The Impact of Pavement Structural Response on Vehicle Fuel ...

4 downloads 0 Views 1MB Size Report
include the effects of pavements on vehicle fuel use (Van Dam et al. 2015). Models for. 89 ...... ,M.A., and Janajreh, I. (2006). “Flexible pavement responses. 522.
1 2 3 4 5

The Impact of Pavement Structural Response on Vehicle Fuel Consumption

Erdem Coleri1; and John T. Harvey, Affiliate M. ASCE2

6 7 8

ABSTRACT

9

In this study, vehicle energy consumption induced by the viscous behavior of asphalt

10

pavement materials was calculated by using 3D viscoelastic finite element (FE) modeling.

11

Five pavement structures with conventional, rubber, and polymer-modified asphalt overlay

12

mixes with old asphalt, cracked concrete, and cement treated base (CTB) underlying layers

13

were modeled to evaluate the impact of structure type on calculated excess fuel consumption

14

(EFC). A fully loaded 18 wheel tractor-trailer was simulated. EFC was defined as the fuel

15

consumption beyond what occurs for an ideal pavement with no energy consumed due to

16

structural response. A total of 234 FE models for a full factorial design were developed to

17

simulate the impacts of three speed levels (8, 52, 105 kph), two subgrade stiffnesses (50 and

18

200 MPa), and nine temperatures (0oC to 80oC at 10oC intervals) on calculated EFC. Results

19

show that the greatest EFC is primarily observed at pavement temperatures higher than 40oC.

20

For an average truck highway speed of 97 kph, average pavement temperatures for three

21

cities in California (Daggett, Sacramento, and San Francisco) result in excess fuel

22

consumptions ranging from 0.1% to 0.35% for the structures analyzed in this study. Although

23

this level of EFC appears to be much less than typical roughness related EFC, models for

24

EFC due to structural response can aid in decision-making processes, particularly where 1

Assistant Professor, School of Civil and Construction Engineering, Oregon State University, 101 Kearney Hall, Corvallis, OR, USA, 97331, Tel: (541) 737-0944, Email: [email protected]. 2

Professor, University of California Pavement Research Center, Department of Civil and Environmental Engineering, University of California, Davis, One Shields Avenue, Davis, CA, 95616, Tel: (530) 754-6409, [email protected].

25

models show it to be greatest. Polymer-modified overlays were observed to have the highest

26

EFC when compared with other structures. Having a concrete layer or a CTB layer under the

27

asphalt overlay rather than having an aged asphalt layer did not create any significant

28

reduction in EFC. Subgrade stiffness was also not observed to be a significant factor

29

affecting EFC.

30 31

Key words: excess fuel consumption; viscoelasticity; energy dissipation; pavement; asphalt;

32

finite element modeling.

33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55

56

INTRODUCTION

57

Pavements can influence the fuel efficiency and greenhouse gas (GHG) emissions of vehicles

58

through three mechanisms, which together can be called the pavement-related rolling

59

resistance (Coleri et al. 2015). These three mechanisms are: i) roughness: consumption of

60

energy through the working of shock absorbers, drive train components, and deformation of

61

tire sidewalls related to pavement roughness; ii) texture: consumption of energy through

62

viscoelastic working of the tire rubber in the tire/pavement contact patch as it passes over

63

positive macrotexture of the pavement surface; and iii) asymmetrical and viscoelastic

64

deformation: Consumption of energy in the pavement itself through asymmetrical (due to

65

any dissipation mechanisms within pavement such as damage and dissipation between

66

layers) and viscoelastic deformation of pavement materials.

67 68

A number of models have been developed for energy dissipation by the deflection

69

mechanism, including models developed by the Massachusetts Institute of Technology

70

(Louhghalam et al. 2013; Louhghalam et al. 2014), the University of Lyons National School

71

of Public Works of the State (Pouget et al. 2011), the Swedish National Road and Transport

72

Research Institute (VTI) (Jonsson and Hultqvist 2009) and Michigan State University. There

73

is active research in the development of these models, and several general approaches are

74

being used by model developers. The MIT model has been used to characterize the Virginia

75

DOT pavement network albeit with very simplified structure, climate and load information

76

(Akbarian et al. 2014). Kim et al. (2016) also developed a new method for estimating the

77

energy dissipation in the tire and the suspension of vehicles traversing rough non-deformable

78

pavements. Shakiba et al. (2016) also developed a mechanics based model for predicting

79

structure-induced rolling resistance. The effects of tire contact stresses, viscoelastic

80

properties, speed, and temperature were evaluated. None of these models has been

81

empirically calibrated yet. Empirical calibration in the field is difficult because most of the

82

response occurs at hotter temperatures limiting the time available, and measurements can

83

only be done on nearly windless days.

84

consistent pavement structures where traffic flow allows for relatively slow constant

85

operation are also needed. A model calibration study is currently collecting calibration data

86

in California.

