Effect of Compaction Mode on the Mechanical Performance and Variability of Asphalt Mixtures

Dow

nloa

ded

from

asc

elib

rary

.org

by

GE

OR

GE

WA

SHIN

GT

ON

UN

IVE

RSI

TY

on

05/1

4/13

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.

Effect of Compaction Mode on the Mechanical Performanceand Variability of Asphalt MixturesAlistair E. Hunter1; Liam McGreavy2; and Gordon D. Airey3

Abstract: This paper investigates the relationship between mode of asphalt mixture compaction and resulting mechanical performancein terms of stiffness modulus, permanent deformation and fatigue resistance. Four modes, comprising three laboratory methods ofgyratory, vibratory, and roller compaction together with site compaction were included in the study. An interlaboratory program wasundertaken using these compaction methods and three asphalt mixture designs. The results showed that mold based compaction methods,such as gyratory and vibratory, generally produce stiffer specimens with higher resistance to permanent deformation when compared toroller compacted or field specimens of comparable air voids. However, the results also showed significant variability within a particularcompaction method between different laboratories. The statistical two-sample t-confidence test was used to compare the sample means ofair voids content and stiffness modulus between the different compaction methods. The statistical analysis showed that producingvolumetrically identical asphalt mixture specimens using different compaction methods does not produce materials with identical me-chanical properties. Statistical analysis was also undertaken on stiffness and permanent deformation results to provide engineers with thenumber of tests required to obtain a mean value within a certain accuracy of the “true” value, for a given confidence level. Stiffnessmodulus was found to have a far higher repeatability than permanent strain with the mode of compaction not affecting the variability ofthe results.

DOI: 10.1061/�ASCE�0733-947X�2009�135:11�839�

CE Database subject headings: Asphalts; Mixtures; Compaction; Stiffness; Deformation; Statistics.

Introduction

Achieving high quality site compaction is essential if superiordurability and performance is to be achieved from a road. Manyfactors influence the final quality of the compaction. Critically theasphalt mixture needs to be mixed at the correct temperature toensure that the aggregate particles are fully coated. Transportationof the hot mix to site needs to be undertaken using insulatedwagons to ensure the compaction temperature is within the suit-able range. Transportation times need to be limited to avoid ex-cessive aging of the bitumen while at high temperatures. Thecompaction equipment used on site needs to be appropriate forthe specific asphalt type and, finally, the compaction equipmentneeds to be in a position to compact immediately behind the paverwhile the asphalt mixture mat is still hot.

All of the aforementioned factors are well recognized and maybe managed successfully if the relevant standards are followedand good practice adhered to. The factors which are more difficultto manage, and generate inherent variability in terms of internal

1Associate, Scott Wilson Ltd, 12 Regan Way, Nottingham NG9 6RZ,U.K. E-mail: [email protected]

2Researcher, Nottingham Transportation Engineering Centre, Univ. ofNottingham, University Park, Nottingham NG7 2RD, U.K.

3Professor, Nottingham Transportation Engineering Centre, Univ. ofNottingham, University Park, Nottingham NG7 2RD, U.K. E-mail:[email protected]

Note. This manuscript was submitted on July 5, 2007; approved onJuly 6, 2009; published online on October 15, 2009. Discussion periodopen until April 1, 2010; separate discussions must be submitted forindividual papers. This paper is part of the Journal of TransportationEngineering, Vol. 135, No. 11, November 1, 2009. ©ASCE, ISSN 0733-
947X/2009/11-839–851/$25.00.
JOURNAL OF TR

J. Transp. Eng. 2009.

asphalt mixture structure and resulting mechanical performanceare aggregate gradation changes, aggregate segregation, differen-tial temperature within the mat, and changes in the degree ofconfinement across the width of the mat. Aggregate gradationvaries depending on production at the quarry and asphalt mixtureplant. Aggregate segregation occurs when the delivery wagonstransfer their loads into the pavers, with the heavier aggregateparticles tending to run to the front. Segregation also becomesmore prominent as the maximum nominal particle size of theaggregate increases. Finally, the degree of confinement, deter-mined by the mat boundaries, limits the compactive effort whichmay be applied to the asphalt mixture.

Many of the factors responsible for variability on site, such astemperature and segregation, can be controlled to a greater extentwithin a laboratory environment, where the production of the as-phalt mixture is on a much smaller scale. Unfortunately, withinlaboratory conditions a new problem presents itself, namely themode of laboratory compaction used to manufacture the asphaltmixture which has a significant influence on the mechanical per-formance of the compacted specimens. The fact that differentmodes of compaction create volumetrically identical but mechani-cally different specimens has been well documented �Vallerga1951; Nevitt 1959�. In addition, this subject has received substan-tial attention in recent years, at least in part due to the StrategicHighways Research Program �SHRP� recommending the use ofthe gyratory compactor as the preferred means of asphalt mixturecompaction �Sousa et al. 1991; Harvey and Monismith 1993�.However, many of the studies undertaken on this subject havereached contradictory conclusions as to which laboratory methodof asphalt mixture compaction best emulates site compaction, andthe strengths of the respective methods. For example, Gibb �1996�
found that in terms of permanent deformation, vibratory com-
ANSPORTATION ENGINEERING © ASCE / NOVEMBER 2009 / 839

135:839-851.

Dow

nloa

ded

from

asc

elib

rary

.org

by

GE

OR

GE

WA

SHIN

GT

ON

UN

IVE

RSI

TY

on

05/1

4/13

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.

pacted specimens generally produced results which were closer tosite cores than steel roller compacted specimens. In contrast,Hartman et al. �2001� compared the indirect stiffness values ofspecimens manufactured from five laboratory methods of com-paction �steel roller, Marshall, vibrating hammer, static compres-sion, and gyratory� with site cores and found that steel rollercompacted specimens were of comparable stiffness with those ofthe site cores. Sousa et al. �1991� also recommended the use ofthe steel roller compactor, whereas Consuegra et al. �1989� andVon Quintas et al. �1991� recommended the use of the gyratorycompactor.

Direct comparison of these studies is difficult for two mainreasons. First, they assess the asphalt mixture performance usingdifferent mechanical tests aimed at assessing different mechanicalproperties, for example, stiffness versus fatigue life. Even withina given physical parameter, for example permanent deformation,there are several different tests and countless test parameterswhich can be used. Second, there is large scope for differencewithin any one generic compaction method. This point was high-lighted in the study by Harvey and Monismith �1993� where twotypes of gyratory compactor �Texas and SHRP� created speci-mens with very different resistances to permanent shear deforma-tion. Despite the different conclusions reached by the variousstudies, there is a general consensus that the mechanical param-eter most sensitive to mode of compaction is permanent deforma-tion. Some studies suggest stiffness is affected by the mode ofcompaction and most studies suggest that the fatigue life is unaf-fected, at least when measured by strain based criteria. The threemechanical tests undertaken in this study were the indirect tensilestiffness modulus �ITSM� test �British Standards Institution �BSI�1993�, the indirect tensile fatigue test �ITFT� �BSI 1995� and therepeated load axial test �RLAT� �BSI 1996�. These were chosen torepresent the most important aspects of asphalt mixture behavior.With this background, the main objective of this paper was toquantify the effect of different compaction modes on the me-chanical properties of asphalt mixture specimens of comparableair voids content. Four modes of compaction �gyratory, vibratory,roller, and site� and three commonly used asphalt mixture typeswere investigated. A statistical assessment was undertaken to de-termine similarities in the volumetric characteristics and mechani-cal properties between the laboratory compaction methods. Thelaboratory compaction data were also compared with field data toanalyze the predictive capabilities of the laboratory specimens.Finally, the variability of the results were analyzed in terms of thenumber of tests required to obtain a mean value within a specified

Table 1. Experimental Program

Material Aggregate Bitum

Base 28-mm continuously graded, limestone Venezuelan

Binder course 20-mm continuously graded, limestone Venezuelan

Surfacing 14-mm gap graded, granite SBS P

Note: nd�nondestructive; d�destructive; G�gyratory; V�vibratory; R�

accuracy of the “true” value for the ITSM and RLAT tests as a

840 / JOURNAL OF TRANSPORTATION ENGINEERING © ASCE / NOVEMB


function of compaction method and asphalt mixture type. This ispotentially useful information for both contractors and clientswho are interested in measuring the mechanical performance of aroad against contractually specified criteria. To calculate the re-quired testing numbers, confidence limit charts were generatedusing the established student t method, adopting a similar ap-proach to that used by Collop et al. �2001�. The testing was run asan interlaboratory program, ensuring the results were representa-tive of the U.K. industry as a whole.

