Mix design and aggregate requirements for stone matrix asphalt mixtures


Material Information

Mix design and aggregate requirements for stone matrix asphalt mixtures
Physical Description:
viii, 286 leaves : ill. ; 29 cm.
West, Randy Clark
Publication Date:


Subjects / Keywords:
Asphalt rock -- Research -- Florida   ( lcsh )
Pavements, Asphalt -- Testing   ( lcsh )
Pavements, Asphalt concrete -- Testing -- Florida   ( lcsh )
Civil Engineering thesis, Ph. D
Dissertations, Academic -- Civil Engineering -- UF
bibliography   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph. D.)--University of Florida, 1995.
Includes bibliographical references (leaves 208-213).
Statement of Responsibility:
by Randy Clark West.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002052091
oclc - 33448215
notis - AKP0052
System ID:

Full Text





n^ t


To my mother and father, who nurtured me, inspired me, supported me, and gave
me the confidence to venture toward my own goals.


To my girls: Ronda, Anna Grayce and Clara, who have lovingly and patiently
helped me reach this goal with my identity intact.


The author thanks all of those individuals who provided assistance during the

progress of this research.

The author is deeply appreciative of the many forms of support

provided by the Florida Department of Transportation including the funding for this study,

the educational leave granted to pursue this degree, and the courtesy of permitting much

of the work to be performed by fellow students at the Bituminous Laboratories of the

State Materials Office.

Those individuals from the State Materials Office which provided significant

contributions to this work were Gale Page, Edd Leitner, Herb Cone, Toby Dillow,

George Lopp, Aaron Turner, and Randy Burnett.

Much of the testing was performed by students from the University of Florida

including Ivo Moroni, Bouzid Choubane,

William Freeman, Raja Subramanian,

Mark Davies, John Veilleux, Howard Eshram, John Bertram and Shin-Che Haung.



ACKNOW LEDGEM ENTS ................... ................... ................... ................... .* 4

ABSTRACT .................................... ............................... .... .. .. -.- .



Significance of Pavement M materials Research .............................................
Description of SM A M ixtures .......................... . .. ............. .... ...............
Issues to be Resolved .................................................................................
Objectives and Approach to Research ........................................................
Organization ................... ................... .................. ....... .................. ... .... ..


Design of SM A M ixtures .................................
The European Asphalt Study Tour ...................
Early SMA Projects in the United States ..........
Development of United States SMA Guidelines

S*. . .. .. . . .. *** ** * *

. . . 111 .. .. .( .m .
. . . . . .



Purpose of M ix D design ...................................
Current Methods of SMA Mix Design ............

Review of Mix Tests for SMA

Introduction ...........................

Selection of Tests ...................................... ................... .............................

. . .. .. .1 i wo .*0 e .. *. ** .* *** *

. . 1 .. ... .. ... .. ... .. ... .. ... .. ... .. ---
. .. .. .. .. .. .. ... .. .. .. .. .. .. ... .. .. .. -- -- .-

Origin of Stone M atrix Asphalt ..................................................................
Comparison of SM A and DGAC M ixtures .................................................
Selection of M materials ................................................................................

EVALUATION OF M IXTURE TEST DATA ...........................................
GTM Compaction ......................................................................................
Design of Experimental M fixtures ...............................................................
GTM Densification/Shear Tests .................................................................

Georgia Loaded Wheel

Triaxial Repeated Load Tests .....................................................................
Correlation of Rutting Tests .......................................................................
Resilient M odulus Tests .............................................................................
Indirect Tensile Strength Tests ...................................................................
Diametral Creep Tests ...............................................................................

Moisture Damage

Tests ... . .... ..... . . . . . .. .. .......... . .


* **. **

. . *T k. f
*. ** .* .* *

The Aggregate Issues .................................................................
Review of Desired Aggregate Characteristics ..............................
Aggregate Characteristics Related to Shear Strength ...................
Aggregate Resistance to Degradation .................................... .....
Tests for Aggregates ...................................................................
Aggregate Specifications ............................................................
Test Procedures ..........................................................................

Traditional Aggregate Properties ................................................
Particle Index .............................................................................
Crushing Strength .......................................................................
GTM Aggregate Strength and Degradation .................................
Correlation of Aggregate Characteristics to Mixture Test Results


* .. ... .. **** *
. .* * . *

S ....

Assessment of Mixture Tests .......
Assessment of Study Mixtures .....

Assessment of Aggregate Tests ................... ........................................... ... . .
Assessment of Florida Aggregates ................... ................... ................... .....

IMPACT OF SMA MIXTURES ON FLORIDA .................... .. ................

Projecting SM A Usage ..............................................................................
Effects on M material Supply .........................................................................

CONCLUSIONS AND RECOMMENDATIONS ......................................

T ests ................................ ......... ... . ...... ..... ... .

. .......... ., .. *. ., > .* -* ** ** . ** ** . -
- . ** ** .* .* .* ** ** . ** .* .(* .

.............. 164

Cost Analysis ...................................






SM A Production ....................................................................................
SM A Placement ......................................................................................

REFERENCES ......................................................................................................

BIOGRAPHICAL SKETCH .....................................................................

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy




May 1995

Chairman: Byron E. Ruth
Major Department: Civil Engineering

Stone Matrix Asphalt (SMA) mixtures were developed in Europe where they have

proven to be very resistant to rutting, cracking, and polishing.

SMA mixtures are

characterized by a stone-on-stone aggregate structure filled with a rich binder-filler mastic.

Implementation of SMA in the United States began in 1991.

The traditional method of

design for SMA mixtures has utilized the Marshall procedure, but recipe formulations of

component percentages and restrictive material guidelines persist.

A better design method

was needed that could provide an engineering basis for the selection and optimization of


In this research study, SMA mixtures from several field projects and experimental
v S I *.1 44

compaction and shear strength testing with the Gyratory Testing Machine; a rutting test

using the Georgia Loaded Wheel Tester; a rutting test using a confined repeated-load

procedure; resilient modulus tests at low, moderate, and high temperatures; the indirect

tensile strength and strain test; wet/dry moisture damage tests, and a low-temperature

diametral-creep compliance test.

Each of the rutting tests do indicate that the SMA mixtures are generally very

resistant to permanent deformation.

However, the procedure used to evaluate fatigue

cracking indicated that many SMA mixtures were less resistant to fatigue than standard

dense-graded mixtures.

The validity of this assessment is uncertain.

Tests on the mixtures

and the aggregates show that some of the softer Florida limestones tend to be more prone

to degradation which may lead to moisture damage and increase the potential for ravelling

or loss of skid resistance. A preliminary assessment of the impact that SMA may have on

the paving and aggregate supply industries indicates that due to limited application of this

technology to heavy-traffic-type roadways the effect will be very minor.


1.1 Significance of Pavement Materials Research

The quest to extend pavement life has great economic importance.

In the United

States, an estimated $10.5 billion of hot mix asphalt is produced each year for highway

construction and resurfacing pavements.

In Florida, approximately $150 million is spent

each year on hot mix asphalt placed on the state's highway system.

Most of that is used

for resurfacing of existing pavements that have reached their useful life .

Highway engineers must continually search for ways to improve the performance

of pavements.

The amount of traffic and the loads carried by our highway pavements has

steadily increased through the years.

It has become apparent over the past decade that

many of the materials and procedures that have been conventionally used to design

pavements are not suitable for the increase in traffic and heavy loads imposed upon them.

One technology that may have potential for improving the performance of heavily

loaded pavements is the European concept of stone matrix asphalt (SMA).

The European

SMA mixtures have been heralded to have better highway performance than conventional

dense-graded asphalt concrete (DGAC) mixtures used in the United States.

SMA mixtures have been said to have three primary advantages: high deformation

These attributes are a result of several factors which distinguish SMA mixtures from

DGAC mixtures.

Description of SMA Mixtures

The most unique characteristic of SMA mixtures is the aggregate gradation.


mixtures are said to be gap-graded with a very high proportion of coarse aggregate and a

high mineral filler content.

The gap in the aggregate gradation of SMA mixtures occurs in

the size range between 4.75mm and 75 im, where little material is retained

the gap-gradation is evident in the core sample shown in Figure 1

The effect of

. The contact between

coarse aggregate particles has been described as a stone-on-stone skeleton which provides

the strength (i.e. rutting resistance) of the SMA mixture.

rca....-nr&'.- S V VS.-Eb


The other unique feature of SMA mixtures is the binder mastic.

In SMA mixtures,

the asphalt binder combines with the mineral filler and added stabilizers to form a stiffened

binder mastic to bond the aggregate particles together and provide durability to the


Several types of stabilizing additives have been used in SMA mixtures including

cellulose fibers, mineral fibers, and polymers.

Without a stabilizing additive the binder will

tend to drain to the bottom of the mixture during storage or transport between the hot-mix

production facility and the paving operation at the roadway.

The filler and stabilizing

additives stiffen the binder at high temperatures without sacrificing flexibility of the binder

to resist cracking.

Issues to be Resolved

There are, however, several issues regarding SMA mixtures that must be resolved

before the technology can be effectively applied in Florida and many other states.

most important issue is cost effectiveness.

European experience has been that SMA

mixtures cost ten to twenty percent more than the conventional DGAC mixtures [2,3].

The increased costs are attributed to several factors such as the use of premium quality

aggregates, the cost of additives, and slower production rates.

Compared with the current

DGAC mixtures, can SMA mixtures provide longer pavement service lives to offset the

higher initial costs?

In Florida, the economics of SMA mixtures will depend primarily upon the


European recommendations adopted by the Federal Highway Administration (FHWA) [4]

have been to limit SMA aggregates to very high quality materials.

The SMA aggregate

specifications suggested by FHWA are so restrictive that no native Florida aggregates

meet the requirements.

If it is determined that Florida's aggregates are not suitable for use

in SMA mixtures, it is expected that importing acceptable aggregates from out of state

would make SMA mixtures cost prohibitive in most of the state.

Another critical area of concern is the current practice of SMA mixture design.

Selection of materials and component proportions for SMA mixtures has been traditionally

based upon recipes and rules-of-thumb developed from experience in Europe.

optimum asphalt content for most SMA mixtures has been determined using only the

volumetric elements of the Marshall mix design method.

Considering the widely

recognized deficiencies of the Marshall procedure, it is not surprising that many SMA

designs are adjusted in the field after the start-up of production.

Possibly the most significant problem with the current practice of SMA mixture

design is the total absence of performance-related, mixture strength or deformation tests

which may indicate how well the mixture resists loading and environmental conditions in

the field.

Without such tests, it is impossible to compare the relative benefits of using

alternative materials or to evaluate expected performance of SMA mixtures versus dense-

graded mixtures. The primary reason for the lack of such testing is because SMA mixtures

L -_- ... -l. -- ,,* ^ k .* ** .. nn. 1 nrh a n f. u a A a a' ,n I.. n r4 a a^ nr^ rn na n a

are inadequate rather than the SMA mixtures.

Good mixture design tests should be able

to predict, or at least indicate, the mixture's performance.

1.4 Objectives and Approach to Research

It is important to research the viability of the SMA mixtures containing currently

utilized materials.

Likewise it is also important to evaluate these SMA mixtures relative to

the standard dense-graded mixtures.

If it can be demonstrated that the current Florida

aggregates are suitable in SMA mixtures, and that SMA mixtures will perform better than

typical DGAC mixtures, the savings earned from longer lasting pavements may amount to

several million dollars each year.

One approach to this research could be to simply construct test pavements with

SMA mixtures and evaluate field performance.

However, this approach is very costly,

limited to the evaluation of only a few of the variables, and would probably take ten to

twenty years to fully evaluate all aspects of performance.

A more sensible approach is to

conduct the research in the laboratory using tests which are representative of field

conditions and clearly related to field performance.

Therefore, the objectives of this

project were developed to address two issues regarding SMA mixtures:

mixture design

and aggregate requirements.

To evaluate improved mix design for SMA mixtures, a combination of mixture

tests were selected to assess the resistance of SMA mixtures to the four major types of

of Engineers Gyratory Testing Machine, rutting tests with the Georgia Loaded Wheel

Tester, and deformation resistance measured in triaxial repeated-load tests.

Fatigue and

stripping resistance of SMA mixtures were evaluated using the indirect tensile strengths

and resilient moduli of conditioned and unconditioned samples.

Thermal cracking

potential was evaluated using indirect tensile creep and resilient moduli tests.