Long, straight, flat sections of roadway with

87 88

Models are needed for the use stage of life cycle assessment (LCA) studies for pavement that

89

include the effects of pavements on vehicle fuel use (Van Dam et al. 2015). Models for

90

calculating the GHG emissions from pavement rehabilitation and maintenance, materials

91

production and construction, and the resulting changes in pavement roughness and

92

macrotexture, have been developed by the University of California Pavement Research

93

Center (UCPRC) and implemented in LCA procedures that are being used to analyze

94

Caltrans practices for potential use in network and project-level decision-making (Wang et

95

al. 2012; Wang et al. 2014). Models considering roughness and texture developed by Chatti

96

and Zaabar (2012) adapted by Wang et al. (2012 and 2014) have also been used by others for

97

similar purposes such as Dass and Mukherjee (2011) and Ozer et al. (2017).

98 99

Models for the roughness mechanism (1st mechanism stated above) are well established and

100

have been empirically validated and calibrated. Chatti and Zaabar (2012) concluded that a 1

101

m/km (63.4 in./mi) decrease in pavement roughness results in a 3% decrease in the fuel

102

consumption for passenger cars at highway speeds, which translates into an annual fuel

103

savings of $24 billion in the U.S. For heavy trucks, this decrease is about 1% at a normal

104

highway speed of 96 km/h and about 2% at low speed at a speed of 56 km/h. This paper

105

focuses on the quantification of energy consumption due to viscoelastic deformation of

106

pavement materials (3rd mechanism stated above) using finite element (FE) modeling. The

107

significance of viscoelastic deformation related fuel consumption when compared to

108

roughness related consumption is investigated. The impacts of temperature, vehicle speed,

109

pavement structure, and material type on viscoelastic deformation related fuel consumption

110

were investigated.

111 112

OBJECTIVES AND SCOPE

113

The major goals of this study are as follows:

114



Compare the pavement structural response related EFC for different pavement

115

structures with conventional dense graded, rubberized and polymer modified

116

overlays.

117



118 119

graded asphalt concrete (DGAC), and cement treated base) on EFC. 

120 121



Develop a method to simplify the thermal gradient simulation in asphalt pavements to facilitate the use of thermal gradients in network level applications.



124 125

Evaluate the effects of temperature and speed on EFC for different pavement material types.

122 123

Determine the impact of underlying bounded materials (cracked concrete, old dense

Determine the accuracy of calculated total vehicle EFC determined by calculating the EFC from one tire and multiplying it by the number of tires on a vehicle.



Using the climate data for three cities in California, determine the approximate

126

contribution of structural response related fuel consumption to total fuel consumption

127

of a fully loaded 18 wheel semi-tractor trailer.

128 129

130

GENERAL PROCEDURE FOR MODEL DEVELOPMENT

131

Materials and Testing

132

Four asphalt mixes were tested in this study to develop viscoelastic FE models: i) New

133

DGAC: a dense graded mixture with PG 64-16 binder with a nominal maximum aggregate

134

size (NMAS) of 19mm, 5.2% binder content, 4% air-void content, 15.5% voids in mineral

135

aggregate (VMA), and 74.9% voids filled with asphalt (VFA); ii) Rubberized Hot-Mix

136

Asphalt (RHMA): a gap-graded rubberized asphalt mix with PG 64-16 binder with an

137

NMAS of 12.5mm, 7.5% binder content, 4% air-void content, 18.7% VMA, and 78.8%

138

VFA; iii) Polymer-Modified (PM): a dense graded polymer modified mixture with PG 64-

139

28 binder with an NMAS of 25mm, 6.0% binder content, 4.3% air-void content, 17.6%

140

VMA, and 74.9 VFA; iv) Aged DGAC: a highly field aged dense graded mixture with PG

141

64-16 binder with an NMAS of 19mm. After mixing, all mixtures [except the Aged DGAC

142

(field mix)] were subjected to short-term oven-aging at 135°C for four hours (AI 2001).

143

Dynamic modulus (compressive stiffness) of these four asphalt mixes were determined by

144

conducting experiments with two replicates of each mix at four temperatures (4, 21, 38,

145

55oC) for a frequency sequence of 25, 10, 5, 1, 0.5, 0.1 Hz. Master curves developed for a

146

reference temperature of 21 oC are shown in Figure 1. It can be observed that Aged DGAC

147

mix has a significantly higher dynamic modulus than all other mixes. PM mix has the lowest

148

dynamic modulus values for all loading frequencies.

149 150

Model Details

151

A viscoelastic finite element (FE) model was developed to calculate the dissipated energy

152

under different conditions. Energy dissipation due to damping in the subgrade and all other

153

layers other than those that are asphalt-bound was not simulated in the model. In the

154

developed viscoelastic FE model, only the linear behavior was considered (small strain

155

domain). The possible impacts of pavement distresses (fatigue, permanent deformations, and

156

cracks) were not taken into account. The material constituting the base is considered isotropic

157

linear elastic.

158 159

The temperature dependency of the asphalt mix was defined by using the Williams-Landel-

160

Ferry (WLF) equation, given as follows (Ferry 1980):

161 162

log( aT ) 

163

where

164 165

 C1 T  Tref 

C 2  T  Tref 

(1)

aT = the time-temperature shift factor, C1 and C2 = regression coefficients,

166

Tref = the reference temperature, and

167

T = test temperature.

168 169

To optimize the regression coefficients C1 and C2, the shear modulus data were first fitted to

170

a sigmoid function, in the form of:

171

log G    

172

where

173



174



1  exp    log 

(2)

= regression coefficients, and reduced time.

