Estimation and uncertainty analysis of energy consumption and CO2 emission of asphalt pavement maintenance

Estimation and uncertainty analysis of energy consumption and CO2 emission of asphalt pavement maintenance

Accepted Manuscript Estimation and uncertainty analysis of energy consumption and CO2 emission of asphalt pavement maintenance Bin Yu, Shuyi Wang, Xin...

1MB Sizes 0 Downloads 6 Views

Accepted Manuscript Estimation and uncertainty analysis of energy consumption and CO2 emission of asphalt pavement maintenance Bin Yu, Shuyi Wang, Xingyu Gu PII:

S0959-6526(18)31089-8

DOI:

10.1016/j.jclepro.2018.04.068

Reference:

JCLP 12649

To appear in:

Journal of Cleaner Production

Received Date: 22 August 2017 Revised Date:

3 April 2018

Accepted Date: 8 April 2018

Please cite this article as: Yu B, Wang S, Gu X, Estimation and uncertainty analysis of energy consumption and CO2 emission of asphalt pavement maintenance, Journal of Cleaner Production (2018), doi: 10.1016/j.jclepro.2018.04.068. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT 1

Estimation and Uncertainty Analysis of Energy Consumption and

2

CO2 Emission of Asphalt Pavement Maintenance Bin Yu1,*, Shuyi Wang1, Xingyu Gu1, 2, * 1. School of Transportation, Southeast University, Nanjing, China 210096 2. College of Engineering, Tibet University, Lhasa, Tibet, China 850011, *Corresponding authors, email: [email protected]; [email protected]

RI PT

3 4 5 6 7

Abstract: Substantial energy consumption and CO2 emission are generated in asphalt pavement maintenance and their quantifications based on individual case are highly contingent on the input data qualities and model parameters, exhibiting high uncertainties. This study established energy consumption and CO2 emission uncertainty assessment statistical methodology to deal with the issue. Two sources of uncertainty, the data quality one and the model parameter one, were identified. The former was captured by converting the input data to probability density function (PDF) using Beta distribution following the definitions of data quality pedigree matrix; the latter was assessed by defining uncertainty factor (UF) to determine the distribution form and parameter. Environmental data of 18 field asphalt pavement maintenance projects were collected, including hot mix asphalt (both base and SBS modified asphalts), hot in-plant recycling (15% reclaimed asphalt pavement), cold in-place recycling and cold in-plant recycling asphalt maintenance plans, covering pavement material production, transportation, mixture preparation and construction phases. The PDFs and statistical parameters (mean, standard deviation and percentiles etc.) were obtained via Monte Carlo simulation for these maintenance measures. This study further developed environmental burden comparative parameter to construct the 95% confidence intervals for the comparisons of different maintenance measures. The methodology proposed in this study captured the uncertainty of energy consumption and CO2 emission calculation results of maintenance activity and allow researchers to evaluate the results in a statistical view.

26 27

Keywords: asphalt pavement maintenance; environmental burden; uncertainty analysis; data quality evaluation; Monte Carlo simulation

28

Highlights

29

1. Data quality uncertainty and model parameter uncertainty were characterized

30

2. Probability density functions of asphalt pavement maintenance measures were built.

31

3. Environmental burdens of each measure were compared at the statistical level.

M AN U

TE D

EP

AC C

32

SC

8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

1

ACCEPTED MANUSCRIPT 1. Introduction

2 3 4 5 6 7

The maintenance activities of asphalt pavement are essential in maintaining pavement performances. Asphalt pavement maintenance is to save cost, reduce environmental burden so as to reap the economic and social benefits. However, maintenance behavior itself also brings significant environmental burden, such as energy consumption and greenhouse gas (GHG) emissions, so the environmental burden evaluation of asphalt pavement maintenance has become an increasingly important proposition (Huang et al. 2009).

8 9 10 11 12 13 14 15 16 17 18 19 20 21 22

Wen et al. (2014) performed quantitative evaluation of the burden of the use of reclaimed asphalt pavement (RAP) on the energy consumption and GHG emission of hot mix asphalt (HMA). They found the energy consumption and GHG emissions are affected by the RAP content in HMA, the moisture in the RAP, and the HMA discharge temperature and established corresponding empirical models. Cheng et al. (2010) compared the air emissions of warm mix asphalt (WMA) and HMA during the paving process. Through field data collection, they observed a reduction of 35.4% of CO, 53.0% of NO2, 63.1% of SO2 and 56.2% of volatile organic compounds (VOCs) for WMA compared with HMA. Aside from conventional engineering performance and economy consideration, environmental burden has now been taken into account in the decision making process of asphalt pavement maintenance (Zhang et al. 2010; Batouli et al. 2017). For instance, Yu et al. (2015) established a multi-objective optimization model of asphalt pavement maintenance integrating the cost, performance and environment elements at the project level. It is found there is an opportunity of reducing the cost and environmental burden to 80.3% and 77.8% and increasing the pavement performance to 146.6% compared to the base case. However these objects can not be realized simultaneously.