Current aggregate requirements for SMA mixtures were thoroughly reviewed.

Arguments of why certain aggregate requirements may or may not be necessary to ensure

good mixture performance are presented.

SMA aggregates from several field projects

were evaluated in terms of traditional aggregate tests and a couple of new tests felt to be

more indicative of performance.

In particular, a new test was developed to assess

aggregate degradation potential by comparing sample gradations before and after

compaction tests of dry aggregates and SMA mixtures in the Corps of Engineers Gyratory

Testing Machine.

A cost-effectiveness comparison of SMA mixtures and conventional asphalt

mixtures can not be accurately assessed until typical production costs and life cycles are


A brief discussion is presented to focus on the implementation of SMA

mixtures and its potential impact on the paving and materials industries in Florida.

In summary,

the specific research objectives of this research are as follows:

Identify mixture tests which can be used to indicate the performance of SMA

4.0 t

Recommend an appropriate testing protocol for the design of SMA mixtures that

will enable the designer to choose the materials and proportions that will provide

the greatest resistance to the major forms of pavement distress.

Establish more rational limits for aggregate quality parameters which are necessary

to assure success in SMA mixtures.

Identify the aspects of the paving industry that will be most affected by the use of

SMA mixtures in Florida.

The scope of the research was established so that test results could be related to


This was accomplished by obtaining eleven mixtures (or materials) from

SMA projects constructed in six states between 1991 and 1993.

The laboratory test

results of these field proven materials and mixtures were used to establish a baseline of

results for use in the evaluation of Florida materials and experimental mixtures.

aggregate sources, typically used in asphalt mixtures and representative of the geologic

and geographic range of aggregates used in Florida, were selected for the research.


This dissertation has been organized to present the literature review, research plan,

testing, analysis, and findings in as clear and logical sequence as possible.

Following this

introductory chapter is a chapter on the historical development of SMA mixtures and their

A rxis7nz 7 nf -^MA mey

Inrr^rrln*/^l /^/^rao r ltfl r mll rta-F^^a^a c^rl a nm; l-i+r iri1/rn i -T-Ql / iciTA

of the materials and the mixture test procedures.

The results of the SMA mixture tests are

presented and analyzed in Chapter 6.

A review of

aggregate characteristics and aggregate

tests is presented in Chapter

. Chapter 8 consists of evaluations of the aggregate

properties and characteristics for the materials used in this study. Chapter 9 presents an

assessment of the mixture and aggregate tests covered in this study. Chapter 10 projects

the impact of SMA mixtures on Florida highway construction, the probable role and

extent of usage, and the changes that will be necessary in the hot-mix asphalt (HMA)

industry and in the aggregate industry.

Conclusions and recommendations are presented

in Chapter 11.



Origin of Stone Matrix Asphalt

SMA mixtures were developed over the past three decades in several of the

northern European countries.

Germany, where the SMA technology originated, has the

most experience, with over 14,000 lane miles of SMA pavement.

Germany placed nearly a million tons of SMA [5].

In 1989 and 1990,

Sweden, Denmark, Norway, the

Netherlands, Finland, Austria, France, and Switzerland all utilize the SMA concept to

some degree.

Each country has different requirements and different terminology.

Splittmastixasphalt, grit mastic asphalt, stabinor, and HABS (hot asphalt with stone) are

all names synonymous with stone matrix asphalt as it has been titled in the United States.

The SMA concept was initially developed in the 1960s as a pavement surface

course to provide improved abrasion resistance under studded tires [2].

When studded

tires were banned from most European countries in the early 1970s, the use of the SMA

concept declined.

However, from the performance of the early SMA pavements, it was

recognized that the SMA mixtures were also very resistant to rutting and shoving under

heavy traffic.

Since most countries now prohibit studded tires, SMA mixtures have

become used primarily to provide a highly stable pavement surface.

In 1984, standard


European countries adopted slightly variant specifications for SMA mixtures in the late


In addition to the wear resistance and rutting resistance of SMA mixtures, several

other attributes have also been cited: improved durability [2,3,4], improved cracking

resistance [2,6], very good skid resistance [2], reduced water spray [7], and reduced

noise [8].

Comparison of SMA and DGAC Mixtures

Like conventional asphalt concrete, SMA mixtures are hot-laid, densely compacted

mixtures composed of about 94 to 95 percent aggregate and about 6 to 7 percent asphalt

by weight of total mix.

However, the primary difference between conventional DGAC

mixtures and SMA mixtures is the aggregate gradation.

As shown in Figures

and 3, a

typical aggregate gradation for DGAC consists of a uniform distribution of particle sizes,

whereas SMA mixtures utilize a gradation which is dominated by a single particle size,

typically between 9.5 mm and 4.75 mm, and contains few aggregate particles between

4.75 mm and 75 jpm.

Figure 4 illustrates the effect of gradation for an SMA mixture and a typical

DGAC mixture containing the same maximum aggregate size.

The obvious and important

detail evident from this figure is that the SMA gradation creates more contact between

coarse aggregate particles.

In the DGAC gradation, the large aggregate particles float in a


75un 300m em0pm 2.36mm 4.75mnm .Smn 12.Smm o.anmm


Plot of Typical DGAC and SMA Gradations


4.75 9.5mm

2.36 4

1.18 2.

0.075 -

2.36 4.75mm


9.5 12.5mm


4.75 9.5mm

- 12.5mm

12.5 19.0mm

< 0.075mm

0.075 0.3mm

0.6 1.18mm

12.5 19.0mm

Figure 3

0.3 0.6mm

Typical Percentages of Aggregate Particle Sizes in SMA and DGAC Mixtures


PH'.*,"MAK'.".'. .
MP '.MMAf.


The high percentage of essentially one-size coarse aggregate creates excellent particle


The durability of SMA mixtures is derived from the rich voidless mastic of asphalt,

filler, and stabilizer. SMA mixtures in Europe generally have higher binder contents than

dense-graded mixtures. The coarse aggregate gradation and the higher binder content

together yield a thicker binder coating on the aggregate particles which holds the particles

tightly together and provides greater resistance to binder aging.

Selection of Materials

In Europe, selection of materials for SMA mixtures is based primarily on


This applies to the selection of aggregates, fillers, stabilizers, and grades of


Most references state that premium-quality, crushed aggregates must be used.

Bellin [2] states that Germany requires "double crushed, tough, premium aggregates of

definite size, soundness, shape, etc." and that the aggregates must have a high Polished

Stone Value (PSV) for good skid resistance.

The term "double-crushed" is unclear, and

quantitative criteria corresponding to the descriptive requirements of "toughness" or

"soundness" were not cited. A requirement that aggregates contain less than 20 percent

flat and elongated particles was given. Aggregates commonly used in German SMA

mixtures are gabbro, diabase, basalt, and granite. Limestones, sandstones, and "soft"

index, and abrasion resistance [9].

Swedish SMA mixtures. Danish s

Quartzite is cited as a commonly used aggregate in

specificationss also require 100 percent crushed rock with

a minimum flakiness value and a minimum brittleness value [10].

In references from other

countries, specific aggregate requirements were not given although "crushed rock" was

often used to describe the coarse aggregate.

Requirements for the fine aggregates are also unclear.

Several references stated

that fine aggregate shall be manufactured sand [11,12], whereas others appeared to allow

manufactured sands, natural sands, or a combination of the two [10,13,14,15].

No test

requirements for fine aggregates were given in the literature.

The most common type of filler used in European SMA mixtures appears to be

limestone dust [2,12].

No specific requirements are provided in the literature.


[11] suggests that the filler should be as fine as possible so that it embeds in the asphalt

According to Bellin [2], baghouse fines are not used much as filler in SMA mixtures.

A variety of stabilizers have been used in European SMA mixtures.


fibers were used in early SMA mixtures before studies indicated potential hazards with this


Cellulose and mineral fibers appear to now dominate the stabilizer market.

Cellulose fibers are used exclusively in Denmark and the Netherlands [10,13].

reportedly utilizes cellulose and mineral fibers on a equal basis [16].


Germany primarily

uses cellulose fibers, but also permits mineral fibers, as well as siliceous acid and polymers

ra- L tl I 11 I. .. .-^ __ 1- t-- ....

European asphalt cements are graded by the penetration test.

utilize a 60 to 70 penetration asphalt for SMA mixtures [6,9,10,12].

Most countries

In some cases a

lower penetration asphalt is used for heavy trafficked pavements [13,14], or higher

penetration asphalts are used for low volume roads or cold climates [9,10,12,13].

Design of SMA Mixtures

The European approach to design of SMA mixtures has been largely based upon

rules-of-thumb and recipes for selecting and proportioning components.

The literature

indicates that proportioning of aggregates has been controlled to a large degree by narrow

gradation bands developed from years of experience in each country.

Most of the

countries have several SMA gradations which are identified by the maximum particle size.

Germany, for example, has gradation specifications for SMA mixtures with

mm, 8 mm,

and 11 mm top size.

Sweden utilizes larger top sizes of 12 mm and 16 mm.

The 30-20-

10 rule (30% passing the 4.75 mm sieve, 20% passing the 2.36 mm sieve, and 10%

passing the 75 Cjm sieve) also appears to be used to determine aggregate proportioning in


The practice of selecting the design binder content varies among the European

countries, although all appear to require at least 6.0 percent binder in SMA mixtures.

some countries, such as the Netherlands [15], a specific binder content for each gradation

is also specified as part of the recipe.

In Sweden [17] and Denmark [10], minimum binder

practice is apparently to fix the binder content and to vary the gradation to achieve

percent voids using the Marshall procedure.

2 to 4

Although none of the available European

references stated how many blows are used in the Marshall compaction, it is assumed by

nearly all U.S. references that 50 blows-per-face is used.

Other than air voids, the

European specifications did not identify other volumetric requirements, except for

Denmark which requires a minimum VMA of 16 and voids filled between 78 and 93


Stability and flow requirements are apparently not utilized by any of the


Several references [2,19],

in fact, cautioned that misleading conclusions could

be made by comparing low stability and high flow values obtained for SMA mixtures to

properties of dense-graded mixtures.

2.5 The European Asphalt Study Tour

In September 1990, a group of U

pavement specialists toured six European

countries to observe and learn about the asphalt pavement technologies, practices, and

experiences that could be brought back to the U.S.

The group consisted of 21 participants

from FHWA, the paving industry, state highway agencies, and research organizations.

The group was very impressed with the European pavements and shamelessly admitted

that U.S. pavements were inferior [3].

One of the reasons that many of the European

pavements performed so well was attributed to the durable SMA surface mixtures.

the return of the study tour, the heralded SMA technology has been aggressively


Early SMA Proiects in the United States

During 1991, four states (Wisconsin, Georgia, Michigan, and Missouri)

constructed SMA test sections. Since then the number of states and test sections has

steadily increased.

Table 1 summarizes the SMA test projects constructed in the United

States. through 1993.

In 1994, the total tonnage of SMA mix

projected to be placed in

the United States was over 240,000 tons.

Development of United States SMA Guidelines

In the United States, guidelines for the selection of materials and the design and

construction of SMA mixtures have been developed by a Technical Working Group

(TWG) sponsored by the Federal Highway Administration and the National Asphalt

Pavement Association.

The SMA TWG is a panel of knowledgeable pavement engineers

that was organized to provide guidelines for the transfer of the European SMA technology

to the United States.

The SMA TWG has developed model specifications for the materials, design

requirements, and construction of SMA pavements.

Early versions of the model

specifications attempted to duplicate, as closely as possible, the specifications used in


However, this was not a simple task due to the fact that each of the European

countries appeared to have unique requirements and often used different test methods.

Since Germany and Sweden have the most SMA experience, their specifications were

contractor "secrets"

. In the United States, it is not appropriate to specify particular


The development of the model specifications has therefore been an evolutionary

process in which experience with the technology and materials in the United States has

provided a basis for adjusting the requirements.

The current version of the model

specifications is divided into three areas: mix design, aggregate requirements, and

construction guidelines.

The specifications for each of these areas are presented in the

respective chapters.