175 176

Shift factors were calculated by fitting the measured modulus to the sigmoidal function (Eqn.

177

2). One shift factor was calculated at each test temperature while the shift factor for the

178

reference temperature (21oC) is set at zero. A MatlabTM code was developed to optimize the

179

regression coefficients and shift factors for all temperatures. Regression coefficients

180

C1 and C2 were calculated by simply fitting the WLF equation to the calculated shift factors.

181

C1 for New DGAC, PM, and RHMA mixes were calculated to be 36.7, 17.2, and 16.6,

182

respectively. C2 for New DGAC, PM, and RHMA mixes were calculated to be 300.8, 149.1,

183

and 131.9, respectively.

184 185

The generalized Maxwell-type viscoelastic model (Prony series) was used in this study to

186

simulate the time dependency. The model consists of two basic units, a linear elastic spring

187

and a linear viscous dash-pot. Various combinations of these spring and dashpot units define

188

the type of viscoelastic behavior. The implementation procedure developed by Pouget et al.

189

(2011) was used to simulate the effects of truck loads, vehicle speed, and temperature on

190

dissipated energy.

191 192

AbaqusTM software was used for model development.

193

represented by a 3.5 m long and 2 m wide slab. The finite-element mesh consists of Lagrange

194

brick elements with a second-order interpolation function. The mesh was refined under the

195

wheel path. International Roughness Index (IRI) and macrotexture are not simulated in the

196

model. Therefore, the effects of vehicle suspension dynamics on the wheel load were not

197

simulated.

The pavement structure was

198 199

The bottom side of the model is clamped. The symmetry condition in the transversal

200

direction imposes a boundary condition on one side. To ensure the continuity of this slab

201

with the rest of pavement, only vertical displacement was allowed for other lateral sides

202

(Pouget et al. 2011). Perfect bonding was assumed between different pavement layers.

203 204

In order to simulate moving wheel loading in the viscoelastic FE model, the trapezoidal

205

impulsive loading method (quasi-static) was used (Yoo et al. 2006). Only one vehicle was

206

simulated in this study, a fully loaded (maximum gross vehicle weight in many states), 18

207

wheel class 9 truck. The tire is assumed to have a square contact area (175 by 175 mm) and

208

the distribution of load on the tire is assumed to be constant. Tire pressure of the truck was

209

assumed to be 670 kPa. For a class 9 truck with 18 tires (one single steering axle and two

210

dual-tandem axles), the simulated total loaded truck weight adds up to be 37.7 tons.

211 212

The dissipated energy per time w(t) was integrated on a d long slice of the asphalt layer,

213

located in the center of the 3.5 m long structure. The dissipated energy for a truck (Wtruck)

214

with Z dual wheels, covering a distance of X can be calculated using the following equation

215

(Pouget et al. 2011):

216 217

Wtruck 

 wt   dt  xd  Z

(3)

218 219

Where w(t) is calculated using the following equation in which E is the phase angle, 0z is

220

the stress curve for the loading time while 0z is the strain for the same loading time for a unit

221

volume dV. Unit volume is the volume of each Lagrange brick element in the center of the

222

3.5 m long structure with a d long slice (Pouget et al. 2011):

223 224



 E   0z   0z  dV

w(t)       sin 

(4)

225 226

Sensitivity Analysis

227

Sensitivity analyses were conducted to evaluate the effects of several factors on calculated

228

deflection basin and excess fuel consumption (EFC). The model developed for the project

229

was revised and improved according to the findings of the sensitivity analysis. The factors

230

considered for the sensitivity analyses were i) Temperature gradient and ii) Number of tires.

231 232

Temperature Gradient Effect

233

In this section, the impact of using an average temperature for the complete asphalt layer

234

rather than simulating the temperature gradients is investigated. Using an average

235

temperature for the complete asphalt layer creates a significant reduction in model run times.

236

Asphalt layer temperature distributions for a summer day were simulated using a random air

237

and surface temperature for Sacramento and the software Enhanced Integrated Climatic

238

Model (EICM) (Zapata and Houston 2008). The distribution of temperatures along the

239

asphalt depth at different times of the day are given in Figure 2. The asphalt layer in the FE

240

model was divided into 10 layers (Figure 3a) and the temperatures given in Figure 2 were

241

assigned to the layers.

242 243

EFC outputs were determined for all six hours of the day used for the analysis. These results

244

were assumed to be the ground truth outputs simulating the effect of temperature gradients on

245

EFC. In addition, four finite element models were developed by using only one temperature

246

for the complete asphalt layer. Temperatures at the surface (0 mm depth), at 75mm depth

247

(one third depth), at 125mm depth, and at 175mm depth were used for the analysis (Figure

248

2). Figure 3(b) shows the distribution of EFC at different times of the selected day for five

249

different cases with: i) actual temperature gradient simulated (ground truth); ii) surface

250

temperatures; iii) temperature at one third depth (75 mm); iv) temperature at 125mm depth;

251

and v) temperature at 175mm depth.