Experimental Testing Program

To maximize the usefulness of the experimental testing program,a wide range of mixtures were covered, including a typical base,binder course, and surfacing “polymer modified” mixture as de-scribed in Table 1. The mixtures were carefully chosen to repre-sent three nominal maximum aggregate sizes and a range ofbituminous binders. In general, the risk of segregation and non-uniform compaction increases with increasing nominal maximumaggregate size. The modes of laboratory compaction were chosento represent those commonly used in the United Kingdom. Wherepossible a total of 30 specimens were tested for each compactionmethod to provide a reasonable number for statistical analysiswhile being practicable in terms of laboratory time.

Material from each of the three asphalt mixtures was distrib-uted to the laboratories involved in the compaction study. Afterthe specimens were manufactured, they were collected and testedat one laboratory �Lab 1�. In this way, the variability of testing atdifferent laboratories was removed and differences in the speci-mens could be attributed directly to the mixing and compactionprocess. The sites which provided the field cores consisted of thesame constituent materials as those used in the interlaboratorytesting program.

Compaction Methods

Gyratory CompactorTwo types of gyratory compactor were used in the study consist-ing of a “Cooper Research Technology” compactor manufacturedin the United Kingdom and used in this study by four laboratories�Lab 1 to Lab 4�, and the French “Laboratoire Central des Pontset Chaussees �LCPC�” gyratory compactor used by Laboratory 5.The gyratory compaction parameters used in the interlaboratory

Testing program Compaction mode and sample size Lab

n Air voids �nd� G �30�, V �30�, R �30� Lab 1

ITSM �nd� G �30� Lab 2

ITFT �d� G �30� Lab 3

G �30� Lab 4

G �30� Lab 5

n Air Voids �nd� G �30�, V �30�, R �30� Lab 1

ITSM �nd� R �30� Lab 4

RLAT �d� S �30�

Air Voids �nd� G �15�, V �15� Lab 2

ITSM �nd� S �18�

RLAT �d�

and S�site.

en

50 pe

35 pe

MB

roller;

testing program are listed in Table 2. The gyratory compaction

ER 2009

135:839-851.

Dow

nloa

ded

from

asc

elib

rary

.org

by

GE

OR

GE

WA

SHIN

GT

ON

UN

IVE

RSI

TY

on

05/1

4/13

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.

parameters represent the actual values used in the various labora-tories and hence the results reflect the true variation occurringwithin U.K. testing laboratories. The specimens were all com-pacted to a target density, as opposed to a set number of gyrations.The base asphalt mixture was compacted in 150-mm diametermolds, whereas the binder course and surfacing mixtures werecompacted in 100-mm diameter molds. During the compactionprogram, silicon spray was used on the inside of the molds to easesubsequent specimen extraction. The mixing and compactiontemperatures were carefully controlled for all compaction modesto ensure high quality compaction and appropriate bitumen vis-cosities in line with the relevant standards �BS EN 12697-31, 32,33 and 35 �BSI 2003a,b, 2004a,b��.

Vibratory CompactorAn electric vibratory hammer was use to compact the specimensin a split mold of internal diameter 152.45 mm�0.5 mm. Thepower consumption of the vibratory hammer was 800 W andoperated at a frequency of 35 Hz. Very similar vibratory hammerswere used at both laboratories �Lab 1 and Lab 2� involved in thispart of the testing program. Consistently achieving a designatedtarget density using vibratory compaction is difficult and as aresult there is some variance in the air void contents of thesespecimens.

Steel Slab Roller CompactorAn identical roller compactor was used by all the laboratoriesinvolved in this study. The roller �slab� compactor consisted of acurved steel segment, pivoting on a hinge and applying the loadvia pneumatic actuators. The asphalt mixture was placed in a steelmold �internal dimensions of 300 mm by 300 mm by 140 mm�which moved back and forth beneath the roller. The partial freeface allowed the aggregates to orient themselves in a mannersimilar to that which occurs on site �Consuegra et al. 1989�. Theprecise depth of the compacted asphalt layer was set at the start ofthe test based on knowledge of the maximum density of the mix-ture and mass of the slab to enable a target air voids content to bespecified. The direct force applied to the slab was not measured,but complied with BS EN 12697-33 �BSI 2003b�, which specifiesa minimum static load, F, that the device should apply

F

2Dl� 10−5 �1�

where F=load applied onto the roller �kN�; l=interior width ofthe mold �mm�; and D=diameter of the wheel or roller �mm�. Thedimensions of resulting slabs were 300 mm�300 mm�100 mm. This enabled two 150-mm diameter specimens orfour 100-mm diameter specimens to be cored from each slab.

Site CompactionThe initial compaction for the 20-mm binder course asphalt mix-

Table 2. Gyratory Compaction Parameters Used in Interlaboratory Study

LaboratoryExternal gyration angle

�°�Compaction pressure

�kPa� RPM

Laboratory 1 1.25 600 30




Laboratory 5 55� 662 30

ture was carried out by means of a dead weight 12/14 t three-

JOURNAL OF TR


wheeled roller. Secondary compaction was undertaken by aBomag 161, with a vibration affected weight of 24 t. The asphaltlaying temperature was around 150°C, with an air temperature ofaround 8°C. For the 14-mm surfacing asphalt mixture, the initialcompaction was undertaken by a 12 t static dual roller HAMMcompactor. Secondary compaction was undertaken by a Bomag120 2.5 t tandem vibratory roller. The laying temperature wasaround 145°C with an air temperature of 13°C.

Mechanical Property Tests

Indirect Tensile Stiffness ModulusThe stiffness moduli of the asphalt mixture specimens were mea-sured using the ITSM test. The test is nondestructive and involvesthe application of load pulses to the vertical diameter of a cylin-drical specimen, with the resultant peak transient deformationmeasured along the horizontal diameter. The stiffness modulus isthen a function of load, deformation, specimen dimensions, andan assumed Poisson’s ratio of 0.35. The ITSM test was under-taken in accordance with recognized guidelines �BSI 1993� usingthe following test parameters:• Test temperature 20°C.• Loading rise-time 124 ms.• Peak transient horizontal deformation 7 �m �150-mm diam-

eter specimen�, 5 �m �100-mm diameter specimen�.The test specimen is initially conditioned through the applicationof five load pulses. A further five load pulses are then appliedfrom which a mean stiffness modulus is obtained. The sample isthen rotated through 90° and a further five pulses are applied anda resulting mean stiffness modulus obtained. The stiffness modu-lus of the asphalt mixture specimen is then calculated as theaverage of these two mean values.

Repeated Load Axial TestThe resistance of the asphalt mixture specimens to permanentdeformation was determined by means of the RLAT using a directuniaxial compression testing configuration. The test consists ofapplying a number of load pulses to the flat faces of the specimenand recording the resulting deformation. The axial strain obtainedat the end of the test is used as a measure of the specimen’sresistance to permanent deformation. The RLAT testing was un-dertaken in accordance with recognized guidelines �BSI 1996�and the following test parameters:• Test temperature 30°C.• Test duration 7200 s �3600 cycles�.• Loading pattern 1 s loading followed by a 1 s recovery period

per cycle.• Axial stress 100 kPa.• Conditioning stress 10 kPa for 120 s.