Table 1. U.S. SMA Test Sections and Projects Through 1993


1991 WISCONSIN 1-94 1,000
GEORGIA I-85 3,000
MICHIGAN M-52 3,000
MISSOURI I 75, St. Louis 1,000
1992 INDIANA 1-70 1,000
CALIFORNIA 1-40 1,000
MARYLAND US-15 10,000
I 1-70 25,000
GEORGIA I1-75 2,000
NEW YORK Rt. 281 1,000
MICHIGAN 1-94 12,000
MISSOURI I 70, Jefferson City 4,000
OHIO US-33 16,000
TEXAS I-35/SH 171 8,000
VIRGINIA US-29 2,000
WISCONSIN 1-43, 3,000
1993 ARKANSAS 1-40 11,000
CALIFORNIA Rt. 152, Santa Clara 1,000
GEORGIA 1 95, Savannah 62,000
ILLINOIS 1- 80 12,000
57 5,000
55 4,000
US 24 8,000
US-36 16,000
Rt. 121 3,000
Rt. 1 5,000
Lamont Rd., 11,000
KANSAS US-54 1,000
MARYLAND 1 95, Toll Rd. 55,000
S- 83 14.000

Table 1. continued


1993 MICHIGAN 1-96 9,000
1 94 9,000
US- 131 22,000
MISSOURI I 70, St. Charles 9,000
I 70, Callaway 7,000
I 70, Saline 16,000
NEBRASKA Hwy 75 27,000
TEXAS US-79 5,000
US 323 7,000
US 60/83 1,000
VIRGINIA 1-66 10,000
OHIO US 23, Sandusky 60,000
I 75, Findlay 20,000
WISCONSIN US- 51 5,000
US 63 5,000
US 45 5,000
_1-43 15,000


3.1 Purpose of Mix Design

The design of any asphalt-aggregate mixture is basically the selection of material

proportions to produce a mixture having certain desired properties or characteristics.

Ideally, a mix design procedure should optimize component materials and proportions to

yield a mixture that is resistant to all forms of distress which may reduce its performance

in the pavement structure.

SMA mixtures are intended to be surface mixtures.

obviously have good skid resistance.

Surface mixtures must

Rutting resistance of surface mixtures is also critical

since the highest shear and compressive stresses occur at or near the pavement surface.

Environmental elements (e.g., temperature changes, rain, sunlight, etc.) are also more

active on the pavement surface, therefore, surface mixtures must be able to withstand

these conditions.

In other words, SMA mixtures must be resistant to polishing, rutting,

thermal cracking, ravelling, moisture damage, and aging.

3.2 Current Methods of SMA Mix Design

In following with the European practice, the Marshall procedure has been used to

I 1 1 1 1 r1 .fi- A

1 /^I ms A ,1 TT ". 1 ttI -

Design Procedure:


Marshall Method (50 blows to each side of specimen)

Marshall Parameters:

Voids in Total Mix,


Voids in Mineral Aggregate,



Stability, lbs.
Flow. 0.01 inch

Asphalt Content, percent:

1400 min.

6 min

Schnellenberg Draindown,


0.3 max.

Gradation Range:

SiPv Iesigntin3/4"

No. 4
No. 50
No. 200

Percent Passing


The Marshall method is also used by most states for the design of conventional

dense-graded asphalt concrete mixtures.

Therefore, it is a well known procedure and

most agencies and contractors are equipped to perform the tests.

Many states have

successfully placed SMA mixtures that were designed by the Marshall procedure, although

field adjustments in asnhalt content have been common

the studies.

Each laboratory used 50 blows per face with the standard Marshall hammer

except for Payne & Dolan Inc. who used a rotating base and bevelled foot hammer.

Optimum asphalt contents were selected at three percent air voids.

robin studies are shown in Table

Results of the round-

Notes from the TWG meeting [21] state that the

variability of optimum asphalt content for SMA mixtures is significantly greater than for

dense-graded mixtures.


Results of the SMA TWG Sponsored Marshall Procedure Round Robin

First Set Second Set
Asphalt Asphalt
Agency Content VMA Stability Flow Content VMA
(%) (%) (lbs) (.01 in) (%) (%)
Asphalt Institute 5.2 14.8 1800 10 7.0 20.6
Turner Fairbank 5.2 14.7 2350 13 6.4 19.4
FHWA OTA 5.2 15.0 2300 12 6.6 19.7
GA DOT 5.9 16.0 1650 10 6.9 20.2
MDDOT 5.2 15.0 1880 11 6.1 18.4
MI DOT 4.7 14.0 2300 9 5.9 18.3
MO DOT 4.5 13.5 1900 10 7.3 21.2
NCAT 5.9 16.1 1615 12 7.3 21.0
Payne & Dolan 4.9 13.9 2500 13 5.5 17.7
VADOT 5.8 16.4 1610 11
KY DOT 6_.4 19.5

primarily to the impact compaction of the Marshall hammer which fails to orient the

aggregate particles as does field compaction and create the aggregate structure which is

important to SMA mixture strength.

Therefore, the procedure is merely a volumetric

exercise that is based on an inappropriate compaction method.

3.3 Review of Mix Tests for SMA

3.3.1 Drain-down Tests

A drain-down test is necessary to determine if the stabilizing additive is effective in

mitigating run-off of the binder between production and placement of SMA mixtures.

Schnellenberg drain-down test, borrowed fiom Germany, was recommended in early

versions of the model specifications [4].

The Schnellenberg drain-down test consists of

pouring about 1000 grams of SMA mixture immediately after mixing at 1500C into an 800

ml glass beaker, then placing the beaker and mixture in an oven at 1700C for one hour.

At one hour, the mixture is dumped out of the beaker without shaking or vibrating, and

the weight loss of the mixture is determined.

In 1993, the National Center for Asphalt

Technology (NCAT) began development on an alternate procedure [20].


drain-down test utilizes a cylindrical wire basket rather than a beaker to contain the

mixture while in the oven.

The advantage of this procedure is that the drained binder drips

out of the container which makes determination of the weight loss more straight forward.

"*> r) / T-1-: 1/",,-T,.


unconfined uniaxial creep test to evaluate field cores from Georgia's first SMA project.

Tests were performed at 40 C with axial stress levels of 20, 70, and 100 psi for one hour.

Using criteria developed in the Asphalt-Aggregate Mixture Analysis System (AAMAS)

study [23], Little concluded that the SMA mixtures had excellent resistance to


He added that the test was very demanding for an SMA mixture, and that

confinement would substantially improve the creep response.

In a separate study [24],

Little used the creep test to evaluate laboratory prepared

SMA mixtures. Four inch diameter by eight inch high specimens were prepared using a

kneading compactor. The compactive effort was selected to produce specimens with

approximately four to five percent air voids.

and with 30 psi confining pressure. In the u

applied; for the confined tests the creep load was 50 psi.


In this study, samples were tested unconfined

confined state, an axial stress of 20 psi was

All tests were performed at

In both cases, samples containing a low density polyethylene (LDPE) modified

asphalt and polyester fibers performed well, whereas samples with only fibers failed.

Brown [25] and Brown and Manglorkar [26] also used confined creep tests, but

using different conditions.

Using samples compacted in the Gyratory Testing Machine,

creep tests were performed at 60C with 20 psi confining pressure and 100 psi axial

deviator stress.

Several SMA mixtures were evaluated with variations in asphalt content,

fiber type and content, filler content, and percent passing the No.4 sieve.


;nrl;ratP t t htlip mnvtilrpv nprpt n-nrr rscctrnt trn JfiMnrn itnn at Inrpr aenhalt r nntrntec

around 30 percent.

Optimum filler contents were dependent on each mixture.

particular fiber type appeared to be superior based on the creep test results.

3.3.3 Repetitive Uniaxial Tests

Several research projects have also used repeated-load uniaxial tests.

Brown and

Manglorkar [26] used a repeated-load sequence as a modification to the confined,

uniaxial, creep test.

Test conditions were the same as used in their creep tests (20 psi

confining stress, 100 psi deviator stress, and 600 C) except that repeated loads were

applied using a haversine waveform over 0.1 second followed by 0.9 seconds unload.

Repeated-loads were applied for 3600 cycles, then the sample was allowed to recover for

15 minutes.

The repeated load condition was much more severe than the static-load


Accumulated permanent deformations were generally

to 3 times greater than

for the static creep tests.

The repetitive load tests showed more sensitivity to asphalt

content, but less sensitivity to filler content, fiber content, and fines content.


and Veizer [27] performed repeated uniaxial tests on SMA mixtures

containing various percentages of ground tire rubber.

unconfined state.

They used Marshall specimens in an

Loading was applied with a square wave with maximum stress of 100

psi for 0.2 seconds followed by 1.8 seconds rest.

Tests were performed at

and 400C.

Based on the accumulated permanent strains and the slopes of the permanent deformation

versus cycles plots, they concluded that five to ten percent rubber by weight of asphalt

3.3.4 Loaded Wheel Tests

The Georgia Loaded Wheel Tester (GLWT) was used by FHWA's Turner-

Fairbank Highway Research Center [28,29] and the Georgia Department of

Transportation [30,31] to evaluate the rutting resistance of SMA mixtures.


studies included tests with the GLWT and the French Pavement Rutting Tester to evaluate

the effects of different stabilizers [29] and to compare SMA mixtures to dense-graded

mixtures [28].

The study to evaluate the stabilizers concluded that the laboratory

prepared SMA mixtures had good rutting resistance in both wheel rutting devices and that

the results were not significantly different for the different stabilizers.

Likewise, in the

study comparing DGAC mixtures to SMA mixtures, the results with both devices

indicated that all mixtures were rut resistant and that no statistical difference was evident

between the DGAC mixtures and the SMA mixtures.

The Georgia DOT also used the GLWT to test plant-produced mixtures and cores

from their first two SMA projects.

conventional mixtures were tested.

On the first project, three SMA mixtures and two

Although the GLWT results indicated that all five

mixtures were very resistant to rutting, the SMA mixtures were clearly superior [30].

the second SMA project, plant mix and cores of both conventional DGAC and SMA

mixtures were again tested.

Both mixture types again were very resistant to rutting, but

for this project, the GLWT rut depths for the SMA mixture were slightly greater than for

+1. nc m r / t raI *a 1 .. r n [I 1 1

3.3.5 Gyratory Tests

Compaction characteristics of SMA mixtures using gyratory-type machines have

been evaluated by several researchers.

Brown and Manglorkar at NCAT [26] used several

tests including the Corps of Engineers Gyratory Testing Machine (GTM) to study the

effects of variations in component proportions in SMA mixtures. This study utilized a

GTM equipped with an oil-filled roller, and samples were compacted to 75 revolutions

with a vertical pressure of 120 psi and a one-degree angle of gyration. Three GTM

parameters were evaluated for each sample: Gyratory Shear Index (GSI), Gyratory Elasto-

Plastic Index (GEPI), and the shear stress produced at a one degree angle.

GSI results

indicated that all mixtures were very stable regardless of variations in asphalt content, filler

content, fine aggregate content, and fiber type.

evident between GEPI and mixture proportions.

one-degree angle was found to be more sensitive

Likewise, no relationships or trends were

Gyratory shear strength produced at a

e. Results indicated that increasing the

asphalt content reduced shear strength slightly, but the small strength loss was evidence

that the SMA mixtures were highly tolerant to changes in asphalt content.

The effect of

changes in fiber type, fiber content, and mineral filler content on shear strength was found

to be dependent on the type of aggregate used in the mixtures.

The two FHWA studies by Stuart et al. [28,29] also used the Corps of Engineers

GTM to compare the effects of stabilizers on SMA properties and to compare DGAC

t-h^ 'roi isn+-i, CAL A A wniv+inra

TI,0 cltn 1, cflltn't [4)01 IIA 1002 n naA r rr^ 1 'rartt^v +


the point where the change in density decreases to less than 16 kg/m3 per 100 revolutions.

This typically required between 260 and 300 revolutions.

were measured.

Shear strength, GSI, and GEPI

All three gyratory parameters for each mixture containing different

stabilizers were found to be statistically equal. It was also found that the three parameters

were not significantly affected by change in binder content. This was interpreted as an

indication that binder contents higher than designed by the Marshall procedure could be

used without detrimentally affecting rutting resistance.

The study comparing DGAC mixtures with SMA mixtures [28] used a slightly

different procedure.

The report states that the mixtures were tested at 600C in

accordance with the National Cooperative Highway Research Program (NCHRP)

Asphalt-Aggregate Mixture Analysis System (AAMAS) procedure [23].

Specimens were

tested in the GTM with an oil roller using 120 psi vertical pressure and a one-degree


Samples were compacted to 300 revolutions.

