Materials and Structures DOI 10.1617/s11527-015-0744-x

Numerical evaluation of pavement design parameters for the fatigue cracking and rutting performance of asphalt pavements Amirhossein Norouzi . Dahae Kim Y. Richard Kim

.

Received: 26 May 2015 / Accepted: 13 November 2015  RILEM 2015

Abstract Over recent years, significant research has been conducted to investigate ways to predict fatigue cracking and permanent deformation (rutting), which are two common distresses found in asphalt pavements. These distresses are affected by material properties, environmental conditions, and the pavement’s structure. This paper investigates common pavement design parameters, including surface mixture type, base layer thickness, base layer type, subbase layer thickness, and an anti-frost layer, with regard to the asphalt pavement performance of the Korea Expressway Corporation (KEC) test road. Test roads are often regarded as the most realistic tools for evaluating the effects of various parameters because they are subjected to real traffic and environmental factors. The KEC test road is 7.7 km long and was constructed with the aim of developing a Korean mechanistic-empirical pavement design guide. According to the findings, the surface layer type, base layer thickness, and base layer material type were found to affect the fatigue cracking and rutting

A. Norouzi  D. Kim (&)  Y. Richard Kim Department of Civil, Construction, & Environmental Engineering, North Carolina State University, Campus Box 7908, Raleigh, NC 27695-7908, USA e-mail: [email protected] A. Norouzi e-mail: [email protected] Y. Richard Kim e-mail: [email protected]

performance, whereas the sub-base thickness and anti-frost layer were found not to affect the amount of distress significantly. The newly developed ‘layered viscoelastic pavement analysis for critical distresses’ (LVECD) program was able to capture the effects of the changes in the aforementioned parameters on the amount of cracking and rut depths. Reasonable agreement was found between the LVECD predictions and the field distress measurements. However, it remains necessary to develop a laboratory-to-field transfer function in order to obtain more accurate field performance predictions. Keywords Asphalt  Pavement  Pavement design  Fatigue cracking  Rutting  LVECD program  Field condition

1 Introduction For decades, the pavement industry has been endeavoring to improve pavement construction practices, extend the life of new pavements, and minimize the need for pavement rehabilitation efforts. It is now well accepted that fatigue cracking and rutting are two major distresses that occur in hot mix asphalt pavements. These distresses have been reported in many parts of the United States as well as Europe and other countries. Two primary types of models that are essential to predict these distresses in mechanistic-empirical

Materials and Structures

pavement design and analysis are pavement response models and performance prediction models. Pavement response models determine the stresses and strains in a pavement system and in turn are used in the performance prediction models to determine the evolution of the critical distresses. Numerous studies have been conducted to determine the effects of different pavement design parameters (e.g., thickness and material type) on pavement responses and performance. For example, a study by [1] showed that the base layer type, base layer thickness, and the subgrade resilient modulus are the key elements that control the strain levels at critical locations in the pavement structure. Therefore, the key step for performance predictions and design purposes is to utilize accurate methods that consider pavement structure, layer material type, traffic loading, and temperature variations to obtain reasonable stress–strain analysis results for both the vertical and horizontal directions throughout the depth of the pavement. Generally, in order to predict asphalt performance, effective models that can reliably represent fatigue damage growth and permanent deformation are still in demand in addition to efficient tools for strain and stress calculations. The Simplified ViscoElastic Continuum Damage (S-VECD) model is a mechanistic approach that has been applied effectively to predict the performance of asphalt concrete mixtures during pre-localization stages under different modes of loading [2–5]. Zhang et al. [6] developed an energy measure that represents the rate of damage growth using the S-VECD model and can predict fatigue failure. Based on Zhang et al.’s work, [7] proposed a new energy-based failure criterion called the GR method, which is independent of mode of loading, test temperature, and strain amplitude. The term GR is the pseudo strain energy release rate and is defined as the term that considers both stress and strain within asphalt concrete specimens. This new criterion has been verified successfully for multiple mixtures that contain reclaimed asphalt pavement, warm mix additives, and modified binders [7, 8]. In this study, the rutting performance of the study mixtures was evaluated using a permanent deformation model developed at North Carolina State University by Choi and Kim [9]. This so-called shift model is capable of expressing the permanent strain growth of asphalt concrete in both the primary and secondary

regions as a function of deviatoric stress, load time, and temperature, and is based on the time–temperature and time-stress superposition principles [10]. These three factors (deviatoric stress, load time, and temperature) are important in predicting rut depths in asphalt pavements because they vary throughout the depth of the asphalt layer. The triaxial stress sweep (TSS) test used in this study was developed to predict the rutting performance of the study mixtures by calibrating the shift model. However, performance predictions that are based solely on material testing cannot provide a comprehensive picture of the pavement’s behavior. Because pavement is a layered structure, the stress and strain distributions vary from point to point, which results in complicated shear and flow zones [11]. The state-ofthe-practice approach for stress–strain analysis is layered elastic analysis, where the pavement layers are considered to be the elastic material under stationary axisymmetric loading [12]. However, layered elastic analysis is not an accurate tool for asphalt pavement because asphalt concrete exhibits viscoelastic behavior, especially under traffic loading. Layered viscoelastic moving load analysis, which is an improvement over layered elastic analysis, is considered also to be more reliable than linear viscoelastic analysis in the viscoelastic domain due to its efficient functions that account for the effects of moving loads and viscoelasticity [13]. The newly released ‘layered viscoelastic pavement analysis for critical distresses’ (LVECD) software program is able to calculate linear viscoelastic pavement responses that can be used for both the shift model and the S-VECD model to predict rutting and fatigue cracking, respectively, under moving loads. Norouzi and Kim [14] verified the LVECD program for multiple pavement sections in the United States and found strong agreement between the simulation results and the field observations. A comparison of pavement performance prediction results from the LVECD program and the Pavement ME software is presented in [15] using the performance data from 33 pavement sections in the US, Canada, South Korea, and China. Given the above considerations, the purpose of this study is to gain a better understanding of the effects of the pavement design parameters, i.e., surface layer type, base layer material properties, base layer thickness, sub-base thickness, and subgrade properties, on the fatigue and rutting

Materials and Structures

resistance of asphalt pavements using experimental material characteristics and the LVECD software. This study used the following test protocols to find the required parameters for the performance predictions: dynamic modulus tests (AASHTO TP 79), uniaxial fatigue tests using the S-VECD model (AASHTO TP 107), and TSS tests for permanent deformation. To verify the LVECD analysis results, field data were collected from the test sections and compared with the program simulations.