23 24 25 26 27 28 29 30 31 32 33

Above mentioned researches drew their conclusions based on the deterministic approach while uncertainty, more or less, exists in the environmental burden calculation results inevitably. The reliability of calculation results is very important for implementing environmental-friendly options in pavement, for instance, application of low GHG emission paving materials (Jamshidi et al. 2012), optimization of pavement construction activities (Zhang et al. 2015), and selection of more sustainable pavement designs (Wang et al. 2016). However, the uncertainty analyses of calculation results in pavement area are limited. Noshadravan et al. (2013) considered the uncertainty associated with the input parameters of pavement life cycle assessment models and evaluated the credibility of conclusions drawn regarding the environmental implications of alternative pavement designs. Yu et al. (2016) estimated the energy consumption variations of pavement material productions based on the available datasets.

34 35 36 37 38 39 40 41 42

The accuracy of calculation result highly depends on the quality of the environmental data in the life cycle modeling. Unfortunately, many available studies (Jullien et al. 2014; Giani et al. 2015; Chen et al. 2016) did not take into account the impact of uncertainty of environmental burden calculation results on the conclusions or decisions. According to the summary of Santero et al. (2011), the energy consumption intensity range of cement production is 4.6-7.3 MJ/kg, and the energy consumption intensity range of asphalt production is 0.7-6.0 MJ/kg. Such a big difference is not surprising because of the differences in system boundaries, in the production processes and technology that depend on local areas, and many other factors that lead to fluctuation. Different researchers choose different energy intensity values (some with justification

AC C

EP

TE D

M AN U

SC

RI PT

1

2

ACCEPTED MANUSCRIPT and some are not), which will have a great impact on the calculation results, and even lead to the opposite conclusions, e.g., comparisons of environmental burden of cement concrete pavement and asphalt pavement. Furthermore, the analyses of the existing studies focus on individual cases. The calculation parameters of different phases in pavement life cycle carry the variations that are pertinent to the corresponding cases, such as aggregate grade design, transport distance and mode, mixture heating and mixing equipment efficiency, which make the conclusion of different researches lack of the same foundation.

8 9 10 11 12 13 14

Facing the above mentioned shortcomings, this study limits the research scope to the quantitative evaluation of energy consumption and CO2 emission in asphalt pavement maintenance. Based on the field data, the uncertainties of the calculation results of asphalt pavement maintenance measures were captured. This study intends to: 1) establish the methodology for evaluating the reliability of environmental burden calculation results; 2) compare the environmental burden of different asphalt pavement maintenance measures in a statistical view.

15

2. Methodology

16 17 18 19 20 21

The calculation of the environmental burden of asphalt pavement maintenance measures requires a large amount of input data, and the accuracy of the calculation results depends on the reliability of the input data. However, the input data has certain degree of uncertainty. In the context of this study, the uncertainties were divided into two categories, namely data quality uncertainty and model parameter uncertainty, and the corresponding methods to evaluate them quantitatively were established.

22

2.1 Data Quality Uncertainty

23 24 25 26 27

Empirical or expert estimates, measurements/human errors, data that are not representative (such as outdated data, data whose time and space, technical differences are too large), and even data missing are sources of data quality uncertainty, for instance the energy consumption intensity of producing unit of road construction materials. Three potential methods can be used to qualitatively represent the data quality uncertainty targeting different sample sizes:

28 29

1. Large size of field data: use parameter estimation techniques and goodness-of-fit test to fit distributions when sufficient data are available;

31

SC

M AN U

TE D

EP

AC C

30

RI PT

1 2 3 4 5 6 7

2. Small size of field data: simply use log-normal distribution to avoid negative values; 3. Paucity of field data:

32 33 34 35

Without field data, the following procedures can be used to characterize data quality uncertainty: 1) perform literature review to obtain secondary data and use them as proxy; 2) build data quality indicator (DQI) of proxy in terms of selected evaluation criterions; 3) convert DQI to probability density function (PDF).

36 37

The data as the input for asphalt pavement maintenance is first evaluated about the quality by the ‘Pedigree Matrix’ method consisting of various DQIs, as listed in Table 1.

38 3

ACCEPTED MANUSCRIPT Table 1 Data Quality Pedigree Matrix (based on May and Brennan 2003; Weidema and Wesnaes 1996) 5

4

3

Verified data partly Reliability

Verified data by

based on assumptions

measurements

or non-verified data by

Non-verified data partly by assumptions

measurements

from a sufficient Completeness

Representative data

Representative data

from a smaller number

from a smaller

from adequate

of sites and shorter

number of sites

number of sites

periods or incomplete

but for adequate

but from shorter

data from

periods

periods

sample of sites over an adequate period to

fluctuations

correlation

Non-qualified estimate

Representativeness Unknown or

incomplete data from a

smaller number of sites and/or from shorter

an adequate number of

periods

sites and periods

Less than 3 years difference to year of study

Less than 6 years

Less than 10 years

Less than 15 years

difference

difference

difference

Average data from Geographical

Data from area

larger area in which the

Correlation

under study

area under study is included

Data from area with

Data from area with

similar production

slightly similar

conditions

production conditions

Data for processes

Data from processes

Technical

enterprises,

and materials under

and materials under

Correlation

processes, and

study but from

study but from

different enterprises

different technology

TE D

Data from

materials under study

by industrial expert)

