their optimum value, the percentage of
the empty vehicles is decreased, and more freight is undertaken by B-doubles. In addition, a simplified pavement performance prediction
model was used as a basis to determine the future pavement maintenance and rehabilitation (M&R) schedules, and therefore help in compar-
ing the long-term pavement treatment costs for different traffic loading scenarios. The outcome from the economic analysis shows that
there would be significant benefits in the pavement M&R costs over the pavement service life by improving the distribution of freight.
DOI: 10.1061/JPEODX.0000099. © 2019 American Society of Civil Engineers.
Author keywords: Road freight transport; Pavement damage; Weigh-in-motion (WIM) technology; Mathematical model.
freight assignment. In this paper, two representative types of large Source: Data from Austroads (2012).
vehicles that operate in Australia are studied: semi-trailers and
B-doubles.
Deer Park Bypass from April 1–30, 2013. This road is a divided
The overall objectives of this paper are to (1) investigate and
freeway located approximately 10 km west of the central city area
compare the two types of freight vehicles on pavement infra-
of Melbourne, which is a good representative for investigating the
structure with various levels of payloads; (2) minimize the total
features of freight vehicles due to the high level of daily truck
pavement damage by optimizing the freight assignment for trucks
movements. After filtering the data records carefully and classify-
based on WIM data; and (3) calculate the cost savings in the multi-
ing the trucks based on the Austroads vehicle classification system
year pavement M&R activities by applying the optimized freight
(Austroads 2008, 2012), it was found that 77% of 1-month road
arrangement.
freight task in terms of weight was carried by semi-trailers and
B-doubles. Table 2 lists the detailed attributes of the WIM recorded
semi-trailers and B-doubles. The freight capacity of B-doubles in
Vehicle Configuration and Truck Factor
terms of weight (46.6 t) was substantially larger than that of semi-
In this paper, ESA was applied to quantify the overall pavement trailers (29.8 t). The 1-month freight task undertaken by the two
damage induced by heavy vehicles. The truck factor, also termed vehicle configurations was approximately 2.34 × 105 t, and the
the number of ESAs per individual truck, was estimated using the percentage of freight carried by more than 11,200 semi-trailers
recorded weights and standard loads for axle groups based on the was 54%, whereas the rest was undertaken by approximately 4,200
fourth power law developed in the United States in the 1950s. To B-doubles. The number of ESAs (overall pavement damage) for
calculate the ESA for the truck, the recorded weights for each axle transporting the total freight was around 3.4 × 104 , 58% of which
group are divided by the standard loads for the corresponding axle was caused by semi-trailers. Although the average ESA of semi-
group and raised by the power of 4 and then summed together. Dual trailers was much lower than that of B-doubles, the average payload
tires were assumed to be present on all axles but the steer axle carried by B-doubles was considerably higher.
(Thompson et al. 2016). The standard axle loads for different As recorded, both vehicle classes carried various amounts of
axle-tire combinations in Australia are listed in Table 1. freight; their distributions are shown in Fig. 2. The normalized fre-
Eqs. (1) and (2) offer examples of how to calculate the ESA for quency of semi-trailers and B-doubles falling into each load group
the two types of trucks investigated in this study. Semi-trailers have was calculated. The distributions tended to be bimodals, represent-
three axle groups and axle pattern “1-2-3,” representing a single ing the prevalence of either unloaded or almost fully loaded ve-
steer axle followed by a tandem axle group and then a triaxle group. hicles. The reason for the existence of these empty trucks might
B-doubles are vehicles with four axle groups and axle pattern be that they were on the way back after completing their freight
“1-2-3-3,” depicting a leading single axle followed by a tandem transport tasks. However, these empty trucks still induced pave-
axle group and then two triaxle groups. Fig. 1 shows the typical ment damage while having no contribution in freight transport.
configurations of semi-trailers and B-doubles. Moreover, there was a considerable proportion of trucks only par-
• Semi-trailers: tially utilizing their weight capacity. This may be due to the limi-
tation in the volume capacity of trucks or the lack of organization of
MoAG1 4 MoAG2 4 MoAG3 4 the freight assignment.