252 253

It can be observed from Figure 3(b) that simulations conducted with the temperature at one

254

third depth (75mm) provides EFC values that are almost identical to the case with the actual

255

temperature gradient. To check the validity of this conclusion for sections with thinner

256

asphalt layers, a model with the same material properties and with a thinner (175mm thick)

257

asphalt layer was developed. A temperature profile for an air temperature of 26.6oC and a

258

surface temperature of 31.8 oC at 10am was developed using the EICM software. The model

259

had seven sublayers for the asphalt layer to be able to input the thermal gradient.

260 261

Results showed that the excess fuel consumption for the model with seven sublayers in the

262

asphalt layer was almost identical to the section with one asphalt layer using the one-third

263

depth temperature (EFC7sublayers=1.44ml/km; EFCone-third

264

suggested that there is no need to simulate temperature gradients in the simulations for the

265

sections considered in this study. These results suggest that for the calculation of annual fuel

266

use (network level application), pavement temperatures at one third depth can be used to

267

simulate the effect of temperature gradients throughout the pavement depth. This assumption

268

can create a significant reduction in model development and run times. Model runtime is

269

longer since additional calculations are required at the boundaries between additional layers.

temp=1.43ml/km).

These results

270 271

The validity of this assumption for predicting vertical strain ( was also investigated.

272

Calculated vertical strain for a surface volume element in the middle of the model structure

273

for the section with seven sublayers was determined to be almost identical to the section with

274

one asphalt layer using the one-third depth temperature (7sublayers=152.8; one-third

275

temp=147.4).

276 277

Number of Tires

278

The effect of number of tires used for the finite element modeling was investigated. The

279

objective was to determine minimum number of tires that will provide total dissipated energy

280

that is close to the dissipated energy calculated from a four-tire model. The objective was to

281

reduce model runtime by reducing the number of tires. Figure 4 shows the displacement field

282

for the finite element models with: i) one tire; ii) two tires; and iii) four tires (one side of

283

dual-tandem axle). Figure 5(a) shows the change in calculated dissipated energy with

284

temperature for models with one, two, and four tires. The increase in dissipated energy with

285

increasing number of tires can be observed. However, when the dissipated energy values in

286

Figure 5(a) are divided by the number of tires, all three curves overlap (Figure 5b). This

287

result suggests that total vehicle dissipated energy and EFC can be accurately calculated by

288

modeling one tire and multiplying the output by the number of tires on the vehicle.

289 290

MODEL DEVELOPMENT

291

Several FE models were developed to evaluate the impacts of temperature, vehicle speed,

292

subgrade stiffness, mix types, and structural properties on pavement structural response

293

related EFC. Figure 6 shows the structures for which models were developed. For structures

294

1 through 4, models for a full factorial design were developed to simulate the impacts of

295

three speed levels (8, 52, 105 kph), two subgrade stiffnesses (50 and 200 MPa), and nine

296

temperatures (0oC to 80oC at 10oC intervals). Structure 5 was used to evaluate the impact of

297

having a cement treated base (CTB) layer under the asphalt layer rather than having an old

298

DGAC layer. For this reason, in order to reduce the number of model simulations, models for

299

structure 5 were only developed for a speed level of 52 kph. Table 1 presents the factorial

300

developed for model simulations. In this study, a total of 234 models were developed (Table

301

1).

302

303

Structures 1 through 5 were selected to simulate typical structures in California. Structures 1

304

through 3 were used to simulate the impact of rubber-modified and polymer-modified asphalt

305

mix use on EFC. Structure 4 was used to evaluate the impact of using an asphalt overlay to

306

rehabilitate cracked concrete structures in California. Cracked concrete and CTB layers were

307

simulated as elastic layers with stiffnesses 25,000 and 4,300 MPa, respectively. Aggregate

308

base (AB) and subgrade (SG) layers were also assumed to be elastic layers. All asphalt layers

309

were assumed to be linear viscoelastic materials. Boundary conditions and truck loads used

310

for modeling were discussed in the “Model details” section.

311 312

RESULTS

313

Stress, strain, and phase angle were integrated over time to calculate dissipated energy in the

314

pavement (Eqn.3 and Eqn.4). Calculated dissipated energy values were converted to energy

315

required to make the vehicle move. Finally, EFC for all temperatures, structures, subgrade

316

stiffnesses, and speeds were calculated by dividing the calculated dissipated energy (MJ/km)

317

by the calorific value of the fuel (MJ/L) (Eqn. 5). The calorific value of diesel for trucks is

318

assumed to be 40 MJ/L (Coleri et al. 2015). Since the engine efficiency of trucks are

319

generally around 40 percent, 16MJ/L (40x0.40) is used as the final calorific value of diesel

320

for trucks (Chatti and Zaabar 2012; Coleri et al. 2015). By using a 40 L/100km (400 ml/km)

321

fuel consumption rate for a 40-ton truck (Sharpe and Muncrief 2015), percent EFC is

322

calculated using Eqn. 6.

323 324

325

excess fuel consumption (

ml ) km

Wtruck  MJ  Fuel calorific    L 

 1000

(5)

excess fuel consumption (ml / km)  100 400 ml / km

326

excess fuel consumption (%) 

327

Results are shown in Figures 7 through 11 for all structures. In general, much higher EFC can

328

be observed at pavement temperatures higher than 40oC. For some cases, EFC starts to

329

decrease at very high temperatures (Figure 7 and 10). This reduction is a result of change in

330

asphalt behavior from viscoelastic to elastic at very high temperatures (Pouget et al. 2011).