The permanent deformation performance of the asphalt mix-tures was quantified by means of the ultimate percentage strainafter 3,600 cycles, although the rate of strain �microstrain percycle� over the linear phase of the deformation response can alsobe used �Brown and Gibb 1996�.

Indirect Tensile Fatigue TestThe fatigue life of the asphalt mixture specimens was assessedusing the ITFT with an experimental arrangement similar to thatused for the ITSM test but under repeated loading conditions. Arange of stress levels were chosen and the number of pulses to
failure �defined as the point at which there is 9 mm of vertical

135:839-851.

Dow

nloa

ded

from

asc

elib

rary

.org

by

GE

OR

GE

WA

SHIN

GT

ON

UN

IVE

RSI

TY

on

05/1

4/13

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.

deformation� was recorded. The range of stress levels should en-sure a wide range in fatigue lives with the test being carried outusing the following test parameters �BSI 1995�:• Test temperature 20°C.• Loading condition: controlled-stress.• Rise time 124 ms.• Pulse repetition 1.5 s�0.1 s �40 pulses/min�.• Range of tensile stress 200 to 700 kPa.

The maximum tensile strain generated at the center of thespecimen is defined as

�x max =�x max�1 + 3��

Sm� 1000 �2�

where �x max=maximum tensile horizontal strain at the center ofthe specimen in microstrain; �x max=maximum tensile stress at thecenter of the specimen in kPa; �=Poisson’s ratio �assumed to be0.35�; and Sm�ITSM at �x max in megapascals. The maximumtensile stress at the center of the specimen is defined as

�x max =2L

dt�3�

where d=diameter of the test specimen �m� and L=vertical ap-plied line load in kN. Linear regression analysis of the ITFTresults was used to determine fatigue functions for the asphaltmixtures using the following relationship:

Nf = a�0−b �4�

where Nf =fatigue life; �0= initial tensile strain �microstrain�; anda, b=experimentally determined coefficients.

Table 3. Volumetric and Stiffness Modulus Results for 28-mm DBM As

Properties

Gy

Lab 1 Lab 2 La

Average air voids �%� 3.1 2.8

Standard deviation air voids �%� 0.4 0.4

CoV air voids �%� 13 13

Average stiffness �MPa� 7,906 6,349 6,7

Standard deviation stiffness �MPa� 596 799 4

CoV stiffness �%� 8 13

Table 4. Volumetric and Stiffness Modulus Results for 20-mm DBM Bi

Properties

20-mm b

Gyro Vibro

Lab 1 Lab 1

Average air voids �%� 4.0 2.9

Standard deviation air voids �%� 1.7 0.9

CoV air voids �%� 42 32

Average stiffness �MPa� 10,600 10,288

Standard deviation stiffness �MPa� 882 1,208

CoV stiffness �%� 8 12



Mechanical Property Results

Asphalt Mixture StiffnessThe main objective of the testing program was to establish whichmechanical parameters were most sensitive to mode of compac-tion. The volumetric proportions �air void contents� together withthe stiffness results from the ITSM are presented in Tables 3 and4 for the 28-mm Dense Bitumen Macadam �DBM�, 20-mm DBM,and 14-mm surfacing materials. The average, standard deviationand coefficient of variation were recorded for the different com-paction methods as well as the different laboratories. In general,the results of the air voids and stiffness moduli for the 28-mmDBM asphalt mixture are relatively consistent with low coeffi-cients of variation within the range of 8% to 15% for the air voidsand 7–13% for stiffness modulus. The coefficients of variation ofthe air voids for the smaller aggregate size 20-mm DBM and14-mm surfacing asphalt mixtures are surprisingly high rangingfrom 2% to 45% and 19 to 30%, respectively, although the coef-ficients of variation for stiffness moduli were comparable to thatof the 28-mm DBM ranging from 6 to 13%. The increase vari-ability associated with the air void contents of the 20- and 14-mmasphalt mixture specimens indicates a less consistent compactionoutcome �increased compaction variability� compared to the28-mm asphalt mixture. This would appear to be counterintuitive,although it is important to note that, first, the binders used in the20- and 14-mm mixtures were either stiffer or polymer modifiedcompared to the 28-mm mixture and, second, the mold and finalspecimen size was reduced from 150- to 100-mm in diameterfrom the 28-mm to smaller sized asphalt mixtures. These twofactors, together with observations that the cores from field com-paction tended to produce high coefficients of variation and that

ixture

Compaction method

Vibratory Roller

Lab 4 Lab 5 Lab 1 Lab 1 Lab 1

5.2 1.9 2.8 9.6 3.0

0.6 0.2 0.2 1.4 0.4

12 9 8 15 14

8,081 11,165 10,088 6,362 7,322

583 850 932 628 540

7 8 9 10 7

ourse and 14-mm Surfacing Mixtures

Asphalt mixture

course 14-mm surfacing

Compaction method

Roller

Site

Gyro Vibro

SiteLab 4 Lab 2 Lab 2

1.3 1.7 3.4 2.2 2.8

0.4 0.8 0.6 0.7 0.6

29 45 19 30 21

6,260 5,475 2,889 2,104 1,565

624 586 228 252 196

10 11 8 12 13

phalt M

ratory

b 3

3.6

0.4

11

22

65

7

nder C

inder

Lab 1

2.7

0.9

33

8,172

493

6

ER 2009

135:839-851.

Dow

nloa

ded

from

asc

elib

rary

.org

by

GE

OR

GE

WA

SHIN

GT

ON

UN

IVE

RSI

TY

on

05/1

4/13

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.

the sample size for the 14-mm surfacing was reduced from 30 to15, probably explain the increased variability associated with the20- and 14-mm asphalt mixtures.

The mean stiffness moduli as a function of mean air voids forthe 28-mm asphalt mixture are plotted in Fig. 1. The results showa high degree of variability in terms of air voids and stiffnessmoduli between the different laboratories as well as betweencompaction methods. In terms of the gyratory compacted speci-mens produced with the U.K. compactor �Lab 1 to Lab 4�, themean air voids and stiffness moduli are comparable with thoseobtained from the roller compactor, although the gyratory com-pacted specimens produced by Lab 4 had a higher air voids con-tent. The fifth gyratory compaction laboratory using the FrenchLCPC compactor �Lab 5� produced specimens with a significantlyhigher stiffness modulus. The reason for this difference betweenthe two gyratory compaction types is difficult to ascertain, withthe most significant difference being the small gyratory angle of55� �1°� used with the French gyratory compactor compared tothe larger angles of 1.25 and 1.50° of the U.K. compactor. Therelationship between stiffness modulus and gyratory angle is com-plex and not fully understood with Peterson et al. �2004� reportingthat the Texas gyratory machine operating at an angle of 6° pro-duced specimens with lower stiffness modulus than specimensproduced by the Superpave gyratory compaction �SGC� operatingat an angle of 1.25°. However, they also showed an increase instiffness modulus when the gyratory angle of the SGC was in-creased from 1.5 to 2°. The vibratory compacted specimens �Lab1� had a higher mean stiffness modulus compared to the roller andU.K. gyratory compactors but a similar value to the LCPC gyra-tory compacted specimens over the air voids range of 1.9–3.6%.It is interesting to note that a second set of vibratory compactedspecimens at a considerably higher air voids content of 9.6%produced a mean stiffness modulus similar to that obtained for theroller and U.K. gyratory compacted asphalt mixtures at air voidcontents of 2.8–3.6%.

