TRANSPORTATION RESEARCH

RECORD
No. 1374
Pavement Design, Management, and
Performance

Pavement
Rehabilitation

A peer-reviewed publication of the Transportation Research Board

TRANSPORTATION RESEARCH BOARD

NATIONAL RESEARCH COUNCIL

NATIONAL ACADEMY PRESS

WASHINGTON, D.C. 1992
Transportation Research Record 1374 Sponsorship of Transportation Research Record 1374
Price: $22.00
GROUP 2-DESIGN AND CONSTRUCTION OF
TRANSPORTATION FACILITIES
Subscriber Category Chairman: Charles T. Edson, New Jersey Department of
IIB pavement design, management, and performance Transportation

TRB Publications Staff Pavement Management Section

Director of Reports and Editorial Services: Nancy A. Ackerman Chairman: Joe P. Malroney, University of Washington
Senior Editor: Naomi C. Kassabian
Associate Editor: Alison G. Tobias Committee on Pavement Rehabilitation
Assistant Editors: Luanne Crayton, Norman Solomon, Chairman: David E. Newcomb, University of Minnesota
Susan E. G. Brown Secretary: Roger C. Olson, Minnesota Department of
Graphics Specialist: Terri Wayne Transportation .
Office Manager: Phyllis D. Barber Paul Autret, Michael C. Belangie, Thomas L. Boswell, James L.
Senior Production Assistant: Betty L. Hawkins Brown, Martin L. Cawley, Denis E. Donnelly, Wade L. Gramling,
Joseph B. Hannon, Don M. Harriott, Ira J. Huddleston, Thomas J.
Kazmierowski, Walter P. Kilareski, Aramis Lopez, Jr., Joe P.
Mahoney, Richard W. May, William G. Miley, Louis G. O'Brien,
Gary Wayne Sharpe, James F. Shook, Eugene L. Skok, Jr., R. N.
Printed in the United States of America Siubstad, Shiraz D. Tayabji, Robert L. White

Library of Congress Cataloging-in-Publication Data Frank R. McCullagh, Transportation Research Board staff
National Research Council. Transportation Research Board.
The organizational units, officers, and members are as of.
Pavement rehabilitation. December 31, 1991. -
p. cm.-(Transportation research record ISSN 0361-1981;
no. 1374)
"A peer-reviewed publication of the Transportation Research
Board."
ISBN 0-309-05416-8
1. Pavements-Design and construction. 2. Pavements-
Maintenance and repair. I. National Research Council (U.S.)
Transportation Research Board. II. Series: Transportation
research record; 1374.
TE7.H5 no. 1374
[TE250]
388 s-dc20 92-44289
[625.8] CIP
Transportation Research Record 1374

Contents
Foreword v

Evaluation of Granular Overlays in Washington State 1

Joe P. Mahoney, Newton C. Jackson, and Daniel J. O'Neil

ROADHOG-A Flexible Pavement Overlay Design Procedure 9

Kevin D. Hall and Robert P. Elliott ·
DISCUSSION, T. F. Fwa, 16
AUTHORS' CLOSURE, 17

Nationwide Evaluation Study of Asphalt Concrete Overlays 19

Placed on Fractured Portland Cement Concrete Pavements
Matthew W. Witczak and Gonzalo R. Rada

Asphalt Concrete Overlay Design Methodology for 27

Fractured Portland Cement Concrete Pavements
Matthew W. Witczak and Gonzalo R. Rada

Revision of AASHTO Pavement Overlay Design Procedures 36

Kathleen T. Hall, Michael I. Darter, and Robert P. Elliott
DISCUSSION, T. F. Fwa, 45
AUTHORS' CLOSURE, 46

Field Testing of AASHTO Pavement Overlay Design Procedures 48

Kathleen T. Hall, Michael I. Darter, and Robert P. Elliott
DISCUSSION, T. F. Fwa, 60
AUTHORS' CLOSURE, 61

Overlay Design Procedure for Pavement Maintenance 63

Management Systems
Arieh Sidess, Haim Bonjack, and Gabriel Zoltan
Pavement Evaluation and Development of Maintenance and 71
Rehabilitation Strategies for Illinois Tollway East-West Extension
Elias H. Rmeili, Kurt D. Johnson, and Michael I. Darter

PARES-An Expert System for Preliminary Flexible Pavement 81

Rehabilitation Design
Timothy Ross, Stephen Verzi, Scott Shuler, Gordon McKeen, and
Vernon Schaefer

ABRIDGMENT
Interlayers on Flexible Pavements 90
Hong-fer Chen and Douglas A. Frederick
Foreword
.Mahoney et al. overview the study of the performance of granular overlays in Washington.
Hall and Elliott describe an overlay design system (ROAD HOG) based on the 1986AASHTO
Guide for the Design of Pavement Structures. Witczak and Rada pre~ent the general approach
and the analysis of performance and structural data for a nationwide study of the fractured
slab approach for concrete pavement rehabilitation. In another paper Witczak and Rada
present design procedures for fractured concrete pavement slab overlaid by asphalt concrete.
These procedures are based on their national study and are in accordance with the flexible
pavement methodology presented in the 1986 AASHTO Guide for the Design of Pavement
Structures.
Hall et al. describe revisions to the AASHTO overlay design procedures. In a second
paper Hall et al. report the results of testing of their revised overlay design procedures using
data from in-service pavements located throughout the nation and found that they produced .
reasonable results. Sidess et al. present a methodology for flexible pavement rehabilitation
and development of overlay thicknes.s design curves for pavement maintenance management
systems. Rmeili et al.. present results from a project to evaluate and develop maintenance
and rehabilitation strategies. ·
Ross et al. describe the development of a computerized integrated system (PARES) to
evaluate and design rehabilitation schemes for flexible pavements. Chen and Frederick report
on the evaluation of stress-relieving interlayers used to reduce reflection cracking in asphalt
concrete overlays of asphalt concrete pavements.

v
TRANSPORTATION RESEARCH RECORD 1374

Evaluation of Granular Overlays in

Washington State
}OE P. MAHONEY, NEWTON C. JACKSON, AND DANIEL J. O'NEIL

Granular overlays have been used by the Washington State De- behavior of granular materials. When a crushed rock layer is
partment of Transportation (WSDOT) for about 30 years. Since subjected to a confining pressure, its stiffness increases. Since
the mid-1980s and along with the full implementation of the _the old pavement surface and the new surfacing confine the
WSDOT Pavement Management System, WSDOT has been in- crushed rock layer in a granular overlay, traffic loads can
terested in examining the performance of granular overlays ..It is
believed by WSDOT that the performance of this rehabilitation provide high confining stresses, which, in effect, increase the
treatment is better than might reasonably be expected. Further, stiffness of the crushed rock layer.
past practice in Washington occasionally required that the preex- As the use of the granular overlay increased in Washington
isting surfacing (often several bituminous surface treatment lay- State, WSDOT realized that to improve and continue to use
ers) be scarified before placement of the crushed rock layer. The granular overlays, it needed to better understand how they
study and conclusions are reviewed. worked, where they were appropriate, and how best to design
and build them.
The granular overlay system (hereafter referred to as "gran- In cooperation with WSDOT, two initial studies were un-
ular overlay") is an alternative type of overlay for rehabili- · dertaken at the University of Washington (1,2). The results
tating mostly low-volume, rural roads. The overlay consists of these studies were encouraging. They led WSDOT and the
of a layer of densely compacted, crushed rock overlain by a associated Washington State Transportation Center (TRAC)
generally thin surface layer. Figures la and lb show typical at the University of Washington to enter into an agreement
granular and asphalt concrete (AC) overlays. with the Federal Highway Administration to prepare a re-
Granular overlays have been used throughout the world as port on this topic, which was the source information for this
a pavement rehabilitation treatment. The reasons for their paper (3).
use fall into four primary categories:

1. To reduce reflective cracking from a preexisting pave- METHODOLOGY

ment structure,
2. To add extra pavement structure thickness to combat The study examined granular overlays in three ways. First,
frost-related effects, previous research on the behavior of confined crushed rock
3. To improve the cross-slope road profile (and ride in gen- laye_rs was reviewed. These studies provided information con-
eral), and cerning the stiffnesses that have been found in crushed rock
4. To strengthen the pavement structure. layers, the actions that can be taken to improve the crushed
rock layer, and the problems that have been encountered in
The last category will be the primary focus of this paper. working with confined crushed rock layers. Next, the usable
Granular overlays have been used by the Washington State life of th~ granular overlay was compared with that of other
Department of Transportation (WSDOT) for about 30 years. types of pavement resurfacing, including AC overlays and
Since the mid-1980s, and afong with the full implementation BST. Finally, the granular overlays were field tested to de-
of the WSDOT Pavement Management System (WSPMS), termine their properties and to measure the effect of different
WSDOT has been interested in examining the performance designs on their performance.
of granular overlays. One reason is that WSDOT engineers
have found that the performance of this rehabilitation treat-
ment has been better than could be reasonably expected.
Further, past practice in Washington State occasionally has LITERATURE REVIEW
required that the preexisting surfacing [often several bitu-
minous surface treatment (BST) layers] be scarified before The behavior of granular overlays depends ·on the condition·
placement of the crushed rock layer. of the crushed rock layer. Both the surface and the old pave-
One view of why the granular overlays have worked well ment serve to protect this layer and to confine it. The crushed
structurally is that they take advantage of the stress stiffening rock layer can provide much of the "strength" of the overlay.
When crushed rock is used as a base course, it generally has
a modulus of elasticity of about 15 to 30 ksi (100 to 200 MPa)
J. P. Mahoney and D. J. O'Neil, University of Washington, Seattle,
(4). When it is subjected to a confining pressure of 125 psi
Wash. 98195. N. C. Jackson, Washington State Department of Trans- (0.9 MPa), its modulus of elasticity can exceed 100 ksi (690
portation, Olympia, Wash. 98504. MPa) .or more (5).
 2 TRANSPORTATION RESEARCH RECORD 1374

Surface Treatment }
Granular Overlay
Crushed Rock Layer

Existing Surface Course

(a)

(b)

FIGURE 1 (a) Typical granular overlaid pavement and (b) typical AC overlaid
pavement.

The stress sensitivity of a granular material will, in general, termined the thicknesses of the crushed rock layer in a gran-
follow Equation 1: . ular overlay that would provide the same pavement perfor-
mance as different thicknesses of AC. In both the granular
(1) and AC overlay analyses, ·he varied the moduli of the subgrade
and the granular overlay crushed rock layer.
where Sibal considered three modes of failure: fatigue cracking of
the surface, fatigue cracking of the preexisting pavement sur-
E = modulus of elasticity (psi),
face layer, and rutting. He determined which of the three
Ki, K 2 = constants,
modes of failure was critical for each model and used the
6 = CTi + CJ2 + CJ3 = bulk stress, and
corresponding number as the number of loads that would
CTi, cr2 , cr3 = principal stresses.
cause failure for the pavement. Finally, he compared the num-
A study by WSDOT and the University of Washington ber of loads that would cause failure for each of the models
found that the crushed rock WSDOT normally uses (crushed to determine the equivalent thicknesses of granular overlays
surfacing top and base course) has the following "typical" and AC overlays. The 1.0 in. of AC on top of the crushed
constants (b): Ki, 8,500 (mean), 2,300 (standard deviation); rock layer was not calculated in the equivalency factor. For
K2, 0.375 (mean), 0.067 (standard deviation). The laboratory example, if a 4-in. AC overlay was to be converted to a
tests used to obtain these constants were conducted at bulk granular overlay with an equivalency factor of 1. 70, the con-
stresses ranging from 4 to 28 psi. version would be as follows: 4-in. AC = 1.0-in. AC + 3.0
In a traditional pavement system, the confining stresses on · in. of AC = 1.0-in. AC + 3.0 x 1. 70 of crushed rock =
the crushed rock base depend on a number of factors, in- 1.0-in. AC + 5.1 in. of crushed ro.ck (or 6.1 in. total thick-
cluding the stiffness of the subgrade. Since the granular over- ness).
lay is sandwiched between two stiff pavement layers, it will Sibal's analyses are shown in Figure 2. His results suggest
be subjected to higher confining pressures. equivalency factors of about 2.0 for the stiffer crushed rock
moduli. ·

Equivalency Factors
Crushed Rock Layer in Inverted Pavements
The stiffness of a granular overlay is provided largely by the
crushed rock layer (assuming that the surfacing is relatively A series of South African studies (5 ,7-9) and related data
'thin). One method for comparing granular and AC overlays (10) investigated the effects of different parameters on the
is to determine the thickness of a granular overlay that would behavior of the crushed rock layer in inverted pavements.
provide the same life as a thickness of AC overlay. This is These studies verified that the modulus of the crushed rock
the technique used by Sibal (2) and Deoja (1) in their studies layer can be high and offered insight into improved designs
of granular overlays. for this layer and optimum gradation.
Sibal used two elastic layer programs to model the behaviors Horak et al. noted that the gradation specifications are
of AC and granular overlays: ELSYM5 and EVERSTR (2). important for achieving the high densities required for optimal
The ELSYM5 program treats all layers as linearly elastic. performance of confined crushed rock layers (8) and pub-
EVERSTR treats the granular layers as nonlinear. Sibal de- lis~ed a paper that dealt with the effects of tightening the
Mahoney et al. 3

6 - - D 10-ksi subgrade model (Case 1)

5
- [El 20-ksi subgrade model (Case 2).

4 .
-
Equiv. Factor 3

. - -
2
- - .... - ,...

0
10 20 40 60 80
Modulus of Elasticity for the crushed roe~ layer (ksi)

FIGURE 2 Equivalency factors versus the modulus of elasticity for the

crushed-rock layer based on Sibal's calculations (2).

grading specifications beyond those normally required for the but with a few additional requirements (9). They found that
Gl (South African) base (9). Although the importance of the the greater effort required to set up the crusher and obtain
strength, durability, shape, and Atterberg limits of the ag- the correct gradation was more than offset by the increased
gregate was mentioned, the report focused on changes to the ease in compacting the material to a higher density.
specifications to produce a better compacted base. As was previously stated, the crushed rock layer is stiffer
For convenience, the gradation bands for the South African and more durable if it is well constructed. The compaction
G 1 material, as well as somewhat similar gradations from and integrity of the crushed rock layer is very important. The
AASHTO M147 (Gradings A and B) and WSDOT (Crushed material for the crushed rock layer must be durable. As was
Surfacing Top Course and Base.Course), are given in Table mentioned by Horak et al., the easiest way to obtain the
1. The gradation band for the G 1 material was obtained di- highest compaction is to use an optimum gradation (8). In-
rectly from a figure in Horak et al. (9). It appears that the terviews during 1990 with WSDOT project engineers on gran-
most similar gradations to the G 1 are AASHTO Grading B ular overlay construction projects indicated that the moisture
and the WSDOT Crushed Surfacing Base Course (11). (The content is also important (12-14). Because the crushed rock
majority of granular overlays constructed by WSDOT to date is spread in a thin layer, the moisture from the rock tends to
have used the Crushed Surfacing Top Course grading.) evaporate rapidly.
The base course on which Horak et al. reported was com- A significant problem that WSDOT has encountered in the
pacted with 99 to 103 percent of modified AASHTO and had construction of the granular overlay is traffic. When there is
a gradation that fell within the specifications for a Gl base, no possible diversion, the traffic has to pass over the crushed

TABLE 1 Gradation Bands for Various Crushed Rock Specifications

Percent Passing
wsoar
South AASHTO 9-03.9 (3)
Sieve Designation African GI* MI47-65 Crushed Surfacing
Standard mm Grading A Grading B Top Course Base Course
2 in. 50 100 10·0 100
I 1/4 in. 32 93 - 97 100
1 in. 25 82-92 75 - 95
3/4 in. 19 72- 85
5/8 in. 15.9 64-78 100 50- 80
3/8 in. 9.5 50 - 67 30 - 65 40- 75
1/4 in. 6.35 40- 57 55 - 75 30- 50
No.4 4.75 35 - 52 25 - 55 30- 60
No. 10 2.00 23 - 39 I5 - 40 20- 45
No. 30 0 . 600 I4 - 26
No. 40 0.425 11 - 24 8- 20 15 - 30 8 - 24 3 - I8
No. 200 0.075 5 - I2 2-8 5- 20 10 max 7.5 max.
*South African GI grading taken from plotted gradation band (10)
 4 TRANSPORTATION RESEARCH RECORD 1374

rock layer during construction. This frequently causes wash- percent of the total expected frost depth. In this manner, the
boarding. Therefore, the granular surface must be rebladed subgrade is at least partially insulated against frost.
immediately before the surface layer is placed. In Washington State, designing the overall pavement depth
to be equal to 50 percent of the total frost depth has worked
well for controlling all but the most severe frost problems ·
Improvement in BST Construction (18). Unfortunately, many of the state's low-volume rural
roads were built before this design procedure was adopted.
· A recent WSDOT study examined the effects that construc- These roads frequently consist of only a thin BST over 6 to
tion practices have on the problems associated witQ BSTs (15). 9 in. (150 fo 225 mm) of base course when the frost design
The main problems that were investigated included flushing thicknesses are 15 to 24 in. (380 to 610 mm). Rehabilitating
of excess asphait, windshield damage due to loose rock, and these roads typically means adding a granular overlay with a
aggregate loss due to poor embedment. Through a review of minimum of 4.2 in. (107 mm) of crushed rock.
the use of BSTs in other western states and an examination To prevent the crushed rock layer from contributing to frost
of BST constructio1_1 projects, a series of design and construc- heave, the amount of material that passes the No. 200 sieve
tion guidelines was developed. (0.075 mm) may have to be limited. Researchers have ob-
The study also recommended several guidelines for the proper served that the finest content of soils is an important indkator
choice of roads to be overlaid with BSTs. The report suggested of frost-susceptible material (19). The pavement agencies sur-
that the BST be applied only to roads that were not considered veyed specified that the maximum percentage of material that
a high traffic risk [i.e., roads with average daily traff~c (ADT) passed the No. 200 sieve be 5 to 15 percent. The lower range
counts in excess of 5 ,000]. If a BST surface was used on of this specification is lower than the range suggested by Horak
granular overlays, these same limitations were applicable. et al. for obtaining the maximum compaction (9).
(WSDOT mostly requires BSTs on routes with ADTs of 2,000
or less and discourages the use of BSTs on routes with ADTs
of 5,000 or more.) Change in Road Geometry

By adding 3 to 6 in. (75 to 150 mm) to the overall pavement

Existing Pavement structure, the granular overlay can alter the road geometry.
It can be used to increase drainage, improve the road profile,
Although granular overlays reduce the rate of reflective crack- and level off inconsistencies in the pavement. The additional
ing, they are also sensitive to moisture infiltration through height makes it unusable in areas where the road geometry
breaks in the existing pavement surface. Infiltration of mois- is restricted by curb height or other considerations.
ture lowers the stress sensitivity of the crushed rock layer (in
terms of K 1 , hence overall stiffness). ·
To circumvent the problem of an inconsistent base for the Resistance to Shear
overlay, the highway department in Zimbabwe rips and spreads
the existing base and shoulder, then stabilizes with 2 percent Little information could be .found concerning the ability of
cement or lime (if required) and recompacts it. In this manner, confined crushed rock layers to resist shear. However, this is
the granular overlay is assured of having a solid base (16). a potential problem. Although granular materials are very
Normally, all that is required is to patch any holes and seal strong in compression, they have little resistance to tension.
any cracks in the existing pavement surface. When subjected to a high confining pressure, the crushed rock
particles can distort, be crushed, shift, roll, or slide. The
amount of movement caused by any of these actions is directly
Other Advantages and Limitations proportional to the confining pressure. If the shear stress
becomes sufficiently large, the combined movement from these
Insulation actions will result in a shear failure (20).