331

Although 40oC and higher pavement temperatures are commonly observed on roadways on

332

summer days, pavement temperature distributions for different climate regions need to be

333

combined with the EFC curves obtained in this study to determine the importance of

334

structural response on fuel consumption. The annual pavement temperature distributions at

335

one-third depth for Daggett, Sacramento, and San Francisco in California were calculated by

336

using the weather station data and the Enhanced Integrated Climate Model (EICM).

337

Distributions for a year are shown in Figure 12. These distributions were used to find

338

temperature averages for a year and simulations for the entire distributions were not

339

conducted. The average one-third depth pavement temperatures for Daggett, Sacramento, and

340

San Francisco are 26.4oC, 24.5oC, and 20.3oC, respectively. For an average truck highway

341

speed of 97 kph in California, these average pavement temperatures result in average excess

342

fuel consumptions ranging from 0.1% to 0.35% across all 5 structures and three locations.

(6)

343 344

Although the level of EFC found in this study is smaller than typical calculations of

345

roughness related EFC, which is most likely approximately independent of EFC from

346

structural response, it can be considered in the decision-making process, and will have

347

greater significance where there are heavy vehicles traveling at slow speeds in hotter climates

348

(Harvey et al. 2015).

349

350

Simulations using this model and models from the Massachusetts Institute of Technology

351

(Louhghalam et al. 2013) and Michigan State University have been used to estimate annual

352

EFC due to structural response as is described in Harvey et al. (2015). These simulations

353

included the joint distributions across representative days in each month of hourly annual

354

pavement temperatures, actual typical hourly urban and rural speed distributions for different

355

types of vehicles, and hourly distributions of car and truck wheel loads based on weigh-in-

356

motion distributions.

357 358

Table 3 also presents the EFC for different speeds and structure types for the average one-

359

third depth pavement temperatures for Daggett, Sacramento, and San Francisco. It can be

360

observed that the section with PM mix has the largest EFC while the section with RHMA

361

mix has the second largest values. Structure 2 with the polymer-modified overlay has the

362

highest EFC when compared with other structures (Figure 8-Table 3). This is a result of the

363

low stiffness of the polymer-modified mix (Figure 1). Calculated EFC for Structure 1 (New

364

DGAC on old DGAC) and Structure 3 (RHMA) were observed to be close (Figures 7 and 9 –

365

Table 3). Although the differences in EFC between Structure 1 and Structures 4 (New DGAC

366

on cracked concrete) (Figure 10) and 5 (New DGAC on CTB) (Figure 11) were not

367

significant, EFC for Structure 1 was observed to be higher than the EFC for Structures 4 and

368

5. This result suggests that having a concrete layer or a CTB layer under the asphalt overlay

369

rather than having an aged asphalt layer does not create any significant reduction in EFC, and

370

that most of the energy dissipation is near the surface in the asphalt layer.

371 372

The analysis of variance (ANOVA) approach was used to quantify the impact of pavement

373

temperature, speed, structure type, and subgrade stiffness on calculated EFC. The results of

374

the ANOVA are given in Table 2. Independent variables with p-values larger than 0.05 can

375

be regarded as insignificant variables. It can be observed that subgrade stiffness does not

376

have any significant effect on EFC with a p-value of 0.82 (significantly larger than 0.05). On

377

the other hand, temperature, speed, and structure type appeared to be important factors

378

affecting EFC. F values given in Table 2 suggested that temperature is the major factor

379

controlling EFC. While speed is also a significant parameter, its effect on EFC is

380

significantly lower than the influence of temperature.

381 382

SUMMARY OF FINDINGS AND CONCLUSIONS

383

In this study, vehicle energy consumption induced by the viscous behavior of asphalt

384

pavement materials was calculated by using 3D viscoelastic FE modeling. Five pavement

385

structures with conventional, rubber, and polymer-modified asphalt overlay mixes with

386

underlying old asphalt, cracked concrete, and CTB layers were modeled to evaluate the

387

impact of structure type on calculated excess fuel consumption (EFC). A total of 234 FE

388

models for a full factorial design were developed to simulate the impacts of three speed

389

levels (8, 52, 105 kph), two subgrade stiffnesses (50 and 200 MPa), and nine temperatures

390

(0oC to 80oC at 10oC intervals) on calculated EFC.

391 392 393

Results of this study can be summarized as follows: 

For the structures analyzed in this study, there is no need to simulate temperature

394

gradients in the simulations. For the calculation of average annual EFC (network

395

level application), pavement temperatures at one third depth can be used to simulate

396

the effect of temperature gradients throughout the pavement depth. This assumption

397

can create a significant reduction in model run times.

398



Total vehicle dissipated energy and EFC can be accurately calculated by modeling

399 400

one tire and multiplying the output by the number of tires on the vehicle. 