1

Representative data but

even out normal

Temporal

Qualified estimate(e.g.,

M AN U

Representative data

2

RI PT

Indicator score

SC

1 2

Data on related

processes or materials

Age of data unknown or more than 15 years difference Data from unknown area or area with very different production conditions Data on related processes or materials

but same technology

but different technology

The qualitative DQI score (average of the five indicator scores of equal weighting) is then transformed to PDF. With different DQI scores, the PDF varies so that one can extract statistical information, such as mean, standard deviation, percentiles, to perform uncertainty analysis. Statistical analysis results reveal more possible scenarios and capture the inherent variability of data in the life cycle modeling.

8 9 10 11 12 13

The modified Beta distribution was used prevalently in previous studies (Tan et al. 2002, Wang and Shen 2013) to build PDF (Eq.1) based on Table 2. The preference of Beta probability function is primarily due to the fact that “the shape parameters and range end points allow virtually any shape probability distributions to be represented” (Canter et al. 2002). The shape parameters establish the shape of the distribution and thus the location of the probability mass, whereas the endpoints limit the range of possible values.

14

15

AC C

EP

3 4 5 6 7

f ( x;α , β , A, B ) = [1/( B − A)] × {Γ (α + β ) / Γ (α ) × Γ ( β )} × [( x − A) /( B − A)]α −1 × [( B − x ) /( B − A)]β −1 (A ≤ x ≤ B ) where α , β are the distribution’s shape parameters; A , B are the selected range endpoints.

16 4

(1)

ACCEPTED MANUSCRIPT 1 2

Table 2 Converting Parameters of Beta Distribution (Based on Kennedy et al., 1996; Canter et al., 2002; Weidema and Wesnæs, 1996) Aggregated DQI

Beta distribution

5.0

4.5

4.0

3.5

3.0

2.5

2.0

1.5

1.0

shape parameter (α , β )

(5,5)

(4,4)

(3,3)

(2,2)

(1,1)

(1,1)

(1,1)

(1,1)

(1,1)

range endpoints ( ± )

10

15

20

25

30

35

40

45

50

2.2 Model Parameter Uncertainty

4 5 6

Model parameter uncertainty refers to the variations of input parameters that may influence the environment burden predictions of the model. An example to calculate the energy consumption for transportation is presented.

EC = L × FE (2)

SC

7

RI PT

3

where EC is energy consumption (MJ); L is transportation distance (L); FE is fuel economy (L/km)

10 11 12 13

Transportation distances vary for different maintenance projects. Fuel economies of vehicles vary within the same project depending on the driving speed, load, and truck age etc. Using constant values for fuel economy of truck neglects the influence of external factors on the energy consumption of transportation.

14 15 16 17

To account for the ranges of model parameters, the uncertainty factor (UF) was defined to build statistical distribution as the input for Monte Carlo simulation. The uncertainty estimates were represented by UF so that 95% of the values of a stochastic variable (X) are within a factor UF from the mean M(X) of a normal or lognormal distribution, which shall satisfy Eq.3-4.

20 21

22

TE D

3

EP

19

M (X )   X lower = UF   X upper = UF × M ( X ) 

where X lower , X upper , M ( X ) represent the lower limit, upper limit and mean values.

AC C

18

M AN U

8 9

p( X lower < X < X upper ) = 95%

4

And the selection of distribution form can refer to the following stipulation.

 P(X<0) ≥ 5%  otherwise

log-normal distribution normal or log-normal distributions

(5)

23

2.3 Uncertainty Propagation

24 25 26 27

Uncertainty propagation consists in propagating input uncertainties to the result’s uncertainty. Defining representation for the input uncertainty is the first step. It is assumed the uncertainty comply with one unique distribution using either pedigree matrix approach or uncertainty factor approach. Once the statistical distribution of each input parameter is characterized, the Monte 5

ACCEPTED MANUSCRIPT Carlo simulation can be performed to propagate the uncertainty into the life cycle inventories (LCIs). For this purpose random values of all the uncertain inputs are sampled based on their corresponding distributions. The set of inputs is fed into each life cycle phase, e.g. transportation, and the corresponding environmental burden is computed which can be summed up to obtain the overall environmental burden. This process is repeated for N set of input samples which leads to N set of environmental burden. From these samples the probability distribution as well as all the statistics of this quantity, such as mean, standard deviation, and percentiles, can be estimated. Fig.1 is a schematic plot of uncertainty propagation process in the context of asphalt pavement maintenance.

RI PT

1 2 3 4 5 6 7 8 9

Paving Material Proudction

Uncertainty PDF

95th

75th

SC

PDF1 Monte Carlo Simulation

Transportation

Uncertainty PDF

PDFn

M AN U

PDF2

Uncertainty PDF

Median

5th

Output PDFs

Construction

Input PDFs

10

25th

Fig.1 Schematic plot of uncertainty propagation process

11

Environmental Database for Asphalt Pavement Maintenance

3.