Gyratory Stability Index (GSI) and

refusal density were reported for each mixture.

The study concludes that GSI values for

the DGAC mixtures and the SMA mixtures were statistically equal and that all of the

mixtures were resistant to rutting.

with the DGAC mixtures. Extrac

Refusal air voids were lower in the SMA mixtures than

tions were also performed on the tested samples to

compare aggregate breakdown in the GTM to the Marshall hammer.

These data indicated

that the breakdown caused by the above GTM procedure was nearly identical to the

lrnniA,1, a, ,r nfl. oat 1, *1 A ,1 1a1n,,i p n ,rcninll nn'mnfn49 ra nT+^

kneading compaction and 50-blow Marshall compaction.

Results showed that at equal

asphalt contents, the Texas gyratory produced the lowest air voids, followed by kneading

compaction and then Marshall compaction.

However, the low air void content gyratory

samples produced the highest Hveem stabilities, which according to Little indicated that

when the high binder content SMA mixtures were compacted to ultimate densification

(such as with the Texas gyratory) the mixtures did not suffer a loss of strength.

He added

that the gyratory compactor apparently produces a better interparticle matrix of the coarse

aggregate for load distribution than the kneading compactor.

3.3.6 Resilient Modulus Tests

Diametral resilient modulus tests have been performed on SMA mixtures by

several researchers.

Generally, most researchers have followed the resilient modulus


ASTM D 4123.

Little [22] used the resilient modulus test in the evaluation of

the SMA mixtures used in Georgia's first SMA project.

Using cores taken from the

Georgia SMA test sections, modulus values were determined at three temperatures: 0,

and 40C.

These results indicated that the stiffness of the SMA cores tended toward the

low end of the modulus range typical of DGAC mixtures.

At 400C, the resilient moduli of

the SMA cores were actually slightly below the range for dense graded mixtures.

prompted Little to postulate that diametral testing of SMA mixtures may not be

appropriate at higher temperatures.

Little also evaluated other tensile properties such as

AAMAS procedure [23].

This evaluation led him to conclude that the SMA mixtures had

a favorable fatigue response which was consistent with the concept of using a rich binder


In another report [24],

Little used the diametral resilient modulus to demonstrate

the effects of a low density polyethylene (LDPE) additive on tensile stiffness of an SMA


His data showed a significant increase in mixture stiffness at

and 400C when

4.5% LDPE was added to the binder.

Another interesting note about his results was that

the binder content did not appear to effect the moduli of mixtures containing a 0.15%

polyester fiber alone, but when the fiber and LDPE were added together, the stiffness

peaked at 5.5% binder content.

Brown and Manglorkar [26] also included diametral resilient modulus testing in

their research to evaluate variations in mix proportions.

They found no consistent

correlations between moduli and fiber content, dust content, fine aggregate content, or

asphalt content.

The lack of any significant trends was attributed to high variability of the

SMA mixtures prepared with a granite aggregate typically had moduli approximately

equal to that of dense-graded mixtures; whereas those containing a gravel aggregate had

moduli lower than typical dense-graded mixtures.

Stuart and Mogawer's comparison of DGAC and SMA mixtures [28] included the

resilient modulus test to evaluating several mixture characteristics.

They performed

ratfl~an rntnnun,. +nn+ nn '112. A/"^ A* A OX A A eb.terd..a-a n* i/"\ A- ri.4-3-a...--S -

moduli were statistically equal for SMA and DGAC mixtures.

Resilient moduli of

mixtures before and after moisture conditioning were also evaluated to



damage potential.

This analysis indicated that the SMA mixtures were less susceptible to

moisture damage than the DGAC mixtures.

Resilient modulus ratios for the SMA

mixtures were greater than 70 percent; whereas resilient modulus ratios for the DGAC

mixtures were below

70 percent.

These conclusions were also confirmed by tensile

strength tests.

This same type of analysis was also performed to evaluate changes

occurring upon aging in accordance with the procedures recommended from the Strategic

Highway Research Program.

The conclusions from this part of the study was that tensile

strength and stiffness of SMA mixtures increased significantly less than DGAC mixtures

due to aging, which indicated that SMA mixtures would be more resistant to cracking

after aging.

3.3.7 Tensile Strength Tests

Indirect tensile strength has also been used by many researchers to evaluate SMA


Most researchers have measured indirect tensile strength at 25 C using a

vertical deformation rate of 2 inches per minute.

Brown and Manglorkar [26] found that tensile strengths of SMA mixtures were

lower than DGAC mixtures.

Increasing the asphalt content and filler content of SMA

mixtures did result in a gradual increase in tensile strengths.

Likewise, shifting the

sieve) also increased the tensile strength.

Fiber content and fiber type, however, did not

appear to have a significant effect on tensile strength.

Little [24] used the indirect tensile strength and the strain (horizontal) at failure to

evaluate several mixtures.

He found that LDPE and fibers added to an SMA mixture

substantially increased tensile strength but no statistical difference was evident in strain at


He also measured tensile strengths and strains at failure at 25 C using a

deformation rate of 0.5 inches per minute and found the same trends.

Compared to typical

DGAC mixtures, Little concluded that SMA mixtures had about the same tensile strength,

but higher strains at failure.

However, he also made several precautionary comments on

the use of the indirect tensile testing mode for SMA mixtures, stating that "this type of

testing may not be indicative of the relative response of SMA mixtures to traditional

densely graded mixtures in the field...Probably the most effective use of the IDT (indirect

tensile) test is to simply compare candidate SMA mixtures and not to compare the results

of SMA and densely graded mixtures."

The indirect tensile test has also been used to evaluate moisture damage potential

of SMA mixtures by several researchers.

This type of evaluation typically consists of

comparing the tensile strengths of two sets of samples: one set conditioned by water

saturation, soaking in a 600C temperature bath and then, in some cases, one or more

freeze-thaw cycles, the other set unconditioned.

A A CTT-rr'." P _-a A P fT'l T' AOCT 'Tl^ /"l

Standardized procedures for this test are

-rnn T ..nnA *L<^ nrannA. .rc an nlnt .'nh,

3.3.8 Diametral Creep

Diametral creep, also known as indirect tensile creep, was measured by Little

[22,24] on cores from the first Georgia SMA project and on laboratory prepared samples.

Results on the cores, apparently at low test temperatures, showed that the SMA mixtures

deformed more than typical dense-graded mixtures which indicated a greater ability of the

SMA mixtures to relax and shed tensile stresses during a decrease of pavement


The laboratory prepared samples were tested at warmer temperatures and

used to evaluate relative differences in fatigue resistance.

Little concluded that the SMA

mixtures are more resistant to fracture and crack propagation than typical dense-graded


He also stated that polymer modification was necessary to keep tensile strains at

a reasonable level to reduce accumulated damage due to distortion [24].


4.1 Introduction

There are several reasons to develop an improved procedure for SMA mix design.

The Marshall procedure, although it is familiar, has some deficiencies including hammer

compaction, high variability, and probably most significant, it lacks a measure of

deformation resistance.

The recipe based formulation of SMA mixtures may provide some

assurance based on historical performance, but this can also be restrictive to alternative

component blends which may perform better or be more economical.

Mix design should

include tests that indicate which materials and combinations are superior in terms of

deformation resistance, cracking resistance, and moisture damage resistance.

The testing

program for this study was set up to identify a set of test methods which better represent

the important mixture placement and distress conditions which exist in the field.

Selection of Tests

Selection of the tests to be evaluated for potential use in SMA mix design was

based on a variety of factors. It was desired to have at least one test to evaluate each of

the major forms of pavement distress. It was important that the test simulate field

-- -: -- .a t. -- t.--: -- -i-:- --*--- ci ,-_ _- -^-- i-- t.. 1 -

From a general review of recent literature, some of the most promising

performance related test procedures were identified.

Many of the tests were identified

from the SMA literature review presented in the previous section.

Other promising tests

had been used for evaluation of dense-graded asphalt mixtures, but had not yet been used

for testing of SMA mixtures.

The final factor, or constraint, used to select tests for this project was equipment


Funding for this research project did not include any funds for the purchase of

equipment or for subcontracting tests to other laboratories.

Therefore, it had to be

possible to perform the selected tests with equipment available at the University of Florida

Civil Engineering laboratories or the Florida Department of Transportation's Central

Bituminous Laboratory or Bituminous Research Laboratory.

The selected tests, categorized by distress mode, are shown in Table 3.

Three tests

were selected to characterize rutting potential: the GTM densification/shear test, the

Georgia loaded wheel test, and the triaxial repeated-load test.

This redundancy was felt to

be beneficial considering the possibility that one of the tests may not be representative of

actual performance, or if all three tests were satisfactory, it would allow other agencies to

determine rutting resistance of SMA mixtures with only one test device.

To provide an

indication of fatigue resistance, two tests were selected: the indirect tensile strength test

and the resilient modulus test.

For evaluation of low temperature or thermal cracking, the

I. 1 1 1 I -

Table 3. Selected Mixture Tests
Distress Test References

Rutting GTM Densification/Shear Ruth et al [32]
Georgia Loaded Wheel Tester Lai [33], West et al. [34]
Triaxial Repeated Load Brown [26], Gabrielson [35]

Fatigue Cracking Indirect Tensile Strength & Strain Von Quintus et al. [23]
Resilient Modulus

Thermal Cracking Diametral Creep Von Quintus et al. [23]
Low Temperature Resilient Modulus

Moisture Damage Wet/Dry Resilient Modulus Von Quintus et al. [23]
Wet/Dry Indirect Tensile Strength AASHTO T 283 [36],
ASTM D 4123 [37]

Compaction of all test specimens, except for the Georgia Loaded Wheel test

beams, was accomplished with the Corps of Engineers Gyratory Testing Machine (GTM).

The GTM compactive effort was adjusted by using different numbers of gyrations so that

the air void content of the compacted samples were as close as possible to the average in-

place air void content for each particular mixture.

For experimental mixtures that did not

have any field data, a standard compactive effort was used that had been selected from the

data generated from the mixtures with field data.

A flow chart showing the general order of testing and number of specimens per

test is shown in Figure

As indicated, test specimens were often used for more than one

This was possible due to the fact that some tests were non-destructive to the


This permitted a significant reduction in the material and sample preparation

T rT ":1L.2I zL ^A--tI.1- --. '-2I ..... 1...... ....... f __

,I..- -


Flow Chart of Testing


compactive effort, the step where the compactive effort is determined (shown as the

shaded box) was eliminated.

Prepare Materials,
Determine Targets


Georgia Loaded
Wheel Tests
3 beams

Compact Samples to
Initial, In-place Density
9 specimens

Triaxial Repeated
Load Tests
2 specimens

GTM Densification
/ Shear Test
3 specimens

2 specimens

Creep Compliance
@ -10oC
2 specimens

Indirect Tensile
Strength @ 25C
2 specimens

Resilient Modulus
@ 525 & 40oC
2 specimens


Resilient Modulus
@ 5,25 & 40C
2 specimens

Indirect Tensile
Strength @ 250C
2 specimens



5.1 Materials

5.1.1 SMA Mixtures from Field Projects

Material components for some of the early SMA mixtures placed in the U.S. were

obtained in cooperation with several highway agencies and paving contractors.


mixtures from eleven field test sections were duplicated in this study.

For each mixture, the average of the field construction data (gradation, binder

content, and in-place density) was used as the target for preparing the laboratory test


Table 4 includes descriptions of the components and the field data used to

prepare the laboratory mixtures.

Complete descriptions of each mixture are included in

Appendix A.

For the purpose of identification, each mixture was given a "Mix I.D." which

includes the state abbreviation for easy recognition.

For states that have had several SMA

mixtures placed, the number following the state abbreviation should correspond to the

chronologic order in which each mix was placed.


This is a coarse SMA mixture that was used as a binder layer on Georgia's

first SMA project which was placed in late July 1991 on Interstate 85 northeast of Atlanta



N r^ oo

bCW P-

\ A^ IT e, da I I
0 0 9Rws^^zRs
1-1 ** yis y w -


consisting of a marble dust mineral, filler mineral fibers, and an AC-20 modified with 5%

low density polyethylene (LDPE).

This mixture is unique among the other mixtures

included in this study due to its gradation and its use as a binder layer.

The top size

aggregate used in this mixture was


Although specific performance data is not

available on this mixture due to its location in the pavement structure, the pavement as a

whole is performing very well.