Fig. 2 shows the ITSM results for the 20-mm DBM asphaltmixture. In this case the roller compacted specimens show thelowest stiffness out of the laboratory compacted specimens, withthe mold based methods �gyratory and vibratory� providinghigher stiffness at higher air voids. There is also a substantialdifference between the two roller compacted mean stiffnessmoduli, especially when the difference in air voids content isconsidered �Lab 4 producing lower stiffness specimens, althoughthey are considerably denser than those produced by Lab 1�. As

4,000

5,000

6,000

7,000

8,000

9,000

10,000

11,000

12,000

1 2 3 4 5 6 7 8 9 10

Air Voids (%)

StiffnessModulus(MPa)

Gyratory Lab #1Gyratory Lab #2Gyratory Lab #3Gyratory Lab #4Gyratory Lab #5Vibratory Lab #1Roller Lab #1

Fig. 1. Mean stiffness modulus against mean air voids for differentcompaction modes—28-mm DBM asphalt mixture

the materials were batched from the same source, had mixing and

JOURNAL OF TR


compaction temperatures within 5°C for the two laboratories, andused the same type of roller compactor, this difference is probablydue to the different compaction procedures followed by the twolaboratories. As described above, the roller compactors havepneumatic actuators that are used to apply compaction pressurethrough the curved steel segment onto the asphalt mixture slab.As the slabs were produced to a target density �air voids� andtherefore a target height of compacted slab, the selection of com-paction pressures and roller passes is usually left to the discretionof the operator. Finally, the stiffness results of the site compactedcores were similar to those obtained from the roller compactedspecimens, particular those produced by Lab 4.

Fig. 3 shows the complete set of volumetric and ITSM resultsfor the 14-mm asphalt mixture. In this plot the vibratory and sitecompacted specimens are of comparable stiffness if the differencein air void content is accounted for �increased air voids for sitecores�. The gyratory compacted specimens, which are of similarair voids content to the site compacted specimens, have almostdouble the mean stiffness modulus. The results from all threeplots �Figs. 1–3� show a degree of scatter with no one laboratorycompaction method consistently producing either extremely highor low stiffness modulus specimens. However, in general themold based compaction methods tend to produce specimens ofhigher stiffness than the roller �slab� and site compacted speci-mens.

4,000

5,000

6,000

7,000

8,000

9,000

10,000

11,000

1 1.5 2 2.5 3 3.5 4 4.5 5

Air Voids (%)

StiffnessModulus(MPa)

Gyratory Lab #1Vibratory Lab #1Roller Lab #1Roller Lab #4Site

Fig. 2. Mean stiffness modulus against mean air voids for differentcompaction modes—20-mm DBM asphalt mixture

500

1,000

1,500

2,000

2,500

3,000

3,500

4,000

1 1.5 2 2.5 3 3.5 4 4.5 5

Air voids (%)

Stiffnessmodulus(MPa)

Gyratory Lab #1

Vibratory Lab #1

Site

Fig. 3. Stiffness modulus against air voids for different compactionmodes—14-mm surfacing asphalt mixture


135:839-851.

Dow

nloa

ded

from

asc

elib

rary

.org

by

GE

OR

GE

WA

SHIN

GT

ON

UN

IVE

RSI

TY

on

05/1

4/13

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.

Assuming the mixture design, mixing procedures, compactiontemperatures, and volumetrics are identical, the reasons for thedifferent stiffness results can be linked to the internal structure ofthe asphalt mixture matrix. Previous research, using image analy-sis and X-ray tomography techniques, has established that a cir-cumferential aggregate orientation occurs in mold basedcompaction �Hunter et al. 2004�, as well as a distribution of ag-gregate sizes between the inside and outside of the specimen�Tashman et al. 2001, 2002; Masad et al. 1999, 2002�. Thesestudies have also revealed a “bath tub” air voids distribution ingyratory compacted specimens i.e., far higher air voids content atthe surfaces in contact with the loading platens. In contrast, linearkneading �vibratory� compactors produce a near linear increase ofair voids content with depth. These distinct particle arrangementswithin a specimen will have a direct influence on the microme-chanical properties as manifested through variations in stiffnessmodulus between compaction methods.

Asphalt Mixture Permanent Deformation

The permanent deformation results for the 20-mm DBM and14-mm surfacing mixtures are presented in Table 5 in terms of theaverage, standard deviation and coefficient of variation of thepermanent strain after 3,600 repeated load cycles. As expected,the coefficients of variation for the mixtures are greater than thosefound for their stiffness modulus results but can nevertheless beaccepted as reasonable. On average there is less variability of theresults for the 14-mm compared to the larger 20-mm asphalt mix-ture.

The permanent deformation results obtained from the RLATfor the 20-mm asphalt mixture are shown in Fig. 4 as a functionof air voids content. The results, if viewed as a whole, show anincrease in the resistance to permanent deformation of the speci-mens with an increase in air voids, leveling out at approximately3.5% air voids content. Gibb �1996� reported a similar relation-ship is his study where poor resistance to permanent deformationwas observed in both vibratory compacted specimens and fieldcores with low air voids, typically below 3% for a range of U.K.asphalt mixtures. He hypothesized that this was due to insufficientvoid space to accommodate the bitumen/fines mortar, resulting inloss of frictional contact between aggregate particles under load-ing. Further more, data relating to continuously graded DBM ma-terials indicated an “optimum void content range” typically 3–8%within which the asphalt mixture performance remained relativelyconstant. Above 8–10% air voids content there is a marked reduc-tion in resistance to permanent deformation. Another interestingobservation from this study was that at high air voids ��10%�

Table 5. Permanent Deformation Results for 20-mm DBM Binder Cour

Properties

20-mm bin

Gyro Vibro

Lab 1 Lab 1 La

Average air voids �%� 4.0 2.9 2

Average permanent strain �%� 0.40 0.32 0

Standard deviation strain �%� 0.13 0.09 0

CoV permanent strain �%� 33 28 16

there is little difference in permanent deformation performance



between laboratory and field compacted specimens indicating thatdegree of compaction is more significant than the method of com-paction �Gibb 1996�.

Looking at the different compaction methods, the roller com-pacted specimens �Lab 4� and the site specimens are comparableboth in terms of axial strain �mean axial strain of 1.26 and 1.41%,respectively� and air voids �mean air voids of 1.3 and 1.7%, re-spectively�. In comparison, the roller compacted specimens fromLab 1 have a higher air voids content �2.7%� and significantlylower axial strain �0.64%�. The difference between the two sets ofroller compacted specimens follows the trend described by Gibb�1996�. There is a further increase in the resistance to permanentdeformation for the gyratory �average permanent strain of 0.40%�and vibratory �average permanent strain of 0.32%� compactedspecimens. It should be noted that the gyratory compacted speci-mens have a higher mean air voids content �4.0%� than the rollerand vibratory compacted specimens �2.7 and 2.9%, respectively�.In general, the data suggests that mold based compaction methods�gyratory and vibratory� produce specimens with greater resis-tance to permanent deformation compared to roller compaction.This can be clearly seen over the air voids range of 2–4% for thegyratory, vibratory and roller compacted asphalt mixtures speci-mens produced by Lab 1 in Fig. 4.

As there is a considerable degree of variability in terms of theaverage air voids content of the gyratory �3.4%�, vibratory�2.2%�, and site �2.8%� 14-mm asphalt mixtures, no direct com-parison of the permanent deformation results is possible. How-ever, the results in Table 5 do show a large increase in permanent

14-mm Surfacing Mixtures

Asphalt mixture

urse 14-mm surfacing

Compaction method

oller

Site

Gyro Vibro

SiteLab 4 Lab 2 Lab 2

1.3 1.7 3.4 2.2 2.8

1.26 1.41 0.56 1.36 1.56

0.36 0.41 0.12 0.18 0.25

29 29 21 13 16

0

0.5

1

1.5

2

2.5

0 1 2 3 4 5 6 7 8

Air Voids (%)

AxialStrain(%)

Gyratory Lab #1Roller Lab #1Roller Lab #4Vibratory Lab #1Site

Fig. 4. Permanent strain versus air voids for different compactionmodes—20-mm DBM asphalt mixture

se and

der co

R

b 1

.7

.64

.10

ER 2009

135:839-851.

Dow

nloa

ded

from

asc

elib

rary

.org

by

GE

OR

GE

WA

SHIN

GT

ON

UN

IVE

RSI

TY

on

05/1

4/13

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.

deformation resistance for the gyratory compacted specimenscompared to the vibratory compacted specimens and site cores.