2 Test protocols 2.1 Dynamic modulus testing Asphalt material is considered to be linear viscoelastic material at specific strain levels; clearly, viscoelastic materials have both viscous and elastic components. The dynamic modulus, |E*|, can be expressed in the form of a mastercurve that exhibits frequency- and temperature-dependent behavior. In this study, dynamic modulus testing was performed in loadcontrolled mode with axial compression following AASHTO TP 79. The tests were carried out for all the study mixtures at 4, 20, 40, and 54 C and at frequencies of 25, 10, 5, 1, 0.5, and 0.1 Hz. The load levels were specified by trial and error so that the strain amplitudes were between 50 and 75 microstrains to prevent damage to the specimens. The dynamic modulus values were used to develop the dynamic modulus mastercurve by simultaneously optimizing the sigmoidal function of the mastercurve and the quadratic time–temperature shift function. After determining the shift factors, the dynamic modulus was converted to the relaxation modulus. Finally, a power term, alpha (a), used in viscoelastic continuum damage (VECD) theory, was calculated from the maximum log–log slope, m, of the relaxation modulus and time using the relationship, a ¼ 1 þ m1 .

Tð CÞ 

2.2 Cyclic testing using the S-VECD model The S-VECD fatigue performance model is the simplified form of the more rigorous VECD model and can be used to characterize the fatigue behavior of asphalt concrete using the elastic–viscoelastic correspondence principle, continuum damage mechanics, and time–temperature superposition principle. The SVECD model has been proven to be independent of mode of loading. Controlled crosshead (CX) cyclic direct tension tests were performed at 10 Hz at different temperatures based on the binder performance grade (PG) following AASHTO TP 107. All the tests were performed at three different strain amplitudes (high, medium, and low). The strain amplitudes were selected in such a way to create a spread of numbers of cycles to failure (Nf) in the range of 1000–100,000 cycles. In addition to taking air void measurements, in order to check the variability of the fatigue test specimens more precisely, fingerprint dynamic modulus tests were conducted at 10 Hz and 50 cycles before running the CX cyclic direct tension tests. The dynamic modulus value measured from this test is specified as |E*|fingerprint and is used to calculate the dynamic modulus ratio (DMR) via Eq. (1). A DMR value in the range of 0.9–1.1 guarantees that the linear viscoelastic properties obtained from the dynamic modulus tests can be used effectively in SVECD analysis. DMR ¼

jE jfingerprint jE jLVE

ð1Þ

To determine failure for each sample, the corresponding cycle that is related to the sudden drop in the phase angle, which typically happens around the failure point, has been specified based on the Reese approach [16]. In order to minimize the effects of viscoplasticity, [7] suggested using the PG of the base binder to determine the proper testing temperature, as shown in Eq. (2).

High temperature binder PG grade þ Low temperature binder PG grade  3  19 C 2

ð2Þ

Materials and Structures

2.3 Fatigue failure criterion (GR method)

where (eR0,ta)i is the pseudo strain amplitude at cycle i and, Fi is the pseudo stiffness at cycle i.

where evp is the viscoplastic strain (i.e., permanent strain), e0, NI, b is the coefficients of the incremental model, Nred is the reduced number of cycles at reference loading conditions (N  10atotal ), N is the physical number of cycles of a certain loading condition, atotal is the total shift factor, which is the summation of two shift factors (anp þ arv ), anp is the reduced load time shift factor, p1, p2 is the coefficients of reduced load time shift factor, np is the reduced load time, arv is the vertical stress shift factor, d1, d2, d3 is the coefficients of vertical stress shift factor, rv is the vertical stress, and Pa is the atmospheric pressure to normalize stress. The MSS tests consist of three loading blocks with increases in the deviatoric stress level (70, 100, and 130 w) while the other loading conditions are kept constant. In this study, the shift factors were obtained by shifting the permanent strain of an individual loading block toward the permanent strain mastercurve, which was obtained from the reference test. The reduced load time shift factor and deviatoric stress shift factor are shown in Eq. (5). The physical number of cycles at a given condition was converted into a reduced number of cycles using the total shift factor, which is the sum of the deviatoric stress shift factor and the reduced load time shift factor. These two shift functions utilize temperature, load time, and vertical stress to calculate the shift factors. Details regarding the TSS test method and shift model can be found elsewhere Choi and Kim [9].