ESA ¼ þ þ ð1Þ
5.4 13.8 18.5
Fig. 3. Trend of ESA 1 standard deviation against the growth of payload for semi-trailers and B-doubles.
Therefore, various variables should be considered when mini- ESA of the vehicles was calculated. The ESA values for other pay-
mizing the total ESA. Among them, two are investigated in depth load levels were obtained similarly. The number of ESA per vehicle
in this paper: the payload-related ESA equations for the two types against the weight of the freight carried is plotted in Fig. 3, and the
of trucks; and individual vehicle number at each payload bin. standard deviations of the individual ESAs were marked. The ESA
grew exponentially when the payload increased, and the equations
and their coefficients of determination (R2 ) are shown in Fig. 3.
Payload-Related ESA Module At higher payload levels, B-doubles appeared to have significant
advantages in ESA compared with semi-trailers when undertaking
The average ESA values of vehicles carrying a variety of loads were the same amount of freight. With the satisfactorily regressed ESA-
calculated based on WIM data. The vehicles were classified into payload equations (high R2 values), the ESA of either type of trucks
integer freight weight groups, and the average ESA was calculated could be estimated based on the freight weight it carries. Therefore,
to diminish the scatter of the data. For example, the vehicles carry- the equations on Fig. 3 are used in the objective function [Eq. (3)]
ing 4.5 to 5.5 t of freight were grouped as 5-t class and the average for optimization.
Fig. 4. Trend of the efficiency of semi-trailers and B-doubles carrying various payloads.
Fig. 5. Total ESA under the optimal payload points and WIM-recorded payload distributions against different proportions of freight carried by
B-doubles (freight amount: W ¼ 2.34 × 105 t).
Fig. 6. Total ESA plotted against various truck payload combinations when the total freight amount and B-doubles’ freight share equal the
WIM-recorded statistics (W ¼ 2.34 × 105 t, p0 ¼ 46%).
limit. The minimum total ESA among all of the payload combina- indicates that the total ESA is not very sensitive to the variations
tions is marked by a circle in the pattern, which corresponds to the of payloads in a specific range. In other words, it may still have
advantageous points given in Fig. 4. The surface of the ESA in the advantages in reducing the total ESA when the payloads for semi-
blue area was flatter and lower compared with other areas, which trailers and B-doubles have some variations.
Fig. 7. Fitted statistical distributions for normalized frequency of payload for (a) semi-trailers; and (b) B-doubles.
Optimization Scenario 2: Vehicles Carrying Different where f1 and f2 = PDF of the theoretical distributions that are
Payloads (p i and p j Follows Statistical Distributions) added together with proper proportions (pe & 1 − pe ) to form the
final PDF for the whole bimodal. Commonly used distributions
Fitted Theoretical Distributions were applied to fit the measured payload distributions. The normal
In this scenario, statistical distributions were used to fit the WIM distribution was selected to model the second peak, whereas log-
recorded payload spectra for semi-trailers and B-doubles, to inves- normal distribution was chosen for the first peak to avoid the neg-
tigate the probability of the vehicles falling into each payload ative values. The PDF for normal and lognormal distributions are as
interval, which are shown in Figs. 7(a and b). The reason for follows:
applying the theoretical distributions instead of using polynomials • Normal distribution PDF:
to fit the curve is that the latter has little statistical meaning, even
though higher R2 values could be achieved. In addition, compared 1 2 2
f1 ðx; μl ; σl Þ ¼ pffiffiffiffiffiffi e−ðx−μl Þ =2σl ð10Þ
with discrete probability function, statistical distributions are more σl 2π
straightforward and computationally efficient than tables of values.