The greatest EFC can be observed at pavement temperatures higher than 40oC. For

401

the fully loaded truck simulated in this study and an average truck highway speed of

402

97 kph, average pavement temperatures for three cities in California (Daggett,

403

Sacramento, and San Francisco) result in EFCs ranging from 0.1% to 0.35% for the

404

structures analyzed in this study. Although this level of EFC appears to be much less

405

than the roughness related EFC (1-3%) (Chatti and Zaabar 2012), which is most

406

likely approximately independent of EFC from structural response, EFC from

407

structural response can be considered in the decision-making process, and will have

408

greater significance where there are heavy vehicles traveling at slow speeds in hotter

409

climates (Harvey et al. 2015).

410



411 412

Polymer-modified overlays showed the highest EFC when compared with other structures.



413

Having a concrete layer or a CTB layer under the asphalt overlay rather than having an aged asphalt layer did not create any significant reduction in EFC.

414



Subgrade stiffness was not observed to be a significant factor affecting EFC.

415



Temperature, speed, and structure type were determined to be important factors

416

affecting EFC. Temperature was observed to be the major factor controlling EFC.

417

While speed is also a significant parameter, its effect on EFC is significantly lower

418

than the influence of temperature.

419 420

In this study, average annual temperatures for three different cities in California were used to

421

provide a first estimate of the average percentage contribution of pavement structural

422

response on EFC. These results have been used with hourly temperature and traffic data to

423

calculate the percentage contributions for different cities. The impact of subgrade damping

424

and layer densities on EFC should also be investigated by using a dynamic model. The

425

impact of subgrade thickness and bedrock location should also be evaluated. The interaction

426

of roughness-induced dynamic loading should be considered.

427 428

The EFCs predicted in this study do not provide an absolute value associated with a

429

particular pavement, and these results are as-yet uncalibrated with field data. However,

430

results will provide important information about the impacts of speed, temperature, material

431

type, and structure configuration on pavement structural response related EFCs. This study

432

also points out the level of significance of pavement structural response on vehicle fuel

433

economy. Models should be calibrated by using field measurements.

434 435

ACKNOWLEDGMENTS

436

This paper describes research activities that were requested and sponsored by the California

437

Department of Transportation (Caltrans), Division of Research and Innovation. Caltrans

438

sponsorship is gratefully acknowledged. The contents of this paper reflect the views of the

439

authors and do not reflect the official views or policies of the State of California or the

440

Federal Highway Administration. The advice regarding model development elements of Dr.

441

Shmuel Weissman of Symplectic Engineering Corporation is appreciated.

442 443 444 445 446 447

448 449 450

REFERENCES

451

Akbarian, M., A. Louhghalam, and F.J. Ulm. (2014). “Mapping the excess-fuel consumption

452

due to pavement vehicle interaction: a Case study of Virginia’s interstate system.” Papers

453

from the International Symposium on Pavement LCA, Davis, CA.

454 455

Asphalt Institute. 2001. SP-2 Superpave Mix Design.

456 457

Cass, D. and Mukherjee, A. (2011) "Calculation of greenhouse gas emissions for highway

458

construction operations by using a hybrid life-cycle assessment approach: case study for

459

pavement operations." Journal of Construction Engineering and Management 137.11: 1015-

460

1025.

461 462

Chatti, K. and Zaabar, I. (2012). “Estimating the effects of pavement condition on vehicle

463

operating costs.” Publication NCHRP Report 720. National Cooperative Highway Research

464

Program, Transportation Research Board, Washington, D.C.

465 466

Coleri, E., Harvey, J.T., Zaabar, I., Louhghalam, A., and Chatti K. (2015). “Model

467

development, field section characterization and model comparison for excess vehicle fuel use

468

due to pavement structural response.” Accepted for publication in the Transportation

469

Research Record: Journal of the Transportation Research Board.

470 471 472

Ferry, J. D. (1980). “Viscoelastic properties of polymers.” John & Sons.

473

Harvey, J., Lea, J.D., Kim, C., Coleri, E., Zaabar, I., Louhghalam, A., Chatti, K., Buscheck,

474

J., and Butt, A. (2015). Simulation of Cumulative Annual Impact of Pavement Structural

475

Response on Vehicle Fuel Economy for California Test Sections. University of California

476

Pavement Research Center Report UCPRC-RR-2015-05.

477 478

Jonsson, P., and Hultqvist, B.Å. (2009). “Measurement of fuel consumption on asphalt and

479

concrete pavements North of Uppsala.” Swedish National Road and Transport Research

480

Institute, Linköping, Sweden.

481 482

Kim, R.E., Kang, S., Spencer, B.F., Al-Qadi, I.L., and Ozer, H. (2016). “New stochastic

483

approach of vehicle energy dissipation on nondeformable rough pavements.” Journal of

484

Engineering Mechanics.

485 486

Louhghalam, A., Akbarian, M., and Ulm, F-J. (2013). “Flügge's conjecture: Dissipation vs.

487

deflection induced pavement-vehicle-interactions (PVI).” Journal of Engineering Mechanics,

488

140 (8), 10.1061/(ASCE)EM.1943‐7889.0000754.

489 490

Louhghalam, A., Akbarian, M., and Ulm, F-J. (2014). “Scaling relations of dissipation-

491

induced pavement-vehicle interaction.” Transportation Research Record: Journal of

492

Transportation Research Board, No.2457, 95-104, Washington, D.C.