An advantage of the granular overlay is that it protects the

existing pavement from daily extremes in temperature. This Design of Granular Overlays
is important during hot summer months and in tropical coun-
tries where pavement surface temperatures can approach and There are two basic techniques for designing granular over-
exceed 160°F (70°C) (16). lays. One is based on mechanistic-empirical calculations. The
studies done by Sibal (2) and Deoja (1) are examples. Another
technique is based on practical construction considerations.
Increased Frost Resistance The second design approach is to use the maximum thick-
ness of granular material that can be easily compacted in one
Frost heave and thaw weakening problems are often a result lift. This suggestion was offered in a report by Maree et al.,
of subgrade freezing, so if the subgrade is insulated from the in which the authors found that a crushed rock layer 12.6 in.
cold, these problems· are reduced. Several states, including (320 mm) thick did not perform significantly better than one
Alaska, Iowa, Oregon, and Washington, use frost protection 5.5 in. (140 mm) thick (both crushed rock layers were on
in their pavement thickness design calculations (17). Most of cement-stabilized subbases) (7). Maree et al. concluded that
these states design the pavement thickness to be at least 50 the crushed rock layer need not exceed the maximum thick-
Mahoney et al. 5

ness that can be placed and compacted in one lift [6 in. (150 weather conditions, and soil support, affect the amount of .
mm)]. time that a pavement lasts, but these were not directly con-
A study by Otte and Monismith produced similar conclu- sidered. Since the study was conducted within a small geo-
sions (21). They used a layer elastic program, PSAD2A, to graphic area, the weather and subgrade effects were assumed
model the behavior of several inverted pavement structures. to be constant for all roads. The effects of higher traffic load-
The computations were made for 9.0-kip (40-kN) dual wheels, ings were expected to be somewhat canceled by the thickness
13.0 in. (330 mm) apart. The pavement that was simulated of the resurfacing.
was an inverted pavement with a 1.4-in. (35-mm) BST surface. Roads overlaid by the different techniques had different
The thicknesses of the crushed rock layer and the cement- characteristics. In general; roads with AC surfaces had higher
stabilized base and subbase were varied over a wide range of traffic counts than roads with BST surfaces. In addition~ gran-
thicknesses. ular overlays were often used as a method for repairing badly
The authors found that the primary stresses in an inverted distressed roads. Most roads that received a granular overlay
pavement were on the surface course and cement-stabilized had suffered from either thermal cracking or roughness prob-
layers. They found that because of stress stiffening, as the lems and therefore had required special treatment.
thickness of the crushed rock layer increased from 5 to 20 in. In comparing the survival times and performance periods,
(125 to 500 mm), the equivalent elastic modulus of the gran- note that the sets of data did not come from the same time
ular base declined about 30 percent. Otte and Monismith period. Since the performance life equations were only cal-
recommended the following (21): culated for t_he present road surface, they represented only
resurfacings built since the late 1970s. The data for calculating
1. The bituminous surfacing .for inverted pavement designs the,survival lives, on the other hand, were equally spread
should not exceed 1.2 to 1.4 in. (30 to 35 mm). from the 1980s to the 1960s, with some dating as far back as
2. For typical highway traffic loads, the granular layer should the 1940s. Because of the problems previously mentioned with
have a thickness of about 5.0 to 6.0 in. (125 to 150 mm). the "old" survival life data, the old information tended to
3. The cement stabilized layers supporting the granular layer increase the average survival life slightly.
should be (a) two layers, each 6 in. (150 mm) thick, if the
subgrade has a CBR of 15 or better, or (b) one layer, 6 in.
Source of Data
(150 mm) thick, for light traffic (rural).
The source of data for this analysis was the WSPMS (22). The
The one thickness approach is also used in Zimbabwe (16).
WSPMS contains records of work done on the roads and the
The highway department in Zimbabwe has found that the
pavement condition analyses. Data from the WSPMS were
practical range for the construction of the crushed rock layer
spot checked against as-built plans and pavement conditions
in an overlay is 5 to 6 in. 020 to 150 mm). Thinner lay~rs
and were found to be accurate.
tend to shear under a roller. If a thicker layer is needed, the
The estimation of survival lives and performance periods
road probably needs to be reconstructed.
was restricted to WSDOT Districts 2 and 6 (eastern Wash-
In Washington State, granular overlays are generally built
ington). These two districts are generally rural areas where
with thicknesses of 3 to 6 in. [most with a 4.2-in. (107-mm)
the topography ranges from mountainous to rolling hills. The
thickness]. ·
average annual precipitation is about 17 in. (432 mm) in Spo-
kane (23), but the area has severe frost in the winter. These
are also the two districts that contain the majority of roads
Costs
with granular overlays.
The WSPMS was searched to locate all roads containing
Typical WSDOT project specific costs for granular overlays
BST, AC, and granular overlay resurfacings. First, the actual
with both AC and BST surfaces _were· estimated. In general,
survival time for the different layers of pavement surfaces was
granular overlays surfaced with AC are about twice as ex-
calculated by subtracting consecutive resurfacing construction
pensive as those surfaced with a BST (about $7.00/yd2 for AC
years from the previous years. Next, pavement condition rat-
versus $3.50/yd 2 for BST).
ing (PCR) (100 = no distress, 0 = extensive distress) data
on the most recent resurfacing were examined to determine
whether the regression equation in the model represented a
SURVIVAL AND PERFORMANCE STATISTICS
"true" regression equation. Finally, the survival live_s and the
performance periods were compared for each type of resur-
One definition of the survival life of a pavement is the amount
facing and among the types of resurfacings.
of time between a pavement's construction and its resurfacing.
The performance period is the amount of time from its initial
construction to the time it reaches· a minimally acceptable Bituminous Surface Treatments
level. Both survival lives and performance periods provide
valuable estimates of pavement life. In this portion of the The survival times for BST resurfacings are approximately
report, the survival lives and performance periods of AC over- normally distributed with a median survival time of 8.0 years
lays, BST resurfacings, and granular overlays are estimated and a mean of 9.2 years (standard deviation = 5.1 years for
and compared. a sample size of 1,310).
This part of the study considered only the ages of the dif- Next, the PCR equations were examined to determine the
ferent types of surfaces. Other factors, such as traffic loadings, performance periods. Although more than 200 road sections
6 TRANSPORTATION RESEARCH RECORD 1374

TABLE 2 Statistics on Performance Periods

PCR Level Median Mean Std. Dev
PCR40 6.1 7.0 2.5
BS Ts I PCR20 7.4 8.2 2.4
PCRO 8.7 9.3 2.4
PCR40 10.6 10.2 1.7
AC Overlays2 PCR20 11.4 11.3 1.7
PCRO 12.2 12.1 1.8
PCR40 8.7 9.2 1.8
Granular Overlays3 PCR20 10.7 10.8 2.1
(All Surfaces) PCRO 12.2 12.2 2.6
PCR40 7.5
Granular Overlays PCR20 8.3
(BST Surfaces) PCRO 11.0
PCR40 9.5
Granular Overlays PCR20 11.3
(AC Surfaces) PCRO- 11.8

I Based on 21 data points

2Based on 29 data points
3Based on 17 data points

were analyzed, only 21 had usable regression equations (sec- on the WSDOT route system; however, projects tend to have
tions less than 5 years since last resurfacing were not used). long lengths (thus a large amount of mileage). In addition,
These results are summarized in Table 2. because granular overlays had only been used with greater
A comparison of the values for the actual survival times frequency since the mid-1980s, few data were available con-
and the performance periods reveals that WSDOT resurfaces cerning survival times. Therefore, survival times were not
BST roads when their PCR is between 20 and 0. analyzed. This left the examination of the performance pe-
riods. The data are summarized in Table 2.
AC Overlays

The survival times for AC overlays were essentially normally

distributed. The survival times based on AC overlays with Performance Summary
thicknesses of less than 1.2 in. (30 mm) were separated out.
According to the 1988 WSDOT specifications, only AC over- Although both methods used for calculating the usable life of
lays thicker than 1.2 in. (30 mm) are subject to compaction the resurfacings have uncertainties, the two methods together
control. The survival times for AC overlays less than 1.2 in. provide reasonable estimates of usable life.
(30 mm) were, again, normally distributed with a median A comparison of BST resurfacings with granular overlays
survival time of 8.0 years and a mean of 8.7 years (standard with BST surfaces shows that the difference in predicted per-
deviation = 2.8 years). For AC overlays greater than 1.2 formance is relatively small (7 percent) at a PCR of 40 but
in. (30 mm), these times were 10.0, 9.7, and 4.5 years, re- increases (18 percent) at a PCR of 0 (the granular overlays
spectively. in both cases last longer than a simple BST). However, for
There were also a significant number of survival times of AC, the granular overlays surfaced with AC do not last as
5 years and less. If only the survival times of AC overlays long as a conventional AC overlay (7 percent less at a PCR
thicker than 1.2 in. (30 mm) and lasting more than 5 years of 40 and 2 percent less at a PCR of 0).
are considered (thus eliminating construction-related and under- These comparisons show that the BST-surfaced granular
designed factors), the average survival life increases to 11.5 overlay performs better than simple BST resurfacings (with-
years. out the crushed-rock layer) and AC-surfaced granular over-
The predicted performance periods are summarized in lays not quite as well as plain AC overlays. Granular overlays
Table 2. are generally used to repair· pavement structures with signif-
icant distresses. BST and AC overlay resurfacings are nor-
Granular Overlay mally placed on pavements early in the distress cycle (usually
due to fatigue cracking). If the granular overlays had been
Unlike the other two resurfacings (or overlays) that were used on pavements in better condition, these comparisons
examined, there are not a large number of granular overlays would have probably been different.
Mahoney et al. 7

NDT TESTING AND EVALUATION CONCLUSIONS

To assess the actual performance of roads with granular over- The following conclusions are based on the entire study of
lays more than 50 centerline mi (80 km) of roads were tested granular overlays (including reviewed literature).
with a Dynatest falling weight deflectometer (FWD) Model
8000. These roads were located throughout WSDOT Districts 1. Granular overlays are effective at reducing reflection
2 and 6 and had a variety of structures, ages, and conditions. cracking, insulating the old pavement surface against ex-
The tests were designed to provide evidence about the com- tremes in temperature, and improving the road geometry.
parative performance and stiffness of granular 9verlays. 2. BSTs are more appropriate surfacings for granular over-
All of the selected roads were in rural eastern Washington. lays than are AC overlays (on the basis of observed perfor-
Traffic on these roads is mostly local, with occasional long- mance).
haul trucks. The topography of eastern Washington ranges 3. The crushed rock layer should have a maximum rec-
from mountainous to rolling hills, and most of the land is ommended thickness of 6 in. (150 mm) (on the basis of struc-
either dry land wheat farms or scrub fields. tural considerations only) and a minimum value of 3.0 in.
A sample of roads with enough different characteristics to (75 mm).
be representative of all the other roads in these two districts 4. The AC equivalency factor for the confined crushed-rock
was sought. Characteristics that were considered included age, layer that is properly constructed and well protecte_d can be
pavement structure, traffic flow, and road location. The roads about 2.0 (refer to Figure 2), but in-service WSDOT pave-
that were selected are summarized in Table 3. Where more ments suggest this value is probably higher (i.e., less support).
than one section was tested on a specific road, each section 5. Consideration should be given by WSDOT to using crushed
was designated by a letter. surfacing base course (maximum aggregafe size= 1% in.) for
The analysis of these test sections was done in two fun- the crushed rock portion of the granular overlay on some
damental ways. First, basic deflection parameters were cal- projects and evaluating its performance. The gradation is sim-
culated for each test section (subgrade modulus, D 0 , and area ilar to (but not the same as) the South African Gl material
parameter). These measured parameters were then compared specification.
with calculated parameters for typical AC-surfaced sections 6. The pavement surfacing (before granular overlay) should
with similar subgrade moduli (modeled with the ELSYM5 be left in place if project conditions permit. This enhances
linear elastic program). Second, the layer moduli were back- the stress stiffening of the crushed-rock layer in the granular
calculated using the program EVERCALC 3.0. The overall overlay system.
goal of this second effort was to estimate granular overlay
moduli by use of FWD deflection data. The backcalculation The full study documentation is contained in the report by
process was difficult and the results contain uncertainties; O'Neil et al. (3). ·
however, reasonable convergence errors of 1.5 percent or less
were used (RMS basis).
On the basis of the analyses of test sections on SR28A and ACKNOWLEDGMENTS
SRI 7, the modulus of elasticity for the crushed rock layer of
the granular overlay was estimated to be approximately 80 The authors wish to express their appreciation to John Liv-
ksi (551.2 MPa) under a 9.0-kip load. The South African ingston and Linda Pierce (WSDOT) for their technical sup-
studies showed elastic moduli ranging from 29.0 to 75.4 (and port and Duane Wright and Ron Porter for their help with
higher). Presumably, one of the reasons for the higher mod- the graphics and word processing. The review comments pro-
ulus of elasticity was that the bulk stress in the crushed rock vided by Keith Anderson (WSDOT) assisted the authors in
layer in the granular overlay was higher than in the inverted improving the original study report. Special thanks are given
pavement. to Jim Sorenson (FHWA), who strongly encouraged this effort.

1'ABLE 3 Road Sections Tested with FWD

Granular Ageat
Surfacing Thickness testing Test Length Average Daily Annual 18kip
Location Type in. mm (years) miles km Traffic 1 (80 kN) ESAL2
District 2
SR17 BST 2.4-123 60-300 6 8.0 12.8 960 107,000
~
SR24 BST 3.6 90 4 11.0 17.6 920 26,000
SR28A AC 3.0 75 6 5.0 8.0 4,700 180,000
SR28B BST 3.6 90 6 5.8 9.3 640 29,000
District 6
SR21A AC 4.2 107 11 5.0 8.0 330 28,000
SR21B BST 3.0 75 10 5.0 8.0 280 13,000
SR231A BST/AC 4.2 107 11 5.0 8.0 180 5,700
SR231B BST/AC 4.8 122 8 5.0 8.0 230 6,400

lWSPMS, 1990 2Calculation is based on WSDOT's estimate of truck count (MIDAL, 1990)
3Tuickness varies
8 TRANSPORTATION RESEARCH RECORD 1374

REFERENCES 10. S. Jackson (ed.). At/tis Climatologique de L'Afrique. South Af-

rican Government Publications, Pretoria, South Africa, 1961.
1. B. B. Deoja. A Comparison· of Cushion Course and Asphalt Con- 11. 1991 Standard Specifications for Road, Bridge and Municipal
crete Overlays on Flexible Pavements. Thesis. University of Wash- Construction. Washington State Department of Transportation,
ington, Seattle, 1986. · Olympia, 1991.
2. V. Sibal: Cushion Course Overlays-An Alternative to Asphalt 12. S. Reister. Telephone interview. Wenatchee, Wash., 1990.
Concrete Overlays. Research paper. University of Washington, 13. B. Stokes. Telephone interview. Wenatchee, Wash., 1990.
Seattle, 1988. 14. B. Murray. Telephone interview. Wenatchee, Wash., 1990.
3. D. J. O'Neil, J. P. Mahoney, and N. C. Jackson. An Evaluation 15. D. C. Jackson, N. C. Jackson, and J. P. Mahoney. Washington
of Granular Overlays in Washington State. Research Report WA- State Chip Seal Study. In .Transportation Research Record 1259,
RD 226.1. Washington State Department of Transportation, TRB, National Research Council, Washington, D.C., 1990.
Olympia, 1991. 16. R. L. Mitchell. The Economics of Overlay Design and Road
4. The AASHTO Guide for the Design of Pavement Structures.· Rehabilitation. Highway Investment in Developing Countries, The
AASHTO, Washington, D.C., 1986. Institute of Civil Engineers, London, 1983.
5. J. H. Maree, N. J. W. van Zyl, and C. R. Freeme. Effective 11. M. S. Rutherford, J. P. Mahoney, R. G. Hicks, and T.
Moduli and Stress Dependence of Pavement Materials as Mea- Rwebangira. Guidelines for Spring Highway Use Restriction.
sured in Some Heavy-Vehicle Simulator Tests. In Transportation Washington State Department of Transportation, Olympia,
Research Record 852, TRB, National Research Council, Wash- 1985.
ington, D.C., 1982. 18. J. P. Mahoney, R. G. Hicks, and N. C. Jackson. Flexible Pave-
6. J. P. Mahoney. Comparison of Laboratory Resilient Moduli of ment Design and Rehabilitation. Course notes. Federal Highway
Recompacted Samples-Revised. Letter to WSDOT, Seattle, Administration, 1988:
1990. 19. P. Irick (ed.) NCHRP Synthesis of Highway Practice 26: Roadway
7. J. H. Maree, C. R. Freeme, N. J. W. van Zyl, and P. F. Savage. Design in Seasonal Frost Areas. TRB, National Research Coun-
The Permanent Deformation of Pavements with Untreated cil, Washington, D.C., 1974.
Crushed-Stone Bases as Measured in Heavy Vehicle Simulator 20. G. Sowers. Introductory Soil Mechanics and Foundations: Geo-
Tests. Proc., 11th Australian Road Research Board Conference, technica/ Engineering. MacMillan Publishing Co., Inc., New York,
Melbourne, Australia, 1982. 1979.
8. E. Horak, J. C. du Pisani, and C. J. van der Merwe. Rehabili- 21. E. Otte and C. L. Monismith. Some Aspects of Upside-Down
tation Alternatives for Typical "Orange Free State" Rural Roads. Pavement Design. The Eighth Australian Road Research Board
Proc., Annual Transportation Convention, National Institute for Conference, Vol. 8, Perth, Australia, 1976.
Transport and Road Research, Perth, Australia, 1986. 22. Washington State Pavement Management System. Washington State
9. E. Horak, E. M. de Villiers, and D. Wright. Improved Perfor- Department of Transportation, 1990.
mance of a Deep Granular Base Pavement with Improved Material 23. M. Hoffman (ed.). The World Almanac and Book of Facts, 1989.
and Control Specifications. 1987. Scripps Howard Co., New York, 1988.
TRANSPORTATION RESEARCH RECORD 1374 9

ROADHOG-A Flexible Pavement

Overlay Design Procedure
KEVIN D. HALL AND ROBERT P. ELLIOTT

ROAD HOG, a practical, easy-to-use system for selecting flexible and, thus; be compatible with AHTD's other pavement design
pavement overlay thicknesses, was developed for the Arkansas practices.
State Highway and Transportation Department. ROADHOG is The completed design procedure developed under TRC-
an NDT-based structural design procedure that follows the 1986 8705 was named ROADHOG to designate that it is a roadway
AASHTO Guide structural deficiency approach to overlay design
with structural capacities expressed as structural numbers (SNs). design tool developed at the University of Arkansas (the
Major departures from the 1986 AASHTO Guide are the meth- "Hogs"). The computerized version of the design procedure
odologies for determining the effective structural number of the is a stand-alone (executable) program that is extremely
existing pavement (SNett) and for estimating the in situ subgrade flexible and user friendly. The programming was done in
resilient modulus (M,). SNeff is determined by a relationship be- CLIPPER, a data base management language accessible to
tween two falling weight deflectometer (FWD) surface deflec- data base file formats, to facilitate the handling of nondes-
tions: the deflection at the center of loading and the deflection
at a distance fr:om loading equal to the pavement thickness. tructive testing data. (All product names are registered trade-
M, is estimated using a regression algorithm developed from marks of their respective development corporations.) AHTD
the ILLI-PAVE finite element structural pavement model. stores and manipulates field NDT results using the dBASE
ROADHOG is contained in a user friendly, stand-alone (exe- data base management software. An attractive feature of the
cutable) computer program that directly uses the FWD field data ROADHOG program is its modular construction. Each ma-
files. In addition to determining overlay thicknesses, the program jor function of the procedure is contained in a separate pro-
can subdivide a project into statistically similar analysis units on
gram module. This will facilitate upgrading the program as
the basis of the overlay thickness required at each NDT test site.
technology improves with little visible effect on the system as
a whole.
In recent years, highway programs nationwide have shifted
their emphasis from new construction to rehabilitation 1 main-
tenance, and preservation. With this shift, a major deficiency
OVERLAY DESIGN METHODOLOGY
in pavement design technology became more significant. That
deficiency was the lack of practical, proven design procedures
The AASHTO approach to flexible pavement design uses a
for selecting the thickness of pavement overlays. The need
structural number (SN) to reflect the combined structural
for a flexible pavement overlay design procedure was partic-
contribution of all the pavement layers (surface, base, and
ularly significant in Arkansas, where, except for the Inter-
subbase). SN is defined by
states, most highways have flexible pavements. Research Project
TRC-8705 was initiated by the Arkansas Highway and Trans-
(1)
portation Department (AHTD) and the University of Ar-
kansas to correct this deficiency.
From the beginning, the major objective of the project was where an is the layer coefficient of layer n and D n is the
to develop a practical, easy-to-use design procedure that was thickness of layer n.
compatible with other AHTD pavement design practices and The 1986 guide uses SN in a "structural deficiency" ap-
that consistently produced reasonable design thicknesses. proach to overlay design. In its simplest terms, the structural
AHTD designs pavements using the AASHTO Guide (1). deficiency approach states that the overlay required is the
When the 1986 guide became available, AHTD adopted it in difference between the total structure needed and the struc-
place of the previous guide. Unlike the previous guide, the ture that currently exists. The guide expresses this with the
1986 guide contained procedures for overlay design. These following equation:
were not complete but did provide a framework around which
a complete design procedure could be developed that would (2)
be compatible with the new pavement portions of the guide
where
SN01 = required structural number of overlay,
K. D. Hall, Department of Civil Engineering, University of Illinois SNY = structural number required to carry future pro-
at Urbana-Champaign, 1208 NCEL, 205 N. Mathews, Urbana Ill.
61801. R. P. Elliott, Arkansas Highway and Transportation Research
jected traffic,
Center, Department of Civil Engineering, University of Arkansas, F,1 = remaining life factor, and
4160 Bell Engineering Center, Fayetteville, Ark. 72701. SNeff = effective structural number of existing pavement.
10 TRANSPORTATION RESEARCH RECORD 1374

Determination of the required overlay thickness involves demonstrated that for all practical purposes, 1.0 is the ap-
converting the required structural number of the overlay into propriate value for F,1 • The concept of r_emaining life should
a thickness of asphalt concrete using the appropriate material not be completely ignored in a comprehensive overlay design
coefficient: methodology; however, its inclusion within the structural de-
ficiency design approach used in ROADHOG was not found
(3) to be reasonable or practical. Thus, for ROADHOG Equa-
tion 2 is reduced to
where
(2a)
D 01 = thickness of overlay,
SN01 = structural number of overlay, and
aac = material coefficient of asphalt concrete. OVERVIEW OF THE ROADHOG PROCEDURE
Within this general approach, the major components lack-
ing for a complete, workable design procedure are specific The basic framework for a structural deficiency overlay design
methodologies for determining SNeff and the subgrade resil- procedure is summarized in the following steps. The ROAD-
ient modulus (Mr) needed for calculating SNY. Although HOG procedure was constructed upon such a framework.
ROADHOG uses the general AASHTO approach to overlay
1. Analysis unit delineation,
design, the methods used by ROADHOG for determining 2. Traffic analysis,
SNeff and Mr are radically different. 3. Materials and environmental study,
In the AASHTO method the values for SNeff and M, are
4. Effective structural capacity analysis (SCxett),
interdependent. Both values are determined on the basis of
5. Future overlay structural capacity analysis (SCy), and
backcalculated moduli from the NDT deflection data. Mr is 6. Overlay thickness selection.
determined from the deflection at a point some distance from
the center of NDT loading. This value is then used in the Step 1, analysis unit delineation, identifies. subsections of
determination of modulus values for the pavement layers. an overlay project (analysis units) with similar features such
SNetr is calculated using these moduli. As a result, an error as cross section, subgrade support, and pavement condition.
in the determination of M, will result in an error (although Unit delineation helps to optimize an overlay design by iden-
opposite in sign) in SNeff· tifying varying overlay requirements along a project, rather
In ROAD HOG the determination of SNerr and M, are inde- than recommending a single "average" overlay requirement.
pendent. As in the AASHTO method, SNeff is assumed to be Unit delineation should be performed both before the overlay
a function of pavement stiffness. However, instead of back- thickness selection on the basis of existing conditions and after
calculating layer moduli (or a single overall equivalent mod- the required overlay is calculated for each NDT point along
ulus), ROADHOG uses a relationship between SN and a the project. ROADHOG includes methodology for the sec-
deflection differential called delta D. Delta D is the difference ond unit delineation; that is, methodology for breaking the
between the deflection measured at the center of loading and project into units based on the overlay thicknesses. Unit de-
the deflection at a point equal to the total pavement thickness lineation based on existing conditions needs to be done by
(surface + base + subbase). M, is determined from the de- the designer before using ROADHOG. Further discussion of
flection measured 36 in. from. the center of loading using an this is given later.
analysis algorithm developed by Elliott and Thompson (2) Steps 2 and 3, traffic analysis and materials and environ-
using the ILLI-PA VE finite element pavement model. mental study, respectively, generate input parameters for use
In addition to independence, these methods of SNerr and in subsequent steps. These parameters may include design
M, determination have other advantages over the AASHTO (future) traffic, traffic history, material coefficients, and data
methods. SNerr determination is relatively independent of depth concerning current material condition. The designer must have
to bedrock, a major concern in a backcalculation scheme such completed these tasks before using the ROADHOG struc-
as that used by AASHTO. Also, the Mr value determined by tural design procedure.
ROADHOG is consistent with the value used in the AASHTO Steps 4 and 5 determine the values of the variables (SNY
Guide design equation to represent the AASHO Road Test and SNerr) appearing in Equation 2a. From the values gen-
subgrade. As discussed by Elliott (3), the M, value backcal- erated in these steps, the required structural capacity (SN01 )
culated by the AASHTO method must be modified to be of the overlay is calculated.
consistent with the AASHO Road Test subgrade value. The Step 6, overlay thickness selection, finalizes the overlay
methods of SNerr and M, determination are discussed in greater design. Equation 3 is used in this step to determine overlay
detail later. thicknesses for the project.
Another component of the AASHTO design methodology ROADHOG uses nondestructive testing (NDT) data to
modified in the ROADHOG procedure was Frt, the remain- evaluate the existing pavement system (pavement and
ing life factor. Fri is an adjustment factor for the effective subgrade). NDT-based design procedures have several po-
structural capacity of the existing pavement. The factor at- tential advantages over other procedures, including speed at
tempts to "reflect a more realistic assessment of the weighted which data may be obtained, cost of obtaining data, and the
effective capacity during the overlay period" (1). Elliott (4) amount of data available for a particular project. In addition,
investigated the concept and' application of remaining life in NbT-based procedures attempt to evaluate existing pavement
overlay design as presented in the AASHTO Guide. He dem- systems in situ instead of attempting to relate laboratory test
onstrated that the Fri relationship is flawed. Elliott's analyses results to field conditions.
Hall and Elliott 11