2.4 Permanent deformation (triaxial stress sweep testing)

3 Materials and pavement sections

According to the S-VECD failure criterion (the GR method), the maximum stored pseudo strain energy at each cycle represents the material’s ability to store energy at that particular time. The material loses its stored energy as the damage grows for the same magnitude of applied pseudo strain due to the reduction in pseudo stiffness. The difference between the maximum stored pseudo strain energy and the corresponding undamaged state is referred to as the total released pseudo strain energy and is denoted as WRC. Sabouri and Kim [7] defined the energy term, GR, as the rate of change of the averaged released pseudo strain energy (per cycle) throughout the entire history of the test. Sabouri and Kim found that a characteristic relationship exists between the GR during fatigue testing and the final fatigue life (Nf). This failure criterion combines the advantages of the S-VECD model and this characteristic relationship, which both originate from fundamental mixture properties. Details regarding the GR method and its corresponding calculations can be found in the paper by [7]. The GR can be calculated using Eq. (3). RNf GR ¼

WCR

0

Nf2

1 2

¼

RNf  R 2 e0;ta ð1  Fi Þ 0

Nf2

ð3Þ

The TSS test is composed of two type of tests: a reference test at the high temperature (TH) followed by three multiple stress sweep (MSS) tests at three different temperatures of low, intermediate, and high (TL, TI, and TH), respectively. The reference test in this study utilizes a 0.4-s pulse with a 10-s rest period. This reference test provides permanent strain mastercurves by fitting the incremental model, which is expressed as Eq. (4). evp ¼

e0  Nred ðNI þ Nred Þb

;

anp ¼ p1 logðnp Þ þ p2 ; arv ¼ d1 ðrv =Pa Þd2 þ d3 :

ð4Þ

ð5Þ

3.1 Description of the KEC Test Road The KEC test road was constructed in December 2002. This test road is composed of 33 types of asphalt pavement; Fig. 1 schematically presents the KEC pavement structures. The field performance data obtained from the test road allow researchers to compare different types of pavement structures and different mixtures under various climate conditions and real traffic loads. ARAN (Automatic Road Analyzer, ROADWARE) has been used to conduct annual pavement condition surveys of the test road [17]. Asphalt concrete overlays were applied to some of the pavement sections in 2006. Therefore, the

Materials and Structures

A1

A2

A2-2

5

A3

A4

A5

A5-2

5

5

A6

A7

A8

A8-2

5

5

A9

A10 A10-2 A11 A11-2 A12 A12-2 A13 A13-2 A14 A14-2 A15 A15-2

5

5

5

5

5

5

5

5

5

5

5

5

5

7

8

8

8

18

18

18

28

28

28

30

30

30

40 30

30

30

30

30

30

40

40

40

40

40

30

30

30

30

20

20

20

20

10

PMA

ASTM

BB5

BB3

BB1

Aggregate Sub-Base

Anti-Frost

Subgrade

Fig. 1 Asphalt pavement sections at the KEC test road

performance data obtained in 2005 were used for the LVECD analysis, because direct comparisons between field measurements and LVECD predictions are only possible using performance data prior to the overlay construction. 3.2 Mixtures For this study, experiments were performed using five laboratory-produced mixtures. Of these mixtures, two types of asphalt mixtures were used at the surface to compare the rutting and crack propagation; these mixtures were an ASTM mix and a 19-mm nominal maximum aggregate size (NMAS) polymer-modified styrene butadiene styrene (SBS) mix, which is designated as PMA throughout the study. The intermediate layer consisted of a 25-mm NMAS BB5 mixture with 70 mm thickness. Mixes designated as BB1 (25 mm

NMAS) and BB3 (40 mm NMAS), which are frequently used in South Korea, were used for the base layers. Table 1 summarizes the volumetric properties of the mixtures used in the KEC test road sections. The sublayers below the base layer are composed mostly of sub-base and anti-frost layers that are placed on top of the subgrade. An anti-frost layer often is used to compensate for the level difference due to the base; however, the antifrost layer was omitted from some sections for comparative purposes to evaluate the effectiveness of anti-frost layers on pavement performance. Figure 1 schematically presents the KEC pavement structures. 3.3 Specimen fabrication To prepare the specimens, aggregate stockpiles were dried and sieved for batching. The aggregate particles were then heated to the mixing temperature for more

Materials and Structures Table 1 Volumetric properties of the KEC mixtures

* The numbers in the figure are all in centimeters (cm)

Type

Surface

Base

Mixture

ASTM

Binder type

Unmodified

Styrene butadiene styrene

Unmodified

Binder grade

PG 64-22

PG 76-22

PG 64-22

Binder content (%) NMAS* (mm)

5.2 19

4.9

4.2 25

4.4 40

4.3 25

PMA

BB1

Intermediate BB3

BB5

% Air void (S-VECD)

5.9

5.7

7.6

7.5

% Air void (Rutting)

5.9

6.0

8.0

9.9

Sieve size

Gradation, % Passing

37.5 (mm)

100.0

100.0

100.0

100.0

25.0 (mm)

100.0

100.0

88.6

100.0

19.0 (mm)

99.6

92.5

71.0

91.0

12.5 (mm)

84.9

72.9

51.1

67.5

9.5 (mm)

71.1

63.9

44.1

55.1

4.75 (mm)

49.3

48.5

38.1

31.2

2.36 (mm)

36.2

36.1

29.1

23.0

0.60 (mm)

18.1

18.0

15.1

12.8

0.30 (mm)

11.6

11.6

10.1

9.2

0.15 (mm)

7.4

7.3

6.8

6.7

0.075 (mm)

4.4

4.2

4.4

4.6

than 6 h before mixing. The asphalt binder and aggregates were mixed together using a bucket mixer. Then, the mixtures were short-term oven-aged for 4 h at the compaction temperature. All the test specimens were compacted to a height of 178 mm and a diameter of 150 mm using the Superpave Gyratory Compactor (SGC). To obtain specimens of uniform air void distribution, these samples were cored to a diameter of 100 mm and cut to a height of 130 mm [18] for direct tension cyclic testing, and to a diameter of 100 mm with the height of 150 mm for the dynamic modulus and permanent deformation (TSS) testing. The air void contents were measured using the CoreLok method for each specimen prior to testing. All the test samples met the target air void content by ±0.5 %. The direct tension test specimens were glued to metal plates using Devcon steel putty. Vertical deformations were measured using four linear variable differential transducers (LVDTs) with the gauge length of 70 mm at intervals of 90 degrees for both the dynamic modulus and direct tension cyclic tests and a 100-mm gauge length for the TSS tests.