493 494

Ozer, H., Yang, R., and Al-Qadi, I.. (2017) "Quantifying sustainable strategies for the

495

construction of highway pavements in Illinois." Transportation Research Part D: Transport

496

and Environment 51: 1-13.

497

498

Pouget S., Sauzéat C., Di Benedetto, H., and Olard, F. (2011). “Viscous energy dissipation in

499

asphalt pavement structures and implication for vehicle fuel consumption.” Journal of

500

Materials in Civil Engineering, 24(5):568-576.

501 502

Shakiba, M., Ozer, H., Ziyadi, M., and Al-Qadi, I.L. (2016). “Mechanics based model for

503

predicting structure-induced rolling resistance (SRR) of the tire-pavement system.”

504

Mechanics of Time-Dependent Materials, 20(4):579-600.

505 506

Sharpe, B. and Muncrief, R. (2015). “Literature review: real-world fuel consumption of

507

heavy-duty vehicles in the United States, China, and the European Union.” ICCT White

508

Paper. Washington, DC: ICCT.

509 510

Van Dam, T. Harvey, J., Muench, S., Smith, K., Snyder, M., Al-Qadi, I., and Ozer, H. (2015)

511

"Towards sustainable pavement systems: a reference document." Federal Highway

512

Administration, HIF-15-002.

513 514

Wang, T., Lee, I.-S., Kendall, A., Harvey, J., Lee, E.B., and Kim, C. (2012) “Life cycle

515

energy consumption and GHG emission from pavement rehabilitation with different rolling

516

resistance.” Journal of Cleaner Production, Volume 33, 86-96.

517 518

Wang, T., Harvey, J., and Kendall, A. (2014) “Reducing greenhouse gas emissions through

519

strategic management of highway pavement roughness.” Environmental Research Letters 9

520

034007.

521

522

Yoo, P.J., Al-Qadi, I.L., Elseifi ,M.A., and Janajreh, I. (2006). “Flexible pavement responses

523

to different loading amplitudes considering layer interface condition and lateral shear forces.”

524

International Journal of Pavement Engineering, 7(1):73-86.

525 526

Zapata, C.E. and Houston, W.N. (2008). “Calibration and Validation of the Enhanced

527

Integrated Climate Model for Pavement Design.” NCHRP Report 602: Transportation

528

Research Board of the National Academies, Washington, D.C.

529

LIST OF TABLES Table 1 Factorial Design for FE Models Table 2 ANOVA Results Identifying the Effects of Speed, Temperature, and Subgrade Stiffness on Excess Fuel Consumption Table 3 Excess Fuel Consumptions for Average Temperatures and Different Speeds (for 200MPa subgrade stiffness)

Table 1 Factorial Design for FE Models Structures

Temp. (oC)

Speed (kph)

STR#1 STR#2 STR#3 STR#4 STR#5

0, 10, 20,…,80 0, 10, 20,…,80 0, 10, 20,…,80 0, 10, 20,…,80 0, 10, 20,…,80

8, 52, 105 8, 52, 105 8, 52, 105 8, 52, 105 52

EAB (MPa) 250 250 250 250 250

ESG (MPa) 50, 200 50, 200 50, 200 50, 200 50, 200

Number of models 54 54 54 54 18

NOTE: Temp. = Temperature; EAB = Aggregate base stiffness; ESG = Subgrade stiffness.

Table 2 ANOVA Results Identifying the Effects of Speed, Temperature, and Subgrade Stiffness on Excess Fuel Consumption Variable Temp. Speed Str Esg Temp:Speed Temp:Str Temp:Esg Speed:Str Speed:Esg Str:Esg

Df 1 1 4 1 1 4 1 3 1 4

SS 1616.6 112.7 147.3 0.29 119.9 162.8 0.02 64.6 0.00 0.09

F value 303.9 21.2 6.9 0.05 22.5 7.7 0.00 4.04 0.00 0.00

p-value 0.00000 0.00000 0.00003 0.81558 0.00000 0.00000 0.94824 0.00795 0.99840 0.99996

NOTE: Temp. = Temperature; Str = Structure; Esg = Subgrade stiffness; Df = degrees of freedom; SS = sum of squares.

Table 3 Excess Fuel Consumptions for Average Temperatures and Different Speeds (for 200MPa subgrade stiffness) City Daggett Sacramento San Francisco Daggett Sacramento San Francisco Daggett Sacramento San Francisco

Average Temp.(oC) 26.4 24.5 20.3 26.4 24.5 20.3 26.4 24.5 20.3

Speed (kph) 8 8 8 52 52 52 105 105 105

EFC STR1(%) 0.38 0.33 0.23 0.21 0.19 0.13 0.17 0.14 0.09

EFC STR2(%) 0.94 0.80 0.50 0.44 0.39 0.26 0.35 0.30 0.18

EFC STR3(%) 0.56 0.48 0.30 0.29 0.26 0.20 0.23 0.20 0.15

NOTE: Temp. = Temperature; Str = Structure; EFC=Excess fuel consumption.