The ROADHOG overlay design procedure uses NDT data 3. MRCALC-The MRCALC module uses the deflection
generated by a falling weight deflectometer (FWD). The FWD measured at 36 in. from the load to estimate the in situ Mr
applies a load pulse to a pavement and measures the resulting of the subgrade at each test point.
surface deflection. Details of the setup and operation of the 4. NEWFLEX-The NEWFLEX module uses designer in-
FWD are available elsewhere (5). The FWD attempts to sim- put and the sub grade Mr determined in MRCALC to calculate
ulate the effect of a moving wheel load on the pavement the structural number required to carry future traffic (SNy).
surface (6). Studies have indicated that the FWD generates 5. OVLTHICK-The OVLTHICK module calculates the
load responses comparable with those produced by moving overlay thickness required to strengthen the existing pave-
wheel loads (5,7). This effect is a major advantage over static ment to carry future projected traffic.
tests when attempting to evaluate the pavement as it exists 6. UNIDEL-The UNIDEL module uses the cumulative
in the field. difference approach with the required overlay thickness at
ROADHOG uses NDT data taken directly from the FWD each FWD test site to subdivide the overlay project into anal-
field storage file. The program reads the file from the floppy ysis units.
disk and creates a data base file that holds both th.e field data 7. OUTPUT - The OUTPUT module allows the user to
and the calculated overlay design parameters. Each record in see the results of the design procedure, sending the results to
the data base represents a single FWD test performed for the either the screen, a printer, or a file.
overlay project. A required overlay thickness is determined
for each FWD test along the project. The project may be
(designer's option) divided into recommended analysis units EFFECTIVE STRUCTURAL CAPACITY ANALYSIS
based on the required overlay thickness at each NDT site.
Output options allow the designer to choose whether to use A number of methods currently exist to estimate the effective
the recommended analysis units and to choose any combi- structural capacity of a pavement, primarily falling into three
nation of available data stored in the data base file. User categories: (a) component analysis procedures, (b) deflection-
inputs, in addition to the NDT data file from the FWD, in- based procedures, and (c) analytically based, or mechanistic,
clude the existing thickness of asphalt concrete surface, the procedures. An excellent synopsis of each type of procedure
total existing pavement thickness, new pavement design pa- is available elsewhere (8).
rameters (reliability, standard deviation, delta PSI, and design ROADHOG uses a deflection-based procedure in which
traffic), asphalt concrete (overlay) material coefficient, and the effective structural capacity of the existing pavement sys-
the minimum acceptable length of an analysis unit. tem is related to the deflection basin generated and measured
Figure 1 is a flow diagram showing the primary modules of by the FWD. The ROADHOG module that performs the
the ROADHOG design procedure. A brief synopsis of the effective structural capacity analysis is termed "SNEFF." The
ROADHOG modules follows. Subsequent sections detail the algorithm forming the basis of SNEFF was developed at the
procedures and algorithms used by each respective module. University of Arkansas by Kong (9). The salient features of
Kong's procedure are reproduced here.
1. XFORM-The FWD device stores data generated dur- The development of the SNeff algorithm began with the
ing a nondestructive test in ASCII format on a floppy disk; concept that at sufficient distances from the center of loading
XFORM transforms these data into a data base file (dBASE the surface deflection is almost entirely due to deformation
format) for use in later modules. within the subgrade. As shown in Figure 2, the zone of in-
2. SNEFF-The SNEFF module uses a relationship be- fluence due to loading extends with depth. Directly below the
tween the deflection basin and structural capacity of a pave- loading plate, all materials "feel" the effect of the load and
ment system to generate SNefffor each FWD deflection basin. deform. At locations beyond the loading plate, only those
materials within the zone of influence are deformed. At some
distance, only the subgrade deforms. This concept serves as

Effective Future Overlay

Structural
Capacity Analysis
XFORM Structural
Capacity Analysis
FWD Load

,MRrLCI
SNEFF
Overlay
I. NEWFLEX I.
Thickness
Selection

--lovLTHJcKI· I
Existing
AC
.d "!:J .....
-,
l Analysis
T

1
Granular
Unit
Base
Delineation

louTPUTj Subgrade

FIGURE 1 ROADHOG flow diagram. FIGURE 2 Conceptual basis for SNEFF.

12 TRANSPORTATION RESEARCH RECORD 1374

the basis for most subgrade resilient modulus backcalculation complicating factors in the backcalculation of subgrade resil-
methods. ient modulus (6). Additional analyses were performed to de-
Viewed from the perspective of the pavement, this concept termine whether this factor might also be significant relative
suggests that the difference between two deflections could be to the delta D-SNeu relationship. Subgrade depths ranging
used as measure of the pavement stiffness. Using the AASHTO from 8 ft to semi-infinite were considered. The delta D-SNeu
assumption that SNeu is a function of stiffness, the deflection relationship was found to be reasonably independent of the
difference becomes a measure of SNeu· If the deflection at subgrade depth (Figure 4). These findings indicate that this
distance T in Figure 2 · is due to sub grade deformation and approach provides a practical method for the determination
the deflection at the center of loading is due to pavement and of SNeff that is independent of the subgrade.
subgrade deformation, the difference between the two de- For the method to be complete, a means was needed for
flections, delta D, should represent the deformation within temperature adjustment. Asphalt concrete is quite tempera-
the pavement alone. If the zone of influence spr~ads at an ture sensitive, exhibiting modulus increases at lower temper-
angle of about 45 degrees, the distance T would be equal to atures and modulus decreases at higher temperatures. As a
the pavement thickness. result, delta D is also temperature sensitive. The elastic mod-
A relationship between the pavement stiffness (delta D) ulus used in the delta D-SNeff analyses was selected as typical
and the effective structural capacity of the pavement (SNeu) of the resilient modulus of an Arkansas asphalt concrete at
was developed using elastic layer theory. Deflection basins 70°F.
were generated for a variety of pavement cross sections using Additional ELSYM5 analyses were conducted to examine
the elastic layer program ELSYM5 (JO). Delta D was cal- the effect of other AC temperatures on delta D. The AC-
culated for each deflection basin and plotted against the struc- modulus temperature relationship shown in Figure 5 was used
tural number of the associated pavement cross section. The to select modulus values for other temperatures. From these
structural number of a pavement was calculated using Equa- analyses temperature adjustment curves were established. The
tion 1. Layer coefficients for new pavements were used in the temperature correction factor is the ratio of delta D at a given
determination: AC coefficient = 0.44; crushed stone base temperature to delta D at 70°F. For testing temperatures other
coefficient = 0.14. Plots of delta D versus SNeu resulted in than 70°F, delta D from the given test is multiplied by the
curves like those shown in Figure 3. temperature correction factor to yield a corrected value of
Total pavement thicknesses were 8, 12, and 24 in. Asphalt delta D. The corrected delta D value is used with the curves
thicknesses ranged from 1 to 17 in. The elastic modulus values in Figure 3 to estimate SNeff of the existing pavement. The
used in ELSYM5 to represent the asphalt and_ granular ma- temperature adjustment was reasonably independent of the
terials were 500 ksi and 30 ksi, respectively. These represent subgrade but depended on both total pavement thickness and
typical values for AC at about 70°F and dense graded granular AC thickness. The temperature adjustment factors for an 8-
base. They also are consistent with the layer coefficients and in. pavement are shown in Figure 6.
modulus relationships used by AASHTO. Sub grade resilient The SNEFF module uses a second-order polynomial equa-
modulus values of 3.5 ksi, 7 ksi, 14 ksi, and 21 ksi were used tion (r2 = 0.98) to approximate the delta D/SNeff relationship.
. for the analyses. These were selected as representative of the For pavement thicknesses other than those shown in Figure
range of values for Arkansas subgrades expected on the basis 3, SNEFF uses linear interpolation to generate the points
of previous work (11). The results of the analyses (Figure 3) necessary to define a delta D/SNeff curve. For the temperature
show the delta D-SNeff relationship to be reasonably indepen- adjustment, SNEFF approximates each curve with two straight-
dent of the subgrade modulus. line segments joined at the 70°F point. For pavement thick-
These analyses, however, incorporated the standard elastic nesses other than those analyzed by Kong, SNEFF calculates
layer assumption of a semi-infinite depth of subgrade. Subgrade a temperature correction using linear interpolation.
thickness (depth to bedro_ck) is believed to be one of the

Delta Deflection (0.01 in)

2.5~--------------------~
Delta Deflection (0.01 in)
2.5.-----------------------, 0
24· Pavement Subgrade Depth
2>-----•-------------1~7Feet
0
o Semi-Infinite
G-B-EJ Mr • 7000 psi e:, 8 Feet
*** Mr • 21000 psi
----------

0.5 1 - - - - - - -
0.5

0 ' - - - - - - ' - - - - - - ' - - - - - , __ _ ___,____ _ ___,

0'------'-----'------'-----'------' 0.00 2.00 4.00 6.00 8.00 10.00
0.00 2.00 4.00 6.00 8.00 10.00 Effective Structural Number
Effective Structural Number
FIGURE 4 Effect of subgrade depth on the delta D-SNetr
FIGURE 3 Delta D versus effective structural number. relationship for a 12-in. pavement.
Hall and Elliott 13

Dynamic Stiffness Modulus, ksi attention. A number of procedures for backcalculating the
1600.--~~--~----~---~~-~--~
subgrade resilient modulus from deflection data have been
1400 --------- ----------------- developed (5,7,8,13). As part of the ROADHOG develop-
ment effort, Morrison (6) studied the types of backcalculation
1200
procedures available and their applicability to the soils and
1000 environmental conditions found in Arkansas. Morrison iden-
tified three general backcalculation proc~dures for determin-
800
ing subgrade modulus: iterative techniques, direct solution,
and empirical response algorithms. Each type of backcalcu-
lation technique has strengths and weaknesses.
400
The primary · goals is selecting a procedure for use in
200 ROAD HOG included accuracy, simplicity, and speed. Some
published backcalculation techniques are extremely elegant
o~~--~--~--'---~-"------L---------'

20 40 60 80 100 120 but were not yet considered_ practical for use in an everyday
0
Mix Temperature, F design procedure due to the equipment and time necessary
to run the analyses. In addition, the inherent variability of
FIGURE 5 AC temperature modulus relationship used in resilient modulus associated with in situ soils makes the expen-
SNEFF (2). diture of large amounts of time and energy to backcalculate
a modulus value to the nearest psi seem unproductive. The
value of subgrade resilient modulus backcalculated using data
The delta D/SNeff relationship developed using elastic layer from an FWD test represents the state of the subgrade soil
theory (e.g., ELSYM5) was verified using an alternate ap- at that particular point along the project and at prevailing
proach-the ILLl-PAVE finite element method (12). The moisture and stress-state conditions. Because of the variability
relationship between delta D and SNeff generated using ILLI- of soil properties in horizontal construction, the procedure
PAVE was virtually identical to that generated using ELSYM5. selected for estimating En for overlay design should be prac-
tical and yield reasonable Eri estimates.
· Another consideration in selecting a backcalculation pro-
FUTURE OVERLAY STRUCTURAL CAPACITY cedure is the appropriateness of using the modulus in the
ANALYSIS AASHTO design equation for flexible pavements. The orig-
inal performance equations developed from AASHO Road
In the ROADHOG design procedure, future overlay struc- Test data did not include any measure of soil support. To
tural capacity analysis consists mainly of two steps: (a) de- modify the AASHO performance equation for design, the
termination of the in situ subgrade resilient modulus and (.b) 1986 guide incorporated a subgrade resilient modulus function
calculation of the structural capacity required to carry future in which a value of 3,000 psi was assigned to the subgrade at
traffic. The subgrade resilient modulus, used in the structural the AASHO Road Test site. This value seems to agree with
capacity calculation, is determined from NDT data. The re- the breakpoint resilient modulus values obtained by Thomp-
quired structural capacity calculation is identical to a new son and Robnett (14) using Road Test soils. The breakpoint
pavement design. A discussion of the procedure used by resilient modulus is defined as the point at which the slope
ROADHOG in each of these steps follows. of the resilient modulus-repeated deviator stress curve (Fig-
Estimation of elastic properties (e.g., modulus) of pave- ure 7) typical of a fine-grained soil "breaks," or changes.
ment layers and subgrade from NDT data has received much
Resilient Modulus, Er

Correction Factor
2.5.----------------------~ Max.
2.25 i - - - - - - - - - - - - - - - - - - - - - - - - - - 1
21------------------
Eri.............K1 / E:i • "Breakpoint" Resilient Modulus

Plastic Yield

0.75
. 0.5

0.25 Sdi
o~----'-----'----"-------'----_J---~
40 50 60 70 80 90 100
Temperature (deg F)
FIGURE 7 Typical representation of the resilient modulus-
FIGURE 6 Temperature adjustment factors for 8-in. repeated deviator stress relationship for fine-grained soils
pavement in SNEFF. (2).
14 TRANSPORTATION RESEARCH RECORD 1374

To be consistent with the development of the AASHTO pavement design. ROADHOG uses the AASHTO flexible
design equation, the subgrade modulus value used in the pavement design procedure in the module NEWFLEX to
equation should be Eri• the breakpoint resilient modulus. Of determine structural requirements (1). User input for the
the backcalculation methods available, only some empirical NEWFLEX module includes the design reliability, design
response algorithms based on the finite element pavement standard deviation, design change in serviceability index (delta
model ILLI-PA VE calculate the breakpoint subgrade resilient PSI), and the number of ESALs for the design period. The
modulus. Modulus values calculated by other methods must subgrade resilient modulus, calculated by the MRCALC mod-
be adjusted to remain consistent with the AASHTO design ule, is supplied by ROADHOG.
equation (3).
ROADHOG uses an empirical response algorithm devel-
oped by Elliott and Thompson (2) to estimate the subgrade OVERLAY THICKNESS SELECTION
resilient modulus. The algorithm was developed from data
generated by the finite element structural pavement model Thickness selection of a flexible (structural) overlay for
ILLI-PAVE. Finite element-based backcalculation proce- an existing flexible pavement is straightforward. The
dures have several advantages over elastic layer- and numeric- ROADHOG procedure uses the structural number relation-
based procedures (6). However, a major obstacle to using a ships shown in Equations 2a and 3 for thickness selection.
finite element analysis in routine design is the complexity of In the ROADHOG procedure, overlay thickness selection
the calculations involved. To date, mini- or mainframe com- is performed by the module OVLTHICK. OVLTHICK ob-
puters are needed to fully exploit the advantages gained by tains the values of effective and required structural number
using the finite element method. Empirical respof!se algo- from the data file and prompts the user for the asphalt con-
rithms like those developed by Elliott and Thompson help to crete material coefficient. A value of aac equal to 0.44 is rec-
bring finite element methods to the routine design level. ommended to the user; however, the user may elect to use
A finite element model used to generate a response algo- another value if it is deemed to be more appropriate. The
rithm must be· valid for the conditions under which it is used. 0.44 value is the average material coefficient for asphalt con-
The data base used to develop the empirical response algo- crete as determined from the AASHO Road Test.
rithm discussed here was comprehensive, covering a wide
range of asphalt concrete and granular base thicknesses and.
subgrade strengths. Elliott and Thompson actually developed ANALYSIS UNIT DELINEATION
three equations for estimating the subgrade modulus beneath
existing flexible pavements: (a) surfac~ treatments-·asphalt Analysis unit delineation is a process by which a length of
concrete thickness equal to 0.0 in., (b) conventional flexible pavement slated for rehabilitation (e.g., overlay) is subdi-
,pavements-asphalt concrete thicknesses ranging from 3 to vided into homogenous sections. Homogeneous sections or
16 in., and (c) full-depth pavements-asphalt concrete rang- analysis units have been defined as "sections of pavement that
ing from 4 to 16 in. The three equations are nearly identical can be considered nearly alike in terms of performance, age,
and produce practically the same subgrade modulus predic- traffic, structural capacity, etc., and for which a single treat-
tion. Therefore for practical purposes the equation selected ment is appropriate" (8). Subdividing a project into analysis
for use in ROADHOG covers the range of AC thicknesses units can greatly increase the efficiency and cost-effectiveness
from 0 to 16 in. The ranges of material properties used in the of an overlay design. The use of analysis units can help to
analysis agree with observed material properties in the state ensure that the optimum amount of overlay is placed where
of Arkansas, validating the use of the response algorithm for it is needed.
the ROADHOG procedure. The subdivision of an overlay project into analysis units.
The calculation of in situ subgrade resilient modulus for the may be performed at a number of occasions in the overlay
ROADHOG procedure is contained in the module MRCALC. design process. Unit delineation should be performed before
The finite element- based response algorithm uses a single any pavement testing or design analysis based on construction
measured deflection to estimate the subgrade resilient mod- records, visible pavement distress, known subgrade condi-
ulus. The regression equation takes the form tions, and so forth The overlay design would then be per-
formed separately on each predetermined analysis unit. Anal-
Eri = 25.0346 - 5.2454D3 + 0.2864m (4) ysis units may also be defined by a material sampling program
or NDT data such as maximum deflection under load, and
where Eri is the breakpoint subgrade resilient modulus and the overlay design performed separately on each predeter-
D 3 is the surface deflection at 3 ft from load. mined analysis unit. ROADHOG performs analysis unit de-
Deflection data from sensors spaced at 0, 1, 2, and 3 ft· lineation on the basis of the actual required overlay thickness
from the load were analyzed during the development of the determined at each NDT test site, making unit delineation
response algorithms. The data from the sensor at 3 ft had the the final step in the ROADHOG overlay design procedure.
highest correlation coefficient (0. 99) with the calculated re- The ROADHOG procedure uses the "cumulative differ~
silient modulus. The standard error of estim~te for the re- ence method" outlined in the 1986 AASHTO Guide to per-
sponse algorithm was 0.64 ksi. The comprehensive data base form unit delineation in the module UNIDEL. A full discus-
and an excellent fit make the algorithm a powerful compu- sion of the statistical method is contained in Appendix J of
tational tool. the AASHTO Guide (1). ROADHOG uses the required ov-
The determination of the structural capacity (structural erlay thickness calculated for each measured deflection basin
number) required to carry future traffic is identical to new as the response variable in the procedure. The actual required
Hall and Elliott 15

overlay thickness is the most reasonable estimate of the struc- the algorithms used by ROADHOG) should be performed
tural deficiency of the pavement at each NDT test site. with the local material properties.
The UNIDEL module allows the designer to set the min-
imum length of an analysis unit. For long projects, a rec-
ommended minimum length of analysis unit is 1,000 ft (8). CONCLUSION
The minimum length should be based on economics and prac-
ticality. UNIDEL establishes "calculated analysis units" based ROADHOG has proven to be a practical, easy-to-use design
solely on the statistical procedure outlined above. "Recom- procedure for determining the overlay thickness needed to
mended analysis units" are determined by combining calcu-. correct structurally deficient flexible pavements in Arkansas.
lated units shorter than the minimum with adjacent units. The procedure follows the general design approach contained
After recommended units are determined, UNIDEL assigns in the 1986 AASHTO Guide but incorporates some improved
each station .along the project a unit number . Output of re- features. SNeff is determined by a method that is independent
sults according to analysis units is based on the assigned unit of the subgrade resilient modulus (Mr) and the depth to bed-
numbers. rock, and Mr is determined in a manner consistent with the
AASHTO design equation and not requiring the modification
needed by other backcalculation methods. The unit deline-
DESIGN RELIABILITY ad.on method used in ROADHOG assists the designer in
optimizing the design by identifying areas needing different
One difficulty in making meaningful comparisons between overlay thicknesses.
ROAD HOG and other overlay design procedures is the method
of applying a reliability level to the design. Reliability is the ACKNOWLEDGMENTS
probability that a design will perform as intended. for the
design period. Thus a design with "50 percent reliability" has This paper is based on Project TRC-8705, "Development of
a 50 percent chance of performing satisfactorily; conversely, a Flexible Pavement Overlay Design Procedure Utilizing Non-
the design has a 50 percent change of failing during the design destructive Testing Data." TRC-8705 was conducted by the
period. Arkansas Highway and Transportation Research Center,
In NDT overlay design, a level of reliability can be applied University of Arkansas. The project was sponsored by the
to the required thickness at each individual NDT test point, Arkansas State Highway and Transportation Department and
to the overall average required thickness, or to both. How- the U.S. Department of Transportation, Federal Highway
ever, the meaning of applying a reliability to the average Administration.
required thickness from thicknesses already determined at a
reliability level is unclear. An in-depth study is needed to
determine a meaningful method of handling reliability in REFERENCES
overlay design.
1. AASHTO Guide for Design of Pavement Structures. American
Association of State Highway and Transportation Officials,
Washington, D.C., 1986.
IMPLEMENTATION 2. R. P. Elliott and M. R. Thompson. Mechanistic Design Concepts
for Conventional Flexible Pavements. Transportation Engineer-
ROADHOG was completed in May 1990 and turned over to ing Series No. 42, University of Illinois, Urbana, 1985.
AHTD for trial implementation. For approximately 1 year, 3. R. P. Elliott. Selection of Sub grade Modulus for AASHTO Flex-
ible Pavement Design. Presented at the 71st Annual Meeting of
the AHTD research staff evaluated ROADHOG by using it the Transportation Research Board, Washington, D.C., 1992.
together with other overlay design approaches to develop 4. R. P. Elliott. An Examination of the AASHTO Remaining Life
thickness recommendations for the Roadway Design Divi- Factor. In Transportation Research Record 1215, TRB, National
sion. After 1 year, ROAD HOG was released to Roadway Research Council, Washington, D.C., 1989.
Design to oe used as the primary, routine overlay design tool. 5. A. 0. Bohn, P. Ullidtz, R. Stubstad, and A. Sorensen. Danish
Experiments with the French Falling Weight Deflectometer. Proc.,
In a recent meeting, design engineers using ROADHOG ex- Third International Conference on the Structural Design of As-
pressed satisfaction with the procedure. The only reservations phalt Pavements, London, 1972.
expressed were relative to some very thin overlay thicknesses 6. N. T. Morrison. Prediction of Su~grade Elastic Moduli Through
from a few projects. However, a review of these projects Nondestructive Testing for Arkansas Sub grade Soils. Masters the-
sis. University of Arkansas, Fayetteville, 1990.
revealed that, whereas the pavements needed rehabilita- 7. H. J. Larsen and R. Stubs tad. The Use of Nondestructive Testing
tion, this need was not due to structural inadequacy; and in Flexible Pavement Rehabilitation Design. International Sym-
ROAD HOG, like other NDT-based procedures, only ad- posium on Bearing Capacity of Roads and Airfields, Trondheim,
dresses structural inadequacy. Norway, 1982.
As stated in previous sections, methods used for SNeff and 8. F. N. Finn and C. L. Monismith. NCHRP Synthesis of Highway
Practice 116: Asphalt Overlay Design Procedures. TRB, National
Mri estimation are based on material properties representative Research Council, Washington, D.C., 1984.
of conditions encountered in Arkansas. To implement the 9. S. H. Kong. Determination of Effective Structural Number in
ROADHOG procedure in other areas, care must be taken Flexible Pavement Overlay Design. Masters thesis. University of
to ensure that the material properties used by ROADHOG Arkansas, Fayetteville, 1989.
10. S. Kopperman et al. ELSYM 5; Interactive Microcomputer Ver-
are representative of local conditions. If local material prop- sion, Users Manual: IBM-PC and Compatible Version. Report
erties vary significantly from those used in ROADHOG mod- FHWA-TS-87-206. Federal Highway Administration, U.S. De-
ules, additional analyses (similar to those used in developing partment of Transportation, 1986.
16 TRANSPORTATION RESEARCH RECORD 1374