This mixture contains the same component materials as GA1,

but has a

much finer aggregate gradation [30].


The top size aggregate used in this mixture was 3/%

This mixture, placed on Georgia's first SMA project in September 1991, was used

as a surface layer.

The available performance data (Table 5) shows that the section is

performing very well with very low rut depths and excellent skid test results.



Performance Data for GA2

This is the third SMA mixture placed on Georgia's first SMA project [30].

This mixture contains a different granite aggregate (Ruby mine) which has a lower LA

abrasion value than the Buford granite.

Compared to GA2, the gradation of GA3 is

Date (Age) April, 1993 (1.6 years) June, 1994 (2.75 years)

Average Rut Depth (inches) 0.01 0.09

Average Friction Number 52 not available

Average Air Void Content (%) 7.3 5.8

section is also performing very well.

Rut depths are practically negligible and skid test

results are very good.

Table 6. Performance Data for GA3
Date (Age) April, 1993 (1.6 years) June, 1994 (2.75 years)

Average Rut Depth (inches) 0.05 0.09
Average Friction Number 48 not available

Average Air Void Content (%) 7.6 8.0


This SMA mixture was placed on Georgia's second SMA project which

was constructed on Interstate

mines in Tyrone and Mt. Viev

values for the Tyrone and Mt.

south of Atlanta in 1992 [31 ].

v, Georgia were used in this mi

View granites were 41 and 37

Granite aggregates from

e. The LA abrasion

, respectively, which are well

above the recommended maximum limit of 30, and are believed to be the highest used in

an SMA project to date.

used in GA3.

The gradation for the GA4 mixture was nearly identical to that

The mixture also contains marble dust mineral filler, domestic cellulose

fibers, and an SBS modified AC-20 special.

A thin overlay was placed over this project in

late 1992 due to motorists overreacting to the change in surface texture of an adjacent test

section. There were no signs of distress in the SMA or other sections at the time of the


Consequently, no performance data was available for this section during

preparation of this dissertation.

filler, and an 85-100 penetration asphalt.


A pelletized polyolefin was added as a stabilizing

The gradation of this SMA mixture is one of the few included in this study which

meets the TWG's recommended gradation band.

occurred in an area where a vibratory roller was t

Brown [1] reported that flushing had

ised. Early friction test results showed

that the SMA had lower skid numbers than the adjacent conventional mixture.


discussions with Wisconsin DOT personnel indicate that the test section is performing


No other performance data was available.


The second SMA mixture from Wisconsin was placed on Interstate 43 in

Waukesha County during August, 1992.

This mixture contains specially screened

limestone aggregates from Crowbar quarry, a limestone dust mineral filler, and an 85-100

pen asphalt.

Several test sections were constructed with this base mixture.

Each section

contains a different stabilizing additive. The mi

polyolefin additive was duplicated in this study.

xture from the section containing 7%

A unique feature of this mixture is that

the gradation has a 3/% inch top size rather than a

inch top size.

Construction data

indicate that this mixture had high in-place pavement air voids.

The average in-place air

void content for this SMA pavement was 10.5 percent, whereas most other SMA

pavements in the U.S. have had initial air void contents in the range of 5.5 to 8.0 percent.

Post construction performance of this pavement has not been reported.


c ',l,-_,1- 1 r

This SMA mixture was placed on Ryan Road in the Milwaukee area during

I C^. a11, ni,..n-n.n.naA., 1n-tr,-,,+nn*-,A n-ran.,fa, o CRrtnrn raynlrlin nntornr lirMaectnn

mixture was very near the same as used in WI2.

No performance information has been

reported for this SMA section.


This SMA mixture, which was also placed in September 1993, was used on

a project in Three Lakes,

Wisconsin on US 45.

This mixture contains a crushed gravel

aggregate from Consolidated mine. A limit

the binder was a 85-100 pen asphalt. Six t

additives, were constructed on the project.

stone dust was used for the mineral filler, and

:est sections, each with different stabilizing

Only the mixture using inorganic fibers was

duplicated in this study since the construction data showed that this mixture had the least

variation during production.


This SMA mixture was placed on a project on US 29 in Lynchburg,

Virginia in the fall of 1992 [38]. This mixture contains specially screened aggregates

quarried from an anorthosite traprock. Agricultural lime from an argillaceous marble was

used as the fine aggregate and mineral filler.

AC-20 was used as the binder.


sections were placed on the project: one containing polyolefin and the other containing

cellulose fibers.

The mixture containing cellulose fibers was duplicated in this study.

August 1993, flushing of the binder to the pavement surface was observed in isolated

areas of both sections.

A brief investigation [39] indicated that the mixtures were close to

the job-mix-formulas and no definitive reason could be given for the flushing.

Despite the

flushing, no rutting had occurred in the SMA sections.

'k1 rim. o' EA _- .1 --TT it m. rE_- -i. 1. ~I.._i-2 2

mineral fill'



Test sections were constructed with cellulose fiber and polyolefin stabilizing

The mixture containing the polyolefin additive was evaluated in this study.

Shortly after the sections were constructed, the surface of the cellulose section appeared

"tight" or seemed to have less texture.

However, no rutting was reported.

More recent

discussions with Maryland DOT personnel indicates that the SMA sections are performing


No performance measurements have been reported.


This SMA mixture was placed in August 1992 on the business loop of


in Alvarado, Texas [40].

The mixture was also placed on Texas State

Highway 171 near Cleburne.

Components of the mixture were limestone coarse

aggregate, limestone dust mineral filler, AC-20, and pelletized cellulose fibers.

gradation of this mixture was within the range recommended by the SMA TWG.

Discussions with Texas DOT personnel indicate that the SMA is performing well.

measured performance data is available.

Field performance. In general, all of the SMA mixtures obtained from other

states have performed well in the field. Some states have reported isolated areas of

flushing of the mastic to the pavement surface, but in each case the flushing was not

associated with any rutting. At the time this dissertation was prepared, very little

performance data was available. Several of the mixtures included in this study have been

in service for over three years.

5.1.2 Experimental SMA Mixtures with Florida Aagregates

It was necessary to include in the study aggregates that are available within the

state of Florida to evaluate their potential for use in SMA mixtures.

Six sources of coarse

aggregate were selected representing the range of properties for materials currently used

in asphalt mixtures in Florida. An attempt was made to include aggregates covering the

geologic and geographic areas of the state. Brief descriptions of the selected coarse

aggregates are given below.


Each description includes a Mix I.D.

The order of the I.D.

to simplify

numbers is not significant; the mixtures were simply

numbered in the order that they were tested.


Limestone aggregates from Dade County (FDOT mine no.

87-339) were

used in this mixture.

This and other similar aggregates mined from the Fort Thompson

and Miami Limestone Formations in Dade County

are widely used as construction

aggregates in southeast Florida and along the east coast of the state.

This aggregate

possesses unique wear characteristics due to the significant silica content of this limestone



Therefore, it is used extensively for friction course mixtures throughout the

The aggregate tends to be chalky and typically has an LA abrasion value of 31,

and water absorption of about 3 percent.


This mixture contains limestone aggregates mined from the Ocala and

Suwannee Limestones in Hernando County (FDOT mine no.




nancr4nra t ,ryCt1 rolotrt nli, nnro mnno 1tnna r iCrn th a noa?_antral naaf+ n0^fthia

abrasion for this aggregate is typically about 36, and water absorption is about 2.3


This aggregate also tends to be chalky.


This mixture contains a dolomitic limestone coarse aggregate mined from

the Ocala and Suwannee formations in Taylor County (FDOT mine no. 38-268).

aggregate source is one of the few limestone mines in the northern part of the state.

abrasion for this aggregate is usually around 40, and water absorption about 2.7 percent.

The aggregate contains more porous particles than the calcium carbonates from mid and

south Florida.

It is presently considered as marginal in quality relative to the other

aggregates mined in the state.


This mixture contains aggregate derived from a crushed concrete pavement

taken from Interstate 75 in Hillsborough County.

FDOT mine no.13-395 in 1991.

This crushing operation was given

LA abrasion for this crushed concrete aggregate has

averaged around 48, and water absorption is about 5.1 percent.

results this aggregate would be considered poor quality. Howe

Based on these test

ver, it is apparently being

used successfully in conventional asphalt mixtures.


The aggregate used in this mixture is from a dolomitic limestone quarry in

the Newala Formation in Shelby County, Alabama (FDOT mine no. AL-149).


aggregate is included in the study because it is commonly used in asphalt mixtures in the

panhandle of the state.

.. ... 1 A- n a..-^ 4- 1

It typically has an LA abrasion value of 23, and water absorption

i .:.l r,. .- .-.: -- i- _. 1_ -_-i -- 1 at,- -- .i


This mixture contains crushed river gravel aggregates obtained from

deposits along the Chattahoochee River in Gadsden County (FDOT mine no. 50-120).

This mostly quartz gravel is mined from the Citronelle formation. Th

approved, non-limestone source of coarse aggregate within the state.

used for asphalt mixtures in the central part of the panhandle. Due to

is is the only FDOT

It has been widely

)its polish resistance,

it was also used as a coarse aggregate in open-graded friction course mixtures for over ten


It was disallowed in this application in 1993 due to problems with ravelling.

aggregate typically has an LA abrasion value of 41, and water absorption of about 0.4


Other components of experimental mixtures.

In order to evaluate the differences

in mixture characteristics which were due only to the aggregates, the gradation and the

other mix components were kept constant.

The gradation used for all experimental mixtures, shown in Table 7

was near the

middle of the gradation range of the early versions of the FHWA's model SMA

specifications [4].

This gradation is similar to that used in mixture WI1.

In July 1993, the

range for passing the No. 4 sieve was adjusted to 20

- 28 percent by the TWG [20].

Therefore, the gradation of the experimental mixtures is slightly above the current

recommended range at the No.

4 sieve.

Generally, to meet this gradation with the different aggregates, it was necessary to

C.~, c ct..2 a n,,1 a a. ay a in is. a pE. .t .a a Aa n .L to a a a aaSr a

Table 7

Interpolated from Standard

. Gradation Used For Experimental SMA Mixtures


The mineral filler used in the experimental mixtures is a marble dust, known as

R04, obtained from Georgia Marble Co.

in Marble Hill

This is the same filler that

was used in all of the Georgia SMA mixtures tested in this study.

Characteristics of this

mineral filler are given in Table 8.

The asphalt cement selected for the experimental mixtures was an AC-30 from

Coastal Fuels' terminal in Jacksonville.

Forty gallons of this

asphalt were obtained on

September 14, 1992. This asphalt was tested according to the SHRP binder specifications

and graded as a PG 64 -22.

Cellulose fibers and FDOT Type A ground tire rubber were used as stabilizing

additives in the experimental mixtures.

Cellulose fibers, CF31005,

from Custom Fibers

Sieve Gradation (Percent Passing) TWG Spec. Range
3/4 (19.0 mm) 100 100

1/2 (12.5 mm) 95 85-95

3/8 (9.5 mm) 70 60 -75

No.4 (4.75 mm) 30 20 -28

No.10 (2.0 mm) 20 15 -23

No.40 (425 pm) 14 12- 15.5'

No.80 (180 pm) 11 10-13'

No.200 (75 pm) 9 8- 10

Table 8.

asphalt cement.

Mineral Filler Properties

Half gallon quantities of GTR and asphalt were preblended at about

F with a high speed paddle mixer for about five minutes to achieve uniform

dispersion and reaction of the rubber and asphalt.

Test Procedures

.2.1 GTM Compaction

Gyratory compaction was selected based on several studies which have shown that

this mode of compaction best simulates field compaction by rollers in terms of the

resulting density, particle orientation, and engineering properties of the compacted

samples [23].

Although these studies were performed using traditional dense-graded

mixtures, it is Presumed that the conclusions apply to SMA mixtures as well.

Due to the

Composition Mg,CaCO3 95% min. Hydrometer Analysis
Acid Insolubles 3% max. Grain Diameter (mm) Percent Finer
pH 9.0-9.5 .075 99
Specific Gravity 2.710 .034 47
.022 31
Sieve Analysis
.013 20

Sieve (Opening) Percent Passing .010 13
No.40 (.425mm) 100 .007 10
No.80 (.180mm) 95 .003 5
No. 200 (.075mm) 71 .001 2


For this research, a gyratory compaction procedure initially developed by Ruth and

Schuab [41] was used. This procedure, which utilizes a GTM equipped with an air roller

(Figure 6),

is intended to compact the samples to the initial in-place density following

construction, typically between 92 and 94 percent of the theoretical maximum density.