13 14 15 16 17 18

Asphalt pavement maintenance measures are numerous, such as crack filling, micro-surfacing, thin overlay etc. The targets of this research mainly include HMA (base asphalt and modified asphalt) overlay, hot in-plant recycling (15% of RAP), cold in-place recycling and cold in-plant recycling of asphalt mixtures. The reasons for the selections are: 1) these maintenance measures are of high energy consumption and CO2 emission; 2) they are widely applied and the field projects are abundant in China.

19

3.1 Environmental Data Preparation

20 21 22 23 24

The phases incorporated for environmental burden estimation of asphalt pavement maintenance are material production, mixing of asphalt mixture, transportation and construction. The study did not include the usage phase, because it has many contributing factors, e.g. rolling resistance, albedo, and carbonation (Araújo et al. 2014), and for different maintenance activities the enhancements of pavement performances are not consistent.

25 26 27

For raw material production, through literature review, theoretical calculation and field investigation, the energy consumption and CO2 emissions for the production of unit of road materials were determined, as shown in Table 3.

AC C

EP

TE D

12

28 29 6

ACCEPTED MANUSCRIPT Table 3 Energy Consumption and CO2 Emission of Road Materials

1

Material

Energy consumption (MJ/tonne)

CO2 emission (kg/tonne)

Data source

Base asphalt

4985.85

287.51

China Energy Statistics Yearbook (2015)

9120.82

485.54

China Energy Statistics Yearbook (2015); Eurobitume (2011)1

(60% asphalt)

3469.96

205.98

China Energy Statistics Yearbook (2015); Eurobitume (2012) 1

Aggregate

108.59

9.02

Field investigation

Cement

3980.72

377.06

China Energy Statistics Yearbook (2015)

Water

2.51

0.24

General Principles for Calculation of the Comprehensive Energy Consumption (2008)

Steel

19727.98

1868.67

China Energy Statistics Yearbook (2015)

(3.5% SBS)

RI PT

Modified asphalt

M AN U

2 3 4 5

SC

Emulsified asphalt

Note: 1 China Energy Statistics Yearbook (2015) only provides environmental data of base asphalt, so the authors referred to the environmental data of the production and processing of SBS and emulsifiers additives in Eurobitume (2012). Then the energy consumption and CO2 emission of SBS modified asphalt and emulsified asphalt were calculated by their recipes.

Asphalt mixture production phase includes heating and mixing of aggregate, binder and RAP if any. For each process, the environmental data were obtained through field investigations.

8 9 10 11 12 13

For transportation phase, the transportation distance of specific projects have significant effects on the energy consumption and CO2 emissions for maintenance measures. Therefore, to eliminate the influence from transportation distance and to make the calculation results representative, standard distances were established. Through the statistics of a large number of project samples, this study obtained the average distance as the standard distances, as provided in table 4.

EP

Table 4 Statistical of Distance of Regulations of Standard distances (km)

AC C

14

TE D

6 7

Origination to destination RAP collection project to stockpile field Stockpile field to plant

Virgin materials transporting to plant

Plant to project field

Maintenance measures

Distance Range1 (km)

Standard Distances2 (km)

Cold in-place recycling

0

0

Others

5~55

30

All

0

0

Cold in-place recycling

0

0

Others

15~188

60

Cold in-place recycling

0

0

Others

7~52

30

7

ACCEPTED MANUSCRIPT Note: 1the distance range refers to the transportation of aggregate because it is overwhelmingly

1 2

dominating; 2standard distances were used in subsequent transportation phase calculation.

3 4

For the transportation and construction phases of asphalt maintenances, the fuel economy and CO2 emission of all the apparatuses were collected in field projects.

5

3.2 Energy Consumption and CO2 Emission Inventory Data of 18 asphalt pavement maintenance projects were collected in Jiangsu Province of China. A questionnaire was developed and distributed to project managers in order to obtain information of road material production, heating and recycling apparatus, paving and rolling of asphalt mixture and transportation of road materials, respectively. One sample questionnaire was given in supplementary file. According to table 3, energy consumption and CO2 emissions at the phase of material production can be calculated. Based on field data, energy consumption and CO2 emissions at the other phases can be quantified. The inventories of energy consumption and CO2 emissions of asphalt pavement maintenance projects are shown in table 5.

14

4. Uncertainty Analysis of Environmental Inventory

15 16 17 18 19

Table 5 shows that the energy consumption and CO2 emissions of different maintenance measures have great differences. The mean energy consumption of the 18 maintenance projects is 692.61MJ; the standard deviation (STD) is 286.67 MJ; and the coefficient of variation (COV) is up to 41%. And for projects adopting the same maintenance measure, there are still differences. While the maximum COV is less than 8% for energy consumption, the relative range (RR=

SC

M AN U

20

RI PT

6 7 8 9 10 11 12 13

max- min

) is up to 15.50%. For the CO2 emissions, the variations are similar. The reason for the Mean difference of the same maintenance measures is that factors including mixture gradation design, material sources, mixing and construction organization are all specific to individual project. In this respect the evaluation of the environmental burden of certain maintenance measure should not be based on the values obtained by a specific project.