11. R. P. Elliott, S. I. Thornton, K. Y. Foo, K. W. Siew, and R. fore a refinement and improvement of the traditional overlay
Woodbridge. Resilient Properties of Arkansas Subgrades. Final equation.
report-TRC-94. UAF-AHTRC-88-002, University of Arkan-
sas, Fayetteville, 1988.
12. ILLI-PAVE: A Pavement Analysis Program Provided by the
Transportation Facilities Group. Department of Civil Engineer- SIGNIFICANCE OF REMAINING LIFE
ing, University of Illinois at Urbana-Champaign, Urbana.
13. F. W. Jung. Numerical Deflection Basin Interpretation and Tem- FACTOR FRL
perature Adjustment in Non-Destructive Testing of Flexible Pave-
ments. Technical Report 880187, Ontario Ministry of Transpor- The remaining life factor can be derived and shown to be a
tation, Ontario, Canada, 1988. .function of the structural capacities of existing and overlaid
14. M. R. Thompson and Q. Robnett. Final Report-Data Sum-
mary, Resilient Properties of Subgrade Soils. Transportation En- pavements and hence a function of the remaining lives of the
gineering Series No. 14, University of Illinois, Urbana, 1976. pavements (1,4). Referring to Appendix CC of the 1986
AASHTO Guide, where an excellent description of the re-
maining life concept is presented, it is obvious that the correct
FRL value to be used in Equation 5 should be determined by
DISCUSSION considering the overlay requirements at all stages in the over-
lay design life. A detailed explanation of how this can be done
is given elsewhere (4), where a procedure for selecting the
T. F. FwA governing FRL value is described. The FRL value so determined
Center for Transportation Research, Faculty of Engineering, National
University of Singapore. will lead to the choice of SN 0 L from Equation 5 that provides
an adequate overlay thickness for the entire period of the
The 1986 AASHTO Guide (1) incorporates a remaining life design service life. It is apparent that this important aspect
concept in overlay thickness design and expresses the overlay of remaining life factor computation is not considered by El-
structural requirement in the following form (Equation 2 in liott (2) and the authors in their analyses of overlay design.
the paper): Not realizing this significance of the concept has probably
resulted in their call for exclusion of remaining life consid-
(5) eration from overlay design and their doubt of the statement
that factor FRL would "reflect a more realistic assessment of
Elliott (2) demonstrated that the overlay thickness compu- the weighted effective capacity during the overlay period."
tations using AASHTO design curves for FRL produced in-
consistent results and recommended that "the AASHTO
overlay design approach be revised to exclude remaining life
considerations." However, subsequent work by Easa (3) and BASIS OF AUTHORS' RECOMMENDATIONS
Fwa (4) illustrated. that the AASHTO concept of remaining
life is fundamentally correct and that consistent results are The authors state that "for all practical purposes, a value of
obtained using corrected FRL. Unfortunately, the current pa- 1.0 is the appropriate value for FRu" and that the inclusion
per continues to adopt the view expressed by Elliott (1989) of the remaining life concept "within the structural deficiency
and states that the inclusion of the concept of remaining life design approach used in ROADHOG was not found to be
"was not found to be reasonable or practical." It is shown in reasonable or practical." The only basis for these recommen-
this discussion that reverting to the use of the traditional dations was the work reported by Elliott (2). An examination
overlay equation SN 0 L = SN 0 - SNxeff (Equation 2a in the of Elliott's paper shows that there is little justification for the
paper, setting FRL = 1.0), as recommended by Elliott (2) and strong recommendations. The recommendations were based
the authors, is conceptually unsound. solely on an analysis using a "simple scale transformation"
that relates RLyx to RLy for a given RLx as follows:

SIGNIFICANCE OF REMAINING LIFE CONCEPT

{Rd {~} (6)

The basic difference between the AASHTO overlay equation

(Equation 5) and the traditional overlay equation lies in the Substituting Equation 6 into the AASHTO equation for cal-
fact that the traditional equation computes SN 0 L required at culating FRL, Elliott (2) concluded that the appropriate value
the time of overlay construction, and no check is made to for FRL was 1.0. Three points can be raised regarding Elliott's
ensure that the overlay provided will be adequate during the analysis:
entire design period. It can be shown (1,4) that, depending
on the structural deterioration rate of an existing pavement 1. The FRL value was computed for only one point (i.e., at
after overlay application, the overlay requirement at a later the end of the service life), which is not a correct way of
stage of the service life of the pavement may exceed the value selecting a design FRL value. For example, FRL values greater
computed by the traditional overlay equation. Using different than 1.0 were mentioned. This would not be the case if the
deterioration curves for old pavements after overlay, Easa proper procedure of selecting a design FRL value according
(3) and Fwa (4) showed that the overlay thickness requirement to the concept of remaining life is followed. At the time of
varied with the remaining lives of the old pavements. Includ- overlay construction, FRL would be 1.0 if evaluated then. This
ing the remaining life consideration in overlay design is there- effectively eliminates one from choosing an FRL value greater
Hall and Elliott 17

than 1 when the overlay requirements at other times during AUTHORS' CLOSURE
overlay service life are checked.
2. No physical meaning or practical significance was given We recognize the validity of remaining life as a concept that
to justify why the relationship in Equation 6 was used. The needs to be considered in overlay design. However, its ap-
"philosophy" given by Elliott (2) was "the concept of the man plication as presented in the 1986 AASHTO Guide is flawed
who each day walks halfway to his destination." This writer and adds complications to the design process that are not
finds it difficult to relate the philosophy to overlay perfor- warranted. Because of this, the 1986 approach to remaining
mance. It is, however, easy to show that Equation 6 has a life was not included in ROADHOG. The concept adopted
highly controversial implication not stated by the authors or for ROADHOG is that SNeff should represent the contribu-
Elliott (2): for a given pavement with known RLu the rate of tion of the pavement to the future performance after it is
change in RLyx is proportional to the rate of change in RLy- overlaid. If SNeff is selected properly, the application of the
This underlying assumplion of Equation 6 is severely restric- . FRL term represents a double penalty for the effects of past
tive in application, and it is not in agreement with common traffic. Research is needed to determine whether the method
understanding of how old and new pavements deteriorate. of selecting SNerr developed for ROADHOG selects the ap-
Easa (3) and Fwa (4) have shown that there are many other propriate value.
more meaningful deterioration relationships that would pro- We are familiar with the papers by Fwa and Easa. These
duce consistent overlay designs according to the AASHTO papers acknowledge the flaw in the 1986 guide remaining life
remaining life design concept, and that the values of FRL first observed by Elliott. Both papers attempt to remedy the
varied from about 0.5 to 1.0 for the cases they considered. flaw; however, the remedies further complicate a process that
3. It is physically meaningful to derive the pavement perfor- is already more complex than is warranted within the empir-
mance relationship in terms of pavement conditions or struc- ical AASHTO approach to pavement design. The AASHTO
tural capacity. Equations involving multiplication and division approach has served for many years as a useful design tool.
of remaining life fractions RL of different pavements are dif- However, it is a tool that has been extrapolated far beyond
ficult to interpret physically. This is because RL is a nonlinear its original data base, often with little justification other than
function of pavement structural condition, and it is a fraction engineering judgment. Its application to overlay design repre-
of load repetitions Nr, which are different for different pave- sents further extrapolation. The addition of a compli-
ments. Fwa (4) has shown that putting different RL values in cated, sophisticated approach to remaining life simply is not
an equation without paying attention to the different base Nr justified.
values can lead to erroneous results. For example, the "sim- In this respect,. the decision to not include the FRL term in
ple'; transformation of (RLx• RLyx) to (1, RLy) as shown in ROAD HOG was not solely because of its flaws. Even without
Equation 6 is deceptively straightforward. However, ex- the flaws there are reasons and sentiment for its removal.
pressed in terms of pavement structural condition or load Other reasons are the removal of unwarranted complications
repetition capacity, the "walking man philosophy" may not as cited above and the elimination of confusion and lack of
make any sense. understanding generated by the remaining life factor when it
was introduced in 1986.
The introduction of the FRL term into overlay design created
much confusion. Designers did not understand the term or
SUMMARY REMARKS
its application. Even veteran pavement researchers had trou-
ble accepting and understanding remaining life as it is pre-
This discussion deals with the overlay design methodology of
sented in the 1986 guide. This is perhaps best demonstrated
the paper. Explanation and findings of other studies have been
by the fact that the FRL term was introduced in i986 (and
presented to show that the AASHTO remaining life concept
reviewed and questioned by knowledgeable pavement engi-
is technically sound and that the recommendations by the
neers before that), yet the flaws in the concept were not noted
authors concerning the use of FRL = 1.0 and the exclusion
until 1989.
of the remaining life concept in overlay design are misleading
The development of the FRL methodology and the modi-
and not justified.
fications proposed by Fwa and Easa suggest a lack of under-
standing of the limitations of the AASHO Road Test perfor- ·
mance equations that serve as the basis for the AASHTO
REFERENCES pavement design procedures. These equations are best-fit
regression equations developed to predict the performance of
1. AASHTO Guide for Design of Pavement Structures (Vols. 1 and the pavement sections at the Road Test. Strictly speaking,
2). American Society of State Highway and Transportation Of- they are only valid within the very limited context of the
ficials, Washington, D.C., 1986. pavement types, axle loads, subgrade, environment, time, and
2. R. P. Elliott. An Examination of the AASHTO Remaining Life
Factor. In Transportation Research Record 1215, TRB, National so forth of the AASHO Road Test. To use them to develop
Research Council, Washington, D.C., 1989. a sophisticated concept of remaining life for overlay design
3. S. M. Easa. Extension of AASHTO Remaining-Life Methodology represents a gross extrapolation, far beyond the original data
of Overlay Design. In Transportation Research Record 1272, TRB, base or intent of the equations. Such use suggests that the
National Research Council, Washington, D.C., 1990.
4. T. F. Fwa. Remaining-Life Consideration in Pavement Overlay Road Test equations are fundamental behavioral relation-
Design. Journal of Transportation Engineering, Vol. 117, No. 6, ships. They are not. To use these empirical equations in this
1991. manner is an interesting academic exercise, but it is not valid.
18 TRANSPORTATION RESEARCH RECORD 1374

There is also a danger in incorporating procedures devel- The contents of this paper reflect the views of the authors, who are
oped in such a manner into routine engineering practice, es- responsible for the facts and accuracy of the data presented herein.
pecially when such procedures are complicated and sophis- The contents do not necessarily reflect the official views of the Arkansas
ticated in appearance. The incorporation and sophistication Highway and Transportation Department or the Federal Highway
Administration. This paper does not constitute a standard, specifica-
suggest a legitimacy that does not exist. Once accepted into tion, or regulation.
practice, the procedures can become "etched in stone" and The use of product names in this paper does not in any way constitute
very difficult to change or correct when more advanced tech- an endorsement, advertisement, or recommendation of the product.
nology becomes available.
TRANSPORTATION RESEARCH RECORD 1374 19

Nationwide Evaluation Study of Asphalt

Concrete Overlays Placed on Fractured
Portland Cement Concrete Pavements
MATTHEW w. WITCZAK AND GONZALO R. RADA

Historically, agencies responsible for pavement rehabilitation have distress constitutes the most frequent cause of the loss of
tried a wide variety of materials, processes, and construction performance for AC overlays.
methods to eliminate or minimize reflective cracking of asphaltic
concrete overlays placed on existing portland cement concrete Reflection cracks in the AC overlays are caused by a com-
(PCC) pavements. Over the last 10 years, the fractured slab ap- bination of thermal and traffic-induced stresses. Expansion
proach using rubblize, crack and seat, and break and seat has and contraction of the PCC pavement results in horizontal
gained increased acceptance. Because the fractured slab approach movements that produce strains in the AC overlay exceeding
has gradually evolved through field demonstration and actual its tensile strength. Traffic loads can cause vertical differential
projects, very little fundamental knowledge concerning design, movements at the location of joints and working cracks in the
construction, and performance models is available. Understand-
PCC slab and induce critical shear stresses at the bottom of
ably, performance to date has been variable. To improve the
state of the art and develop a better understanding of these tech- the AC layer. The overlay immediately over the joints and
niques, a nationwide study was undertaken. A literature review working cracks in the PCC is not able to accommodate these
resulted in the location of nearly 500 highway projects throughout localized movements, resulting in the development of reflec-
the United States. From this generalized data base, approximately tion cracks.
100 sites were selected for detailed field studies. Field crews con- A wide variety of rehabilitation techniques aimed at pre-
ducted visual distress surveys to assess pavement performance venting the formation of, or minimizing, reflection cracking
and nondestructive deflection testing to assess the in situ char-
acteristics of the pavement layers. The general approach used for have been attempted. They include thick (conventional) over-
the research study and the analysis of field performance and lays, crack relief layers, the saw and seal technique, special
structural data obtained is presented. Performance predictive overlay and interface materials, and the fractured slab ap-
equations are presented along with the evaluation of the back- proach. Of these, the technique that has been used increas-
calculated effective moduli of fractured PCC slabs for each ingly over the last 10 years has been the fractured slab approach.
technique. Analysis of within and between project variability is The major objective of the fractured slab approach is to
presented.
reduce the effective in situ slab length before the overlay is
placed. If this is effectively accomplished, the likelihood of
The selection of optimal rehabilitation procedures and strat- having reflective cracks appear is significantly reduced or
egies for deteriorating highway pavements requires a knowl- eliminated. The probability of reflective cracking is propor-
edge of the type and cause of the distress, determination of tional to the horizontal movement at joints and cracks, which
candidate rehabilitation procedures, and selection of an op- in tum is directly proportional to the spacing between joints
timal strategy based on economic and other considerations. and cracks.
For portland cement concrete (PCC) pavements, the array of The fractured slab category is generally subdivided into
possible rehabilitation procedures includes nonoverlay meth- three major types of rehabilitation: rubblize, crack and seat,
ods such as undersealing, grinding of the surface, and removal and break and seat. Rubblize is a fractured slab process in-
and replacement of distressed areas; fulf reconstruction by tended to transform the existing PCC layers into fragments
replacement or recycling; PCC overlays; and asphaltic con- having textural and gradational characteristics similar to those
crete (AC) overlays. of a large aggregate size crushed stone base. It is most effec-
Review of current practice indicates that AC overlays are tively accomplished with a resonant pavement breaker, which
the most commonly used PCC rehabilitation procedure, with has been successfully used on all types of existing PCC pave-
about $1 billion of AC overlays placed· each year, and this ments [i.e., jointed plain (JPC), jointed reinforced (JRC),
amount will likely increase in the future (J). Even though and continuously reinforced (CRC) concrete pavements].
they are commonly used, the performance of AC overlays on Crack/seat and break/seat are fracture techniques intended
PCC pavements is often hampered by the occurrence of re- to produce very short rigid slabs whose effective lengths vary
flection cracks over existing joints and cracks. This type of from 12 to 48 in. The techniques are similar, with guillotines
or spring-arm (whip) hammers being used to develop reduced
crack spacings in the existing PCC pavement. There is, how-
M. W. Witczak, Department of Civil Engineering, University of ever, a significant distinction between the two techniques.
Maryland, College Park, Md. 20740. G. R. Rada, PCS/Law Engi- Crack/seat is associated with the fractured slab process con-
neering, 12240 Indian Creek Court, Suite 120, Beltsville, Md. 20705. ducted solely on JPC pavements. For these pavements, the
20 TRANSPORTATION RESEARCH RECORD 1374

objective of the crack/seat process is to develop closely spaced, federal and state highway agencies, the Transportation Re-
tight cracks that permit load transfer across the crack through search Board, the National Asphalt Pavement Association,
aggregate interlock with little loss of structural value. Fracture the Asphalt Institute, and other highway-oriented organiza-
or cracking through the entire depth of the PCC layer is the tions. During the conduct of this study, several extremely
ultimate goal. relevant studies were completed, including NCHRP Synthesis
Break/seat is associated with the fractured slab process on 144 (2), FHWA Contract DTFH 61-86-C-00079 (3,4), and
JRC pavements. The ultimate objective of this technique is individual state highway agency investigations. A complete
to physically fracture the distributed steel or completely de- list of the information sources may be found in the final report
bond the steel from the concrete. Whereas cracking may result of this study (5).
through the entire PCC layer depth, if steel fracture or de- The collected information was used to prepare a synthesis
bonding is not accomplished, the effective slab length is not of current practice. This included design, construction, spec-
reduced in the construction process, and what remains is a ifications, costs, and performance experience for projects us-
series of smaller slabs tied together into a longer effective ing the crack-controlling techniques being studied. The data
slab by the bonded distributed steel. base was also used to identify AC overlay rehabilitation proj-
A corequisite to the slab fracturing process is the seating ects that used slab fracture techniques to control reflection
portion of the construction. For both cracking and breaking, cracking as candidates for the field investigations.
it is customary to have five to seven passes of a 35- to 50-ton A total of.454 field projects in 34 states were identified for
rubber-tired roller seat the fractured slab fragments. This pro- which at least partial design, construction, and performance
vides a relatively smooth and uniform grade for paving op- information was available. Of the 454 projects, 250 were crack/
erations and serves as an excellent means of proof-rolling seat, 150 break/seat, 19 rubblize, and 35 unknown (steel in-
before the AC overlay is placed. For rubblized projects, steel formation was not available; thus, they could not be grouped
vibratory rollers (generally 10 tons) are normally used for the into the crack/seat or break/seat categories). The geographic
compaction or seating process. distribution of the projects by rehabilitation type is shown in
Figure 1. In general, crack/seat projects were concentrated
in the upper midwest and western states, break/seat projects
NATIONAL EVALUATION STUDY in the northeastern portion of the country, and rubblize proj-
ects in the eastern half of the country.
Objectives and Overview Of interest from a practical viewpoint is a summary of sev-
eral key variables given in Table 1 for each of the three re-
Over the past 10 years, there has been a dramatic increase in
the use of the fractured slab techniques for the rehabilitation
of deteriorating PCC pavements. Much field experience has
been gained during this time. However, little technical guid-
ance, relative to the design and construction of these tech-
niques, is available to adequately predict their performance
in minimizing reflective cracking under specific traffic and
climatic conditions for a particular pavement structure and
existing condition.
In recognition of the critical need for a sound technical
basis to support the use of the fractured slab approach, a Crack and Seat
major state-of-the-art research study was undertaken. The
overall objective of the study was to develop national guide-
lines and methodologies for the use of these three rehabili-
tation techniques.
This paper presents the general approach used for the re-
search study and concentrates on the analysis of field perfor-
mance and structural data obtained. Performance predictive
equations, using the pavement condition index (PCI), are
presented along with the evaluation of the backcalculated
effective moduli of fractured PCC slabs for each rehabilitation Break and Seat
category. Analysis of within-project and between-project
variability is presented and forms the basis for an overall
design methodology described in a companion paper in this
Record by Witczak and Rada.