The settings used to achieve this level of compaction with the GTM Model 6B/4C are 100

psi ram pressure, 3 degree initial angle of gyration, and 9 psi initial air roller pressure.

Work by Ruth et al. [32] determined that 18 gyrations generally produced the desired

density for many dense-graded mixtures.

To establish the compactive effort necessary for

the SMA mixtures in this study, compaction curves were developed for each mixture.

Three samples for each mixture were prepared and compacted in the GTM to

determine the number of gyrations necessary to achieve the average initial in-place density.

Each sample was compacted to a different number of gyrations.

Sample heights, recorded

with each gyration, and the bulk specific gravity of the sample measured following

compaction were used to backcalculate densities at intermediate numbers of gyrations

using Equation 1.


Gmbl (-)


Gmb, = bulk specific gravity of compacted mixture at x gyrations
,,,.., L .. 1. 1.. 1SA a .. : :,n aC1a aAaa. aan.







Figure 6.

Schematic of the Gyratory Testing Machine

An example of the compaction results is shown in Figure 7

. This graph of %Gmm

versus gyrations shows how the number of gyrations to achieve the average, initial, in-

place field density was selected.

GTM Densification/Shear Test

The GTM densification/shear test, which was also initially developed by Ruth and

Schuab [41], is intended to simulate traffic densification.

This procedure was

recommended as a part of the AAMAS study [23] for the evaluation of rutting potential

for dense-graded mixtures.

The test was performed at 600C to be representative of the

pavement during the hottest part of the year.

The densification/shear test was performed

on previously compacted samples that were left in the molds and using the same GTM

settings as used in compaction: 100 psi ram pressure, 3 degree initial angle of gyration,

and 9 psi initial air roller pressure.

Generally, the samples were densified to 300 gyrations

or until a significant loss of shear strength occurred.


2 was used to calculate the

gyratory shear strength of the mixture during the test.





= gyratory shear strength at gyration

ARP-= Air Roller Pressure at gyration

- 1. atF1 nnIUY~ 1 aC n*~C rn

x, psi

x, psi

* C.. ...



r* ^^i
U*) a)


CV )

(N Ir;I .


| 1 | |co 3

O ~ti
(0 H

o SQ
w "4 U*
;~ i I :r i ;

O 'F I I O

C\ ) 0 a
l I : k-

O 4-


E r3

\ i : 'a~ ,Q

The change in the percent of air voids (i.e. Percent Densification) during the test

was calculated using equation 3.

= %Grmm -


= %AV-




percent densification at gyration


= percent of maximum specific gravity at start of the densification test


nx = percent of maximum specific gravity at gyration
= percent air voids at start of the densification test

= percent air voids at gyration

Typical results from the GTM densification/shear test are shown in Figure 8.


Georgia Loaded Wheel Tester

The Georgia Loaded Wheel Tester (GLWT), shown in Figure 9, was developed by

Lai [33,42] for the Georgia Department of Transportation to evaluate the rutting potential

of asphalt mixtures.

Several studies have shown that results of the GLWT can accurately

discriminate the rutting potential of dense-graded mixtures relative to their actual known

field performance [33,34].

The Georgia DOT has reported very good results from GLWT

tests on their SMA mixtures [30,31].

The GLWT test specimens in this study were 1.5"x3"x15" asphalt concrete beams.

The prepared materials were heated to 300F, mixed, then placed in molds and returned to

the oven prior to compaction.

Compaction was accomplished with a hydraulic

^rnnr~f^"nn ractsr n mot'h*n 1nrrr .~nnIvna I r *k r nina1 fl1n4n^ nn* vn.. 4nn. a^-frt*^ L.La t


8 0
2o gn 40



Figure 9.


Illustration of the Georgia Loaded Wheel Tester


0 1 2 3 4 5 6 7 8 9

CYCLES (Thousands)

fifth application, the load was held at 60,000 lbs. for five minutes.

The compacted beams

were removed from the molds the next day and allowed to cure at room temperature for

one day. During the curing period, bulk density and air void content for each beam was

determined. Prior to testing, each beam was placed in the preheating chamber at 105F

for 24 hours.

The beams were rigidly confined on the sides and ends with angle brackets during

the test.

A stiff hose pressurized to 100 psi was mounted along the top of the beam acts

as a tire to transfer the load from the wheel of the moving chassis to the beam.

chassis, loaded with 122 pounds, was driven by the motor / gear-box assembly through the

articulated arms.


One loading cycle consisted of a forward and return pass of the loaded

The entire GLWT assembly was housed in a heated and insulated temperature

chamber which maintained the test temperature at

within the chamber to circulate the air. Rut depth

105 +3 F. Small fans were placed

measurements were made at three

locations near the center of the beams with a dial gauge and reference template at 0, 1000,

4000, and 8000 cycles.

Figure 10 shows how typical results for the Georgia Loaded Wheel Tester are

graphed to illustrate rut depths versus cycles.

A criteria for maximum acceptable rut

depth for dense-graded mixtures was tentatively set at 0.15 inches at 8000 cycles by

West et al. [34].

5.2.4 Triaxial Repeated-Load Test

Triaxial repeated-load tests have been used by a number of researchers to evaluate

the permanent deformation (rutting) potential of asphalt mixtures.

Gabrielson [35]

evaluated several rutting tests using dense-graded mixtures with known performance and

concluded that the triaxial repeated-load test best models in-situ permanent deformation.

A good correlation was shown to exist between laboratory strain and rutting rate (rut

depth/square root of ESAL's in millions).

The same test configuration was used by Brown

and Manglorkar [26] to evaluate SMA mixtures with variations in mixture components.

The triaxial repeated-load test consisted of applying a constant confining pressure

to a standard cylindrical specimen and a repeated axial stress for 3600 cycles to simulate

traffic loading.

The test conditions were set to realistically simulate the critical field

conditions which lead to rutting as close as possible.

The test temperature was 600C to

simulate high pavement temperatures. The axial deviator stress was applied using a 0.1

second haversine load period followed by a 0.9 second rest period to simulate 60 mph


Although a single confining pressure cannot replicate the horizontal stress

distribution existing in a pavement, Gabrielson found that 20 psi confining pressure gave

very low variation in test results.

The deviator stress level was set at 100 psi to give a

total axial stress of 120 psi which is typical of the high contact pressures existing beneath

the pavement and tires of heavy trucks.

Figure 11 is an illustration of the triaxial repeated-

1 1 1 1




Figure 11. Illustration of the Triaxial Load Test Apparatus

Specimens for the triaxial repeated-load test were standard 4 inch diameter by 2.5 inch

thick specimens compacted in the Gyratory Testing Machine.

smoothed with sandpaper.


Specimen ends were

Prior to testing, specimens were preheated to 600C for 4 to 6

Thin Teflon sheets were also placed between the specimen and platens to eliminate

shear stresses on the specimen faces.

A rubber membrane was placed around the

specimen and platens and held by o-rings at the top and bottom platen.

Once the rubber

membrane was in place, the triaxial cell was placed over the specimen and the confining

pressure was applied.

The entire triaxial assembly was then placed in the test

environmental chamber and allowed to stabilize at 60C for 20 to 30 minutes. Following

the temperature equilibration time, 100 cycles of conditioning load was applied. The

conditioning load was set to produce 20 psi of axial deviator stress.

The cyclic deviator stress was applied with a MTS servo-hydraulic test station.

Axial deformations were measured by two independent LVDT's mounted inside the

triaxial chamber on opposite sides of the specimen.

Test control and data acquisition was

accomplished through a MTS Model 458.20 microconsole using modified MTS software

on a 386 P(

Figure 12.

An example plot of results from the triaxial repeated-load test is shown in

The slope of the permanent strain versus cycles (i.e. permanent strain rate) was

used to characterize rutting potential of the study mixtures.

Permanent strain rates were

calculated using Equation 4.

C .


Sp = Permanent Strain Rate, in/in/cycle
hx = height of sample at cycle x, inches
h, = height of sample at end of conditioning, inches
cx = cycle number

Permanent Strain (in/in)

800 1200 1600 2000 2400 2800 3200





5.2.5 Indirect Tensile Strength Test

Tensile properties of the SMA mixtures were determined using the indirect tensile

(diametral) test for evaluating cracking potential and moisture damage resistance.

indirect tensile strength tests were performed at 77F

(25C) following the loading

guidelines of ASTM D 4123 with a ram rate of

2 inches/min (50 mm/min).

Tests were

performed on GTM compacted samples using an MTS servo-hydraulic testing machine

with a model 458.20 controller. Horizo,

model 632.94E-20 extensometer fixture.

illustrated in Figure 13.

ntal deformations were measured using an MTS

The test set-up for the indirect tensile test is

Test control, data acquisition and calculations were performed by

software originally developed by MTS.

The program for indirect tensile strength was

modified to capture horizontal strain data and to calculate fracture energy.

strengths and strains at failure were calculated using Equations


and 6, respectively.

Fracture energy for each test was determined as the area beneath the stress-strain curve to

the point of failure.

2P dt


r = tensile strength, psi
I = ultimate applied load required to fail sample, lbs
'= diameter of sample, inches
C C fl S 4

0T =


f= strain at failure, %
6h = horizontal deformation corresponding to Pu,,, inches





Figure 13. Illustration of Diametral Testing Configuration

Resilient Modulus

Total resilient moduli of the GTM compacted mixtures were determined in

accordance with ASTM D 4123. Mod

77F and 1040F (5C, 25C, and 400C).

uli were determined at three temperatures: 41F,

Assumed Poisson's ratios were input for each

temperature to avoid errors resulting from inaccurate vertical deformations caused by

.A1 I 1I I I 1*



1 -_~ 1.. 1 _- -_ _I 1iir


+ 0.27)


MTr = total resilient modulus, psi
P = repeated load, lbs
v = Poisson's ratio, assumed v=0.2 at 50C, v=0.35 at 25C, and v=0.5 at

= total recoverable horizontal deformation, inches

5.2.7 Creep Compliance

Static diametral creep tests were performed using guidelines taken from several

sources including the AAMAS procedure [23] and the provisional AASHTO standard

TP9 [43] developed in the Strategic Highway Research Program.

A program was

developed to provide test control, data acquisition, and perform calculations for the

creep compliance test.

All tests were performed at 14F (-10C) by applying a static

load to the specimen's vertical diameter through 0.5 inch loading strips as used in the

indirect tensile and resilient modulus tests.

The magnitude of the load was set to

produce between 50 and 200 horizontal microstrains at 60 seconds.

The load was

maintained for 1000 seconds, then removed and recovery of the horizontal strain was

recorded for an additional 1000 seconds.

Creep compliance was calculated using



= Creep Compliance at time t, square inches per pound
, = horizontal deformation at time t, inches

= 100

- 2000

- 1000


= percent recovery

h t=10oo00

= horizontal deformation at end of unload period
= horizontal deformation at end of load period

5.2.8 Moisture Conditioning

To evaluate the potential for moisture damage, a.k.a. stripping, of asphalt

mixtures, wet/dry tensile strength tests or resilient modulus tests are commonly performed.

Standardized procedures for the wet/dry indirect tensile strength tests include

ASTM D 4867 [37] and AASHTO T 283 [36].

In this study, comparisons were made between wet (moisture conditioned) and dry

(unconditioned) resilient moduli and indirect tensile strengths.


showed how the

A --A-_. -----------t-- -t. .- a i -- -- n nna. n4 pa


Moisture conditioning consisted of vacuum saturation in tap water until specimens

reached 60 to 100 percent saturation.

vacuum of 100 mbars.

Generally, this required one to two minutes under a

For specimens with lower air voids, the vacuum time was extended

to meet the saturation requirement. Following vacuum saturation, the specimens were

placed in a 600C water bath for 24 0.5 hours. Testing at 25 C was initiated within 24

hours after removal from the water bath.


6.1 GTM Comoaction

Compaction results of the eleven duplicated field mixtures are shown in Table 9.

The number of gyrations to achieve the field density (N1.) was estimated from the

compaction curves generated from three samples typically compacted to 12, 15 and 18


Estimated values of N, ranged from

to 20, with an average of 12.1 and a

standard deviation of 4.5.