25 26 27 28 29

A question was raised herein: is there significant difference between cold in-plant recycling and cold in-place recycling and between hot in-plant recycling and HMA-base asphalt overlay? To answer the question, it is necessary to build statistical distribution to determine the degree of uncertainty of environmental burden of t maintenance measures. The two sources of uncertainty, data quality one and model parameter one, in environmental burden calculations were quantified.

30

4.1 Determination of Data Quality Uncertainty

31 32 33

The energy consumption and CO2 of road material production in this study used the value in Table 3, which are proxies of field data. Therefore the quality of environmental data of road material was evaluated in terms of the criterions in Table 1 and the PDF was built using Eq.1.

34 35 36 37 38

Taking base asphalt as an example, the energy consumption of based asphalt production is based on the data of Chinese Statistical Yearbook 2015. Therefore, according to Table 1, the indicator scores are {3.0, 3.0, 5.0, 3.0, 4.0} and the DQI is 3.6 (rounded to 3.5). Referring to the Beta function of Table 2, the shape function is (2, 2), and the interval endpoint is (-25%, +25%). The PDF of energy consumption for the production of base asphalt is shown in Figure 2.

AC C

EP

TE D

21 22 23 24

8

ACCEPTED MANUSCRIPT

Table 5 Energy Consumption (MJ /tonne) and CO2 Emission (kg/tonne) of Various Maintenance Projects Production

CO2 emission2

Total energy consumption Transportation

Construction

RI PT

Material Maintenance measure Cement

Aggregate

Heating

Mixing

Sample

HMA-modified asphalt

589.50

0

86.05

348.02

7.32

65.56

35.12

1131.58

HMA-modified asphalt

546.76

0

86.35

316.70

9.07

65.86

10.24

1077.43

Mean=1043.24

90.43

Mean=87.36

HMA-modified asphalt

537.98

0

86.64

338.36

7.32

65.86

20.78

1056.65

COV=6.59%

87.69

COV=7.18%

HMA-modified asphalt

442.27

0

87.52

342.46

7.32

66.44

45.66

82.21

RR=17.70%

HMA-modified asphalt

442.27

0

87.52

330.75

9.07

66.44

23.12

HMA-base asphalt

237.38

0

87.52

321.68

5.27

66.44

37.17

HMA-base asphalt

247.04

0

87.22

240.60

7.32

66.44

0

87.81

322.56

5.27

66.44

HMA-base asphalt

223.33

0

87.81

322.56

5.27

66.44

HMA-base asphalt

213.67

0

87.81

322.85

5.27

66.74

HMA-base asphalt

199.33

0

88.10

323.73

5.27

66.74

175.91

0

65.56

301.19

14.34

1

151.03

0

71.13

317.29

5.27

1

Hot in-plant recycling

151.03

0

62.35

288.60

5.27

Cold in-plant recycling

141.96

80.79

35.12

2.93

5.27

Cold in-plant recycling

126.45

72.59

26.93

3.51

Cold in-place recycling

107.71

72.59

0

Cold in-place recycling

110.93

72.59

0

0.88

AC C

Hot in-plant recycling

RR=11.30%

Sample

80.51

755.75

62.43

665.60

26.64

732.04

23.42

728.53

23.42

720.04

23.42

706.87

55.87 Mean=718.13

Mean=60.60 64.44

COV=4.22%

COV=4.81% 61.16

RR=12.55%

Mean=692.61 STD=286.67

26.34

643.94

Mean=626.57

62.64

37.17

644.82

60.59

23.42

52.98

COV=41.39%

RR=14.14% 60.42

56.05

Mean=53.26

COV=4.92%

54.12

COV=6.21%

590.96

RR=8.59%

49.61

RR=12.09%

18.44

337.78

Mean=313.48

27.35

Mean=26.81

7.61

289.19

0

0

39.81

225.09

4.98

0

0

21.37

210.16

COV=41.02%

COV=2.01% 26.27

RR=15.50%

4.98

Mean=58.35 STD=23.94

59.32

COV=7.75%

51.22

Statistical parameters

95.98

959.18

60.88

EP

Hot in-plant recycling1

991.37

M AN U

223.33

TE D

HMA-base asphalt

17.27

Statistical parameters

SC

Asphalt

Mean=217.62

RR=4.03% 18.89

COV=3.43%

Mean=18.26 COV=3.42%

17.64 RR=6.86%

RR=6.85%

Note: 1the RAP content is 15%; the RAP is considered to be environmental burden free except for the recycling phase; 2 to save space, only the total CO2 emissions are given.