4 Test results and discussion 4.1 Dynamic modulus values Figure 2a, b present the averaged dynamic modulus values of the replicates for the KEC mixtures on the same graph in semi-log space and log–log space, respectively. The comparison between the PMA and ASTM mixes clearly shows that the PMA mix exhibits lower stiffness values at high reduced frequencies (low temperatures) and higher stiffness values at low reduced frequencies (high temperatures). In other words, it seems that the PMA mix presents favorable characteristics for fatigue resistance at low temperatures and for rutting resistance at high temperatures. In general, the other unmodified mixtures have similar stiffness values, given the sample-to-sample variability, except for the ASTM mix that shows low stiffness values at high temperatures. Table 2 presents the time–temperature shift factors and Table 3 presents the dynamic modulus Prony coefficients for the KEC mixtures as obtained from experimental analysis.

Materials and Structures 25000

(a)

|E*| (MPa)

20000 15000 ASTM

10000

PMA BB1

5000

BB3 BB5

0 1E-08

1E-06

1E-04

1E-02

1E+00

1E+02

Reduced Frequency (Hz)

|E*| (MPa)

100000

(b)

10000

4.3 Fatigue failure criterion lines

1000

100 1E-08

1E-06

1E-04

1E-02

1E+00

1E+02

Reduced Frequency (Hz)

Fig. 2 Dynamic modulus mastercurves for KEC test road mixtures: a semi-log space and b log–log space

Table 2 Time–temperature shift factors for KEC mixtures Time–temperature shift factors Mixture

compared to the PMA mix. Another reasonable explanation for this observation may be related to the physical nature of the base binder and SBS modifier used in the PMA mix. The other important point to make from Fig. 3 is that the curve for the BB1 mix is below the other curves, which could be due to its low asphalt content and large aggregate size. Generally, a comparison of damage curves cannot yield reliable information about different mixtures’ fatigue behavior, because the energy that is input by mechanical force is consumed not only in creating and propagating the cracks, but also in deforming the material. Therefore, it is important to include both stiffness and damage characteristics of the material when determining a mixture’s fatigue cracking resistance.

a1

a2

a3

ASTM

8.02E-04

-1.64E-01

7.99E-01

PMA BB1

1.08E-03 6.75E-04

-1.77E-01 -1.62E-01

8.58E-01 7.94E-01

BB3

7.80E-04

-1.61E-01

7.85E-01

BB5

4.69E-04

-1.43E-01

7.03E-01

4.2 S-VECD characterization curves Figure 3 presents the averaged damage characteristic curves for the KEC test mixtures. As shown, the corresponding curve for the ASTM mixture is slightly above that of the PMA mixture. It can be concluded that both the PMA and ASTM mixes follow the same behavior even though their dynamic modulus values are considerably different. This outcome could be related to the higher asphalt content in the ASTM mix

The S-VECD failure criterion was applied to all of the mixtures in this study, and the results are presented in Fig. 4. The position of the failure criterion line can be used to make a relative comparison of a mixture’s expected fatigue resistance. That is, a line that has a larger Nf value for the same GR value indicates more resistance to fatigue cracking. Also, not only the position but also the slope of the failure criterion line plays an important role in the pavement system’s fatigue behavior. For example, the slopes of the three BB mixes are steeper than those of the surface mixes, and therefore, the poorer cracking resistance of the BB mixes demonstrated by the lower position of their GR lines becomes even worse when the GR values are lower (e.g., when the pavement thickness becomes thicker). As observed in Fig. 4, the line for the PMA mix is parallel and slightly above that of the ASTM mix, indicating better fatigue cracking resistance of the PMA mix than the ASTM mix. Because these two test mixtures follow the same gradation, the difference in their failure criterion lines, which also was observed in their damage characteristic curves (Fig. 3), is due to the modified binder that is used in the PMA mix. The BB5 mix curve is located above the BB3 mix curve and has a lower slope than the BB3 mix, which indicates that mixtures with smaller aggregate particles perform better than ones with larger aggregate particles. Table 4 presents the KEC mixture fatigue performance in terms of the number of cycles to failure for different GR values.

Materials and Structures Table 3 Prony series coefficients for the KEC mixtures

Prony coefficients

Mixtures ASTM

PMA

BB1

BB3

BB5

E0

1.44E?05

1.91E?05

1.60E?05

2.39E?05

1.56E?05

E1

3.52E?03

2.98E?03

9.55E?03

5.98E?03

4.98E?03

E2 E3

3.77E?03 7.90E?03

3.20E?03 6.71E?03

8.41E?03 1.62E?04

6.17E?03 1.27E?04

5.11E?03 1.05E?04

E4

1.69E?04

1.42E?04

3.21E?04

2.66E?04

2.22E?04

E5

3.77E?04

3.09E?04

6.65E?04

5.77E?04

4.90E?04

E6

8.97E?04

7.02E?04

1.47E?05

1.32E?05

1.16E?05

E7

2.35E?05

1.71E?05

3.50E?05

3.24E?05

3.00E?05

E8

6.67E?05

4.48E?05

8.59E?05

8.38E?05

8.27E?05

E9

1.80E?06

1.17E?06

1.93E?06

2.02E?06

2.10E?06

E10

3.67E?06

2.50E?06

3.45E?06

3.69E?06

3.95E?06

E11

4.93E?06

3.75E?06

4.45E?06

4.85E?06

4.91E?06

E12

4.26E?06

3.53E?06

4.20E?06

3.90E?06

4.27E?06

E13

2.70E?06

2.31E?06

3.40E?06

2.41E?06

2.80E?06

E14

1.47E?06

1.26E?06

2.26E?06

1.29E?06

1.60E?06

E15

7.41E?05

6.29E?05

1.41E?06

6.41E?05

8.44E?05

E16

3.61E?05

3.01E?05

8.25E?05

3.09E?05

4.31E?05

E17

1.72E?05

1.41E?05

4.72E?05

1.46E?05

2.16E?05

E18 E19

8.19E?04 3.87E?04

6.57E?04 3.04E?04

2.66E?05 1.48E?05

6.88E?04 3.23E?04

1.07E?05 5.29E?04

E20

1.83E?04

1.40E?04

8.23E?04

1.51E?04

2.61E?04

1.0

10000

ASTM

ASTM

PMA

0.8

PMA

BB1

1000

BB1

BB3

0.6

BB3

C

GR

BB5

BB5

100

0.4

10

0.2 0.0 0.0E+00

2.5E+05

1 1.E+02

5.0E+05

1.E+03

1.E+04

1.E+05

1.E+06

Nf (Cycle)