Excluding the data from WI2, which was apparently

undercompacted in the field, does not significantly alter these statistics. The average

initial in-place air void content for the field SMA mixtures was 6.9 percent. Excluding

WI2 does drop the average to 6.5 percent.

The air void content and VMA data for each mixture, shown in Table 9, are the

results of six samples compacted to N e. Several mixtures had very small variations in

air void contents, for others the repeatability was poor. In most cases, the estimated value

of Nf~ did yield sample air void contents that were reasonably close to the target field air

void content for the six compacted samples.

Using the data generated from the six

samples from each mixture, the adjusted estimate of Nh was made, but these values

were only slightly different.

Table 9. Compaction Results for the Duplicated Field Mixtures

Average Compacted Compacted
Mix I.D. Field Air Estimated Air Voids VMA Adjusted
Voids (%) N (%) (%) Ni

GA1 5.0 12 4.3 4.3 16.5 16.5 10
4.1 4.2 16.4 16.5
3.7 4.0 16.0 16.3

GA2 7.0 5 7.0 5.2 18.2 16.7 4
6.8 6.6 18.1 17.9
6.3 6.3 17.7 17.6

GA3 6.3 16 6.5 6.3 19.6 19.3 15
5.7 6.0 18.8 19.1
5.8 6.5 18.9 19.5

GA4 5.7 15 6.2 5.7 18.5 18.0 16
6.0 5.9 18.2 18.2
6.3 6.1 18.6 18.4

WI1 6.8 20 6.8 7.5 17.4 18.0 22
7.2 6.7 17.7 17.3
7.5 7.9 18.0 18.4

WI2 10.5 5 10.1 9.9 20.6 20.4 4
10.2 9.9 20.7 20.4
9.8 9.2 20.4 19.8

WI3 7.6 12 6.9 4.7 18.5 16.6 9
6.4 5.5 18.1 17.3
5.7 7.0 17.4 18.6

WI4 6.5 11 6.5 4.7 18.1 16.5 11
6.9 6.0 18.4 17.7
6.4 5.4 18.0 17.2

MD1 5.9 12 5.9 6.5 20.9 21.5 13
6.5 5.8 21.5 20.9
7.1 6.1 22.0 21.1

TX1 7.0 10 7.3 6.8 20.3 19.9 10
7.0 8.3 20.1 21.3
6.9 6.3 20.0 19.5

VAl 7.3 15 6.2 6.4 19.8 20.2 9
4.9 5.6 18.7 19.3
5.9 4.6 19.5 18.4__


One purpose of the compaction tests on the duplicated field mixtures was to

develop a standard compactive effort to be used for the design of experimental mixtures.

When the first few mixtures of this study had ben tested, a quick, rough analysis indicated

that about 15 gyrations would generally provide the appropriate level of compaction.

was used as the standard compaction effort for the experimental SMA mixtures using

Florida aggregates.

Based on more complete data, it now appears that about 12 gyrations

would be closer to the average N1t for the mixtures evaluated.

6.2 Design of Experimental Mixtures

Since the experimental mixtures had fixed gradations and components, the only

variable in the mix design was selection of the design asphalt content.

Samples for each

experimental mixture were prepared with a range of asphalt contents and compacted in the

GTM using 100 psi vertical pressure, 3 degree initial angle of gyration, 9 psi initial air

roller pressure and 15 gyrations. Sets of samples were prepared with cellulose fibers and

with ground tire rubber as stabilizing additives. The design asphalt content for each

mixture was selected at 6.5 percent air voids. This was the average initial in-place air void

content of the field mixtures included in this study. Tensile strengths and failure strains

were also measured on each of the samples prepared for the asphalt content determination


Results are shown in Table 10.

Conclusions drawn from this data are:

A half percent increase in asphalt content generally produces about a one percent


For most mixtures, the highest tensile strengths occur at asphalt contents that yield

between 6.0 and 7.0 percent air voids.

Tensile strains at failure generally tend to

increase with increasing asphalt contents and decreasing air void contents.


Results of Experimental Mix Design Tests at

Mix I.D. Asphalt Air Tensile Failure Air Tensile Failure
Content Voids Strength Strain Voids Strength Strain
(%) (%) (psi) (%) (%) (psi) (%)
FLI 5.0 7.7 83.7 0.75
6.0 6.7 83.7 1.41 5.4 105.3 1.14
5.0 8.9 82.4 0.79 7.5 106.4 0.62
FL5.5 7.9 85.3 0.85 6.3 108.2 0.64
6.0 6.6 95.2 0.85 5.8 105.6 0.73
5.0 11.2 65.9 0.79 9.0 86.3 0.62
5.5 9.1 71.0 0.96 8.0 89.1 0.79
6.0 8.0 78.0 0.83 7.2 90.6 0.77
6.5 7.7 73.8 0.92 5.6 105.2 0.75
5.5 8.1 73.5 0.92 7.0 89.4 0.75
6.0 6.8 74.8 0.85 6.2 100.5 0.77
6.5 5.2 87.9 0.89 4.8 105.5 0.71
7.0 4.4 84.0 1.08 3.9 107.9 0.89
4.5 9.1 86.5 0.79 7.9 109.8 0.75
5.0 8.0 97.0 0.92 5.8 151.1 0.71
5.5 7.2 105.3 0.79 5.8 126.9 0.71
6.0 5.1 113.1 0.81 4.2 130.0 0.96
4.0 10.8 62.1 0.79 7.0 110.2 0.71
4.5 9.5 62.8 0.73 5.8 107.6 0.73
5.0 7.3 72.5 0.87 4.9 113.2 0.85
5.5 6.7 71.8 0.98 4.2 110.7 0.87

Cellulose fibers tend to bulk mixtures more than ground tire rubber (and most

other additives as well) and therefore, will allow for more asphalt in mixtures at

equal air void contents.

Increased asphalt contents will likely result in mixtures

that are more resistant to thermal and fatigue cracking.

At equal asphalt contents, mixtures with ground tire rubber have higher tensile

strengths but lower failure strains than cellulose mixtures.

Also, at approximately

equal air void contents (equal densities), mixtures containing ground tire rubber

have higher tensile strengths and lower failure strains.

These effects are related to

conclusion number

After selecting the design binder content, additional samples were compacted for

the other tests.

Average weight-volume data from these compacted samples are shown in

Table 11.

Table 11.

Weight-Volume Data for Experimental Mixtures Compacted 15 Gyrations

Stabilizer Cellulose Ground Tire Rubber
Mix I.D. A.C. Content Air Voids VMA A.C. Content Air Voids VMA
(%) (%) (%) (%) (%) (%)
FLI 7.0 6.4 17.4 6.0 7.1 16.0
FL2 5.9 6.8 16.1 5.3 7.3 15.4
FL3 6.7 6.2 17.9 6.2 6.7 17.2
FL4 6.1 6.2 17.0 5.8 6.4 16.6

6.3 GTM Densification/Shear Tests

Results of the GTM densification/shear tests for all mixtures are summarized in

Table 12.

The shear strength trends (in the first row) indicate the general pattern of the

gyratory shear strength versus gyrations plots between 200 and 300 gyrations.

Trends are

reported as rising (increasing shear strength), stable (little or no change in shear strength)

or declining (shear strength loss and tending toward failure).

Trends for most of the SMA

mixtures are either stable or rising which indicates that the mixtures would provide good

shear resistance under traffic.


Three mixtures (GA2, GA4, and WI1) had declining shear

Mixtures GA2 and WI1 were retested under the same conditions but with a

different operator.

Results of the retests are shown in parentheses.

Insufficient material

was available to retest GA4.

The retest results for GA2 and WI1 show that these mixtures

performed much better than indicated by the original test results.

All data for these

mixtures were carefully reviewed, but no reason was evident for the difference between

the original and retest results.

Based on the available field data it is expected that all of

the duplicated field mixtures would have stable shear strengths.

The value of the shear strength at 200 gyrations is shown in the second row.

Shear strengths for the SMA mixtures vary considerably, ranging from 50.4 psi for one

WI3 sample to 67.0 psi for one of the FL1 samples.

replicate samples of the same mixture.

There is also some variation within

An analysis of variance was performed to

JA --- :-. *. -*a. -.-. :C -Z_ _e *- aI' 1_1 .1-- .---

oI od c4
VIf V- V0

r \OO

r S od
enri c

*/ m^
t r-:v;

S 5

0I ei VO

I* *
Nn00 ^
nfn If V

ON *n 00

t t^ vi (^-

vrninv in min

nr-' in
In 'n vn

0r\ 00 m (
ft ft
0 c.-4 OR

S' S Sn
oc V) V

'n vt il

<^ o- c
n m in v, ov so

Nt flO
s so S

*n *
ri -tN(

r- ooo\
(r; r; r;

00 00 *v ft f

QS C^ S~
en cr U

00 00
&t ti

00O In
* *

no NO

0o o0 t
* 5 5
en eN -

o -to
en enr e n

'n nte
* C S

^t O\ ON O

ac a
-en 0
*r *

C Ot
(-4 ('4 e

0C 0 00 ('
co- 4

C\ So

00 VI t^
n \o \d

*i r-: r-

r- od \d

o n I

0\ 00 l
v-I v r

SO O00 O
so .r ei


0 000
*; *t

^ \0 00t

in i^ so in

'C vi

rd C -^
so l^' so soV in

I OS 0

(S1 ^^
\ov ^\o
_. _, _,

00 In 0

'o on
"N'l r

*- --

v; \d i


Q\ V\ -r
00 (
tS S (

C400 0
*e *
00T in v

* so os ro

t So S

-tCS t 00
o6 r; (^ o6
Sl S S

r -
od e

00 0 (
06 rS o

r~oo rlr t^l
D.- -

In n<
so *n so


j t

mr rJ <
r0 iri \o

\o \o s


Since unbalanced data exist in some cases, SAS procedure PROC GLM was selected and

used throughout this study. A Duncan's multiple range test with a significance level (a) of

0.05 was applied to compare the means for the different mixtures.

Results of this statistical analysis are illustrated in Figure 14.

This bar graph shows

the average gyratory shear strength for each mixture, with the mixtures ordered from

highest to lowest shear strength.

Letters on the bars indicate mixtures which have mean

shear strengths that are not significantly different.

For example, the shear strength,

for FL6 is not significantly different than that of GAl (shown with letter "A") and also not

significantly different than FL3 (shown with letter "B").

However, FL6 is significantly

different than FL1 and all other mixtures with lower shear strengths.

Based on criteria established by Ruth et al. [32] using dense-graded mixtures, the

shear strength at 200 gyrations should be above 54 psi to assure good rutting


Application of that criteria to these SMA mixtures indicates that most of the

mixtures are very rut resistant.

The exceptions are GA4, TX1, and FL4.

The original

results of GA2 and WII were also below 54 psi, but the retest results were well above the


For the mixtures which failed the criteria, only GA4 displayed a declining shear

strength, yet this trend was gradual and not indicative of a catastrophic failure.


TX1 and FL4 were actually still gaining strength when the test was terminated.

The other parameter evaluated from the GTM densification test is the percent

Th;E rtnaromnar ic cimrnl^, tb rinno in air tmrArlc nfthA c2mflla at ani nnrfnt



General Linear Models Procedure

Dependent Variable: Gyratory shear strength at 200 gyrations


Type I

Mean Square

Root MSE

F Value

MIX 15 968.53308 64.5688725 2.231 12.97 0.0001
a = 0.05



Duncan GrouPing

Means with the same letter are not significantly




40 -
i B B B B


10 -
o ---_-----





Mixture I.D.


The general criteria used for dense-graded mixtures has been:

< 2.5%

is good,

3.5% is acceptable,

4.0% is undesirable.

Results for many of the SMA mixtures (Tablel2, Row 4) indicate a potential for

excessive densification. This data was analyzed with the same statistical procedures used

to evaluate shear strengths. The results shown in Figure 15 show a more visually dramatic

spread in the percent densification results.

Yet due to the variation among replicates,

mixtures having as much as 1.0 percent difference in means are not significantly different.

Initial air void contents and VMA's of the test samples are shown in the fifth and

sixth rows, respectively.

These parameters were believed to possibly have some affect on

shear strength and percent densification results.

Several plots were made to assist in

identifying relationships between results and with other mix parameters.

shown in Figures 16 through 20.

These results are

It can be seen that the only apparent correlation is

between percent densification and initial air voids. This intuitive relationship indicates that

mixture densification (consolidation rutting) will be lower when the as-compacted air void

content is low.