9

ACCEPTED MANUSCRIPT 3.5 DQI=5 DQI=3.5 DQI=1

3

2

RI PT

Probability

2.5

1.5

1

0 1.5

2

SC

0.5

2.5 3 3.5 4 4.5 5 5.5 Energy consumption for base asphalt production (MJ/kg)

Fig.2 PDF of Energy Consumptions of Base Asphalt Production under Various DQIs

3 4 5 6 7 8

Fig.2 presents the PDFs of various DQI values. Different DQI values represent different data quality, and the higher the DQI, the higher the data quality and the lower the corresponding uncertainty. According to the PDF established in Fig.2, the energy consumption value and its statistical information of base asphalt material can be calculated. Similarly, the PDFs can be established for CO2 emissions for base asphalt and other road materials, which provide data sources for Monte Carlo analysis.

9

4.2 Determination of Model Parameter Uncertainty

TE D

EP

Based on the survey data of Table 5, the PDF and the corresponding UF of each process of pavement maintenance can be determined. Because the environmental value of mixing process is small, and for the same maintenance measure, large differences occur. In order to avoid high negative probability, lognormal distribution was used to depict the mixing process. The other processes were assumed to obey normal distribution. According to the definition of Eq.3 and 4, the corresponding UF values were calculated, as shown in Table 6.

AC C

10 11 12 13 14 15

M AN U

1 2

Table 6 UF Parameters of Various Maintenance Measures

16

Measures HMA-modified asphalt

HMA-base asphalt

Hot in-plant recycling Cold in-plant recycling Cold in-place recycling

Asphalt

Cement

Aggregate

Heating

Mixing

Transportation

Construction

Sample

UF

1.297

NA

1.015

1.074

1.273

1.011

3.06

5

UF

1.161

NA

1.007

1.241

1.345

1.004

1.628

6

UF

1.004

NA

1.203

1.141

1.116

1.047

1.895

3

UF

1.175

1.160

1.160

1.469

1.291

11.05

1.114

2

UF

1.041

1.131

NA

1.291

NA

NA

2.364

2

10

ACCEPTED MANUSCRIPT 1

4.3 Quantification of Uncertainty of Maintenance Measure

2 3 4

The determined PDFs of data quality and model parameter uncertainties were used as input for Monte Carlo simulation of different maintenance measures. As an example, the simulation results of energy consumption of HMA-modified asphalt are shown in figure 3. 1

800

Mean=1041.64

0.9

STD=207.23

0.8

Cumulative density function

RI PT

900

700 0.7

400 300

0.5 0.4 0.3

200

0.2

100

0.1

600

900 1200 1500 Energy consumption (MJ)

1800

3a. PDF of HMA-modified asphalt

0 300

M AN U

0 300

SC

Probability

0.6 500

2100

600

900 1200 1500 Energy consumption (MJ)

1800

6 7 8 9 10 11 12

Figure 3a depicts the energy consumption and its frequency under the 50,000 runs for HMA-modified asphalt. Given the data quality and model parameter uncertainties, the average energy consumption is 1041.64 MJ/tonne, and the standard deviation (STD) is 207.23 MJ/tonne. Fig.3b is the cumulative density function [CDF] of energy consumption, which can be used to determine the energy consumption intensity of the maintenance measure. According to the CDF diagram, the 2.5th and 97.5th percentiles of HMA-modified asphalt are 635.42 MJ/tonne and 1447.82 MJ/tonne respectively.

13 14

For the other maintenance measures, the Monte Carlo analysis was also performed to obtain the relevant statistical parameters, as shown in table 7.

15

Table 7 Energy Consumption and CO2 Emission of Various Maintenance Measures

EP

TE D

Fig.3 Probability Distribution of HMA-modified asphalt

Energy consumption (MJ/tonne)

Measure

Mean

HMA-modified asphalt HMA-base asphalt Hot in-plant recycling Cold in-plant recycling

2100

3b. Cumulative density function of HMA-modified asphalt

5

AC C

Frequency

600

2.5th

97.5th

percentile

percentile

STD

CO2 emission (kg/tonne)

Mean

2.5th

97.5th

percentile

percentile

STD

1041.64

207.23

635.42

1447.82

2545.32

507.54

1550.54

3540.10

717.41

93.08

534.97

899.84

1764.10

228.01

1317.20

2211.01

618.48

61.76

497.43

739.52

1541.36

158.64

1230.42

1852.30

316.41

82.25

155.20

477.62

780.05

203.13

381.90

1178.19

11

ACCEPTED MANUSCRIPT Cold in-place

208.11

recycling

66.74

77.31

338.91

536.81

172.40

198.91

874.72

1 4.4 Comparisons of Environmental burden of Maintenance Measures

3 4 5 6

Table 7 provides statistical parameters to describe the variations of environmental burden of maintenance measures. However whether there exists significant statistical difference between different maintenance measures need further investigation. The environmental burden comparison parameter R is defined in Eq.6 to test the significance.

Ru =

Eu , I

(6)

Eu , II

SC

7

RI PT

2

where Ru is the comparison parameter for environmental burden category u; Eu ,1 and Eu ,2 are

9

the environmental burden relating to maintenance measures 1 and 2, respectively.