S

Fig. 3 Averaged damage characteristic curves for KEC test road mixtures

Fig. 4 Fatigue failure criterion lines for the KEC test road mixtures

4.4 Triaxial stress sweep tests

Table 4 Mixture performance for different GR values

The TSS test method requires four tests with two replicates for each test to calibrate the shift model. The TSS test temperatures for the KEC mixtures are specified as follows: 22 C as the low temperature (TL), 36 C as the intermediate temperature (TI), and

Nf (number of cycles to failure) GR

PMA

ASTM

BB1

BB3

BB5

10

53,948

37,867

14,059

9106

25,767

100 1000

10,447 2023

7608 1528

3664 955

2675 786

5471 1162

Materials and Structures

46 C as the high temperature (TH). Details regarding the temperature selection method can be found elsewhere Choi and Kim [9]. Figure 5 presents the results of the TSS tests. The dotted lines show the reference curves and the solid lines correspond to the averaged permanent strains of the MSS tests at each temperature. The first important observation regarding the surface mixtures is the difference in permanent strain levels between the

ASTM and PMA mixtures. The corresponding curves for the ASTM mix indicate higher permanent deformation levels than the PMA mix because the PMA mix contains SBS-modified asphalt binder (PG 76-22). Also, the temperature susceptibility of the PMA mix, which can be evaluated by the amount of increase in the permanent strain from the low to intermediate to high temperatures, is much less than for all the other mixes. These observations provide strong evidence for

2%

2%

(b) KEC-PMA

Reference

Permanent Strain

Permanent Strain

(a) KEC-ASTM

TH

1%

TI

Reference

1%

TH TI

TL

TL

0%

0% 0

100

200

300

400

500

600

0

100

Number of Cycles

300

400

500

2%

(c) KEC-BB1

(d) KEC-BB3 Permanent Strain

TH

1%

TI

Reference

TH

Reference TI

1%

TL

TL 0%

0% 0

100

200

300

400

500

600

0

100

Number of Cycles

(e) KEC-BB5 TH

Reference

1%

TI

TL 0% 0

100

200

300

400

200

300

400

Number of Cycles

2%

Permanent Strain

600

Number of Cycles

2%

Permanent Strain

200

500

600

Number of Cycles

Fig. 5 TSS test results of KEC mixtures: a ASTM, b PMA, c BB1, d BB3, and e BB5

500

600

Materials and Structures

the benefits of polymer modification for rutting resistance. For the base course (BB1 mix and BB3 mix) comparison, the BB1 mix exhibits lower permanent deformation levels than the BB3 mix due to the BB1 mixture’s smaller aggregate particles (25 mm) and lower target air void content than the BB3 mix. The averaged permanent strain values presented in Fig. 5 were used to characterize the shift model. The model coefficients were then applied to the LVECD program to evaluate the rutting performance of a pavement structure.

pavement depth. In order to simulate the pavement temperature in South Korea, the climatic data for Washington, D.C. in the United States were selected for LVECD analysis. 5.1 Fatigue performance In order to evaluate a pavement’s fatigue resistance, the LVECD program calculates the damage growth and the damage factor using Eq. (6) based on Miner’s law. If the damage factor is equal to zero, the element has not experienced any damage. A damage factor of one indicates failure of the element.

5 Pavement analysis Damage factor ¼ In this study, LVECD program simulations were used to evaluate the effects of the pavement design parameters, i.e., the surface layer type, base layer thickness, base layer material, sub-base thickness, and the impact of an anti-frost layer on pavement performance [19]. developed the LVECD program to calculate the stresses and strains throughout the pavement depth. The LVECD program, by combining time-scale separation and layered viscoelastic analysis, uses fast-Fourier transforms to perform threedimensional viscoelastic calculations under moving loads in a rapid manner. The assumption behind using the transforms is that loading occurs in the same repeated manner each time, so the behavior is cyclic and has a steady-state response. The asphalt layer is modeled as viscoelastic material with damage. Therefore, the asphalt layer is represented by the Prony series of the dynamic modulus values, time–temperature shift factors, S-VECD model coefficients, and TSS model parameters. The aggregate base, anti-frost layer, and subgrade were modeled using linear elastic properties with modulus values of 350 MPa, 88 MPa, and 75 MPa, respectively. The other inputs required for the LVECD simulations are design time, pavement structure, traffic, and climate. The design time for this study was assumed to be 20 years. A single tire with standard loading of 80 kN at the center of the pavement was assumed. The average annual daily truck traffic (AADTT) was 935 based on the traffic data obtained from the site. Pavement temperatures were obtained from the Enhanced Integrated Climate Model (EICM) software. The EICM program provides hourly temperatures of asphalt pavements in terms of