Synthesis of Current Practice and Data Base

Development

Sources of information initially collected and evaluated were FIGURE 1 Geographical distribution of projects by
obtained from various state asphalt pavement organizations, rehabilitation type.
 Witczak and Rada 21

habilitation types. As can be observed, the average dates of driving lane, as well as the adjoining five segments in the
the rehabilitation options clearly show the relative "youth- passing lane, made up the 10 sample units used for recording
fulness" of the methods discussed, particularly for the rubblize PCI data. For each sample unit, the types of distress present
technique. The table also indicates that the average PCC were identified and their severity and extent were quantified.
thicknesses range between 8 and 10 in., which are typical of The deflection testing was conducted every 25 ft for the
highway pavements. Finally, the resulting average statistics 1,000-ft length, in the outside wheelpath of the driving lane.
for the AC overlay thicknesses show that the rubblize tech- Each test point consisted of three drops at a target 9,000-lb
nique had the largest average overlay thickness. This is con- load with a Phonix MlOOOO FWD. Deflections were measured
sistent with the fact that the rubblization process is intended for every drop by means of six geophones located at distances
to truly transform an existing rigid layer into a conventional ranging from 0 to 60 in. from the center of the load plate.
flexible layer. Deflection data were. also received for 42 test sections in Il-
A computerized data base was produced by compiling in- linois, 8 test sections in Michigan, and 4 test sections in Ken-
formation on these projects from available documents and tucky. In all, 4,700 NDT test points on 140 sections were
reports plus follow-up contacts with state highway agency obtained and subsequently used in the structural analyses-
personnel to search files for additional information. In this 1,019 points on 24 rubblized sections, 1,776 points on 64 crack/
study, two types of data bases were developed: (a) general seat sections, and 1,905 points on 52 break/seat sections.
data base and (b) detailed data base. The general data base
contains available data· and information found previously
through the literature search. Wherever possible, this infor-
mation was used during the analysis portion of the study. Data Analysis and Interpretation
However, the major use of this data base was to select pave-
ment test sections for the field investigation program. These Once the field testing was completed on all test sections and
projects formed the detailed data base. data incorporated into the detailed data base, an extensive
A primary consideration in the selection of the field test analysis was initiated. The major thrust of the analysis was
sections was to include a range of variables that influence to evaluate the rehabilitated pavements from both a functional
pavement performance-for example, climate, AC overlay and structural point of vie.w. The results of the PCI visual
thickness, rehabilitation age, and age of existing PCC pave- condition surveys were used to estimate performance trends .
. ment before overlay. A total of 93 pavement sections-17 Nondestructive deflection data, coupled with available cross
rubblize, 35 crack/seat, and 41 break/seat projects-were section information, were used to backcalculate the effective
eventually selected from the general data base for the field in situ modulus for the pavement section layers.
investigation program. An additional 54 sections were in- The PCI values were determined according to the U.S.
cluded in the detailed data base as a result of deflection testing Army Corps of Engineers procedure (6). In addition, a new
and other pertinent d~ta provided by the Kentucky, Illinois, index value was introduced into the analysis: PCI,t. The PCI,t
and Michigan highway agencies. value reflects the PCI value due only to the presence of lon-
gitudinal/transverse cracking in the overlay-it excludes all
distress types other than the cracking. Because prevention of
Field Investigations reflective cracking is a major concern to the efficiency of the
rehabilitation activity investigated, it was believed that the
Because of the limited availability of documented field perfor- PCU,t value would perhaps be an-other appropriate statistic
mance for the fracture slab techniques, a primary activity of to analyze.
this research study was a field investigation of existing highway Effective moduli of the pavement layers were backcalcu-
pavement sections where these techniques had been used. The lated from elastic layer theory and the measured deflection
two major field testing activities were visual distress surveys basins for each specific test location. One of the major under-
made in accordance with the PCI highway methodology and lying hypotheses of the backcalcufation study deals with the
nondestructive deflection testing (NDT) using a falling weight fundamental concept that a direct relationship exists between
deflectometer (FWD). the fractured PCC modulus (Epcc) value and the overall ef-
The test sections generally consisted of a 1,000-ft strip in fectiveness of the construction operation in reducing the ef-
one direction. For the PCI data collection effort, this strip fective in situ moduli. Thus, the lower the Epcc value, the
was divided into five 200-ft segments. The segments in the greater the effectiveness of the construction operation in min-

TABLE 1 Summary of Average Statistics of Field Test Sections

Rehabilitation Average Date of Average AC Average PCC

Type Rehabilitation Overlay, in. Thickness, in.
Crack and Seat 1984 4.4 8.3
Break and Seat 1985 5.6 9.4
Rubblize 1986 6.0 8.9
Saw and Seal 1983 3.4 8.3
22 TRANSPORTATION RESEARCH RECORD 1374

imizing the potential for eventual reflective cracking of the tions. The rubblized and crack/seat techniques appear to be
AC overlay. the best-performing systems, whereas the typical break/seat
All sections were analyzed as three-layer structures, with sections do not respond quite as well. On the basis of these
the bottom (subgrade) layer being of semi-infinite thickness. results, it can be stated that the ranking of techniques, in
These structures were analyzed with a PCS/LAW in-house order of decreasing typical performance life, appears to be
program called EMOD. This program uses the Chevron rubblization (best), crack/seat, and break/seat (worst).
N-Layer elastic solution as a subroutine within the backcal- The multivariate PCI predictive equations indicate that rel-
culation analysis, and the respective moduli given in the out- atively good models were developed for each rehabilitation
put are those that result in the minimum cumulative residual technique except the break/seat. The recommended models
square error at all deflection readings. Required thickness are summarized in Table 3 for both PCI and PCI1t. As can be
data were obtained from the respective state DOTs. Poisson's observed, the R 2 values for the rubblize and crack/seat equa-
ratios were assumed to be 0.35 for the AC overlay, 0.30 for tions are good. However, this fact must be tempered with the
the fractured PCC layer, and 0.4 for the subgrade. <. small data base coupled with the relative youth of these tech-
On the basis of the resulting PCI, PCI10 and fractured PCC niques. Furthermore, the use of these equations in a design
modulus (Epcd data, three analyses were conducted: (a) de- sense is. not recommended because of the significant degree
velopment of PCI and Epcc predictive models, (b) investi- of extrapolation outside the data variable levels that would
gation of the influence of crack spacing on the Epcc. and (c) be required.
investigation of the variability of the Epcc values between and Whereas the rubblize and crack/seat equations are limited
within projects. This paper cannot address all of the results in use, the rationality and sign of the coefficients for both
and findings that were generated from these analyses, but time and overlay thickness are reasonable; PCI decreases with
several key results are presented that form the basis for the time and increases with overlay thickness. The subbase mod-
design methodology described in the companion paper in this ulus is likewise considered to be a logical variable, reflecting
Record. a portion of the existing pavement support capacity. Because
of the small data base, only a linear variable in time was found,
although it is reasonable to expect that a nonlinear time var-
iable would best model the behavior. Attempts to incorporate
MAJOR STUDY RESULTS traffic and other variables proved fruitless due to the lack of
accurate information in the data base. Finally, the presence
PCI Predictive Equations of the precipitation and temperature terms in the models can-
not be conclusively rationalized because they may either be
One of the major objectives of the data analysis was the false indicators of other more important geographical factors
development of equations to predict both the pavement con- not studied or they may indeed reflect the significant influence
dition index (PCI and PCI1t) and the in situ modulus of the of the environment on the performance of these sections.
fractured PCC layer (EPcd· To accomplish this, multiple Their inclusion, however, was found to significantly increase
regression techniques were used to develop hundreds of models, the overall R 2 value of the predictive models.
using as many as 15 independent variables. Only the best For the break/seat technique, the models given in Table 3
regression models are presented in this paper. The ultimate were the best models developed. However, they are consid-
criteria used to select the best models were the correlation ered to be poor performance predictors, as judged by the low
coefficient (R 2 ) value and engineering reasonableness of the R 2 values obtained. A significant effort was expended toward
significant parameters (independent variables) coupled with obtaining models with greater predictive accuracy, but this
the respective sign of the coefficients. effort proved fruitless, even though the maximum perfor-
Before discussing the results of the multivariate PCI pre- mance period and number of sections within the data base
dictive equations, it is both important and revealing to present were the greatest for this technique. Several equations were
global trends of PCI versus time from the rehabilitation con- developed with R 2 values in the 0.40 to 0.57 range, but they
struction. Table 2 summarizes the PCI1t-time predictive equa- were rejected due to the unreasonable coefficient signs. Thus,
tion developed for each of the three fractured slab techniques the high variability of the performance could not be explained
studied and the average time to reach typical failure condi- in an analytical manner.

TABLE 2 Global PCI-Time Models and Times to Failure

Rehab i1 i ta ti on General PCI-Time Model Time to Time to

Type PC! - 50 PC! - 40
Rubblize PCiu - 100 - 1. 613t + 0.092t2 * *
Crack/Seat PCiu - 100 - 0.343t - 0.136t2 18.0 years 19.8 years
Break/Seat PCiu - 100 - 0.050t - 0. 316t2 12.5 years 13.6 years
All Fractured PCiu - 100 - 0.149t - 0. 252t2 13.8 years 15.2 years
Slabs

(*) Unable to project time as PCI>90 at t-8 to 10 years.

Witczak and Rada 23

TABLE 3 PCI Predictive Equations

Rehabilitation Type PC! Prediction Model R2

Rubblize PC! - 173.77 - 0.878t + 0.753
0.389h0 - 0.744P - 0.719T
PCI 1 t - 217. 28 - 0.615t + 0.665
0.310h0 - 1. 14P - 1. 13T
Crack/Seat PC! - 127.02 - 3.03t + 2.4h0 - 0.755
0.34P - 0.44T + 0.09EsUB
PCilt - 116.8 - l.42t + 0.658ho 0.511
- 0.13P - 0.25T + O.OOlEsUB
Break/Seat PC! - 100.0 - 0.249t 2 - 0.0027 0.393
(T*JS) + 0.072EsUB + 0.465h0
PCilt - 100.0 + 5.6lt - 0.77t 2 0.443
- 0.0032(T*JS) + 0.8lh0
Note: t - time in years; h 0 - HMA overlay thickness (inches); P - annual
average precipitation (inches); T - annual average temperature range
(F°); EsUB - subgrade (subbase) modulus (ksi); JS - joint spacing·
(feet).

Fractured PCC Modulus Predictive Equations encing the modulus are the subbase stiffness, crack spacing,
and seating load used in the construction process. As one
Like the PCI analysis, reasonable predictive models for the would expect, the Ercc value increases as these variables are
fractured PCC modulus (Ercd were obtained for the rub- increased.
blized and the crack/seat techniques. The recommended model Finally, the statistical analysis to develop an Ercc predictive
for the rubblized technique is model for the break/seat technique did not result in a positive
conclusion. More than 35 model forms were tried, but in no
Ercc = 1690 - 15.4P - 17.34T + 2.2Esua case was a model found with an R 2 value greater than 0.20.
This finding is consistent with the large scatter and range of
+ 1.9JS + 34.7CS R 2 = 0.603 (1) modulus values determined in the study. It was concluded
that the primary cause of this problem is the large variability
whereas that for the crack/seat technique is within the construction process to fully achieve steel debond-
ing or steel fracture, or both.
968.39 + 20.34Esua + 34.89CS
+ 5.37SL R2 = 0.776 (2) Influence of Crack Spacing on the Fractured PCC
where Modulus Values
P annual average precipitation (in.),
The effective modulus of a fractured slab is a function of the
T = annual average temperature range (°F), nominal fragment size actually achieved during the construc-
Esua = subgrade (subbase) modulus (ksi), tion process. In practice, this concept has been applied to
CS = crack spacing (iri.), and specifications that substitute the proposed crack spacing to
SL = seating load (tons).
be achieved as an indirect indicator of the effective modulus
For the rubblized data, the relatively good R 2 value prob- of the fractured layer. Whereas general correlations exist be-
ably reflects the small range of Ercc values (200 to 700 ksi). tween crack spacing and modulus, the use of visual assess-
The presence of the precipitation and temperature variables ments is not always accurate for estimating the fractured mod-
cannot be conclusively rationalized; they may indirectly re- uli of the PCC layer. As the spacing is reduced, the PCC layer
flect other geographic or environmental variables not consid- behaves less like a slab having a sound modulus of approxi-
ered in the analysis. The joint spacing variable is not of real mately 5,000 ksi and more like a flexible layer with a signif-
significance as indicated by the small coefficient. The subbase icantly lower modulus value.
modulus variable is believed to be reasonable and important; It is common to denote the relationship between spacing
it implies that as the sub base becomes stiffer, more energy is and reduced effective PCC modulus by the use of a modular
required to achieve a given Ercc value. Finally, the influence ratio parameter (E,):
of the crack spacing appears to be significant at first glance,
but a closer look at the data reveals that this variable only E = Ercc (3)
ranges from 6 to 12 in. r Esound
On the basis of the significant data base size and the rel-
atively high R 2 value obtained, it is believed that the predictive where Ercc is the effective PCC modulus of fractured slab
model developed for the crack/seat technique is quite good. and Esounct is the modulus of elasticity of sound PCC (E
The model indicates that the three primary variables influ- 5,000 ksi).
24 TRANSPORTATION RESEARCH RECORD 1374

Using the fractured PCC moduli found in this study, av- deviati,on, and other statistics were calculated. Whereas the
erage Er values as a function of crack spacing were developed results appeared to be highly variable at first glance, further
and compared with the recommended 1986 AASHTO rela- detailed analyses of the data led to important conclusions
tionship (7). Figure 2 shows that relatively good agreement regarding the two major forms of variation encountered: "be-
exists for the rubblized and crack/seat sections. However, tween" and "within" project variability.
major differences are present for the break/seat sections. As The between project variability reflects the variation be-
shown, the modular ratio results are approximately two to tween the average project predicted EPcc values. As such,
three times as large as those for the crack/seat sections at the the standard deviation (crb) or variance (er~) reflects the varia-
same specified crack spacing. This fact, reinforcing previous tions attributable to each construction process on a national
results and analyses, led the authors to. conclude that the scale. Specifically, factors such as the type of equipment,
break/seat process may not effectively achieve full debonding specific breaking energy, specified crack spacing, and the spe-
of the steel from the PCC or successful fracturing of the steel. cific site factors and pavement cross section are all reflected
Because of this, the actual effective slab length after fracturing within the erb (er~) parameter.
is much greater than one would conclude by looking at the The analysis of the statistical frequency distribution pat-
actual crack spacing. terns of the average project EPcc value was used to assess the
In the previous section, predictive equations for the Epcc variability (crb or er~) of a particular rehabilitation type for the
values were presented for both rubblized and crack/seat sec- spectrum of construction projects examined across the United
tions. These models illustrate the importance of crack spacing States. Figure 3 shows the between project EPcc frequency
on the fractured PCC modulus. As the spacing is increased, distributions for each of the three techniques in question.
EPcc increases. In the case of the crack/seat technique, historic Table 4 summarizes the between project EPcc statistics.
information has shown that crack spacings typically range Figure 3 shows that the rubblized EPcc project distribution
from 12 to 60 in. However, on the basis of the results of this is somewhat normal (actually bimodal) and contained within
study (i.e., Epcc predictive equation), the following crack a relatively small range of EPcc values. Similarly, the fre-
spacings as a function of the foundation type are recom- quency distribution for the crack/seat sections investigated is
mended: 30 in. for subgrade soils, 24 in. for granular subbase, normally distributed and is contained within a small range of
and 12 in. for stabilized subbase. Target spacings decrease as Epcc values. There are, however, several outliers of data found
the stiffness of the underlying foundation is increased. Rub- at high modulus values.
blization generally results in fragment sizes in the range of 6 Unlike the rubblized and crack/seat projects, the break/
to 12 in., which appears satisfactory for this technique. seat Epcc values are widely distributed and highly indicative
of the variable success in fracturing or debonding distributed
steel in the concrete. Typical values of as low as 250 ksi to
Variability of Fractured PCC Modulus Values as large as 2,750 ksi were obtained. The distribution appears
to be more uniform across these limits rather than normally
The analysis of the deflection test data consisted of 4,700 distributed.
backcalculated estimates of the in situ fractured PCC modulus In contrast to the between project variation, the within
(Epcc) on 140 sections (64 crack/seat, 52 break/seat, and 24 project variability (crw or er~) reflects the variation of the EPcc
rubblize). For each section, the average EPcc value, standard values obtained within a given rehabilitation project. As such,

IW
0
~
LEGEND:
! 0.6 * Crack I Seat
o Break I Seat
"3
• Rubblized
"8
:::IE
i
us
0
~ 0.4

it 0.2
~
<(

Project Specified Crack Spacing Onches)

FIGURE 2 Relationship of average PCC modulus ratio versus specified project crack spacing
for fractured slab techniques.
Witczak and Rada 25

The results comparing the within project standard deviation

(crw) to the average EPcc (project) modulus value for all of
30 the fractured slab project data are shown in Figure 4. The
line passing through the origin represents the average best-
fit relationship to the data points. In turn, the slope of this
20
line represents the average coefficient of variation (CVw) for
the within project variation.
10 As observed, a typical coefficient of variation of CVw =
40 percent may be viewed as appropriate for all construction
0 projects studied. Although not .shown, the distribution of these
Crack and Seat
~ values was also normally distributed for all three rehabilitation
! 30 types. These important findings gave way to defining guide-
j lines for project construction uniformity for all fractured slab
techniques, regardless of the actual average Epcc value achieved
0 20
c at the project site.
~ Recommended construction control categories for various
!. 10
levels of project uniformity are given below:
Category Range
0
Break and Seat Good to excellent CVw < 30 percent
Fair to good 30 percent ::;; CV w ::;; 50 percent
30 Poor to fair CVw > 50 percent
For the "good to excellent" and "poor to fair" categories,
20
the percentage of the projects evaluated was approximately
22 percent. The "fair to good" CVw values contained about
10 56 percent of all computed within project CVw values found.

0
0 500 1000 1500 2000 2500 3000 SUMMARY AND CONCLUSIONS
Average Section Modulus of Fractured PCC Layer (ksi)
This paper presented the results of a nationwide study on
FIGURE 3 Frequency distribution of in situ PCC modulus three new and innovative methodologies for rehabilitation of
values by rehabilitation values.
PCC pavement involving the fracturing of the slabs before
the placement of an AC overlay. The major objective was to
the magnitude of this variation within a given site reflects th~ develop guidelines to eliminate or minimize the occurrence
ability of the contractor to develop a uniform fractured slab of reflective cracks in the overlay. The specific techniques
"product" after cracking, breaking, or rubblization has taken evaluated were rubblization, crack and seat, and break and
place. seat.
To determine whether meaningful trends in the magnitude On the basis of the results of this study, the following major
of the within project variation were evident, studies examining observations were made:
both the standard deviation (crw) and coefficient of variation
(CVw) were undertaken. The results indicated that a wide • The relative ranking of the fracturing techniques, in order
range of <Tw values existed for all three rehabilitation types of decreasing performance life, appears to be rubblization
and that no single value of the crw was typical for a particular (best), crack/seat, and break/seat (worst). Reasonable PCI
rehabilitation activity. predictive models were developed for the first two techniques.

TABLE 4 Summary of Between Project Ercc Statistics

Type of No. of Between Project Results

Rehab Sections Remarks
Ei>cc ob CVb(%)
Rubblized 22 412.5 ksi 154.4 ksi 37.4% Recommended (excludes ·2 outliers)
24 501.8 ksi 338.9 ksi 67.5% All data
Crack/Seat 46 409.0 ksi 140.7 ksi 34.4% Recommended (excludes all values
greater than 1000 ksi; i.e.'
Crack spacings greater than 48")
64 780.6 ksi 665.6 ksi 85.3% All data
Break/Seat 52 1271.5 ksi 548.7 ksi 43.2% All data; Recommended
All Fractured 120 783.4 ksi 377.4 ksi 48.2% All Recommended results
Slabs 140 915.1 ksi 578.0 ksi 63.2% All data
·-
26 TRANSPORTATION RESEARCH RECORD 1374

1200 coupled with the tendency to use smaller crack spacings, should
C"
result in improved performance. However, it is strongly rec-
l ommended that the suggested minimum crack spacing guide-
i 1000
Avg. Within Section
CVw•40.7% 0 lines presented in the paper be followed.
~ N•40
<>
• The currently used construction techniques for break and
seat rehabilitation of JRC pavements result in a high degree
~
0
800 <> 0
0 of variability in effective moduli, indicating inadequate break-
j <>
0
ing or debonding of the reinforcing steel. Until improvements
c 0 0
<>
0 600 <> are made in the breaking operations, this rehabilitation option
! 0
<> <>
<>
<>
0 should be used with extreme caution and coupled with field
quality control measures based on deflection testing during
I3:
400 +o
~<>
+<*'8
0.
<>
<>
<>
<>o<> <>
<>
<>

0 Crack/Seat
the construction process unless local experience shows
otherwise.
0 ~ <>
200 <> <>
0 Break/Seat Finally, whereas much useful information was obtained from
+ Rubbllze this initial nationwide study, additional research is required
to further refine and improve the recommended guidelines
and methodologies. This can only be accomplished through
Average Section Fractured PCC Modulus, EPCC(ksl) a combined process of data collection and periodic analysis
of long-term pavement performance information. The de-
FIGURE 4 Within section variability of fractured PCC slab
tailed data base developed in this study can serve as the basic
modulus.
framework upon which additional projects can be added to
expand the total number of experimental sections and perfor-
• Reasonable predictive models for the fractured PCC
mance data.
modulus were obtained for the rubblized and crack/seat tech-
niques. These models clearly show the importance of crack
spacing and the foundation support of the existing PCC pave-
ACKNOWLEDGMENTS
ment. As both of these variables are increased, the EPcc value
of the fractured slab increases. The development of a similar
The work described in this paper was performed by PCS/Law
model for the break and seat technique was not possible be-
Engineering for the National Asphalt Pavement Association
cause of extreme variation in PCC values apparently due to
(NAPA) and the State Asphalt Pavement Association Exe-
inefficient fracturing or debonding of the distributed steel.
cutives (SAPAE). The authors gratefully acknowledge the
• Some of the most significant and important findings of
cooperation and assistance of the NAPA, SAPAE, and PCS/
the study involve the statistical frequency distributions of the
Law Engineering staffs.
effective Epcc values for each rehabilitation technique. Both
between project and within project variability were analyzed.
•For the crack/seat and rubblized pavement sections, the
REFERENCES
resulting frequency distributions of the project mean Epcc
value were quite similar: average EPCc = 400 to 500 ksi and 1. Highway Statistics 1987. Federal Highway Administration, U:S.
a between project coefficient of variation value of approxi- Department of Transportation, 1988.
mately 35 percent. 2. M. R. Thompson. NCHRP Synthesis of Highway Practice 144:
• In contrast, the break/seat distribution was uniformly dis- Breaking/Cracking and Seating Concrete Pavements. TRB, Na-
tional Research Council, Washington, D.C., 1989.
tributed across a wide range of EPcc values (i.e., 250 to 2,750 3. W. Kilareski and R. Bionda. Performance/Rehabilitation of Rigid
ksi). This clearly reinforces the conclusion that the break/seat Pavements. Draft final report, FHWA Contract DTFHG l-86-
process on JRC pavements is not uniformly efficient in fully C-0079. The Pennsylvania Transportation Institute, 1988.
debonding or fracturing the distributed steel. 4. W. Kilareski and S. Stoffels. Improved Design and Construction
•On the basis of the analysis results of the within project Procedures for Crack and Seat and Overlay of Rigid Pavements.
Draft final report, Phase III, Task B, FHWA Contract DTFHGl-
variability, guidelines for project uniformity were developed. 86-C-00079. The Pennsylvania Transportation Institute, 1988.
5. Guidelines and Methodologies for the Rehabilitation of Rigid High-
From these and other observations, the following major way Pavements Using Asphalt Concrete Overlays. Draft final re-
conclusions and recommendations were developed: port. PCS/Law Engineering, 1991.
6. M. Shahin and S. Kohn. Pavement Maintenance Management for
Roads and Parking Lots. Final Report CERL-TR-M-294. U.S.
• Rubblization of deteriorating PCC pavements followed Army Construction Engineering Research Laboratory, Cham-
by an AC overlay is an excellent rehabilitation method 'that paign, Ill., 1981.
is equally effective for all types of existing PCC pavements. 7. AASHTO Guide for Design of Pavement Structures. American
This technique is the preferred rehabilitation method for all Association of State Highway and Transportation Officials; Wash-
ington, D.C., 1986.
types of PCC pavements.
•The crack and seat technique followed by an AC overlay All opinions, conclusions, and recommendations reported in this paper
is a very effective rehabilitation method for deteriorating JPC are strictly those of the authors and do not necessarily represent the
pavements. Improvements in equipment over the past years, views of NAPA or SAPAE.
TRANSPORTATION RESEARCH RECORD 1374 27