Asphalt Mixture Fatigue

The fatigue results from the ITFT for the 28-mm DBM asphaltmixture are shown in Fig. 5 in the form of the fatigue relationshipdefined in Eq. �4� and using a strain criterion. The seven sets offatigue data, representing the three compaction methods �gyra-tory, vibratory, and roller� and five laboratories �all gyratory com-paction�, all lie on the same fatigue function. It is arguablewhether any significantly different fatigue lines exist between thedata sets beyond the natural variation associated with a fatigue�failure� test. This is substantiated by an R2 value of 0.85 for thefatigue line of best fit for the aggregated data sets.

Discussion

The mechanical property results show that in general mold basedcompaction methods �gyratory and vibratory� form asphalt mix-ture specimens of greater stiffness and resistance to permanentdeformation. Site cores appear to lie at the other end of the spec-trum in terms of mechanical performance and general have lowerstiffness and increased susceptibility to permanent deformation.Conversely, the mode of compaction does not appear to influencethe fatigue response of an asphalt mixture. However there areseveral exceptions to these general trends which may be seenfrom the data presented in this paper. Indeed within a given com-paction method there appears to be considerable scope for varia-tion, as demonstrated by the difference in stiffness between thetype of gyratory compactor �28-mm asphalt mixture� and opera-tion of roller compactors �20-mm asphalt mixture�. Sousa et al.�1991� also recognized that compactors within a given mode ofcompaction may manufacture specimens with quite differentproperties. They used the example of a kneading compactor, stat-ing that the compaction parameters �size of foot, number oftamps, size of specimen, thickness of layer, tamping pressure� andthe variation of these parameters will almost certainly effect manymixture properties linked to asphalt performance. This view wasfurther reinforced by the finding of Harvey and Monismith�1993�.

Within any one generic compaction mode there would appearto be compaction variables which have an effect on the stiffnessand permanent deformation of the specimens for a given air void

10

100

1,000

100 1,000 10,000 100,000 1,000,000

Number of Load Cycles to failure (Nf)

InitialStrain( μεμε)

Gyratory Lab #1 Gyratory Lab #2

Gyratory Lab #3 Gyratory Lab #4

Gyratory Lab #5 Vibratory Lab #1

Roller Lab #1

R2 = 0.85

Fig. 5. Fatigue data for 28-mm asphalt mixture—various compactionmethods

content. For example in gyratory compaction it could be the angle

JOURNAL OF TR


of gyration or the compaction pressure. For roller �slab� compac-tion probable parameters are compaction pressure and rollerspeed. In addition, the importance of sample preparation, bitumenand aggregate heating temperatures and duration and finally mix-ing procedures cannot be overlooked. All these parameters un-doubtedly have an influence on the micromechanical behavior ofthe coated aggregate particles.

Statistical Analysis

Variability between Compaction MethodsTo substantiate the observations relating to the influence of dif-ferent compaction methods on the mechanical properties of as-phalt mixtures, a statistical analysis was undertaken to comparethe results of the different compaction techniques with each other.Due to there being at least 15 specimens for each compactionmethod and asphalt mixture, it was possible to use the data sets tocompare the means and identify the differences between thesevalues. This process was undertaken using the air voids contentand stiffness modulus results of each specimen to assess the simi-larities in the volumetric characteristics and the mechanical prop-erties between each of the compaction methods. The laboratorycompaction data was also compared to the field data to analyzethe predictive capabilities of laboratory specimens.

The statistical analysis was carried out on the test data usingthe statistical software package “Minitab” �Davis 2005�, which isa numerical analysis program used to manipulate and analyze datausing a spreadsheet style interface. Hypothesis testing was used tomake inferences about the difference between the values of twoindependent populations using an initial assumption and testingthe validity of the claim �null hypothesis �H0�� versus the alter-native �research hypothesis �H1�� using the sample data. A prede-termined level of significance ��-level� was set to assess the nullhypothesis based on a probability value �p-value�. If the p-valueis greater than the �-level, then there is no support for the alter-native hypothesis and the null hypothesis can be accepted. If thep-value is less than or equal to the �-level then the null hypoth-esis is rejected and there is support for an alternative hypothesis.For this analysis the �-level was set at 0.025 giving a 95% con-fidence of the result.

The data was analyzed using the two-sample t-test to deter-mine the difference between two population means. As the test isonly applicable if the data are normally distributed, the normalityof the air voids and stiffness modulus results was initially ac-

20mmDBM A1 - ITSM

Percent

246024402420240023802360

99

95

90

80

70

60504030

20

10

5

1

Mean

<0.005

2418StDev 16.96N 30AD 1.478P-Value

Probability Plot of 20mmDBM A1 - ITSMNormal

Fig. 6. Probability plot for rejected hypothesis


135:839-851.

Dow

nloa

ded

from

asc

elib

rary

.org

by

GE

OR

GE

WA

SHIN

GT

ON

UN

IVE

RSI

TY

on

05/1

4/13

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.

cessed using the Anderson-Darling normality test function inMinitab. The Anderson-Darling test can be used for sample sizessmaller than 25, but is often seen to reject the assumption ofnormality for large data sizes having only slight imperfections.This was not considered to affect the results in this study due tothe sample size being limited to a maximum of 35 data points.The test evaluates the null hypothesis that the volumetric andmechanical data follow a normal distribution. The results of thetest are given by Minitab as a p-value which can be compared tothe �-level in the manner explained previously as well as in theform of a graph illustrating the distribution. Examples of prob-ability distribution graphs are shown in Figs. 6 and 7 for a set ofdata where the normal distribution assumption can be rejected andaccepted, respectively.

Once the null hypothesis has been accepted for each set ofdata, the t-test can be carried out to compare the independentsample means. Only data sets that were observed to be normallydistributed were used for the t-test with two samples of compac-tion data being compared against each other and the process re-peated so all compaction methods were compared for eachindividual material. No comparisons were conducted between thedifferent materials as it is known there would be little similaritybetween their population means. For each t-test a null hypothesiswas set stating that the difference between the mean values of thetwo samples was zero and an alternative hypothesis that the dif-ference between the mean values of the samples was not equal tothis reference value of zero. The test provides the result of thishypothesis test as a p-value, which can be compared to the target�-level. In order to provide a confidence of 95%, the �-level waschosen at 0.025. The two-sample t-test when completed in

Table 6. Normality Test for 28-mm DBM Base Asphalt Mixture

Compactionmethod

Number ofspecimens

Air voids

Mean�%�

Standarddeviation

G-Lab 1 26 3.12 0.49

G-Lab 2 30 2.77 0.36

G-Lab 4 30 5.20 0.63

G-Lab 5 32 1.88 0.18

V-Lab 1 30 2.76 0.21

R-Lab 1 30 2.99 0.42

20mmDBM Slab - Total - Voids

Percent

876543210

99

95

90

80

70

60504030

20

10

5

1

Mean

0.677

4.121StDev 1.422N 30AD 0.263P-Value

Probability Plot of 20mmDBM Slab - Total - VoidsNormal

Fig. 7. Probability plot for accepted hypothesis



Minitab provides a p-value as well as the confidence interval andan estimate of difference for the two populations. The confidenceinterval is an estimate of the likely range of values for the differ-ence between the two population means based on the sample data.The upper and lower bound figures determine the width of theconfidence interval. For the �-level of 0.025, it is 95% certain thatthe true difference between population mean values lies withinthe confidence interval. The estimate for difference is the differ-ence between the population mean values as calculated from thesample data.

28-mm DBM Asphalt MixtureThe results of the normality tests on air voids and stiffness modu-lus for the 28-mm DBM mixture are shown in Table 6. The airvoids within the asphalt mixture specimens for all of the compac-tion methods can be assumed to be normally distributed as thenull hypotheses was accepted. The same is true with the ITSMresults where all of the p-values for the normality test are greaterthan the �-level of 0.025 resulting in the null hypothesis beingaccepted for all the compaction methods.