However, the gyratory shear strength of the SMA mixtures is not affected

by initial air void content or percent densification.

This supports findings by others [7,26]

that the deformation resistance of SMA mixtures is not sensitive to low air void contents.

This departure from the conventional wisdom used with dense-graded asphalt mixtures

:- ~ r + Lt 4+l- t n !* +.. nt Ci A A ean --.4. -.. ... 2 g 1- -- _- -_

General Linear Models Procedure

Dependent Variable: Percent densification


Type I

Mean Square

Root MSE

F Value

MIX 15 62.185188 4.14567923 0.679 9.00 0.0001
a = 0.05



= 17.897

Duncan Grouping

Means with the same letter are not significantly different






Mixture I.D.

Percent densification

Figure 16.

Gyratory Shear Strength versus Percent Densification Results at 200 Gyrations



o a

a a

8 00
cP B O

SD0 0

o C
0 D
o OD



o o

0O 0

0 C
0 0 D

A _. '

16 18 20 22 24

Voids in mineral aggregate, %

Figure 18.



Gyratory Shear Strength at 200 Gyrations versus Initial VMA

. c atJ




0 0

o m


Voids in mineral aggregate, %

Figure 20

Percent Densification versus Initial VMA

The bottom row of Table 12, average weight loss, was included in the results as a

potential indicator of flushing due to excessive asphalt content.

During the

densification/shear test, some samples would consolidate to a degree at which some mastic

would flush out of the sample and squeeze out of the mold during the test. F

mastic often occurred well before the calculated sample air voids were below

lushing of the


The amount of extruded binder was determined indirectly from the difference in average

I;rrhc* 4nf ctnnlc rA r+ntrlntA r4r A +an 0x t,, ntrornn n10;r^+0o af T+rit A ; 0C.t/A


0 0
C C]

0 [0 a

o cc 0
D a

losses, but their shear strengths were high and very stable.

As mentioned earlier, some

flushing has occurred on several field SMA projects, yet these SMA mixtures have also

remained very stable.

6.4 Georgia Loaded Wheel Tests

Results of the Georgia Loaded Wheel Tests (GLWT) are shown in Table 13

to shortages of some materials, triplicate samples were not possible for all mixtures.

Nearly all of the SMA mixtures performed very well in the GLWT

. Visual

inspections of the samples after the test indicated that the low magnitude ruts were due

primarily to wearing away of the mastic, reorientation or seating of surface aggregate

particles and minor consolidation.

No lateral displacement was evident for any mixture

except for TX1

which was very minor.

Little rutting was apparent at the ends of the

beams, which are often greatly distorted in DGAC samples.

TX1 easily met the maximum rut depth criteria of 0.1

Table 13.

All of the mixtures except for


Results of the Georgia Loaded Wheel Tests at 8000 Cycles


Initial 3.9 7.4 6.9 6.0 5.1 7.2 5.3 5.1 6.1 8.9 6.0 5.7 5.8 5.6 7.3
Air 4.0 7.3 7.4 6.5 3.8 5.5 4.7 5.4 8.5 6.4 4.7 5.0 6.1 7.4
Voids 4.0 7.4 6.0 4.3 5.2 8.9 5.0 4.4 5.6 6.1 6.0

Rut .079 .077 .102 .071 .095 .111 .124 .150 .104 .073 .092 .143 .108 .110 .124
Depth .038 .061 .098 .100 .094 .143 .115 .150 .104 .079 .142 .097 .143 .130

The air void data shown in Table 13 indicates that a reasonable level of

compaction was achieved with the crude static compaction method.

Nearly all sample air

void contents were within the range expected to exist in the field after construction.

As can be seen in Figure 21, some mixtures were clearly more rut resistant than


Statistical analysis of the rut depth data revealed a large number of statistical

groupings for the fifteen mixtures. In general, the Florida experimental SMA mixtures

compare well with the duplicated field mixtures. FL3 and FL5 were among the worst

mixtures in terms of GLWT mean rut depths, yet they were statistically the same as GA3

which is known to be a very rut resistant mixture.

grouped with this baseline mixture. However, as

The other FL mixtures can also be

indicated by the coefficient of variation,

repeatability of this test is poor.

A plot of air void contents versus rut depths, shown in Figure

reveals no

apparent correlation. This provides additional support to the conclusion that rutting

resistance of SMA mixtures is not affected by air void contents.

6.5 Triaxial Repeated Load Tests

Results of the Triaxial Repeated Load Tests are shown in Table 14.

According to

this data, most of the mixtures appear to have very good deformation resistance.


developed by Gabrielson [35] based on the field performance of 16 dense-graded mixtures

was a maximum strain of 0.13 in/in for a test of 3600 cycles, or 3.61 E-5 strain/cycle.

General Linear Models Procedure

Dependent Variable: Rut depth


Tvoe I

Mean Square

Root MSE

F Value

MIX 14 0.02939337 0.00209953 0.015 9.65 0.0001



= 0.844
= 14.16

Duncan Grouping

Means with the same letter are not significantly different








Mixture I.D.


ccc c c c


GLWT Rut Depth (inches)


Table 14

Plot of GLWT Rut Depths Versus Air Voids

Results of the Triaxial Repeated Load Tests

below the criteria, these mixtures were considered as having marginal deformation

Mix GAl GA2 GA3 GA4 Wll WI2 W13 WI4 MD1 TX1 VAl FLI FL2 FL3 FL4 FL5 FL6

Initial 4.2 6.6 6.0 5.9 6.8 10.1 6.5 4.7 5.9 6.8 5.1 8.8 6.9 6.3 6.4 7.0 5.5
Voids 4.0 6.3 6.0 6.1 6.7 10.2 6.4 6.0 5.8 8.3 6.2 8.7 6.7 6.6 6.7 7.0 6.3
c rate 1.25 1.31 1.31 2.50 2.14 1.82 1.65 3.69 2.93 6.85 2.50 1.52 2.24 1.86 1.93 4.22 2.31
(E-5/ 1.27 1.29 1.30 2.51 2.65 1.94 2.12 2.58 3.70 6.47 3.49 1.63 1.63 1.77 1.49 3.76 3.95

The statistical comparison of the TRL results for the study mixtures is shown in

Figure 23.

Mixture TXI is obviously significantly different from the other mixtures.

Otherwise, there are broad groupings of results.

Variability of the test is high as indicated

by the high coefficient of variation.

6.6 Correlation of Rutting Tests

Comparisons were made between the results from the three tests which were used

to indicate rutting potential to infer the adequacy of the tests and the data.

First, if one or

more of the tests results agree that a particular mix is good or bad then the confidence that

the results are reasonable is increased.

Conversely, if the results of one test are clearly out

of the trend, it may be an indication that the test results are in error.

one of the tests fails to consistently

Also, if results from

correlate with the results of the others, then it may

indicate that the test is not representing the conditions critical for rutting or that the test

affected by other variables.

To aid in the comparisons, results were plotted and least-square linear regressions

were performed.

First, shown in Figure 24, is the plot of GTM shear strength (rG200)

versus GLWT rut depth at 8000 cycles.

The expected trend of increasing rut depth for

lower shear strength is apparent, however, as indicated by the R2 value of 0.34 the

correlation is weak.

The regressions were run again without the data from the FL

mixtures and a much improved coefficient of determination was given (R2

= 0.79).


General Linear Model


Dependent Variable: Permanent strain rate


Tvoe I SS

Root MSE

F Value

MIX 16 0.00000001 0.000004 19.71 0.0001
a = 0.05


= 0.949
= 16.939

Duncan Grouping

Means with the same letter are not significantly different.






Mixture I.D.


GLWT Rut Depth (inches)



Plot of Gyratory Shear Strength at 200 Gyrations versus GLWT Rut Depth

M .4

o GA3




0.08 0.1 0.12 0.14 0.16

GLWT Rut Depth (inches)

Figure 26. Plot of GLWT Rut Depth versus TRL Permanent Strain Rate

regression excluding the FL mixtures had an R2 of 0.53.

These two tests were expected to

have a better correlation since they both utilize samples compacted in the GTM and are

performed at the same temperature.

The plot of GLWT rut depth versus TRL permanent

strain rate, Figure 26, shows that higher rut depths correspond to higher strain rates.

second order regression indicated a good correlation with an R2 of 0.92.

A linear

regression excluding the FL mixtures produced an R2 of 0.78.

Table 15. Output of Rut Test Regressions
All Mixtures Excluding FL Mixtures
X Y Enuatinn R2 Enmuation R2

It is apparent from the regressions that the FL mixtures diminish the correlations

involving GTM results.

It is observed that FL5 and FL6 have higher shear strengths than

may be expected from the results of the other tests. It is probable that the shear strengths

for these two mixtures and possibly also FL1 are in error. The cause of such an error may

have been

due to improper alignment of the mold and sample in the GTM or incorrect

readings of the air roller pressure gauge.

Exclusion of this maverick data does indicate a fairly good correlation between the

different rutting tests.

Also, comparison of results with previously reported criteria tend

to indicate that most of the SMA mixtures have good rutting resistance.

the rutting tests indicate that GA1, GA2, and GA3 are very rut resistant.

Results for all of

The very stiff

low-density-polyethylene modified asphalt used in these mixtures probably provided a

substantial increase in their deformation resistance<

these mixtures also corroborates the test results.

mixture TX1 has marginal to poor rutting resistar

The excellent field performance for

Each of the tests also indicate that

ice. However, no field performance data

is available for this mixture, even though it reportedly "appears" to be doing well.

6.7 Resilient Modulus Tests

Resilient Modulus results at

25 and 40C are shown in Table 16.

Four results

arP rp nnrted fnr patrh mixrhtre

The tePt wan nerfnrmed on dunlicate samnles and on two

graphically illustrated in Figures

and 29.

Reasonably good coefficients of variation

were obtained at each of the temperature conditions.

At low temperatures, a low elastic modulus is beneficial to prevent a rapid

accumulation of tensile stresses and strains during temperature drops.

However, the

viscous element of the mixture's stiffness is more important to low temperature cracking.

From Figure 27

have the lowest moduli.


it can be seen that the Florida experimental mixtures, except for FL5,

In diametral tests, the properties of the binder control mixture

The low moduli of the experimental mixtures is probably due to the properties

of the asphalt-rubber binder which permits greater elastic deformations.

Although rutting resistance is primarily dependent on the aggregate structure, the

stiffness of the binder also plays a role in the mixture's deformation resistance.


at high temperatures, a higher stiffness (i.e. higher resilient modulus) is beneficial to

minimize rutting. From Figure 29, it is evident that a wide range of moduli exist among

the SMA mixtures. The three Georgia mixtures which have the highest resilient moduli all

contain a very stiff modified asphalt.

Fatigue resistance of the SMA mixtures was evaluated with resilient moduli and

tensile strain at failure using the technique presented in the AAMAS study [23].

analysis is presented with the strain at failure data in section 6.8.

Plots of resilient moduli versus temperature are shown in Figures 30, 31 and 32.

The reference lineP which pYtpnd the lpnoth nf thp aranhlc ar, rearnmmt ndAonA nnar oan,

Table 16.

Results of Resilient Modulus Tests

Mix I.D. MR at 50C MR at 25oC MR at 400C
(ksi) (ksi) (ksi)
GA1 982 981 424 449 188 183
1051 1110 458 503 198 207
GA2 964 939 452 451 174 173
957 992 444 436 184 176
GA3 1128 1066 514 502 195 201
1178 1110 508 497 206 208
GA4 913 912 267 251 72 68
953 864 260 244 78 82
WI1 1293 1279 525 507 139 146
1227 1177 449 445 120 130
WI2 1080 1050 394 369 103 100
1017 997 357 345 109 118
WI3 1038 973 381 347 81 76
962 948 383 365 90 90
WI4 953 904 250 264 43 50
961 953 299 297 57 70
MDI 906 987 210 217 69 64
885 943 218 211 73 66
TX1 1038 1072 332 336 67 85
976 951 323 334 60 76
VA1 989 935 273 256 88 81
1022 992 270 261 81 92
FL1 809 905 375 370 115 117
752 848 373 370 105 101
FL2 952 858 362 363 85 90
916 912 387 353 94 98
FL3 941 875 342 330 60 74
845 812 308 292 66 72
FL4 774 766 308 307 100 92
787 775 328 319 92 89