Asphalt Concrete Overlay Design

Methodology for Fractured Portland
Cement Concrete Pavements
MATTHEW w. WITCZAK AND GONZALO R. RADA

Little technical information or guidance is presently available for ment perform in a flexible manner. Because of this, use of
engineers to properly design asphalt concrete (AC) overlays over overlay methodologies based on flexible pavement perfor-
existing PCC pavements that have been fractured to minimize or mance models may be erroneous.
eliminate the problem of reflective cracking. Such construction
techniques as crack and seat, break and seat, and rubblization Use of flexible models for the overlay process will lead one
have been used by the industry at an increasing rate over the last to conclude, in the vast majority of design situations, that no
10 years. However, most design procedures have been highly structural overlay is required. Whereas this is technically true
subjective, extremely conservative, and based on a lack of en- from the viewpoint of protecting the subgrade against exces-
gineering principles related to the actual construction process sive repetitive shear displacements, the use of thin AC over-
used as well as an accurate assessment of the in situ physical lays will invariably lead to the major problem of reflective
properties of the fractured slab. To improve the state of the art
and develop a better understanding of these rehabilitation tech- cracking in a relatively short time. The design of an AC over-
niques, a nationwide evaluation study of these rehabilitation types lay over an existing PCC pavement must always take into
was conducted. The study led to the field evaluation of perfor- account the potential to minimize or eliminate the reflective
mance and structural in situ properties of more than 100 actual cracking problem.
construction projects where these techniques had been used with Various AC overlay techniques used on existing PCC pave-
AC overlays. On the basis of the study results, design procedures ments with the primary view of the reflective crack problem
were developed for highway pavements and are presented. The
include thick (conventional) overlays, crack relief layers, the
design procedures are based on the flexible pavement perfor-
mance methodology presented in the 1986 AASHTO Guide. saw and seal technique, special overlay and interface mate-
rials, and the fractured slab approach-rubblize, crack/seat,
and break/seat. This paper focuses on the design of AC over-
In many respects, the rehabilitation of pavement systems is lays for fractured PCC pavements.
a more complex engineering task than the design of new ones. At present, little technical information or guidance is avail-
Many factors within the rehabilitation process are beyond the able for engineers to properly design AC overlays over ex-
current state of the art. As a result, the engineer must use a isting PCC pavements that have been fractured to minimize
great deal of judgment. in the overall design process. The or eliminate the problem of reflective cracking. Such con-.
engineer should also approach the rehabilitation process with struction techniques as crack and seat on plain PCC pave-
the viewpoint that several technically feasible solutions may ments, break and seat on reinforced PCC pavements, and
be present for any given project. rubblization have been used at an increasing rate in the last
From a general viewpoint, there are several major cate- 10 or more years. However, most design procedures have
gories of possible rehabilitation activity available to the en- been highly subjective, extremely conservative, and based on
gineer dealing with existing portland cement concrete (PCC) a lack of engineering principles related to the actual construc-
pavements. Th~y include do nothing, concrete pavement res- tion process used as well as an accurate assessment of the in
toration, PCC overlays, asphaltic concrete (AC) overlays, and ·situ physical properties of the fractured slab.
reconstruction. Because the overall objective of this paper To improve the state of the art and develop a better under-
concerns the rehabilitation of rigid pavements using AC standing, a nationwide study of these rehabilitation tech-
overlays, details regarding the other techniques are not niques was undertaken. The study led to the field evaluation
addressed. of performance (distress) and in situ structural properties of
AC overlays have been used for many years to rehabilitate more than 100 actual construction projects where the fracture
existing PCC pavements, but their successful design affords techniques had been used with AC overlays. On the basis of
a more difficult challenge to the engineer compared with the the study results, design procedures were developed for high-
rehabilitation of existing flexible pavement systems. Placing way pavements and are presented in this paper.
an AC overlay on an existing PCC pavement, even ihignif- The design procedures are based on the use of the flexible
icantly cracked, does not necessarily make the overlaid pave- pavement performance methodology presented in the 1986
AASHTO Guide (1). An analysis of the effective distribution
M. W. Witczak, Department of Civil Engineering, University of of the in situ fractured PCC slab modulus results has led to
Maryland, College Park, Md. 20740. G. R. Rada, PCS/Law Engi- the selection of realistic AASHTO layer coefficients (a; val-
neering, 12240 Indian Creek Court, Suite 120, Beltsville, Md. 20705. ues) in determining overlay requirements. The selection of
28 TRANSPORTATION RESEARCH RECORD 1374

the appropriate ai value is based on reliability values estab- of the postfractured PCC layer are selected with a fair degree
lished from an analysis of the expected variability within and of certainty. Thus, without the precise knowledge of the ef-
between projects. In addition, an alternate design approach fectiveness of the fracturing operation, AC overlay designs
using nondestructive deflection testing (NDT) equipment for can be developed on the basis of the results of this study. In
quality control/assurance (QA/QC) during the construction contrast, field approach designs imply the absolute need to
(fracturing) phase is presented. Finally, example problems measure the as-constructed effectiveness of the construction
are used to illustrate both design approaches. operation to determine the required thickness of the AC over-
lay. Whereas either method can be used for both the rubbli-
zation and crack/seat options, the use of only the field ap-
INITIAL DESIGN CONSIDERATIONS proach is currently recommended for the break/seat option
because of the highly variable field results of the n·ational
Whereas many factors influence the final rehabilitation tech- ·study.
nique selected for any given project, two of the most fun-
damental and important considerations are the pavement type
and its existing condition. The fractured slab techniques are BASIC DESIGN PHILOSOPHY
generally recommended for pavements in fair to failed con-
dition-present serviceability index (PSI) .::; 2.5, pavement The term "fractured slab techniques" relates to those reha-
condition index (PCI) .::; 50, AASHTO structural condition bilitation options directly associated with the reduction of the
factor (Cx) .::; 0.78, or pavement remaining life (RLx) .::; 20 original PCC slab lengths to smaller effective lengths, to min-
percent. However, the engineer should not completely rule imize or eliminate the reflective crack problem. A funda-
out the potential economy of these options for pavements in mental relationship governing these techniques is that as crack
fair. to moderate condition. Whereas these options may be spacing decreases, the likelihood of reflective crac!,cing de-
economically unfeasible for this condition category, detailed creases. In companion papers (2; paper by Witczak and Rada
studies should be conducted within a life cycle approach to in this Record) it has been shown that regardless of the type
ensure that this is the case. of rehabilitation considered, general relationships exist be-
The process of rubblization is the only form of slab fracture tween the effective in situ fractured PCC modulus (Epcd and
recommended for all PCC pavement types: jointed plain (JPC), nominal fragment size. This fact has been shown by numerous
jointed reinforced (JRC), and continuously reinforced (CRC) other researchers as well. Because of this, the selection and
concrete pavements. The crack/seat option is only applicable use of the EPcc must likewise be directly related to the prob-
to JPC pavements, whereas the break/seat technique is only ability of reflective cracking.
recommended for JRC pavements. For each technique, two This consideration is shown in Figure 1. As the Epcc of the
alternative design approaches have been developed from the pavement is decreased, the probability of obtaining reflective
nationwide study: office and field design methods. cracking at any thickness of AC overlay, h 0 1' is also decreased.
In the first approach, the design methodology should be In addition, as the thickness of overlay is increased, at any
viewed as an office type of design in which fairly typical values unique value of the EPcc, the probability of reflective cracking

100
AC 0Ver1ay Thickness

Low EPCC (PCC Modulus) --------High

Small - - - - - - o, (Nominal Fragment Size) -------4- Large

FIGURE 1 Influence of fractured PCC modulus and AC overlay thickness

on structural and reflective crack failure.
Witczak and Rada 29

must decrease. This implies that the best solution for the
reflective crack problem is to ensure that the smallest possible Between Project Variation

effective slab length, nominal fragment size, and effective

PCC modulus exist.
There is, however, another major consideration in the de-
sign and construction process. It should be clearly recognized
that as PCC pavements are fractured, they become and act _..-- Project 2

more like flexible pavement systems rather than PCC pave-

ments designed for rigid slab action. The implication of this
fact is also shown in Figure 1. An opposite relationship exists I

to the reflective crack problem in that as the Epcc is increased, ~

the structural capacity of the existing pavement is increased, Project 1
at any thickness of overlay h01 • As a result, the probability of
structural failure is decreased. Likewise, at any given EPcc EP2
value, as the AC overlay thickness is increased, the structural E PCC (Average Project Modulus)
(a)
failure probability must likewise decrease.
If both of these considerations are viewed together, an
important fact concerning fractured slab techniques is re-
vealed. Figure 1 shows that at a given thickness of overlay, Within Project Variation
the intersecting point of the two relationships (reflective cracking
and structural distress) yields a critical EPcc value; which min-
CVw •40.7%
CD Good - Exe Project Control
imizes both possible distress modes. This critical effective ® Fair - Good Project Control
modulus (Ee,) represents a threshold minimum modulus of
the fractured slab such that the probability of both potential @ Poor - Fair Project Control

distress modes is the minimum possible for any given project.

In the development of the design methodologies, a provisional
critical modulus value of Ecr = 1,000 ksi was established
independent of the AC overlay thickness. Furthermore, to
incorporate the influence of the normal project variation, it
is recommended that no more than 5 percent of the project's
EPcc values be greater than the Ecr = 1,000 ksi value.
30 % 50 %
At this point, it is important to recall the major findings CVw (Within Project Coeff. of Variation)
and conclusions that were presented in the companion paper (b)

(2) regarding the importance of between and within project

variation of the backcalculated EPcc values found from the FIGURE 2 Between and within project Epcc variability: (a)
frequency distribution of average project means; (b) frequency
field NDT study. Figure 2 shows these results for the between
distribution of within project CV value.
project variability and the frequency distribution of the within
project coefficient of variance. Because the within project
variation is highly indicative of the project uniformity or abil-
ity of the contractor to provide a uniform product, the zones the Ecr level. It can therefore be concluded that the ability of
shown in Figure 2b are indicative of the three types of degree a given project to satisfy the Ecr criteria is not only a function
of uniformity that were found from the national study. of the project average Epcc value but also highly dependent
The combination of both variability forms must be jointly on the within project variation attained in the construction
viewed to gain full appreciation of the design methodology process by the contractor. From a structural viewpoint, a
that will be presented. In Figure 2a, the average project EPcc greater thickness of AC overlay would be required for Project
means for two typical projects (EP 1 and EP2 ) are shown. For 1 than for Project 2 because Project 1 has a lower modulus
each project mean, a range of within project CVw values may (Epcd·
affect the actual distribution of the EPcc values for any given Whereas the previous discussion has primarily focused on
project. For purposes of the following explanation, it is as- the Epcc distributions and their within project variability rel-
sumed that three levels of the within project variability (CVw) ative to the critical Ecr for minimizing or eliminating reflective
exist: cvge> good to excellent control; cvfg> fair to good cracking, implications for the Epcc distribution must also be
control; and CVPr' poor to fair control. The actual project considered relative to the structural overlay design. As dis-
Epcc frequency distributions for the six possible combinations cussed later in this paper, the overlay methodology is based
are shown in Figure 3. on the use of the AASHTO structural number (SN) concept
For Project 1, the three frequency distributions reflecting for flexible pavements (1). An important parameter in SN
the range of project uniformity are shown in relation to the computations is the structural layer coefficient (a;).
critical Ecr level for reflective cracking. Because the average Analytically, the a; value can be related to the elastic mod-
EP 1 is small, the probability of any combination ·of within ulus of a material (E;) through the following relationship:
project variation exceeding the critical threshold Ecr value is
very remote. However, if the resulting frequency distributions
for the second project (Ep2) are observed, as the project non-
(1)
uniformity is increased, a significant area for Curve 3 exceeds
30 TRANSPORTATION RESEARCH RECORD 1374

Witbin Er.oje.cl control where the subscript i represents the material in question and
EP1 IProject 1 j the subscripts represents an arbitrary standard material whose
G) Good - Excellent
8c as and Es were established for AASHO Road Test materials.
(g ® Fair- Good
Using a dense-graded crushed stone base as the standard,
~
0
® Poor - Fair it has been found that as = 0.14 and Es = 30,000 psi. Sub-
stituting these values into Equation 1 yields
ECrtt1ca1
Ci'
c
Q)
::I a; = 0.0045-\1£; (2)
~
.
LL
Thus, a direct transformation between the in situ fractured
~ modulus (E; or EPcc) and the AASHTO layer coefficients
(a;) for the fractured material can be easily made.
Figure 4 shows a typical frequency distribution of the a;
Ep1 Ea
~(Project Modulus) value for a given project, which was developed using the E;
to a; transformation. Also shown on Figure 4 are two separate
E; (a;) values. The first (Ecr or acr) has been fully discussed
EP2 as the critical EPcc for reflective crack control. The second
IProject 21 value (ad) represents a design value selected by the engineer
for the structural overlay process; the area under the curve
8c
that is less than this value (i.e., ad) represents the probability
(g
of structural failure for the project. Clearly, as the engineer
~
0
Ecrtt1ca1
desires a higher design reliability, a smaller value of ad must
be selected. This, in turn, will result in thicker AC overlays
~· Probability of
cQ)
Exceeding Ea
being required as the reliability level is increased. It can there-
::I
Value fore be concluded that within project variability is a very
~
LL significant parameter influendng both the probability of re-
~- flective cracking and the probability of structural failure.

EP2
EPCC (Project Modulus) GENERAL OVERLAY DESIGN PRINCIPLES

FIGURE 3 Fractured slab modulus frequency distributions All three rehabilitation options within the fractured slab cat-
reflecting between and within project variability. egory behave more like flexible systems than rigid ones. The

PR {Success}

a 1 (AASHTO Layer Coefficient)

FIGURE 4 Frequency distribution of the AASHTO layer coefficient for

fractured PCC slabs.
Witczak and Rada 31

classical flexible pavement performance models are therefore construction data or obtained from drilling/coring operations,
more applicable and accurate as the basis of any overlay meth- the most significant factor to be determined is the ad value
odology. Another important consideration is the fact that the for the fractured slab.
fractured slab process turns the existing rigid pavements into
new flexible pavements. The new pavement, in turn, can be
viewed as being the AC overlay equivalent of the placement
of an AC surface course on new construction. Because of this, STRUCTURAL DESIGN LAYER COEFFICIENT, ad
it is believed that the use of the remaining life factor (FRL)
present in the AASHTO guide (1) is not applicable to the The selection of the appropriate ad value is a critical and
fractured slab process (i.e., FRL = 1). sensitive part of the overlay analysis. This parameter relates
The AC overlay methodology on which the rehabilitation to the structural failure of the overlaid pavement, and thus it
of fractured slabs is presented is based on the well-known and is necessary to apply design conservatism to the design pro-
widely used structural capacity deficiency approach. The cess. However, the within project variability (CVw) also plays
AASHTO guide (1) flexible performance models using the a key role in the selection of ad in that the optimum construc-
SN approach are used as the equivalent parameter of the tion process should yield an average EPcc as large as possible,
structural capacity. Thus, the general overlay equation is based with as low a CVw value as possible, to ensure that the Ecr
on the simple expression level is met.
As with any design analysis, the engineer must select a
(3) typical design value in the absence of site-specific data. This
"office design?' must obviously be based on a relatively con-
where servative approach which, in turn, is based on a high degree
of reliability. This classical engineering design approach is
SNY = future structural capacity required of a new flex- referred to as the office design method in this paper.
ible pavement constructed over the existing An alternative approach is to use a methodology based on
subgrade to accommodate the traffic within the the site-specific construction information obtained through
life of the overlay, deflection testing. This information, when analyzed, can serve
SNxeff = effective capacity of the existing pavement struc- as a QNQC measure and provide actual in situ response for
ture after fracturing has taken place, and the fractured slab process to develop dynamic design values.
SN01 = additional structural capacity that will be required Whereas there may be practical restraints on the implemen-
from the AC overlay. tation of this design approach, the potential for saving con-
Recognizing that siderable money in the rehabilitation process should not be
overlooked by the design engineer. This approach is referred
(4) to as the field design method in this paper. The following
sections define each of these recommended design methods
and using the commonly accepted a01 = 0.44 for AC, the in further detail.
required overlay thickness can be expressed by

h = SNY - SNxeff (5) Office Design Method

0
0.44
The office design method represents the development of a
The solution of the h value involves the solution of the
0
typical ad value for design of pavements without site-specific
two variables: SNY and SNxerr· The solution of SNY is very
information. In determining this value, the results of all EPcc
direct because it is based solely on the AASHTO guide so-
values obtained in this national study, for a particular type of
lution for new flexible pavements (1). The computation of
fractured slab analysis, were used. The procedure described
the SNxerr value should incorporate not only the fractured slab
is based on not only the between project variability but the
but also any subbase layers present in the existing pavement.
Thus within project variability as well. The overall design reliability
(R) associated with the particular ad found from the ensuing
analysis is related to the joint probabilities associated with
each frequency distribution.
where Using the principle of normal probability, the between proj-
ect distribution of the EPcc value can be characterized by
ad = design layer coefficient of the fractured PCC layer,
asb = layer coefficient of any existing subbase layer ma- (7)
terial,
D0 = original thickness of the PCC slab, and where
hsb = subbase layer thickness.
EP = average EPcc for a given project,
The reader is referred to the AASHTO guide (1) for further Epcc = average of all project means,
details regarding the selection of the appropriate asb values ko.b = standardized normal deviate associated with be-
for a variety of materials that may be present. Because layer tween project probability of failure (cxb), and
thicknesses (D 0 and hsb) can usually be found from historic <Tb = between project standard deviation.
32 TRANSPORTATION RESEARCH RECORD 1374

Likev.1ise, the following relationship exists for the within For each specific fractured slab process, the Ed value can be
project distribution: computed from three EPcc distribution parameters:·EPcc. ab,
and CVw at any desired reliability level, R. In turn, the design
(8) layer coefficient (ad) can be computed from Ed by means of
Equation 2. Therefore, it is possible to develop relationships
where of ad as a function of the overall reliability for the results of
this study.
Ed = design Epcc for a given project, Table 1 summarizes the key between and within project
kaw = standardized normal deviate associated with the statistics found for the fracture techniques; the reader is re-
within project probability of failure (aw), and
ferred to the companion paper for a more detailed discussion
aw = within project standard deviation. of these parameters. Using these statistics as input into the
For the within project variability, the coefficient of varia- equations presented earlier, Ed and ad values were developed
tion (CVw) was a constant value regardless of the EP value as a function of the design reliability for rubblize and crack/
(2). Therefore seat projects. Table 2 summarizes these computations. A com-
parison of these results indicates that the ad values are prac-
tically identical for the rubblize and crack/seat techniques.
(9a) Accordingly, the final recommended ad relationship is shown
in Figure 5. ·
or For typical values of design reliability encountered in prac-
tice, a value of ad = 0.28 is recommended. This is equivalent
(9b) to a reliability value slightly in excess of 90 percent. However,
the engineer must use judgment in selecting the appropriate
Substituting Equation 9b into Equation 8 yields design reliability level for any given project; as the relative
importance of the project increases, a higher R value and
(10) hence lower ad value may be selected.
Whereas Table 1 also summarizes the key project statistics
The value of Ed represents the design value of the fractured for the break/seat projects, the office design method is not
slab technique existing at an overall design reliability, R, de- recommended for this rehabilitation technique-the analysis
fined by of both performance data and in situ structural properties
obtained from the field study indicates that a wide range of
(11) breaking efficiency actually occurs. This finding strongly sup-

TABLE 1 Summary of EPCc Statistics

BETWEEN PROJECT VARIABILITY
Type of No. of Between Project Results
Rehab Sections Remarks
EPCC ab CVb(\)
Rubblized 22 412.5 ksi 154.4 ksi 37.4% Recommended (excludes 2 outliers)
24 501. 8 ksi 338.9 ksi 67.5% · All data
Crack/Seat 46 409.0 ksi 140.7 ksi 34.4% Recommended (excludes all values
greater than 1000 ksi i.e., Crack
spacings 48" or greater)
64 780.6 ksi 665.6 ksi 85.3% All data
Break/Seat 52 1271. 5 ksi 548.7 ksi '43.2% All data; Recommended
Combined 120 783.4 ksi 377 .4 ksi 48.2% All Recommended results
140 915.l ksi 578.0 ksi 63.2% All data

WITHIN PROJECT VARIABILITY

Type of No. of Within Project Results
Rehab Sections Remarks
cv. aC\'v CVevv
Rubblized 24 44.4% 12.9% 29.1% All data
Crack/Seat 64 41.2% 12.8% 31.0% All data

Break/Seat 52 38.4% 12.6% 32.7% All data

Combined 140 40.7% 12.7% 31. 3% All data
Witczak and Rada 33

TABLE 2 Computational Summary for Reliability-Based ad Values

RUBBLIZED PROJECTS
Between Project Within Project Design Values
a:b K.b EP - Epec -K.bab a:. K.. cv. ' % Ed ad I.Ed R, %
--
0.50 0.000 412.5 0.50 0.000 44.4 412.5 .335 2.39 75.00
0.25 0.671 308.9 0.25 0.671 44.4 216.8 .270 1.93 93.75
0.15 1.036 252.5 0.15 1.036 44.4 136.3 .232 1.66 97.75
0.10 1.282 214.6 0.10 1.282 44.4 92.4 .203 1.45 99.00
0.05 1.645 158.5 0.05 1.645 44.4 42.7 .157 1.12 99.75