The t-test was therefore conducted on each combination ofcompaction method for both air voids and stiffness modulus withthe results being shown in Tables 7 and 8, respectively. It can beseen from the p-values that the hypothesis stating that the differ-ence between the mean air voids is zero, can be rejected for allbut three of the combinations. These three combinations are gy-ratory �Lab 1�/roller �Lab 1�, gyratory �Lab 2�/vibratory �Lab 1�,and gyratory �Lab 2�/roller �Lab 1�. Although the two-samplet-test was conducted on all the stiffness combinations in Table 8,only the three combinations showing statistically identical volu-metric proportions were considered for further analysis. The re-sults show that although gyratory �Lab 2�/vibratory �Lab 1� andgyratory �Lab 2�/roller �Lab 1� produced specimens with the sameair void content, the null hypothesis stating that their stiffnessmean values are similar was rejected. Only the gyratory �Lab1�/roller �Lab 1� combination can be statistically proven to pro-duce specimens with the same stiffness modulus if their volumet-rics �air void contents� are the same.

20-mm DBM Asphalt MixtureThe results for the normality testing on the air voids and stiffnessmodulus results for the 20-mm DBM asphalt mixture are shownin Table 9. The results show that for all the compaction methodsthe null hypothesis stating the results are distributed normally canbe accepted for air voids and stiffness modulus �except for gyra-tory �Lab 1��.

The two-sample t-test was completed on each combination of

Property

Stiffness modulus

Normalityp-value

Mean�MPa�

Standarddeviation

Normalityp-value

0.157 7,684 841 0.890

0.065 6,349 799 0.038

0.038 8,082 583 0.772

0.490 11,165 850 0.173

0.165 10,089 932 0.762

0.028 7,323 540 0.696

ER 2009

135:839-851.

Dow

nloa

ded

from

asc

elib

rary

.org

by

GE

OR

GE

WA

SHIN

GT

ON

UN

IVE

RSI

TY

on

05/1

4/13

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.

compaction method for air voids with the null hypothesis statingthat the population means are equal being accepted for vibratory�Lab 1�/roller �Lab 1�, vibratory �Lab 1�/Site and roller �Lab 1�/Site as shown in Table 10. For all of the other combinations thenull hypothesis can be rejected and the mean air voids can beconsidered to be different. These three combinations were then

Table 7. Two-Sample t-test for 28-mm DBM Base Asphalt Mixture for

t-test data groupsDegrees of

freedom p-valu

G-Lab 1 G-Lab 2 45 0.004

G-Lab 1 G-Lab 4 53 0


G-Lab 1 V-Lab 1 33 0.001

G-Lab 1 R-Lab 1 49 0.276



G-Lab 2 V-Lab 1 46 0.851

G-Lab 2 R-Lab 1 56 0.037


G-Lab 4 V-Lab 1 35 0

G-Lab 4 R-Lab 1 50 0



V-Lab 1 R-Lab 1 42 0.010

Table 8. Two-Sample t-Test for 28-mm DBM Base Asphalt Mixture for


freedom p-valu


G-Lab 1 G-Lab 4 43 0.049



G-Lab 1 R-Lab 1 41 0.067










V-Lab 1 R-Lab 1 46 0

Table 9. Normality Test for 20-mm DBM Binder Course Asphalt Mixtu

Compactionmethod

Number ofspecimens

Air voids

Mean�%�

Standarddeviation

G-Lab 1 30 4.01 1.60

V-Lab 1 30 2.90 0.94

R-Lab 1 30 2.67 0.88

R-Lab 4 35 1.32 0.38

Site 30 2.49 0.65

JOURNAL OF TR


analyzed in terms of their stiffness modulus in Table 11 with thep-values demonstrating that the null hypothesis can be rejectedfor all combinations of compaction method. In addition, the esti-mate of difference in terms of stiffness modulus between the dif-ferent compaction method combinations can be considered to befairly large �greater than 2000 kPa for all three combinations�.

ids

Estimation ofdifference

95%lower bound

95%upper bound

0.349744 0.116736 0.582751

2.07426 2.37315 1.77536

1.24433 1.03856 1.4501

0.364077 0.154965 0.573189

0.13441 0.111004 0.379825

2.424 2.68966 2.15834

0.894583 0.7468 1.042367

0.14333 0.138635 0.167301

0.215333 0.416904 0.013763

3.31858 3.07626 3.5609

2.43833 2.19305 2.68362

2.20867 1.93218 2.48515

0.88025 0.981232 0.779268

1.10992 1.27785 0.94199

2.29667 0.402154 0.057179

ss Modulus


95%lower bound

95%upper bound

1,335.78 893.56 1,778.01

397.549 793.479 1.618

3,480.21 3,927.88 3,032.53

2,404.38 2,879.83 1,928.94

361.451 26.825 749.728

1,733.33 2,095.52 1,371.15

4,815.99 5,235.15 4,396.84

3,740.17 4,189.23 3,291.11

974.333 1,328.126 620.541

3,082.66 3,451.65 2,713.67

2,006.83 2,410.33 1,603.33

759 468.48 1,049.52

1,075.83 621.36 1,530.29

3,841.66 3,480.91 4,202.42

2,765.83 2,369.89 3,161.78

Property

Stiffness modulus

Normalityp-value

Mean�MPa�

Standarddeviation

Normalityp-value

0.088 10,601 882 0.016

0.388 10,288 1,208 0.030

0.161 8,173 493 0.041

0.395 6,260 624 0.383

0.201 5,476 586 0.487

Air Vo

e

Stiffne

e

re


135:839-851.

Dow

nloa

ded

from

asc

elib

rary

.org

by

GE

OR

GE

WA

SHIN

GT

ON

UN

IVE

RSI

TY

on

05/1

4/13

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.

14-mm Surfacing Asphalt MixtureThe results for the normality testing of the three compactionmethods �gyratory, vibratory, and site� for the 14-mm surfacingmaterial are shown in Table 12. The p-values for the air voidnormality testing show that the null hypothesis stating that thedistribution is normal can only be accepted for the gyratorysamples. The field samples have a p-value of 0.024 which islower than the �-level of 0.025 therefore the distribution is notnormally distributed to a 95% confidence, although if this confi-dence level were reduced for example to 90%, the hypothesiswould be accepted. The ITSM populations all provide p-valuesabove 0.25 resulting in the null hypothesis of the results beingnormally distributed being accepted with a 95% certainty. Thet-test was therefore only analyzed on the air voids and stiffnessfor the gyratory �Lab 2�/site combination. The results of the t-testin Tables 13 and 14 show that the null hypothesis stating the

Table 10. Two-Sample t-Test for 20-mm DBM Binder Course Asphalt M


freedom p-valu

G-Lab 1 V-Lab 1 46 0.002



G-Lab 1 Site 38 0

V-Lab 1 R-Lab 1 57 0.339


V-Lab 1 Site 51 0.053

R-Lab 1 R-Lab 4 38 0

R-Lab 1 Site 53 0.359

R-Lab 4 Site 45 0

Table 11. Two-Sample t-Test for 20-mm DBM Binder Course Asphalt M


freedom p-valu



V-Lab 1 Site 41 0

R-Lab 1 R-Lab 4 62 0

R-Lab 1 Site 56 0

R-Lab 4 Site 62 0

0

1

2

3

4

5

6

7

-4 -3.5 -3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 4E-Emean/s

Frequency

Frequency

Normal Distribution

Fig. 8. Normalized stiffness modulus results for 28-mm DBM mix-ture compacted using gyratory compaction �Lab 1�



difference between the mean values of the two populations waszero can be accepted for air voids but rejected for stiffness modu-lus.

The two-sample t-test statistical analysis of air voids and stiff-ness modulus for the three asphalt mixtures, three compactionmethods, and four laboratories has confirmed the observation thatproducing volumetrically identical asphalt mixture specimensusing different compaction methods does not assure that the ma-terials will have comparable mechanical properties. Similar ob-servations were found for permanent deformation parameterssuch as permanent strain under uniaxial loading.