Because the payload distributions for both truck types tended to be • Lognormal distribution PDF:
bimodal, they were decomposed into the sum of two theoretical
distributions. This can be expressed as follows: 1 2 2
f 2 ðx; μe ; σe Þ ¼ pffiffiffiffiffiffi e−ðln x−μe Þ =2σe ð11Þ
f ¼ pe f 1 þ ð1 − pe Þf2 ð9Þ σe x 2π
Figs. 7(a and b) were separated into two parts by a dash line. Rel- had a considerable impact on the total pavement damage. More-
atively unloaded trucks are considered at the left of the dash line, over, the impact of the number of empty B-doubles was greater
and loaded trucks to the right. than that of semi-trailers on pavement damage. Therefore, it would
be meaningful if the decision makers could lower the percentage of
Impact of Proportion of Relatively Unloaded Piece on empty trucks by resource integration and optimization.
Pavement Damage ESA Calculation Equations
The relatively unloaded piece accounted for a considerable percent-
age for both semi-trailers and B-doubles, which also induced pave- X
m
ESAs ¼ f s ðxi ; pe ; μe ; μl ; σe ; σl ÞdV s Es ðxi Þ ð18Þ
ment damage with little contribution to freight transportation. The i¼1
influence and sensitivity of the proportions of relatively unloaded
trucks on the ESA for the two types of trucks are investigated. X
n
To study the unloaded part, other factors should remain constant, ESAb ¼ f b ðxj ; pe0 ; μe0 ; μl0 ; σe0 ; σl0 ÞdV b Eb ðxj Þ ð19Þ
such as the total vehicle number, the percentage share of freight j¼1
by each truck type, the deviations of the payload, and the first
peak position of payload distribution. To keep the total amount of Optimization Model: Case Study
freight the same, the mean value of the normal distribution will vary After studying the impact of the unloaded part on the total ESA, a
as the proportion of the relatively empty part changes. Calculation mathematical model was developed to minimize the pavement
of the mean value of the second normal distribution is given in damage by optimizing the payload distributions for the two types
Eqs. (12) and (13) for semi-trailers and B-doubles, respectively. of trucks. The percentages of relatively empty parts for semi-trailers
Second Peak Position Calculation Equations and B-doubles were set to be 35% and 20% (half of the WIM re-
• Semi-trailers: corded percentages) as a case study. Although these percentages
were arbitrarily chosen, they reflect reality, acknowledging that in
X
m
practice empty runs cannot be completely eliminated. In the math-
f s ðxi ; pe ; μe ; μl ; σe ; σl ÞdV s xi ¼ ð1 − p0 ÞW ð12Þ ematical modeling, only the right-side distributions (i.e., loaded
i¼1
trucks) were optimized. Specifically, the mean values for relatively
• B-doubles: loaded semi-trailers and B-doubles were the decision variables. In
addition, the average load of the left-side lognormal distributions
X
n
f b ðxj ; pe0 ; μe0 ; μl0 ; σe0 ; σl0 ÞdV b xj ¼ p0 W ð13Þ and the standard deviations for mixed distributions in the objective
j¼1 function were same as the values in Figs. 7(a and b). The previous
objective function can be updated as follows:
where μe ; σe ; σl ; μe0 ; σe0 ; σl0 = parameters of fitted statistical • Objective function:
distributions (listed in Fig. 7). Xm
Subject to: Min ESA ¼ fs ðxi ; pe ; μe ; μl ; σe ; σl ÞdV s Es ðxi Þ
i¼1
d ¼ 0.1 ð14Þ
X
n
þ fb ðxj ; pe0 ; μe0 ; μl0 ; σe0 ; σl0 ÞdV b Eb ðxj Þ ð20Þ
W; po ; V s ; V b ¼ WIM recorded values ð15Þ j¼1
where
Subject to:
pe ∈ ½0; 0.1; 0.2; 0.3; 0.4; 0.5; 0.6 ð16Þ
xi ∈ ½0; ls ∀i ð21Þ
pe0 ∈ ½0; 0.1; 0.2; 0.3; 0.4 ð17Þ xj ∈ ½0; lb ∀j ð22Þ
Solve: X
n
fb ðxj ; pe0 ; μe0 ; μl0 ; σe0 ; σl0 ÞdV b xj ¼ p0 W ð23Þ
μl & μl0 j¼1
Fig. 8. The second peak position of payload distribution for (a) semi-trailers; and (b) B-doubles.