M AN U

8

10

Then P(R < 1) represents the probability of the environmental burden of maintenance

11 12 13

measure 1 is less than that of maintenance measure 2. The HMA-base asphalt and HMA-modified asphalt measures were compared in terms of the energy consumption under 50,000 runs of Monte Carlo simulation, with the results shown in Fig.4.

1200

1

HMA –base asphalt vs. HMA-modified asphalt

0.8 0.7

EP

600

400

200

0.5

0.6

0.7 0.8 Energy consumption ratio R

0.9

Probability

P(R<1)=99.9%

AC C

Frequenc

800

0 0.4

0.9

TE D

1000

0.6

P(R<0.854)=0.95

0.5 0.4 0.3 0.2 0.1

1

1.1

4a. PDF of R

0 0.4

0.5

0.6

0.7 0.8 0.9 Energy consumption ratio R

4b. CDF of R

14

Fig.4 Energy Consumption Comparison of HMA-base Asphalt vs. HMA-modified Asphalt

15 16 17 18 19

In Fig. 4a, as expected, the energy consumption of HMA-modified asphalt is significantly higher than that of the HMA-base asphalt at the 99.9% confidence level. The 95% confidence intervals of the comparison parameter R of the two maintenance measure can be extracted from Fig.4b. For the comparison of other measures, the relevant statistical parameters were also extracted and summarized in table 8.

20 12

1

1.1

ACCEPTED MANUSCRIPT 1 2

Table 8 Energy Consumption Ratio of Various Maintenance Measure 95% confidence Benchmark measure

Compared measure

P(R<1)

P(R>1)

Mean

STD interval

HMA-modified HMA-base asphalt

99.9%

NA

0.70

NA

99.7%

1.15

NA

99.9%

2.27

NA

99.9%

0.08

(0.56, 0.87)

Hot in-plant HMA-base asphalt recycling Cold in-plant HMA-base asphalt

recycling

recycling

recycling

Cold in-plant

Hot in-plant

recycling

recycling

0.18

(1.92, 2.63)

0.31

(2.82,4.04)

NA

99.9%

1.52

0.15

(1.24,1.81)

99.9%

NA

0.51

0.06

(0.41,0.65)

TE D

Cold in-place

(1.00, 1.32)

Table 8 provides statistical parameters for comparisons of different maintenance measures, which can assess the variation and reliability of comparison results at the statistical view. For example, the mean energy consumption of cold in-plant recycling is higher than that of the cold in-place recycling by 52%. And the energy consumption ratio R fluctuation range at the 95% confidence level is (1.24, 1.81), that is, the former measure consumes energy consumption that is 124%-181% of the latter measure at the 95% confidence level. The analysis of CO2 emissions is similar and will not be discussed.

EP

3 4 5 6 7 8 9

Cold in-plant

3.46

M AN U

Cold in-place HMA-base asphalt

0.08

SC

recycling

RI PT

asphalt

5. Summary

11 12 13 14 15 16 17 18

Environmental burdens of asphalt pavement maintenance measures are conventionally calculated using data that is of inevitable uncertainty. To capture the uncertainty, a methodology for evaluating the reliability of environmental burden of asphalt pavement maintenance was constructed in this study. The uncertainty originating from data quality variation and model parameter variation were quantified. The data quality uncertainty was transformed into the probability density function by the data quality pedigree matrix. The model parameter uncertainty was characterized by the uncertainty factor. Environmental data of 18 asphalt pavement maintenance projects were collected as database.

19 20 21

Through Monte Carlo simulation, the probability density distribution of energy consumption and CO2 emission of each maintenance measure were obtained. The statistical parameters (mean, standard deviation, percentiles, etc.) of energy consumption and CO2 emission of different

AC C

10

13

ACCEPTED MANUSCRIPT maintenance measures were calculated. At the 95% confidence level, the probability density distribution of environmental burden comparison parameter R shows that there are significant differences in the environmental burden of different maintenance measures, and the corresponding confidence intervals were calculated.

5 6 7 8

For the same maintenance measure, the environmental burden calculation result based on the project level data fluctuates considerably. The methodology proposed in this study can evaluate the environmental burden and its reliability of the asphalt pavement maintenance measures at the statistical level.

9 10 11

Acknowledgement: the authors acknowledge the funding supports of the National Natural Science Foundation of Jiangsu Province (BK20171359) and the National Natural Science Foundation of China (51408114).

RI PT

1 2 3 4

SC

12

AC C

EP

TE D

M AN U

13

14

ACCEPTED MANUSCRIPT References

2 3 4

1. Araújo, J.P.C., Oliveira, J.R.M., Silva, H.M.R.D. The importance of the use phase on the LCA of environmentally friendly solutions for asphalt road pavements. Transportation Research Part D 32 (2014) 97–110.

5 6 7

2. Batoli, M., Bienvenu, M., and Mostafavi, A. Putting sustainability theory into roadway design practice: Implementation of LCA and LCCA analysis for pavement type selection in real world decision making. Transportation Research Part D 52 (2017): 289-302.

8 9

3. Canter, K., et al. Screening stochastic life cycle assessment inventory models. The International Journal of Life Cycle Assessment, 2002, 7(1), 18–26.