Variability within Compaction Methods

In addition to quantifying the statistical difference between themechanical properties of specimens produced by means of differ-ent compaction methods, it is also important to quantify the natu-

for Air Voids


95%lower bound

95%upper bound

1.11167 0.43125 1.79208

1.33867 0.66757 2.00976

2.68924 2.08029 3.29819

1.52467 0.88706 2.16227

0.227 0.244324 0.698324

1.57757 1.20658 1.94857

0.413 0.006093 0.8232093

1.35057 0.99829 1.70286

0.186 0.216833 0.588833

1.16457 1.43824 0.89091

for Stiffness Modulus


95%lower bound

95%upper bound

2,115.87 1,633.43 2,598.30

4,025.54 3,534.66 4,522.43

4,812.67 4,317.43 5,307.90

1,912.68 1,635.36 2,190.00

2,696.80 2,416.53 2,977.07

784.124 483.69 1,084.56

-100

-80

-60

-40

-20

0

20

40

60

80

100

1 10 100

Sample Size n

Accuracy(%)

95th Percentile

80th Percentile

Fig. 9. Confidence limit plot for ITSM data for 28-mm DBM mix-ture compacted using gyratory compaction �Lab 1�

ixture

e

ixture

e

ER 2009

135:839-851.

Dow

nloa

ded

from

asc

elib

rary

.org

by

GE

OR

GE

WA

SHIN

GT

ON

UN

IVE

RSI

TY

on

05/1

4/13

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.

ral variability in mechanical properties for these sets of data. Thisinformation enables the number of specimens which need to betested in order to obtain a mean value �for a given confidenceinterval� within a certain accuracy �when compared to the truemean obtained if an infinite number of specimens were tested� tobe specified for a particular compaction method and mechanicalproperty. For example, it might be the case that to achieve anaccuracy of �10% of the true mean stiffness, the average of 5ITSM test results using a 95% confidence interval will be needed.Hence for 19 times out of 20 �95% confidence interval� if theaverage of 5 ITSM test is calculated, the resulting stiffness will bewithin �10% of the true value. If a higher degree of confidenceor an improved accuracy is required, then a greater number ofspecimens would need to be tested. This study uses the student tanalysis to determine the number of tests required to achieve aspecified accuracy for a given confidence interval for the ITSMand RLAT tests as a function of compaction method. For thestudent t method to be valid the data needs to be of approximatelynormal �Gaussian� distribution, as this statistical approach utilizesthe central limit theory. A normalized distribution of 30 ITSMresults is shown in Fig. 8 where E is the stiffness �MPa� and s isthe local standard deviation. A similar approach was used by Col-lop et al. �2001� in assessing the variability of in situ pavementmaterial stiffness moduli.

Using Eq. �5�, the confidence limit plot shown in Fig. 9 can beobtained

x̄ � t�/2� s�n

� �5�

where x̄=mean �in this case mean ITSM, E�; t�/2 �t statistic�=standard statistical function based on the confidence intervalselected and the sample number minus one �degrees of freedom�.Fig. 9 expresses the accuracy as a percentage of the meanstiffness.

ITSM Testing AccuracyThe ITSM repeatability results are given in Table 15. For the28-mm asphalt mixture, a number of laboratories and different

Table 12. Normality Test for 14-mm Surfacing Asphalt Mixture

Compactionmethod

Number ofspecimens

Air void

Mean�%�

Standarddeviation

G-Lab 3 15 3.35 0.63

V-Lab 3 15 2.16 0.66

Site 18 2.84 0.58

Table 13. Two-Sample t-Test for 14-mm Surfacing Asphalt Mixture for


freedom p-value

G-Lab 2 Site 28 0.023

Table 14. Two-Sample t-Test for 14-mm Surfacing Asphalt Mixture for


freedom p-value

G-Lab 2 Site 27 0

compaction methods achieved similar repeatability. The interlabo-

JOURNAL OF TR


ratory testing program indicated an average of 7 specimens wasrequired to achieve an accuracy of �10% �95% confidence inter-val� and 16 specimens for �5% accuracy. The gyratory com-pacted specimens from Lab 2 had the poorest repeatability, whichon inspection of the data was attributed to one “rogue” stiffnessresult which skewed the standard deviation. Removing this datapoint brought the repeatability into line with the vibratory com-pacted specimens. Overall the mode of compaction did not appearto significantly affect the variability of the results.

The 20-mm asphalt mixture had similar repeatability to the28-mm asphalt mixture, although it was expected that the smallersized mixture would have less variability. It is generally recog-nized that the smaller the maximum nominal aggregate size, themore homogenous the mixture and the less variance in the result-ing mechanical properties. Within this group, on average 7 speci-mens would need to be tested to achieve an accuracy of �10%accuracy �95% confidence interval� and 18 specimens for �5%accuracy. As with the 28-mm mixture, the mode of compactiondoes not appear to have a significant effect on the variability ofthe results.

The 14-mm mixture repeatability results are slightly poorerthan the 28- and 20-mm mixtures, again in contradiction to nor-mal expectation. In this group on average 9 specimens wouldneed to be tested to achieve an accuracy of �10% accuracy �95%confidence interval�, and 22 specimens for a �5% accuracy. Asthe accuracy and confidence interval decreases, the difference inspecimens required for testing between the different data sets alsodecreases �maximum of 5 to a minimum of 3 for �10% accuracy,80% confidence interval�.

RLAT Testing AccuracyThe RLAT repeatability results are given in Table 16. It is imme-diately apparent that the variability in the permanent deformationtest �final strain parameter� is far greater than that of the ITSMtest. Only one set of results �roller compaction in Laboratory 1�for the 20-mm mixture was able to deliver a specified number oftests to achieve an accuracy of �10% accuracy �95% confidenceinterval� using a test sample number of 30. Even at the lower

Property

Stiffness modulus

ormality p-valueMean�MPa�

Standarddeviation

Normalityp-value

0.198 2,889 228 0.370

0.005 2,104 252 0.498

0.024 1,565 196 0.299

ids


95%lower bound

95%upper bound

0.50889 0.07419 0.943588

ss Modulus


95%lower bound

95%upper bound

1,323.44 1,169.81 1,477.08

s

N

Air Vo

Stiffne

confidence interval of 80%, large numbers of tests are required to


135:839-851.

Dow

nloa

ded

from

asc

elib

rary

.org

by

GE

OR

GE

WA

SHIN

GT

ON

UN

IVE

RSI

TY

on

05/1

4/13

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.

achieve an accuracy of �10%. The reasons for the high variabil-ity can be attributed, first, to the inherent variability associatedwith large strain testing and, second, to the restraints of the RLATtest �unconfined, short specimens�. It is of interest to note that thedata sets which performed well in terms of ITSM variability didnot necessarily perform will in terms of RLAT variability.

Conclusions and Discussion

The compaction study has shown that in general mold based com-paction methods �gyratory and vibratory� tend to produce asphaltmixture specimens of greater stiffness and permanent deformationresistance than specimens cored from roller compacted slabs. Thisis probably due to the boundary conditions of the molds and theirinfluence on the aggregate-bitumen-voids matrix �internal struc-ture� of the compacted asphalt mixture specimen. In addition,altering the compaction parameters in the gyratory has an effecton the mechanical properties of the asphalt mixtures. Site com-pacted specimens tend to have mechanical properties that aremore aligned with roller compacted specimens than gyratory orvibratory compacted specimens particularly with regard to their

Table 15. Repeatability of ITSM for Three Asphalt Mixtures and Variou

Mode of compaction Asphalt mixture Laboratory�10%

9

Gyratory 28 mm Lab 1





Vibratory 28 mm Lab 1


Roller 28 mm Lab 1



Roller 20 mm Lab 1

Roller 20 mm Lab 4

Site 20 mm —



Site 14 mm —

Table 16. Repeatability of RLAT for Two Asphalt Mixtures and Various

Mode of compaction Asphalt mixture Laboratory�10%

95



Roller 20 mm Lab 1

Roller 20 mm Lab 4

Site 20 mm —



Site 14 mm —



stiffness modulus. In terms of fatigue cracking resistance, the re-sults indicate that mode of compaction does not significantly af-fect this mechanical property.