T X Ni N i¼1 fi

ð6Þ

where T is the total number of periods, Ni is the traffic for period i, and Nfi is the allowable failure repetitions under the conditions that prevailed in period i. It should be noted that the fatigue performance predicted from the LVECD program has not yet been fully calibrated with the field performance data, and the development of a transfer function to convert the damage in the cross-section of a pavement predicted from the LVECD software to the percentage of surface cracking areas is still in process. To compare the fatigue life of the mixtures, the numbers of failure points (elements with the damage factor of ‘1’) were counted and divided by the total number of elements throughout the section, and the resultant value was specified as an index value for the amount of damage (in percent, %) in the pavement section. To verify the cracking data obtained from the LVECD program, field distress data from [17] were used. In the following subsections the impact of each design parameter is discussed briefly. The effects of the various parameters on fatigue cracking for the 24 pavement sections are presented in Figs. 6, 7 for the simulation results and field-measured results, respectively. 5.1.1 Surface layer type The effect of the PMA mixture as a surface layer on fatigue cracking performance is presented in Fig. 6a. It is noted that Sections A13, A14, and A15 are pavements with an aggregate base layer. The beneficial effect of the PMA mixture on fatigue cracking

Materials and Structures 50

50 ASTM

40

Damaged Area (%)

45

PMA

35 30 25 20 15

Damaged Area (%)

45

Surface Layer Type

(a)

BB3

35

Agg Base

30 25 20 15 10

5

5 A2

A5

A8

A10

A11

A12

A13

A14

BB1

40

10

0

0

A15

Base Layer Type

(b)

8 cm

Section 50

45

8 cm

40

18 cm

35

28 cm

30 25 20 15

Subgrade

35

Anti-Frost

30 25 20 15 10

5

5 BB1

BB3

0

Aggregate

Subgrade Type

(d)

40

10

0

28 cm

50

Base Layer Thickness

(c)

Damaged Area (%)

Damaged Area (%)

45

18 cm

Base Layer Thickness

8 cm

18 cm

Base Layer Type

28 cm

Base Layer Thickness

50 45

(e)

Subbase Layer Thickness 30 cm

Damaged Area (%)

40

40 cm

35 30 25 20 15 10 5 0

8 cm

18 cm

28 cm

Base Layer Thickness

Fig. 6 Effects of different parameters on fatigue cracking performance as obtained from LVECD simulations: a surface layer type, b base layer type, c base layer thickness, d subgrade material, and e sub-base thickness

Materials and Structures 30

30

(a) 25

Base Layer Type

(b)

Surface Layer Type

BB1

25

ASTM

20 15 10

Cracked Area (%)

Cracked Area (%)

BB3 PMA

5

Agg Base

20 15 10 5

0

0 A2

A5

A8

A10

A11

A12

A13

A14

A15

8 cm

18 cm

Section

28 cm

Base Layer Thickness

30

30

(c)

Base Layer Thickness

25

Subgrade Type

(d)

8 cm

25

Subgrade

28 cm

20 15 10

Cracked Area (%)

Cracked Area (%)

18 cm

5

Anti-Frost

20 15 10 5

0

0 BB1

BB3

Aggregate

8 cm

18 cm

Base Layer Type

28 cm

Base Layer Thickness

30

(e)

Subbase Layer Thickness

Cracked Area (%)

25

30 cm 40 cm

20 15 10 5 0 8 cm

18 cm

28 cm

Base Layer Thickness

Fig. 7 Effects of different parameters on pavement fatigue cracking in the field: a surface layer type, b base layer type, c base layer thickness, d subgrade material, and e sub-base thickness

Materials and Structures

performance is demonstrated clearly in these three aggregate base pavements, whereas the effect of the PMA surface layer is much less in the full-depth asphalt pavements. However, the field observations presented in Fig. 7a indicate the beneficial effects of the PMA mixture for all of the pavement sections, regardless of the base layer type, as evidenced by low ‘cracked area’ percentages for all the sections. 5.1.2 Base layer type Figure 6b shows the amounts of damage for the different base layers (BB1, BB3, and aggregate base mixtures). A comparison of the results clearly suggests better performance for the pavements that have an asphalt base layer than the ones that have an aggregate base layer. Between the pavements that have an asphalt base layer, it seems that the BB1 mix, which shows better fatigue resistance based on the GR lines, exhibits less damage than the BB3 mix. This outcome is in strong agreement with the field cracking data, as presented in Fig. 7b. Also, the difference in performance between the BB1 and BB3 mixes may be due to the lower air void content and smaller aggregate size of the BB1 mix.

cracked area’ decreased much less in the pavements with asphalt base layers (i.e., the BB1 and BB3 mixes) when the base layer thickness increased. Figure 7c clearly shows that the reduction in ‘percent cracked area’ (i.e., an increase in fatigue cracking resistance) is not proportional to the increase in the aggregate base layer thickness. In short, the LVECD program simulations did not capture this trend properly. It is noted that the LVECD program treats an aggregate base and subgrade as elastic materials with no damage. This assumption may be the reason for the discrepancy observed between the predicted trend and the observed trend in the field. 5.1.4 Subgrade type The LVECD program analysis results presented in Fig. 6d briefly indicate that the subgrade type (antifrost or subgrade) does not appear to play an important role in the fatigue behavior of the different pavements. In other words, similar pavement performance can be expected for pavements that have either anti-frost or subgrade material. The field observations presented in Fig. 7d also confirm this finding. 5.1.5 Sub-base layer thickness

5.1.3 Base layer thickness Figure 6c shows the same data that are presented in Fig. 6b, except that the data are organized to show the effect of the base layer thickness on the fatigue cracking performance. As expected, the BB3 mix was not as effective in resisting fatigue cracking as the BB1 mix, because the BB3 mix performed the worst among all of the mixtures used in the KEC test road. Another interesting observation that can be made from Fig. 6c is the interaction between two factors: the base layer material type and the base layer thickness. For example, in the case of the aggregate base pavements, with an increase in the base layer thickness from 8 to 28 cm, the damage decreased from 36 percent to 30 percent, whereas the same increase in the BB1 mix layer led to a corresponding decrease in damage from 20 % down to less than 5 %. However, this trend was not found from the field data presented in Fig. 7c. As a matter of fact, the ‘percent cracked area’ reduced from 25 percent to 6 percent when the aggregate base layer thickness changed from 8 cm to 18 cm, whereas the ‘percent