10 11 12

4. Cheng, L., Yan, G.J., Chen, D.Z, et al. Quantitative Investigation of Energy Conservation and Emission Reduction for Warm Mix Asphalt. Chinese Journal of Environmental Engineering, 4(9): 2151-2155, 2010. (in Chinese)

13 14 15 16 17 18 19 20 21 22 23 24 25 26 27

5. Chen, F., Zhu, H.R., Yu, B., and Wang, H.P. Environmental burden of regular and long-term pavement design: a life cycle view, International Journal of Pavement Engineering, 2016, 17(4): 300-313. 6. China Energy Statistics Yearbook. Department of Energy Statistics, National Bureau of Statistics, China, 2015. 7. Eurobitume. Life Cycle Inventory: Bitumen. (2nd edition) Brussels: The European Bitumen Association, 2012. 8. General Principles for Calculation of the Comprehensive Energy Consumption. China standard GB/T2589-2008, 2008. 9. Giani, M.I.m Dotelli, G., Brandini, N. and Zampori, L. Comparative life cycle assessment of asphalt pavements using reclaimed asphalt, warm mix technology and cold in-place recycling. Resources, Conservation and Recycling 104 (2015) 224–238. 10. Huang, Y., Bird, R., and Heidrich, O. Development of A Life Cycle Assessment Tool for Construction and Maintenance of Asphalt Pavements. Journal of Cleaner Production, 17(2):283-296, 2009.

28 29

11. Jullien, A., Dauvergne, M., and Cerezo, V. Environmental assessment of road construction and maintenance policies using LCA, Transportation Research Part D 29 (2014) 56–65.

30 31 32 33 34 35 36 37 38 39 40 41

12. Jamshidi, A., Hamzah, M.O., and Shahadan, Z. Selection of reclaimed asphalt pavement sources and contents for asphalt mix production based on asphalt binder rheological properties, fuel requirements and greenhouse gas emissions. Journal of Cleaner Production 23 (2012) 20-27. 13. Kennedy, D.J., Montgomery, D.C., Quay, B.H. Stochastic environmental Life Cycle Assessment modeling: a probabilistic approach to incorporating variable input data quality. Int. J. LCA , 1996, 1 (4), 199-207. 14. May, J.R. and Brennan, D.J. Application of Data Quality Assessment Methods to an LCA of Electricity Generation. Int. J. Life Cycle Assess., 2003(8):215-225. 15. Noshadravan, A., Wildnauer, M., Gregory, J., & Kirchain, R. Comparative pavement life cycle assessment with parameter uncertainty. Transportation Research Part D: Transport and Environment, 2013(25):131-138.

AC C

EP

TE D

M AN U

SC

RI PT

1

15

ACCEPTED MANUSCRIPT 16. Santero, N.J., Masanet, E., and Horvath, A., 2011. Life-cycle assessment of pavements. Part I: critical review. Resources, Conservation and Recycling, 55(2011):801–809.

3 4

17. Tan, R.R., Culaba, A.B., Purvis, M.R.I. Application of possibility theory in the life-cycle inventory assessment of biofuels. Int. J. Energy Res., 2002(26)737-745.

5 6 7

18. Wang, E.D. and Shen, Z.G. A hybrid Data Quality Indicator and statistical method for improving uncertainty analysis in LCA of complex system application to the whole-building embodied energy analysis. Journal of Cleaner Production, 43(2013):166–173.

8 9 10 11 12 13 14 15

19. Wang, H., Thakkar, C., Chen, X., and Murrel, S. Life-cycle assessment of airport pavement design alternatives for energy and environmental impacts. Journal of Cleaner Production, 2016(133): 163-171. 20. Wen, H.F., Lu, J. and VanReken, T.M. Modeling the Effects on Energy and Carbon Dioxide from the Use of Recycled Asphalt Pavement in Hot Mix Asphalt. International Journal of Sustainable Transportation, 2014(8):249–261. 21. Weidema, B.P., Wesnæs, M., 1996. Data quality management for life cycle inventories-an example for using data quality indicators. J. Cleaner Prod. 4 (3-4), 167-174.

16 17 18

22. Yu, B., Lu, Q., and Xu, J. An improved pavement maintenance optimization methodology: integrating LCA and LCCA. Transportation Research Part A: Policy and Practice, 2013 (55):1–11.

19 20 21

23. Yu, B., Liu, Q., and Gu, X.Y. Data quality and uncertainty assessment methodology for pavement LCA. International Journal of Pavement Engineering, 2016, DOI:10.1080/10298436.2016.1176166

22 23

24. Zhang, H., Keoleian, A.G., Lepech, D. M., and Kendall, A. Life-Cycle Optimization of Pavement Overlay Systems. Journal of Infrastructure Systems, 2010, 6(4): 310-322.

24 25

25. Zhang, H. Simulation-Based Estimation of Fuel Consumption and Emissions of Asphalt Paving Operations. Journal of Computing in Civil Engineering, 2015, 29(2): 04014039.

AC C

EP

TE D

M AN U

SC

RI PT

1 2

16