The statistical two-sample t-confidence test was used to com-pare the sample means, in terms of air voids content and stiffnessmodulus, between the different compaction methods. No com-parisons were made between asphalt mixture types and only datasets that were found to be normally distributed where analyzed.The statistical results confirm the mechanical property observa-tions that producing volumetrically identical asphalt mixturespecimens using different compaction methods does not assurethat the materials will have comparable mechanical properties.

The student t analysis was used to determine the number oftests required to achieve a specified accuracy for a given confi-dence interval for the ITSM and RLAT tests as a function ofcompaction method. The results from the interlaboratory testingprogram indicated that to achieve �10% accuracy for an ITSMtest �using a 95% confidence interval� the mean of at least 7 testsis required for the 28-, 20-, and 14-mm asphalt mixtures. Thestatistical analysis also showed that mode of compaction does notappear to have a significant effect on the variability of the me-chanical property results. The variability associated with perma-

paction Modes

Number of samples required for testing

racy,I.

�5% accuracy,95% C.I.



12 4 6

28 5 13

11 3 6

12 3 6

13 4 7

19 4 9

17 4 8

12 3 6

14 4 7

25 5 12

9 3 5

19 4 9

21 5 10

13 4 7

26 5 12

28 5 13

action Modes

acy, �20% accuracy,95% C.I.



28 26 8

11 15 6

6 7 4

11 16 6

12 17 6

8 11 5

5 6 3

6 7 4

s Com

accu5% C.

6

10

5

6

6

7

7

6

6

9

5

7

8

6

9

10

Comp

accur% C.I.

�30

�30

13

�30

�30

23

11

13

ER 2009

135:839-851.

Dow

nloa

ded

from

asc

elib

rary

.org

by

GE

OR

GE

WA

SHIN

GT

ON

UN

IVE

RSI

TY

on

05/1

4/13

. Cop

yrig

ht A

SCE

. For

per

sona

l use

onl

y; a

ll ri

ghts

res

erve

d.

nent deformation testing using the RLAT was found to be higherthan that found for the stiffness modulus results. To achieve�20% accuracy for an RLAT test �using a 95% confidence inter-val� an average of 14 specimens need to be tested for the 20-mmasphalt mixture and 7 specimens for the 14-mm asphalt mixture.

References

British Standards Institution �BSI�. �1993�. “Method for determination ofthe indirect tensile stiffness modulus of bituminous mixtures.” Draftfor Development Rep. No. 213, London.

British Standards Institution �BSI�. �1995�. “Method for the determina-tion of the fatigue characteristics of bituminous mixtures using indi-rect tensile fatigue.” Draft for Development Rep. No. ABF, London.

British Standards Institution �BSI�. �1996�. “Method for determining theresistance to permanent deformation of bituminous mixtures subjectto unconfined dynamic loading.” Draft for Development Rep. No. 226,London.

British Standards Institution �BSI�. �2003a�. “Bituminous mixtures—Testmethods for hot mix asphalt—Part 32: Laboratory compaction of bi-tuminous mixtures by vibratory compactor.” BS EN 12697-32, Lon-don.

British Standards Institution �BSI�. �2003b�. “Bituminous mixtures—Testmethods for hot mix asphalt—Part 33: Specimen prepared by rollercompactor.” BS EN 12697-33, London.

British Standards Institution �BSI�. �2004a�. “Bituminous mixtures—Testmethods for hot mix asphalt—Part 31: Specimen preparation by gy-ratory compactor.” BS EN 12697-31, London.

British Standards Institution �BSI�. �2004b�. “Bituminous mixtures—Testmethods for hot mix asphalt—Part 35: Laboratory mixing.” BS EN12697-35, London.

Brown, S. F., and Gibb, J. M. �1996�. “Validation experiments for perma-nent deformation testing of bituminous mixtures.” Electron. J. Assoc.Asph. Paving Technol., 65, 255–299.

Collop, A. C., Armitage, R. J., and Thom, N. H. �2001�. “Assessingvariability of in situ pavement material stiffness moduli.” J. Transp.Eng., 127�1�, 74–81.

Consuegra, A., Little, D. N., Quintas, H. V., and Burati, J. �1989�. “Com-parative evaluation of laboratory compaction devices based on theirability to produce mixtures with engineering properties similar tothose produced in the field.” Transportation Research Record. 1228,Transportation Research Board, Washington, D.C., 80–87.

Davis, R. �2005�. Minitab lab manual, 2nd Ed., Duxbury Press, Calif.Gibb, J. M. �1996�. “Evaluation of resistance to permanent deformation

JOURNAL OF TR


in the design of bituminous paving mixtures.” Ph.D. thesis, School ofCivil Engineering, Univ. of Nottingham.

Hartman, A. M., Gilchrist, M. D., and Walsh, G. �2001�. “Effect of mix-ture compaction on indirect tensile stiffness and fatigue.” J. Transp.Eng., 127�5�, 370–378.

Harvey, J., and Monismith, C. L. �1993�. “Effects of laboratory asphaltconcrete specimen preparation variables on fatigue and permanentdeformation test results using strategic highway research programmeA-003A proposed testing equipment.” Transport Research Record.1417, Transportation Research Board, Washington, D.C., 38–48.

Hunter, A. E., Airey, G. D., and Collop, A. C. �2004�. “Aggregate orien-tation and segregation in laboratory compacted asphalt samples.”Transportation Research Record. 1891, Transportation ResearchBoard, Washington D.C., 8–15.

Masad, E., Jandhyala, V. K., Dasgupta, N., Somadevan, N., and Shash-idhar, N. �2002�. “Characterization of air void distribution in asphaltmixes using x-ray computer tomography.” J. Mater. Civ. Eng., 14�2�,122–129.

Masad, E., Muhunthan, B., Shashidhar, N., and Harman, T. �1999�.“Quantifying laboratory compaction effects on the internal structureof asphalt concrete.” Transportation Research Record. 1681, Trans-portation Research Board, Washington, D.C., 179–185.

Nevitt, H. G. �1959�. “Some sources of stability measurement variations.”Proc., Association of Asphalt Paving Technologists, 28, 16–34.

Peterson, R. L., Mahboub, K. C., Anderson, R. M., Masad, E., and Tash-man, L. �2004�. “Comparing Superpave gyratory compactor data tofield cores.” J. Mater. Civ. Eng., 16�1�, 78–83.

Sousa, J. B., Deacon, J. A., and Monismith, C. L. �1991�. “Effect oflaboratory compaction method on permanent deformation characteris-tics of asphalt-aggregate mixtures.” Electron. J. Assoc. Asph. PavingTechnol., 60, 533–585.

Tashman, L., Masad, E., D’Angelo, J., Bukowski, J., and Harman, T.�2002�. “X-ray tomography to characterize air void distribution inSuperpave gyratory compacted specimens.” Int. J. Pavement Eng., 3,19–28.

Tashman, L., Masad, E., Peterson, B., and Saleh, H. �2001�. “Internalstructure analysis of asphalt mixes to improve the simulation of Su-perpave gyratory compaction to field conditions.” Electron. J. Assoc.Asph. Paving Technol., 70, 605–645.

Vallerga, B. �1951�. “Recent laboratory compaction studies of bituminouspaving mixtures.” Proc., Association of Asphalt Paving Technologists,20, 117–153.

Von Quintas, H. L., Scherocman, J. A., Hughes, C. S., and Kenneday, T.W. �1991�. “Asphalt-Aggregate Mixture Analysis System, AAMAS.”National Cooperative Highway Research Program Rep. No. 338,Transportation Research Board, Washington, D.C.


135:839-851.

Documents

Effect of Compaction Mode on the Mechanical Performance and Variability of Asphalt Mixtures