EFC STR4(%) 0.19 0.16 0.09 0.09 0.07 0.05 0.07 0.06 0.03

EFC STR5(%) 0.13 0.12 0.08 -

LIST OF FIGURES Figure 1 Dynamic modulus master curves for all asphalt mixes. Figure 2 Asphalt layer temperature distributions for a summer day in Sacramento, California. Figure 3 Simulation of temperature gradient in the asphalt layer (a) Asphalt layer divided into 10 segments to simulate the effect of temperature gradient (b) Calculated excess fuel consumption for the case with temperature gradients and temperatures at different depths. Figure 4 Displacement field for the finite element models with: a) one tire b) two tires; c) four tires (one side of dual-tandem axle). Figure 5 Change in calculated dissipated energy with temperature for models with one, two, and four tires (a) total dissipated energy (b) total dissipated energy normalized by the number of tires. Figure 6 Cross sections of structures used for FE modeling. Figure 7 The impact of speed, temperature, and subgrade stiffness on fuel consumption excess - Structure 1: New DGAC on old DGAC. Figure 8 The impact of speed, temperature, and subgrade stiffness on fuel consumption excess - Structure 2: PM on old DGAC. Figure 9 The impact of speed, temperature, and subgrade stiffness on fuel consumption excess - Structure 3: RHMA on old DGAC. Figure 10 The impact of speed, temperature, and subgrade stiffness on fuel consumption excess - Structure 4: New DGAC on cracked concrete. Figure 11 The impact of temperature and subgrade stiffness on fuel consumption excess Structure 5: New DGAC on CTB. Figure 12 Cumulative distributions of pavement temperatures at one-third depth for Daggett, Sacramento, and San Francisco in California.

Temperature (C) 20

25

30

35

40

45

50

0 mm

0 50

Depth (mm)

75 mm 100

125 mm 150 0:00

200 250 300

4:00 8:00 12:00 16:00 20:00

175 mm

Excess fuel consumption (ml/km)

5 4.5 4 3.5 3 2.5 2 With gradient 0 mm 75mm

1.5 1

125mm 175mm

0.5 0 0.00

4.00

8.00

12.00

16.00

Time of day

20.00

24.00

 

3.0 One tire two tires

2.5

four tires

Energy (mJ)

Car

2.0 1.5 1.0 0.5 0.0 0

20

40

60

80

100

Temperature (C)

  Energy/Number of Tires (mJ)

0.8 One tire two tires

0.7

four tires

0.6 0.5 0.4 0.3 0.2 0.1 0.0 0

10

20

30

40

50

Temperature (C)

60

70

80

90

100 mm

New DGAC

100 mm

PM

150 mm

Old DGAC

150 mm

Old DGAC

200 mm

EAB = 250 MPa

AB

ESG = 50 and 200 MPa

SG

200 mm

EAB = 250 MPa

AB

ESG = 50 and 200 MPa

SG

100 mm

RHMA

100 mm

150 mm

Old DGAC

200 mm

EPCC = 25,000 MPa

200 mm

EAB = 250 MPa

AB

ESG = 50 and 200 MPa

SG

200 mm

EAB = 250 MPa

AB

ESG = 50 and 200 MPa

SG

100 mm

New DGAC

150 mm

ECTB = 4,300 MPa

200 mm

EAB = 250 MPa

AB

ESG = 50 and 200 MPa

SG

CTB

New DGAC Cracked PCC

Fuel consumption excess (%)

14

Speed=8kph & Esg=50MPa

12

Speed=8kph & Esg=200MPa Speed=52kph & Esg=50MPa

10

Speed=52kph & Esg=200MPa Speed=105kph & Esg=50MPa

8

Speed=105kph & Esg=200MPa

6 4 2 0 0

20

40 60 Temperature (C)

80

100

14 Fuel consumption excess (%)

Speed=8kph & Esg=50MPa

28.3% at 80oC

Speed=8kph & Esg=200MPa

12

Speed=52kph & Esg=50MPa

10

Speed=52kph & Esg=200MPa Speed=105kph & Esg=50MPa

8

Speed=105kph & Esg=200MPa

6 4 2 0 0

20

40 60 Temperature (C)

80

100

14 Fuel consumption excess (%)

Speed=8kph & Esg=50MPa Speed=8kph & Esg=200MPa

12

Speed=52kph & Esg=50MPa

10

Speed=52kph & Esg=200MPa Speed=105kph & Esg=50MPa

8

Speed=105kph & Esg=200MPa

6 4 2 0 0

20

40 60 Temperature (C)

80

100

Fuel consumption excess (%)

14

Speed=8kph & Esg=50MPa Speed=8kph & Esg=200MPa

12

Speed=52kph & Esg=50MPa

10

Speed=52kph & Esg=200MPa Speed=105kph & Esg=50MPa

8

Speed=105kph & Esg=200MPa

6 4 2 0 0

20

40 60 Temperature (C)

80

100

Fuel consumption excess (%)

14 Speed=52kph & Esg=50MPa

12 Speed=52kph & Esg=200MPa

10 8 6 4 2 0 0

20

40 60 Temperature (C)

80

100

1.0 Daggett

Cumulative distribution

0.9

Sacramento

0.8

San Francisco

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0

-10

0

10

20

30

40

50

Pavement temperatures at one third depth (C)

60

70