pe0 ¼ 0.20 ð27Þ The optimization results of the mathematical model and the
minimum ESA values when changing the percentage of freight car-
where μe ; σe ; σl ; μe0 ; σe0 ; σl0 = parameters of fitted statistical distri- ried by B-doubles are shown in Fig. 10. Similarly, the change in the
butions (listed in Fig. 7). total ESA under the WIM-recorded payload distribution is plotted
Decision variables: as a comparison. The minimum ESA was also obtained when all of
the freight was carried by B-doubles, which decreased by 27.2%
μl & μl0 when applying the optimal payload distribution compared with the
current ESA state marked in the pattern. In addition, when the
After running the optimization model, it was found that the freight share by the two truck classes followed the WIM-recorded
value of the decision variables remained the same when changing values (semi-trailers: 54%; B-doubles: 46%), the total ESA can be
the proportions of freight carried by B-doubles. This is because dur- decreased by 15.8%. According to Fig. 10, there is a significant
ing optimization, the parameters of the most efficient load spectra advantage in reducing the pavement damage when the percentage
would remain the same (Fig. 10), which would not be influenced of empty trucks was decreased by 50%, so the total freight task
by the percentages of freight undertaken by B-doubles. The can be distributed around the optimum payload points. The opti-
changes in the freight shares of the two sorts of trucks only had mized distributions indicate how much pavement damage could be
an effect on their individual vehicle volumes (V s & V b ). avoided through a more judicious distribution of freight. The values
Fig. 9. Total ESA plotted against the variations of the relatively unloaded piece for semi-trailers and B-doubles.
Fig. 10. Total ESA under the results of the optimal model and WIM-recorded payload distribution against different percentages of freight carried by
B-doubles (freight amount: W ¼ 2.34 × 105 t).
of ESA1 , ESA2 , and ESA3 marked in the pattern are the three traffic pavements for constructing highways and freeways than rigid pave-
loading scenarios selected for the following economic analysis. ments. A common structure of a flexible pavement normally con-
sists of multiple layers: surface course, base course, subbase course,
and subgrade. Each layer is constructed with different materials.
Economic Analysis
The subgrade is the natural formation of existing soils, whereas the
subbase course is a layer of material beneath the base course, which
Pavement Performance Model for Flexible Pavements is usually made up of treated subgrade. The base course consists
To calculate the economic savings in pavement treatments resulting of a mixture of granular materials, such as sand, crushed stone, or
from the reduced ESA, a typical flexible pavement structure was crushed slag. The top wearing course is commonly a layer of
selected for predicting the pavement deterioration. This is be- asphalt.
cause the studied WIM road section belongs to the flexible pave- Pavement performance prediction models should be considered
ment category. In addition, there are wider applications of flexible when determining the M&R schedules. For this study, a simplified
ESA1 34,048 × 12
PSRt ¼ PSRi − AF × a × STRbt × Agect × CESAdt ð28Þ ESA2 28,662 × 12
ESA3 24,776 × 12
where PSRt = pavement serviceability at year t; PSRi = initial Note: ESA1 ¼ WIM recorded distribution, B-doubles’ share: 46%; ESA2 =
value of PSR after construction or major rehabilitation; a, b, c, optimization distribution, B-doubles’ share: 46%; and ESA3 = optimization
d = pavement performance coefficients that are dependent on the distribution, B-doubles’ share: 100%.