The effects of the sub-base layer thickness are demonstrated in Fig. 6e. In general, the pavements with a 30-cm sub-base layer behave similarly to the ones with a 40-cm sub-base layer. It seems that, unlike the case of the base layer thickness that significantly alters fatigue resistance, the sub-base layer does not have a major impact on the amount of cracking. Although this observed trend was not unexpected, the quantification of the changes in fatigue cracking can be useful for an optimal pavement design. The field data shown in Fig. 7e actually show an unexpected trend; i.e., pavements with a thicker subbase layer performed worse than those with a thinner sub-base layer. It is noted that the ‘percent cracked area’ is relatively small in Fig. 7e, and thus, this unexpected trend may be due to the variability typically found in field experiments. 5.2 Rutting performance In order to help determine rutting resistance, effects of the deviatoric stress, load time, and temperature on

Materials and Structures

permanent deformation of asphalt concrete are calculated using the LVECD program. The deviatoric stress changes throughout the depth of the pavement depending on the load applied. In addition, load time and temperature change along the depth of the pavement. In the LVECD program, each asphalt layer is divided into multiple sublayers, and the permanent strain is calculated for each sublayer based on the deviatoric stress, load time, and temperature determined for that specific sublayer. Then, the permanent strain is multiplied by the sublayer thickness to calculate the permanent deformation in that sublayer. The permanent deformation value that is the sum of the permanent deformation values of all the sublayers is the total surface rut depth. The LVECD rutting simulation results also were compared to the measured rut depths in the field, as described in the following subsections for each parameter and presented in Fig. 8.

Figure 8b presents the comparison of the rut depths of the different base layer types (BB1, BB3, and aggregate base). As shown, the 8-cm aggregate base layer exhibits extreme rutting deformation, whereas an increase in the thickness to 18 cm resulted in a significant decrease in rut depth. However, increasing the thickness up to 28 cm did not lessen the rut depth compared to the 18-cm layer. This outcome led to the conclusion that a thicker aggregate base layer provides additional rutting resistance for a pavement, but the rate of the increase in rutting resistance due to the

Surface Layer Type ASTM PMA

8 6 Measured Predicted

2 0

5.2.2 Base layer type and base layer thickness

20

(a)

4

Figure 8a presents both the measured and predicted rut depths and indicates better rutting resistance of the polymer-modified mixture than the ASTM mixture.

8 cm

18 cm

Total Rut Depths (cm)

Total Rut Depths (cm)

10

5.2.1 Surface layer type

(b)

BB1 BB3 Aggregate

15

10

5

0

28 cm

8 cm

Base Layer Thickness 10

8 6 4 2 0

Subgrade

Anti-Frost

Subgrade Type

Total Rut Depths (cm)

Total Rut Depths (cm)

Subgrade Layer Thickness 20 cm 30 cm 40 cm

18 cm

28 cm

Base Layer Thickness

10

(c)

Base Layer Type

(d)

Base Layer Thickness 8 cm 18 cm 28 cm

8 6 4 2 0

30 cm

40 cm

Subbase Layer Thickness

Fig. 8 Effects of different parameters on rutting performance: a surface layer type, b base layer type, c subgrade type, and d sub-base thickness

Materials and Structures

increase in aggregate base layer thickness decreases as the aggregate base layer becomes thicker. The LVECD program was not able to simulate the poor prediction of the thinner aggregate base layer because the rutting coefficient inputs in the LVECD program were not accurate enough to capture the true performance of the aggregate base and because the material model used for the aggregate base layer in the LVECD program might not be capable of capturing the permanent deformation behavior of a granular aggregate base layer. The asphalt base mixtures (BB1 and BB3) provided better rutting resistance than the aggregate base layers. Overall, the BB1 mix showed better rutting resistance than the BB3 section, as demonstrated in Fig. 8b. 5.2.3 Subgrade layer thickness The subgrade and anti-frost layers constituted the sublayers of the asphalt sections. Three Sects. (1, 3, and 6) were selected among six sections that have no anti-frost layer (2, 4, and 8), and their rut depths were compared against those sections with an anti-frost layer. Figure 8c illustrates the effects (or lack thereof) of an anti-frost layer on rutting performance. In general, no significant correlation between the antifrost layers and rutting was found, even though slightly greater rut depths were apparent at sections that had an anti-frost layer. The LVECD prediction results also show very close agreement with and without anti-frost layers.

base and the subgrade. The rut depths of the unbound materials are modeled using the linear elastic properties proposed by [20], which is implemented in the current Pavement-ME software. However, it is possible that this model for unbound materials is inadequate for capturing the relative magnitude of permanent deformation as compared to models for asphalt layers. Permanent deformation from each layer using trench cut of pavement structures is needed to verify the permanent deformation model for unbound materials.

6 Conclusions In this study, the effects of common pavement design parameters on asphalt pavement performance have been investigated through laboratory-produced asphalt mixture testing, numerical simulations, and field verification. The fatigue cracking and rutting performance of 24 asphalt pavement KEC test sections with different structures were evaluated using validated models. The S-VECD model was used to evaluate the fatigue properties of the mixtures, and TSS tests were performed to assess the rutting behavior of the mixtures. The results were then input in the LVECD program to predict the long-term performance of the test pavement sections. A summary of the findings and the conclusions that can be drawn from this study are as follows: •