Epcc - 412.5 ksi; ob - 154.4 ksi; CVv - 44.4%

CRACK/SEAT PROJECTS
Between Project Within Project Design Values
cv.
a:b

0.50
K.b

0.000
~ - Epec -IC.bob

409.0
a:.,

0.50
K..
0.000
'
41.2
' Ed

409.0
ad

.334
I.Ed
2.38
R, %
75.00
0.25 0.671 314.6 0.25 0.671 41. 2 227.6 .275 1.96 93.75
0.15 1.036 263.2 0.15 1.036 41.2 150.9 .239 1. 71 97.75
0.10 1.282 228.6 0.10 1.282 41. 2 107.9 .214 1.53 99.00
0.05 1.645 177.5 0.05 1.645 41.2 57.2 .173 1.24 99.75

Epcc - 409.0 ksi; ob - 140.7 ksi; CVv - 41.2%

ports the fact that this approach may yield highly variable and experience with break/seat techniques, this process should be
uncertain performance. Because of this, extreme care must viewed as a viable rehabilitation approach.
be exercised during construction to ensure that a minimum
effective PCC modulus of the fractured slab occurs. Without Field Design Method
such field verification, the technique of rubblization is cur-
rently recommended rather than break/seat pending future The field design approach is based on the use of deflection
studies. Alternatively, where states have had successful prior basin data collected during the construction operation to en-

0.4
-
-
E
G>
-
15 - ~~
i
8
a;
0.3
- ~ -- -~

~
80..
~
0.2
=/
L
~
u..
-
-
cC)
u;
-
.
0
.
G>

'1:1
0.1
....
-
"' ....
....
0
100 95 90 85 80 75 70

R - Desired Design Reliability Level(%)

FIGURE 5 Recommended AASHTO structural layer coefficient for rubblized

and crack/seat PCC layers as a function of desired reliability.
34 TRANSPORTATION RESEARCH RECORD 1374

sure that both design criteria (ad and Ee,) are met in the relative to the original AC design, a "do nothing" option may
fracturing process before placement of the overlay. Though be selected. However, it is possible to compute the possible
this approach may have some initial practical implementation reduction in AC overlay thickness (~h 0 ) that may be imple-
problems, its use on all future fractured slab projects is highly mented directly in the field. This is accomplished by
recommended because of the significant advantages that can
be gained in the design and construction process. The major
advantages include the potential for increased project uni- (14)
formity, better future pavement performance and life, sig-
nificant cost savings in the initial overlay construction, and a If ~h 0 is greater than 1.0 in. or more, every consideration
better procedure to more accurately assess whether the slab should be given to adjusting the initial design recommenda-
fracturing process is as efficient as desired (i.e., compared tion of h 0 (overlay thickness) by the ~h 0 value. Conversely,
with either visual crack studies or limited coring). This meth- if adf < ado• the ~h 0 equation can be used to determine how
odology is the same for all types of fractured slab rehabili- much more overlay would be necessary for the actual frac-
tation options and is applicable to all types of existing rigid tured conditions achieved in the field.
pavements (JPC, JRC, and CRC).
The general approach to the design method is as follows.
Immediately after the contractor has completed his initial
round of "slab fracture" on a defined section, deflection read- EXAMPLE PROBLEMS
ings should be taken on at least 30 random points within the
section limits. The deflection basin data should then be used Example 1
to calculate the in situ EPcc value for each test point, the
project average Epcc value, and the within project standard An example of the rubblized overlay rehabilitation option is
deviation, CJP. presented to summarize the design methodology recom-
Next, check to see whether the Ecr has been met. This is mended. For this example project, an existing JPC, with ex-
done by finding isting joint spacing of 20 ft, has PSI = 2.1. More than 25
percent of the slabs exhibited extensive cracking indicating
K = (1,000 - ~) fair to poor pavement condition. The existing PCC pavement
(12)
a (JP is 9.0 in. thick and has a subbase (unbound) of 6.0 in. The
AASHTO layer coefficient for the subbase has been found
Using this value as input into the normal probability table to be asb = 0.09. The use of tJ;ie AASHTO new flexible
contained in most statistical and probability textbooks, the a pavement performance model for the overlay life and traffic
value or probability of exceeding the Ee, = 1,000 ksi criterion has indicated that a SNY = 4.82 will be required. An office
,can be determined. If the computed a value is greater than design solution is desired for a rubblized AC overlay.
5.0 percent, the Ecr criterion has not been met, and the con- From the problem description, the following values are
tractor should be instructed to refracture the area. If this is known: SNY = 4.82, a01 = 0.44, D 0 = 9.0, ad =/(reliability
done, the sequence goes back to the beginning. level, R), and hsb = 6.0. Substituting these inputs into Equa-
On satisfying the Ee, criterion, the next step is to check the tion 5 yields
design ad value. Using the normal probability table, a value
of K 13 can be selected for any given design level of reliability h0 = 9.73 - 20.45ad
(e.g., K 13 = 1.037 for R = 85 percent). The field-derived adf
value can be then computed from Because ad is a function of the design reliability level, the
solution of h0 is presented in Table 3 for several levels of R
as well as the recommended values of ad = 0.28. It can be
observed that the design h0 is affected by the selection of the
The final step deals with the comparison of the "field" adf desired R value. For typical reliability levels between 85 and
value with the "office" ado value used to establish the prelim- 95 percent, the overlay thickness requirements vary between
inary AC overlay design thickness. If adf > ad0 , the engineer 3.5 and 4.5 in. The typical recommended value of ad = 0.28
has two options. First, because this condition is conservative results in a design h 0 = 4.0 in.

TABLE 3 Required Overlay Thickness as a Function of Reliability Level-

Example Problem 1

Reliability Layer Overlay Thickness,

Level Coefficient, a h (inches)

75% 0.34 2.8

85% 0.30 3.6
90% 0.29 3.8
95% 0.26 4.4
99% 0.20 5.6
(Recommended) 0.28 4.0
Witczak and Rada 35

Example 2 The basic design philosophy is that as fractured slab frag-

ments become smaller, the EPcc value becomes less. This has
The following example is based on the field design method two important ramifications. To minimize or eliminate re-
applied to a break/seat project on an existing JRC pavement. flective cracking, it is desirable to have the effective Epcc
The pavement to be rehabilitated has a 40.0-ft joint spacing value as small as possible. However, in so doing, the strength
and is 10.0 in. thick. It rests on a 4-in. cement-treated base of this fractured layer decreases, which in turn requires a
having an AASHTO layer coefficient of asb = 0.15. The PSI thicker overlay. As a consequence, the overall philosophy of
of the pavement is 2.3, and it is in fair to poor condition. The the fracture techniques should be to obtain as large an in situ
required future structural capacity needed in the overlay pe- Epcc value possible to minimize the required overlay thickness
riod has been found to be SNY = 5.95. Because the facility but ensure that there is a small probability of having within
receives heavy traffic, the engineer has selected a design re- project EPcc values exceed a certain upper or critical value
liability of 95 percent for the project. For the preliminary (Ecr)•
design, an h value of 5.0 in. was selected by the design team
0
In development of the overlay methodologies, reliability
on the basis of experience. levels of 95 percent have been used as the basis for the rec-
After the contractor conducted a preliminary breaking of ommendations. In addition, the critical level of EPcc to ensure
a given section of the project, NDT testing was used to de- that reflective cracking will not occur has been provisionally
termine ·the statistics associated with the Epcc values. They selected to be Ee = 1,000 ksi.
were Epcc = 1196 ksi, <Tw = 385 ksi, and CVw = 32.2 percent. Two design approaches were presented in the paper: office
These results indicate that the "broken" section does not and field design methods. The office design approach was
satisfy the Ecr criterion of having less than 5 percent of the based on the selection of a conservative estimate of the
EPcc values exceed the threshold limit of 1,000 ksi because AASHTO structural layer coefficient or a; value to be used
the average Epcc is much greater than the threshold. for each rehabilitation technique. Information obtained from
The contractor was then instructed to conduct further the between and within EPcc variability studies was used to
breaking. The NDT backcalculated EPcc statistics were EPcc determine appropriate levels of a; as a function of the desired
= 526 ksi, <Tw = 129 ksi, and CVw = 24.5 percent. For the design reliability for the rehabilitation.
criterion of ex = 5 percent for the Ecr limit, the value of Kaw The second approach, the field method, is predicated on
= 1.645 is found from the normal distribution table. Thus, the use of nondestructive deflection testing at the construction
the upper limit of the actual EPcc distribution at a 5 percent site to monitor and control the final design thickness. At a
level is given by given project site, the deflection test results are used to de-
termine the in situ frequency distribution of the backcalcu-
Eu = EPCC + Ka<Tw = 526 + 1.645(129) = 738 ksi lated EPcc values. This distribution is checked to ensure that
no more than 5 percent of the EPcc results exceed the critical
Therefore, the pavement meets the Ecr reflective crack cri- 1,000 ksi upper limit value. Once this criterion is satisfied,
terion and the actual field adf value can be now determined the actual project EPcc distribution is then used to determine
from the Ed value: the final design project a; value so that the final AC overlay
thickness can be determined.
313.8 ksi Whereas either design approach can be used for both the
rubblization and crack/seat options, the use of only the field
and approach is recommended for the break/seat option.

adt = 0.0045E~ 333 = 0.0045(313,800) 0 ·333 = 0.31

ACKNOWLEDGMENTS
Once the adf value has been established, the required over-
lay thickness check can be performed: The work described in this paper was performed by PCS/Law
Engineering for the National Asphalt Pavement Association
h = SNY - (adfDo + asbhsb) (NAPA) and the State Asphalt Pavement Association Ex-
0 ecutives (SAPAE). The authors gratefully acknowledge the
cooperation and assistance of the NAPA, SAPAE, and PCS/
Thus, the actual broken JRC pavement would require an Law Engineering staffs.
overlay of h0 = 5.1 in. Because the preliminary design was
based on h 0 = 5.0 in., no modification (either + or - Llh0
REFERENCES
adjustment) is required for the final design cross section.
1. AASHTO Guide for Design of Pavement Structures. American
Association of State Highway and Transportation Officials, Wash-
SUl\:fMARY AND CONCLUSIONS ington, D.C., 1986.
2. Guidelines and Methodologies for the Rehabilitation of Rigid High-
In this paper, AC overlay design procedures for fractured way Pavements Using Asphalt Concrete Overlays. Final report.
PCS/Law Engineering, 1991.
PCC pavements were presented. These procedures were de-
veloped from the results of a nationwide evaluation study and All opinions, conclusions, and recommendations reported in this paper
are based on the use of the AASHTO flexible pavement are strictly those of the authors and do not necessarily represent the
performance methodology. views of NAPA or SAPAE.
36 TRANSPORTATION RESEARCH RECORD 1374

Revision of AASHTO Pavement Overlay

Design Procedures
KATHLEEN T. HALL, MICHAEL I. DARTER, AND ROBERT P. ELLIOTT

The AASHTO overlay design procedures have been extensively tural capacity of the existing pavement is subtracted from the
revised to make them easier to understand and use, more adapt- required future structural capacity as determined from the
able to calibration by local agencies, and more comprehensive. AASHTO flexible and rigid pavement design equations. This
The revised overlay design procedures described use the concepts
of structural deficiency and required future structural capacity concept was retained to maintain compatibility between Parts
determined from the AASHTO flexible and rigid pavement de- II and III of the guide and to keep the procedure relatively
sign equations. The procedures provide detailed guidelines on simple. Development of a more sophisticated mechanistic ap-
several important topics related to overlay design, including over- proach to overlay design was not within the scope of the
lay feasibility, structural versus functional overlay needs, preover- revisions. NDT is recommended for characterization of the
lay repair, reflection crack control, and overlay design reliability . existing pavement, to the extent appropriate within the frame-
level. Detailed guidelines were also developed for pavement eval- work of these empirical design procedures.
uation for overlay design, including distress surveying, nonde-
structive testing, and destructive testing (coring and materials Three approaches were developed for characterizing the
testing). Seven separate overlay design procedures were devel- effective structural capacity of existing pavement (SNem Deff).
oped, encompassing all of the combinations of overlay and pave- Not all three approaches are appropriate for all pavement
ment types. Each of the design procedures follows a sequence of types. The approaches are
eight steps, by which the required future structural capacity for
the design traffic, effective structural capacity of the existing pave- 1. Visual condition survey and materials testing,
ment, and required overlay thickness are determined.
2. NDT (where appropriate), and
3. Remaining life (where appropriate).
Chapter 5 of Part III of the 1986 AASHTO Guide for Design
of Pavement Structures (1) addresses overlay design. These This paper presents an overview of the revised AASHTO
overlay design procedures were recently revised to make them overlay design procedures. The procedures are presented in
easier to use, more adaptable to calibration by local agencies, detail by Darter et al. (2). The development of the procedures
and more comprehensive (2-4). The proposed procedures is documented by Darter et al. (3). In addition, the revised
are currently under consideration by AASHTO. procedures were extensively tested using data from many ac-
A complete, step-by-step overlay design procedure was tual in-service pavements located throughout the United States.
developed for the following combinations of pavement and The results of this field testing are provided by Darter
overlay type: et al. (4).
Overlay Existing Pavement
AC AC
AC Crack/seat, break/seat, and rubblized PCC
AC JPCP, JRCP, and CRCP IMPORTANT CONSIDERATIONS IN OVERLAY
AC AC/JPCP, AC/JRCP, and AC/CRCP DESIGN
Bonded PCC JPCP, JRCP, and CRCP
U nbonded PCC JPCP, JRCP, and CRCP Overlay design requires consideration of many important items
PCC AC
in addition to required overlay thickness. Each of these is
Guidelines were also provided for overlay feasibility, pre- very briefly discussed in this section and is addressed in more
overlay repair, reflection crack control, overlay design reli- detail elsewhere (2).
ability level, and several other important considerations in Overlay feasibility: The feasibility of any type of overlay
overlay design. Detailed guidelines were developed for pave- depends on availability of adequate funds, construction fea-
ment evaluation for overlay design, including distress survey- sibility (including lane closure restrictions, materials and
ing, nondestructive deflection testing (NDT), and destructive equipment availability, overhead clearances, and other fac-
testing. tors), and the required future performance life of the overlay.
The revised AASHTO overlay design procedures use the Preoverlay repair: Much of the deterioration that occurs in
structural deficiency approach, in which the effective struc- overlays results from deterioration that was not repaired in
the existing pavements. The amount of preoverlay repair needed
is related to the type of overlay selected. If the existing pave-
K. T. Hall, 1206 Newmark Lab, 205 North Mathews Ave., Urbana,
ment is severely deteriorated, selecting an overlay type that
Ill. 61801. M. I. Darter, 1212 Newmark Lab, 205 North Mathews
Ave., Urbana, Ill. 61801. R. P. Elliott, 4150 Bell Engineering Center, is less sensitive to existing pavement condition may be more
Fayetteville, Ark. 72701. cost-effective than doing extensive preoverlay repair.
Hall et al. 37

Reflection crack control: The revised AASHTO overlay liability using the procedures described in Part I of the guide
design procedures do not consider reflection cracking in the for new pavements. This is done by determining the structural
overlay thickness design. Additional steps must be taken to capacity (SNf or Dt) required to carry traffic over the design
reduce the occurrence and severity of reflection cracking. period at the desired level of reliability.
Traffic loadings: The 18-kip equivalent single-axle loads Reliability level has a large effect on overlay thickness.
(ESALs) expected in the design lane over the design life of Varying the reliability level used to determine SNf or Dt be-
the overlay must be calculated using the appropriate flexible tween 50 and 99 percent may produce overlay thicknesses
pavement or rigid pavement load equivalency factors from varying by 6 in. or more (4). On the basis of field testing, it
Part II of the guide. Failure to use the correct type of ESALs appears that a design reliability level of approximately 95
will result in a significant error in the overlay design. percent gives overlay thicknesses consistent with those rec-
Subdrainage: Existing subdrainage conditions usually have ommended for most projects by state highway agencies when
a great influence on how well an overlay performs. A sub- the overall standard deviations recommended in Part I and
drainage evaluation should be conducted as described in Part II are used (4). There are, of course, many situations for
III of the guide. which it is desirable to design at a higher or lower level of
Rutting in AC pavements: The cause of rutting in an ex- reliability, depending on the consequences of failure of the
isting AC pavement must be determined before an AC overlay overlay. The reliability level to be used for different overlay
is designed. An overlay may not be appropriate if severe types may vary and should be evaluated by each agency for
rutting is occurring because of instability in any of the existing different highway functional classifications or traffic volumes.
pavement layers. The designer should be aware that some sources of uncer-
Milling AC surface: Removal of a portion of an existing tainty are different for overlay design than for new pavement
AC surface frequently improves the performance of an AC design. Therefore, the overall standard deviations recom-
overlay. Significant rutting or other major distortion should mended for new pavement design may not be appropriate for
be removed by milling before another overlay is placed. overlay design. The appropriate value for overall standard
Recycling the existing pavement: Recycling a portion of an deviation may vary by overlay type as well. An additional
existing AC layer may be considered as an option in the design source of variation is the uncertainty associated with estab-
of an overlay. Complete recycling of the AC layer or recycling lishing the effective structural capacity (SNeff or Deft) of the
of a PCC slab necessitates designing the reconstructed pave- existing pavement. However, some sources of variation may
ment according to procedures for new pavement design. be less significant for overlay design than for new pavement
Structural versus functional overlays: The revised AASHTO ·(e.g., estimation of future traffic).
overlay design procedures provide an overlay thickness to Pavement widening: Many AC overlays are placed over
correct a structural deficiency. If no structural deficiency ex- PCC pavements in conjunction with pavement widening (either
ists, an overlay thickness less than or equal to zero will be adding lanes or adding width to a narrow lane). This situation
obtained. This does not mean, however, that the pavement requires coordination between the design of the widened
does not need an overlay to correct a functional deficiency. pavement section and the overlay so that both the existing
Shoulders: If an existing shoulder is in good condition, the and the widening sections will be structurally and functionally
shoulder may be overlaid to match the grade of the traffic adequate.
lanes, after patching of deteriorated areas on the shoulder.·
If an existing shoulder is in such poor condition that it cannot
be patched economically, it should be removed and replaced. PAVEMENT EVALUATION FOR OVERLAY
Existing PCC slab durability: The durability of an existing DESIGN
PCC slab greatly influences the performance of AC and bonded
PCC overlays. If D cracking or reactive aggregate distress It is important that an evaluation of the existing pavement be
exists, the deterioration of the existing slab can be expected conducted to identify any functional or structural deficiencies
to continue after overlay. and to select appropriate preoverlay repair, reflection crack
PCC overlay joints: Bonded or unbonded jointed concrete treatments, and overlay designs to correct these deficiencies.
overlays require special jointed design that considers the char- Figure 1 shows the concepts of structural deficiency and
acteristics of the underlying pavement. Factors to be consid- effective structural capacity. The structural capacity of a pave-
ered in overlay joint design include joint spacing, saw cut ment when new is denoted SC0 • For flexible pavements, struc-
depth, sealant reservoir shape, and load transfer requirements. tural capacity is the structural number, SN. For right pave-
PCC overlay reinforcement: Jointed reinforced and contin- ments, structural capacity is the slab thickness, D. For existing
uously reinforced concrete overlays require an adequate amount composite pavements (AC/PCC), the structural capacity is
of reinforcement to hold cracks together. Friction between expressed as an equivalent slab thickness.
the overlay slab and the base slab should be considered in The structural capacity of the pavement declines with time
the reinforcement design. and traffic. At any time that an evaluation is done for the
PCC overlay bonding/separation layers: Bonded overlays purpose of overlay design, the ·structural capacity has de-
must be constructed to ensure that the overlay remains bonded creased to SCeff· The effective structural capacity is expressed
to the existing slab. Unbonded overlays must be constructed by SNeff for flexible pavements and by Deft for rigid and com-
to ensure that the separation layer prevents reflection cracks posite pavements.
in the overlay. If a structural capacity of scf is required for the future
Overlay design reliability level and overall standard devia- traffic expected during the overlay design period, an overlay
tion: An overlay may be designed for different levels of re- with a structural capacity of SC01 (where SCt - SCeff = SC01 )
38 TRANSPORTATION RESEARCH RECORD 1374

survey to ensure that all significant pavement conditions are

Pl represented. If NDT is done, the data from that testing should
also be used to select appropriate sites for coring.
Specific recommendations for assessing effective structural
.c capacity from distress survey and materials testing informa-
~ P2 tion are given elsewhere (2) for each overlay type .
.~
;::
Q)
(/) 1.5
Np
Structural Capacity Based on NDT

NDT is an extremely valuable and rapidly developing tech-

N Load Applications
nology. When properly applied, NDT can provide a vast amount
of information with a reasonable expenditure of time, money,
~ and effort. The analyses, however, can be sensitive to un-
"(j
0 known conditions and must be performed by knowledgeable,
0.
u
0 experienced persons.
Within the scope of these overlay design procedures, NDT
structural evaluation differs depending on the type of pave-
ment. For PCC pavements, NDT serves three functions: (a)
SCetf
to examine load transfer efficiency at joints and cracks, (b)
to estimate the effective modulus of subgrade rection (k value),
and (c) to estimate the PCC modulus of elasticity (which
N Load Applications provides an estimate of flexural strength). For AC pavements,
NDT serves two functions: (a) to estimate the roadbed soil
FIGURE 1 Structural capacity loss over time and resilient modulus and (b) to directly estimate SNeff· Some
with traffic. agencies use NDT to backcalculate the moduli of the indi-
vidual layers of an AC pavement and then use these moduli
to estimate SNeff· This approach is not recommended in the
must be added to the existing pavement structure. Obviously, revised AASHTO overlay design procedures because it im-
the required overlay structural capacity can be correct only plies and requires a level of sophistication that does not exist
if the required future structural capacity and the assessment with the structural number approach to design.
of the existing structural capacity are correct. The primary
objective of the structural evaluation is to determine the ef-
fective structural capacity of the existing pavement. Three Structural Capacity Based on Remaining Life
methods are described for determining effective structural
capacity. The remaining life approach to structural evaluation is based
on the concept shown in Figure 1. This concept is that re-
peated loads gradually damage the pavement and reduce the
Structural Capacity Based on Visual Survey and remaining number of loads the pavement can carry. To de-
Materials Testing termine the remaining life, the designer must determine the
actual amount of traffic the pavement has carried to date and
A key component in determination of effective structural ca- the total amount of traffic the pavement could be expected
pacity is observation of existing pavement conditions. In ad- to carry to "failure" (when serviceability equals 1.5, to be
dition to information on the pavement's original design, con- consistent with the AASHO Road Test equations). Both traffic
struction, and maintenance history, information on the amounts must be expressed in 18-kip ESALs. The difference
pavement's current condition must be obtained. A distress between these values, expressed as a percentage of the total
survey should be conducted to identify the type, amount, traffic to failure, is the remaining life:
severity, and location of distresses present. The key distress