type of the pavement; AF = adjustment factor that is relevant to the
climate conditions; STRt = structural number of the pavement at
year t; Aget = age of the pavement at year t after initial construction growth rate for truck traffic were presumed and considered to be
or a major rehabilitation; and CESAt = cumulative 80-kN equiv- constant over the pavement service life. To investigate the potential
alent single-axle loads applied to pavement in the heaviest traffic advantage of optimized ESA in pavement rehabilitation costs, two
lane in year t (unit: millions). ESA scenarios from the optimal payload distributions were se-
At different levels of pavement conditions, the corresponding lected. One was obtained when 46% of the total freight task was
pavement treatment activities should be carried out to avoid the pre- carried by B-doubles, whereas the other was from B-doubles trans-
mature pavement failures and provide acceptable ride quality for porting all of the freight. The WIM-recorded ESA value was also
road users. There are different pavement repair activities that can considered as a comparison. The source of the three ESA values can
be categorized as routine maintenance, preventive maintenance, mi- be found in Fig. 10.
nor maintenance, and major rehabilitation. Generally, routine, pre- The analysis period and the terminal serviceability index were
ventive, and minor maintenance have only a small effect on the set to be 50 years and 2.5, respectively. Major rehabilitation would
improvement of pavement condition, whereas major rehabilitation be triggered when the PSR was close to 2.5, to guarantee that the
has noticeable effects (Rajagopal and George 1991; Abaza and pavement condition would always be above an acceptable level.
Abu-Eisheh 2003; Chootinan et al. 2006). For the long-term pave- Based on the parameters given in Table 3, the pavement service-
ment maintenance horizon, the selection of the types of treatments ability can be predicted. Hence, the timing and frequency of pave-
as well as the activity timing is critical for maximizing the pave- ment treatments were determined. Fig. 11 depicts the PSR curves
ment performance and minimizing the pavement repair costs. How- resulting from different scenarios of ESA. After optimizing the
ever, this study focuses on the economic benefits of the reduced payload spectra, within a 50-year period, the number of major reha-
ESA resulting from the improved assignment of freight for heavy bilitation activities required decreased from three to two. Moreover,
vehicles. Therefore, only one type of treatment/major rehabilitation if a larger percentage of freight was undertaken by B-doubles, it
was considered, to simplify the calculation and put emphasis on the would be more beneficial in pavement treatments, which was re-
variable, CESAt . The major rehabilitation would be triggered when flected by the remaining life of the pavement at the end of the
the predicted pavement performance reaches the terminal service- analysis period, also called the salvage value.
ability, which would restore the pavement to the initial condition. To quantify the pavement M&R activities, the difference in costs
for the three rehabilitation schedules shown in Fig. 11 was com-
Calculation of Cost Savings in Pavement M&R pared using the present value (PV), which was determined through
Activities discounting all future project costs to the base year, normally the
year of construction (Walls and Smith 1998). The equation for
According to Eq. (28), CESAt should be calculated to predict the calculating the pavement treatment costs as follows:
pavement performance, which is determined according to the base-
X
n t T
year ESA and the compound yearly ESA growth rate, as follows: 1 j 1
PV ¼ M −S ð30Þ
ð1 þ GRÞt − 1 i¼1
1þr 1þr
CESAt ¼ ESAi × ð29Þ
GR
where PV = present value of all M&R costs; M = cost of the indi-
where ESAi : = initial annual ESA at base year for scenario i (unit: vidual M&R activity (unit: $); r = discount rate; n = total number of
millions); and GR = annual growth rate (unit: %). M&R activities over the analysis period; tj = number of years from
Table 3 lists the specific values of the variables for the pavement the present to the jth M&R activity; S = salvage value at the end
performance model. The initial PSR is 4.5, which is a common of the analysis period (i.e., the pavement remaining life); and T =
value applied in practice (Lee et al. 1993; Chootinan et al. 2006; length of the analysis period (unit: years).