5.2.4 Sub-base layer thickness Figure 8d presents the effect of the sub-base layer thickness on the rutting performance of sections without an anti-frost layer. Although the rut depth values in the 30-cm base layer are slightly greater than those in the 40-cm base layer, no significant relationship is evident in terms of sub-base layer thickness. The LVECD prediction results show that the rut depths of the 30 and 40-cm sub-base layers are the same. The rut depths increase with an increase in base layer thickness because the deformation calculated in a given layer is the permanent strain multiplied by the thickness. The total rut depth is the sum of the permanent deformations of all the rut-susceptible layers in a pavement, including the unbound materials for the

•

•

As expected, utilization of the mixtures that contained modified binders improved the pavement performance. The sections with the PMA mix exhibited less cracking and permanent deformation than the ASTM sections due to the lower stiffness values at low temperatures and higher stiffness values at high temperatures of the PMA mix. However, this improvement was not captured well by the LVECD program for the full-depth asphalt pavements. One interesting observation is that the base layer material played the most important role in affecting both rutting and fatigue cracking in the KEC test road sections. The aggregate base layer exhibited greater rut depths and more cracking than the asphalt base layers. The data obtained from the pavement analyses of the different base layer thicknesses briefly indicate that an increase in the base layer thickness reduces

Materials and Structures

•

•

the pavement distresses, but the amount of improvement is dependent on the material type. According to the observations made from this study, increasing the aggregate base layer thickness from 8 to 28 cm can provide up to 70 % reduction in pavement distress. The pavement simulations indicate no significant change in the fatigue resistance or permanent deformation when using an anti-frost layer or increasing the sub-base layer thickness. The field observations are in strong agreement with these findings. The key point that has emerged from the comparison of the field measurements and the simulation predictions is that the pavement performance rankings, which are based on the damage area (%) index and the total rut depths predicted from the LVECD program, are generally in good agreement with the field observations. However, the magnitudes of these simulated results do not correspond directly to the observed field measurements. This finding suggests the need for developing a laboratory-tofield transfer function. The need for such a function is not unexpected for a numerical simulation program such as the LVECD software.

Acknowledgments This research is sponsored by the Federal Highway Administration under project No. DTFH61-08-H00005. The authors gratefully acknowledge the support of the FHWA. Also, the authors would like to thank the Korea Expressway Corporation for providing the original materials used in the KEC test road pavement sections, as-constructed information, and performance data. The KEC’s help was essential to this project.

References 1. Behiry AE (2012) Fatigue and rutting lives in flexible pavements. Ain Shams Eng J 3:367–374 2. Chehab GR, Kim YR, Schapery RA, Witczak M, Bonaquist R (2003) Characterization of asphalt concrete in uniaxial tension using a viscoelastoplastic model. J Assoc Asphalt Paving Technol 72:315–355 3. Hou T, Underwood BS, Kim YR (2010) Fatigue performance prediction of north carolina mixtures using simplified viscoelastic continuum damage model. J Assoc Asphalt Paving Technol 79:35–80 4. Underwood BS, Baek CMN, Kim YR (2012) Use of simplified viscoelastic continuum damage model as an asphalt concrete fatigue analysis platform. Transp Res Record: J Transp Res Board 2296:36–45

5. Underwood BS, Kim YR, Guddati MN (2010) Improved calculation method of damage parameter in viscoelastic continuum damage model. Int J Pavement Eng 11:459–476 6. Zhang J, Sabouri M, Kim YR, Guddati MN (2013) Development of a failure criterion for asphalt mixtures under fatigue loading. Road Mat Pavement Design 14(Supplement 2):1–15 7. Sabouri M, Kim YR (2014) Development of failure criterion for asphalt mixtures under different modes of fatigue loading. Transp Res Record: J Transp Res Board 2447:117–125 8. Norouzi A, Sabouri M, Kim YR (2014) Evaluation of the fatigue performance of high rap asphalt mixtures. Proceedings of 12th international society for asphalt pavements, Raleigh 9. Choi Y, Kim YR (2012) Development of calibration testing protocol for permanent deformation model of asphalt concrete. Transp Res Record: J Transp Res Board No. 13-2555: 34–43 10. Kim YR, Guddati M (2011) Hot mix asphalt performancerelated specifications based on viscoelastoplastic continuum damage (VEPCD) Models. Quarterly Research Progress Report: October–December 2011, Project Number DTFH61-08-H-00005 11. Gibson NH, Kutay ME, Keramat D, Youtcheff J (2009) Multiaxial strain response of asphalt concrete measured during flow number performance test. J Assoc Asph Paving Technol 78 12. Huang YH (2003) Pavement analysis and design. 2nd edn. Englewood Cliffs, Prentice-Hall 13. Eslaminia M, Guddati MN (2010) Fourier-finite element analysis of pavements under moving vehicular loading. Int J Pavement Eng 14. Norouzi A, Kim YR (2014) Mechanistic evaluation of the fatigue cracking in asphalt pavements. Int J Pavement Eng 1–17 15. Wang Y, Norouzi AH, Kim YR (2015) Comparison of fatigue cracking performance predictions in asphalt pavements using pavement ME and LVECD. Transp Res Record: J Transp Res Board 1–17 16. Reese R (1997) Properties of aged asphalt binder related to asphalt concrete fatigue life. J Assoc Asphalt Paving Technol 66:604–632 17. Seo Y (2010) Distress evolution in highway flexible pavements: a 5-year study at the Korea Highway Corporation Test Road. ID JTE 102107. J Test Eval 38(1): 1–10 18. Lee JS, Norouzi A, Kim YR (2014) Determining specimen geometry of cylindrical specimens for direct tension fatigue testing of asphalt concrete. ASTM J Test Eval 19. Eslaminia M, Thirunavukkarasu S, Guddati MN, Kim YR (2012) Accelerated pavement performance modeling using layered viscoelastic analysis. Proceedings of the 7th international RILEM conference on cracking in pavements, Delft 20. Tseng K, Lytton R (1989) Prediction of permanent deformation in flexible pavement materials. Implic Aggreg Design Const Perform Flex Pavements. ASTM STP 1016:154–172