Batouli and Mostafavi 2014). The calibrated parameters consisting Therefore, according to Eq. (30), the PV costs for the three dif-
of AF, a, b, c, and d, which are dependent on different climate ferent scenarios can be obtained. As a case study, the discount rate
zones and types of pavements, can be traced in Lee et al. (1993). was assumed to be 5% and the results are found in Table 4. As
For this study, the pavement was defined as flexible and the climate indicated, this would have a significant advantage in decreasing
zone for Melbourne was selected as intermediate and no freeze. The pavement treatment costs through optimizing freight assignment.
values of the pavement structure number as well as the annual When the optimal payload distribution was applied, 14.99% of
Fig. 11. Comparison of pavement rehabilitation schedules resulting from different payload spectra and freight shares.
Table 4. Present value of the three pavement rehabilitation schedules 3. To distribute the freight task more reasonably, the approach of
(r ¼ 5%) applying mathematical models to optimize the load spectra
ESA scenario PV was considered. Mixed distributions were used to fit the WIM-
recorded complicated payload spectra. When decreasing the
1 0.5618M
2 0.4776M (−14.99%)
proportion of empty trucks on the road, the freight can be as-
3 0.4297M (−23.51%) signed more evenly (i.e., decreasing the mean payload value),
thus reducing the total pavement damage. To obtain the opti-
mum payload spectra, a case study was developed, which shows
significant advantages in decreasing pavement damage but al-
lowing more flexibility in freight assignment compared with
pavement repair costs could be saved if 46% of the total freight was all vehicles carrying the same payload; and
carried by B-doubles in comparison to costs based on WIM- 4. A simplified pavement performance prediction model was ap-
recorded ESA. Moreover, when B-doubles undertook the entire plied to investigate the potential benefits of the optimized spec-
amount of freight, the benefit in the cost savings was greater, tra in cost savings in pavement M&R activities. Through the
decreasing by 23.51%. case study, it was shown that the pavement treatment costs over
its analysis period will be reduced considerably by optimizing
the freight assignment.
Conclusions The study gives the decision makers an indication of the poten-
tial use of WIM data resources and a prototype of how to assign the
This paper focused on the freight transportation of two popular freight task more efficiently in terms of pavement performance,
types of trucks in Australia, semi-trailers and B-doubles, investigat- thus decreasing the pavement repair costs. Other factors can also
ing their vehicle efficiency in terms of pavement performance be included in the future to make the freight assignment more prac-
and how to distribute the freight task more properly to minimize tical, such as vehicle operating costs, environmental costs, and
their induced pavement damage. In addition, the cost savings in social costs.
pavement-treatment activities by applying the optimized freight as-
signment were studied. WIM data, which can provide detailed
weight information about the trucks, were used to study their ef- References
fects and current load spectra. The following is a summary of the
main results: AASHTO (American Association of State Highway and Transportation
1. When carrying the same load, B-doubles have a greater advan- Officials). 1993. Guide for design of pavement structures. Washington,
tage in preventing pavement damage than semi-trailers at higher DC: AASHTO.
Abaza, K. A., and S. A. Abu-Eisheh. 2003. “An optimum design approach
load levels;
for flexible pavements.” Int. J. Pavement Eng. 4 (1): 1–11. https://doi
2. The best vehicle efficiency occurrs when the weights of .org/10.1080/1029843031000087464.
freight carried by semi-trailers and B-doubles are 13 and 19 t, Austroads. 2000. Weigh-in-motion technology. AP-R168/00. Sydney,
respectively, and the benefit of reducing the total ESA in- Australia: Austroads.
creases with the growth of the percentage of freight carried by Austroads. 2008. Guide to pavement technology. II: Pavement structural
B-doubles; design. AGPT02/08. Sydney, Australia: Austroads.