RL ~ 100 (1 ~ (:;,) J
types for each pavement type that should be considered in
determining the effective structural capacity are described (1)
elsewhere (2). Recommendations for preoverlay repair for
each overlay type are given elsewhere (2). where
A drainage survey should be coupled with the distress sur-
RL = remaining life (percent),
vey. The objective of the drainage survey is to identify moisture-
NP = total traffic to date, 18-kip ESAL, and
related problems and locations where drainage improvements
Nl.5 = total traffic to pavement failure (P2 = 1.5), 18-kip
might be effective in reducing the influence of moisture on
ESAL.
the performance of the pavement after overlay.
A coring and testing program should be coordinated with With RL determined, the designer may obtain a condition
the distress survey to verify layer thicknesses, obtain material factor (CF) from Figure 2. The remaining life method as
samples for testing, and investigate the causes of the observed presented in the revised AASHTO overlay design procedures
distress. Coring locations should be selected after the distress mak~s use of a thorough examination of the relationship be-
Hall et al. 39

Condition Factor, CF
1.0
--------:·---._:
: : :
:

~
0.9
~~

~ ..........
0.8
~

0.7 ~ "\

0.6 \
0.5
100 90 80 70 60 50 40 30 20 10
\
0
Remaining Life, RL, percent

FIGURE 2 Relationship between condition factor and remaining life.

tween remaining life and condition factor done by Elliott (5). little load-associated distress is present. If load-related crack-
CF is defined by ing is present in small amounts and at a low severity level,
the pavement has considerable remaining life, regardless of
what the traffic-based remaining life calculation suggests. At
(2) the other extreme, the remaining life estimate may be very
high even though a substantial amount of medium- and high-
where SCn is the pavement structural capacity after Np ESAL severity load-related cracking is present. In this case, the
and SC0 is the original pavement structural capacity. pavement really has little remaining life. At any point between
The existing structural capacity may be estimated by mul- these two extremes, the remaining life computed from past
tiplying the original structural capacity of the pavement by traffic may not reflect the amount of fatigue damage in the
CF. For example, the original structural number (SN0 ) of a pavement, but discerning this from observed distress may be
flexible pavement may be calculated from material thick- more difficult. If the computed remaining life appears to be
nesses and the structural coefficients for those materials in a clearly at odds with the amount and severity of load-associated
new pavement. SNeff based on a remaining life analysis would distress present, the remaining life method should not be used
be to compute the structural capacity of the existing pavement.
The remaining life approach to determining structural ca-
(3) pacity is not directly applicable, without modification, to
pavements that have already received one or more overlays.
The structural capacity determined by this relationship does
not account for any preoverlay repair. The calculated struc- SUMMARY OF REVISED AASHTO OVERLAY
tural capacity should be viewed as a lower limit value and DESIGN PROCEDURES
may require adjustment to reflect the benefits of preoverlay
repair. The revised AASHTO overlay design procedure actually con-
The remaining life approach to determine SNeff or Deff·has sists of seven separate, stand-alone design procedures, one
some serious deficiencies associated with it. There are four for each of the overlay/pavement combinations listed earlier.
major sources of error: The procedures were developed in this fashion to enhance
their clarity and ease of use.
1. The predictive capability of the AASHO Road Test The design procedure for each type of overlay begins with
equations, a description of the construction tasks involved, conditions
2. The large variation in performance typically observed under which that type of overlay may not be feasible, detailed
even among pavements of seemingly identical designs, preoverlay repair recommendations, and considerations for
3. Estimation of past 18-kip ESALs, and reflection crack control.
4. Inability to account for the amount of preoverlay repair For each type of overlay, the thickness design process fol-
to the pavement. lows eight steps.

As a result, this method of determining the remaining life 1. Determine existing pavement design and construction:
of the pavement can in some cases produce erroneous results. The layer thickness and material inputs required are identi-
The remaining life estimate may be very low even though fied.
. _____ 49 TRANSPORTATION RESEARCH RECORD 1374

2. Traffic analysis: ·Predicted future 18-kip ESALs in the development, and their field testing can be found elsewhere
design lane over the design period are required. The type of (2-4).
ESALs (rigid or flexible) appropriate for the overlay/pave.:
ment combination is required. The remaining life method
AC OVERLAY OF AC PAVEMENT
of determining SNeff or Deff also requires past cumulative
ESALs.
The required thickness of an AC overlay for an AC pavement
3. Condition survey: Distress types, severities, and quan-
is given by the following equation:
tities required for determination of the effective structural
capacity of the existing pavement are specified.
SNol (SNf - SNeff)
4. Deflection testing (strongly recommended): Specific pro- Doi = ~- = ~~~~~ (4)
cedures for deflection testing for determination of inputs to aol aol

the overlay design procedure are described. For AC pave-

ments, deflection testing provides an estimate of the design where
subgrade resilient modulus needed to determine SNf, and also SN01 = required overlay structural number,
a direct estimate of SNeff· For PCC pavements, deflection a01 = structural coefficient for the AC overlay,
testing provides estimates of several parameters needed to D 01 = required overlay thickness (in.),
determine Df, including the effective k value, the PCC elastic SNf = structural number required to carry future traffic,
modulus, the PCC modulus of rupture, and the J load transfer and
factor. A heavy-load deflection device such as the falling weight SNeff = effective structural number of the existing pavement.
deflectometer (FWD) is recommended. Guidelines on NDT
load levels, sensor locations, and testing locations are given The design subgrade resilient modulus, which is required
as appropriate for each existing pavement type. to determine SNf, may be determined from deflection testing
5. Coring and materials testing (strongly recommended): using the following equation:
Guidelines for laboratory testing and visual examination of
materials samples are given.
6. Determination of required structural capacity for future
Design MR = C ( 0 ·24
dr r
p) (5)

traffic (SNf or Df): Each of the inputs required to determine

SNf or Df according to the flexible or rigid pavement design where
equation in Part II of the guide is described. Guidelines for
Design MR = design subgrade resilient modulus (psi),
determining these inputs and the ranges of their reasonable
P = applied load (lb),
values are given. With each overlay design procedure, a work-
dr = deflection at a distance r from the center of
sheet is provided for determination of the future structural
the load (in.),
capacity required. .
r = distance from center of load (in.), and
7. Determination of effective structural capacity of the ex-
C = 0.33 (recommended).
isting pavement (SNeff or D eff): Of the three available methods
for determining effective structural capacity, those appropri- This method of determining the subgrade modulus was pro-
ate for the existing pavement type are described. For AC posed by Ullidtz (6,7) and is based on Boussinesq's·deflection
pavements with no previous overlay, all three methods are equation (8). Its derivation is provided elsewhere (3). This
applicable. For bare PCC pavements, the condition survey equation may be applied to deflections measured at a suffi-
method and remaining life method are applicable. For AC/ cient distance from the applied load that the deflection is due
PCC pavements, only the condition survey method is appli- only to subgrade deformation. A correction factor C no greater
cable. With each overlay design procedure, a worksheet is than 0.33 is required to. make the subgrade resilient modulus
provided for determination of the effective structural capacity consistent with the laboratory-measured value of 3,000 psi at
of the existing pavement. a deviator stress of 6 psi, which was used for the AASHO
8. Determination of overlay thickness: In each procedure, Road Test soil in the development of the flexible pavement
an equation is given for the overlay thickness required to design equation. The need for this correction· was verified
satisfy the pavement's structural deficiency and support the using field and laboratory subgrade modulus data from the
predicted future traffic over the design period. AASHO Road Test and other sites (3). The design subgrade
resilient modulus may also require seasonal adjustment, in
Each overlay design procedure also includes a discussion accordance with Part II of the guide. The subgrade resilient
of other relevant items, such as subdrainage, shoulders, and modulus may also be determined by laboratory testing or from
widening (for all overlay/pavement combinations), surface relationships developed between resilient modulus and other
milling (for AC overlays of AC pavements and existing AC/ soil properties.
PCC pavements), overlay joints and overlay reinforcement The NDT method of SNeff determination follows an as-
(for all PCC/pavement combinations), and bonding proce- sumption that the structural capacity of the pavement is a
dures and separation layers (for bonded and unbonded PCC function of its total thickness and overall stiffness. The re-
overlays, respectively). lationship between SNem thickness, and stiffness is
Highlights of the individual overlay design procedures are
described in the following sections. This summary does not, SNeff = 0.0045DW. (6)
of course, provide the details necessary to apply the design
procedures. Complete information on the procedures, their where D is the total thickness of all pavement layers above
Hall et al. 41

the subgrade (in.) and EP is the effective modulus of all pave- the following techniques: break/seat (for JRCP), crack/seat
ment layers above the subgrade (psi). (for JPCP), or rubblize/compact (for JRCP, JPCP, or CRCP).
The pavement's effective modulus may be determined by The design procedure for AC overlays of fractured PCC slab
trial and error using the following equation: pavements follows the same basic approach used for AC over-
lay of AC pavements. Deflections measured on the PCC slab
d0 = l.5pa before fracturing may be used to determine the subgrade
modulus. A smaller C factor (0.25) is recommended for ad-
justment of the backcalculated subgrade modulus to a design
value, because the stress state in the subgrade is much lower
beneath an intact PCC slab than beneath an AC pavement.
The structural properties of fractured PCC slabs are difficult
to characterize. Backcalculated modulus values ranging from
100,000 to 800,000 psi, and within-project coefficients of var-
iation of 40 percent or more, have been reported for rubblized
where
pavements (11,12). Backcalculated modulus values ranging
d0 = deflection measured at the center of the load plate from a few hundred thousand psi to a several million psi, and
(and adjusted to a standard temperature of 68°F) within-project coefficients of variation of 40 percent or more,
(in.), have been reported for cracked/seated and broken/seated slabs
p = NDT load plate pressure (psi), (11-16).
a = NDT load plate radius (in.), SNeff is determined for fractured PCC slabs by component
D = total thickness of pavement layers above the subgrade analysis using the following structural number equation:
(in.),
MR = subgrade resilient modulus (psi), and (9)
Ep = effective modulus of all pavement layers above the
subgrade (psi). where
This equation is based on Odemark's method for deter- Dz, D 3 = thicknesses of fractured slab and base layers,
mination of deflection in a two-layer system (9), using Bous- az, a 3 = corresponding structural layer coefficients, and
sinesq's one-layer deflection (8) and the concept of "equiv- mz, m 3 = drainage coefficients for fractured slab and gran-
alent thickness" described by Barber (10). Its derivation is ular subbase.
provided elsewhere (3).
The recommended ranges of values for az are 0.20 to 0.35
The condition survey method of SNeff determination
for crack/seat JPCP and break/seat JRCP and 0.14 to 0.30 for
involves a component analysis using the structural number
rubblized PCC of any type. Since the layer coefficient rep-
equation:
resents the overall performance contribution of that layer, it
is likely that it is not related solely to the modulus of that
(8)
layer, but to other properties as well, such as the load transfer
capability of the pieces. The large variability of layer moduli
where
within a project is also of concern. This extra variability should
D 1 , Dz, D 3 = thicknesses of existing pavement surface, ideally be expressed in an increased overall standard deviation
base, and subbase layers; in designing for a given reliability level.
a1 , az, a 3 = corresponding structu~al layer coefficients;
and
mz, m 3 = drainage coefficients for granular base and
sub base. AC OVERLAY OF JPCP, JRCP ~ AND CRCP

Suggested layer coefficients for existing AC pavement layer The required thickness of an AC overlay of a bare PCCpave-
materials are given elsewhere (2). The values suggested are ment is given by the following equation:
less than or equal to the values that would be assigned to the
materials if new, ·depending on the quantity and severity of (10)
distress present, and evidence of pumping, degradation, or
contamination by fines. Guidelines for selection of drainage where
coefficients are given in Part II of the guide. It is emphasized
in the overlay design procedure that the poor drainage situ- D 01 = required thickness of AC overlay (in.),
ation at the AASHO Road Test would be expressed by drain- A = factor to convert PCC thickness deficiency to AC
age coefficient values of 1.0 for granular layers. overlay thickness,
Dt = slab thickness to carry future traffic (in.), and
Deft = effective thickness of existing slab (in.).
AC OVERLAY OF FRACTURED PCC SLAB
PAVEMENT The A factor, which is a function of the PCC thickness
deficiency, is given by the following equation:
This procedure addresses the design of AC overlays placed
on PCC pavements after they have been fractured by any of A = 2.2233 + 0.0099(Df - Deff)2 - 0.1534(Df - Deff) (11)
42 TRANSPORTATION RESEARCH RECORD 1374

A is used to convert PCC thickness deficiency to required where

AC overlay thickness. A value of about 2.5 has been used
for many years in various overlay design procedures. For ek = dense liquid radius of relative stiffness (in.)'
example, a 2-in. bonded PCC overlay is considered roughly Epcc = PCC elastic modulus (psi),
equivalent to a 5-in. AC overlay. However, for greater PCC Dpee = PCC thickness (in.),
thickness deficiencies, using a value of 2.5 for A produces AC µpee = PCC Poisson's ratio, and
overlay thicknesses that are not realistic. This concern was k = effective k value (psi/in.).
addressed by an investigation of the A factor for design of The following equation for ek as a function of AREA was
AC overlays of PCC pavements. developed by Hall (12):
Examination of the Corps of Engineers' field data (17,18)
from which the A value of 2.5 was derived revealed that this 4 387009
36 _ AREA) ]
value is overly conservative (3). To investigate further what ·

A factor should be used in design of AC overlays of PCC = In ( 1812.279133

[ (14)
pavements, the elastic layer program BISAR was used to ek -2.559340
compute stresses in PCC slabs with a range of PCC and AC
overlay thicknesses. The A factor required (for an AC overlay The effective k value may be obtained from Westergaard's
thickness that would produce the same stress in the base slab deflection equation (24) using the measured maximum de-
as a given thickness of bonded PCC overlay) decreased as the flection and the ek corresponding to the computed AREA:
PCC thickness deficiency increased. The development of
Equation 11 is described elsewhere (3).
The overlay design procedures in the 1986 guide proposed
that the effective k value be determined using backcalculated
k ~ (s:, q) { 1 + (z~) HzaeJ
~ - 1.25] (t)'}
elastic moduli for the base and subgrade. This approach is
not recommended in the revised overlay design procedures. + (15)
Rather, direct backcalculation of the effective k value is rec-
ommended, and a simple procedure for doing so is provided.
The effective dynamic k value of the foundation and the where
elastic modulus of the PCC slab may be directly determined
from the maximum deflection d0 measured beneath an NDT
d0 = maximum deflection (in.),
P = load (lb), and
load plate and the deflection basin AREA, defined as follows:
'Y = Euler's constant (0.57721566490).
The effective static k value, used in the rigid pavement
design equation for determination of Dr, is estimated by di-
viding the effective dynamic k value by two (3,21,25).
The elastic modulus of the PCC slab may be determined
from the slab thickness, the k value, and the radius of relative
where d0 is the deflection in center of loading plate (in.) and stiffness. The PCC modulus of rupture may be estimated from
di = deflections at 12, 24, and 36 in. from plate center (in.). the backcalculated PCC elastic modulus or from indirect ten-
AREA has units of length, rather than area, since each of sile strengths of cores. For CRCP, the modulus of rupture
the deflections is normalized with respect to d0 in order to should be determined from backcalculated E values only at
remove the effect of different load levels and to restrict the points that have no cracks within the deflection basins.
range of values obtained. AREA and d0 are thus independent For JPCP and JRCP, deflection testing is recommended to
parameters, from which the two unknown values Epee and k measure load transfer at tranverse joints. The overlay design
may be determined for a known slab thickness. This approach procedure provides guidelines for load transfer measurement
to direct backcalculation of pavement and foundation moduli and selection of the J factor. For CRCP, a J factor of 2.2 to
in two-layer pavements was first proposed by Hoffman and 2.6 is recommended for overlay design, assuming that working
Thompson (19) and adapted to E and k backcalculation for cracks are repaired with continuously reinforced PCC.
PCC pavements by ERES (20) and Foxworthy (21). Further The effective thickness of the existing slab (Den) is com-
investigation of this concept by Barenberg and Petros (22) puted from the following equation:
and by Ioannides (23) has produced a forward solution pro-
cedure to replace the iterative and graphical procedures used (16)
previously.
For a given load radius and sensor arrangement, a unique where
relationship exists between AREA and the "dense liquid"
te
radius of relative stiffness of the pavement system k), in D = existing PCC slab thickness (in.),
which the subgrade is characterized by a k value (24): Fje = joints and cracks adjustment factor,
Fdur = durability adjustment factor, and
Frat = fatigue damage adjustment factor.
(13) The joints and cracks factor Fie adjusts for the extra loss in
PSI caused by deteriorated reflection cracks in the overlay
Hall et al. 43

that will result from any unrepaired deteriorated joints, cracks, ing (ASTM D4123) of cores from the AC surface at two or
punchouts, and other discontinuities in the existing slab before more temperatures to establish a curve of AC modulus versus
overlay. Full-depth repair of these distresses before overlay temperature. Laboratory-measured AC modulus values must
is strongly recommended. The overlay design procedure pro- be adjusted to the frequency of the NDT device used (3).
vides guidelines for assigning Fie on the basis of the number The Asphalt Institute's equation for AC modulus applies
of deteriorated joints, cracks, punchouts, and other major to new mixes. AC that has been in service for some years
discontinuities left unrepaired. may have either a higher modulus (due to hardening of the
The durability factor Fdur adjusts for an extra loss in PSI of asphalt) or lower modulus (due to deterioration from stripping
the overlay when the existing slab has durability problems or other causes) at any given temperature.
such as D cracking or reactive aggregate distress. Guidelines The effective thickness of the existing slab (Deff) is com-
provided for assignment of Fdur on the basis of the severity puted from the following equation:
and quantity of durability-related distress.
The fatigue damage factor Ffat adjusts for past fatigue dam-
(18)
age in the slab. Guidelines are provided for assignment of Ffat
on the basis of the severity and quantity of load-related
distress. Guidelines are provided elsewhere (2) for assignment of Fie,
Fdun and the AC quality adjustment factor Fae• for existing
ACIPCC pavements.
AC OVERLAY OF AC/PCC PAVEMENT

This procedure addresses the design of second AC overlay BONDED PCC OVERLAY OF JPCP, JRCP,
for JPCP, JRCP, and CRCP with an existing AC overlay. AND CRCP
The design procedure follows the same basic approach used
for AC overlays of bare PCC pavements. The equation for This procedure follows the same basic approach used for AC
AC overlay thickness is the same. overlays of bare PCC pavements. The following equation for
The effective dynamic k value of the foundation and the bonded PCC overlay thickness is used:
elastic modulus of the PCC slab may be determined using the
procedure described for bare PCC pavements, except that (19)
adjustments must be made to the measured maximum de-
flection d 0 and basin AREA. The compression in the AC The k value, PCC elastic modulus of rupture, and J load
surface under the NDT load plate must be subtracted from transfer factor for the existing PCC pavement should be used
the maximum deflection measured at the AC/PCC pavement to determine Dr. The effective thickness of the existing slab
surface to obtain the deflection of the PCC layer. The AC (Derr) is computed from the following equation:
compression is a function of the AC modulus, AC thickness,
and AC/PCC interface condition (as determined from cores) (20)
(12). For AC/PCC bonded,
1.0798

do compress = -0.0000328 + 121.5006 ( ~:: )

(17a)
UNBONDED PCC OVERLAY OF JPCP, JRCP,
AND CRCP

For AC/PCC unbonded, This procedure follows the same basic approach used for AC
overlays of bare PCC pavements. The following equation for
D )o.94551 unbonded PCC overlay thickness is used:
do compress = -0.00002132 + 38.6872 ( E:: (17b)
D 01 = \/Of - D~ff (21)
where
The elastic modulus, modulus of rupture, and load transfer
d 0 compress = AC compression at center of load (in.), factor for the overlay PCC should be used to determine Dr.
. Dae = AC thickness (in.), and The effective thickness of the existing slab is computed from
Eac = AC elastic modulus (psi). the following equation:
Direct measurement of the AC mix temperature at three
or more times during deflection testing is recommended to (22)
assign an AC mix temperature to each deflection basin. A
relationship between AC elastic modulus and temperature Field surveys of unbonded concrete overlays have shown
may then be used to assign an AC modulus to each deflection that durability distress and fatigue damage in the existing slab
basin. The Asphalt Institute's equation for AC modulus as a have very little effect on the performance of the unbonded
function of temperature, mix parameters (percent fines, vol- overlay. Therefore, the Fdur and Ffat factors are not used to
ume of voids, percent asphalt, and asphalt viscosity), and determine Deff for design of unbonded concrete overlays.
loading frequency (approximately 18 Hz for the FWD load Field surveys of unbonded jointed concrete overlays have
duration of 25 to 30 msec) may be used for this purpose (26). also shown little evidence of reflection cracking or other prob-
An alternative is to conduct diametral resilient modulus test- lems caused by deteriorated joints and cracks in the existing
44 TRANSPORTATION RESEARCH RECORD 1374

slab. Therefore, the Ficu factor, which is used for design of 3. A complete step-by-step overlay design procedure for
unbonded overlays, makes a smaller adjustment to the exist- each overlay type;
ing slab thickness than the Fie factor, which is used for design 4. Guidelines for pavement evaluation for overlay design,
of bonded PCC and AC overlays. Although the thickness including distress surveying, nondestructive testing, and de-
design procedure is the same for jointed and CRC overlays, structive testing;
unbonded overlays are not intended to bridge areas of poor 5. Guidelines for selecting inputs for determination of re-
support, and in particular CRC overlays may require more quired future structural capacity (SNf, Df);
preoverlay repair in some situations. 6. Guidelines for characterization of effective structural ca-
pacity of existing pavement (SNeff• Deff) using three ap-
proaches [condition survey and materials testing, NDT testing
JPCP, JRCP, AND CRCP OVERLAY OF AC (where appropriate), and remaining life (where appropriate)];
PAVEMENT and
7. Improved adaptability of the overlay thickness design
A PCC overlay of an AC pavement is designed using the procedures to local conditions to produce more reasonable
following equation: answers.
(23)
ACKNOWLEDGMENT
The effective k value to be used for design of a PCC overlay
of an existing AC pavement may be estimated from the subgrade This work was sponsored by the American Association of
modulus and the effective pavement modulus, determined State Highway and Transportation Officials in cooperation
from deflection testing as described previously, using the k with the Federal Highway Administration and was conducted
value nomograph provided in Part II of the guide. This dy- in the National Cooperative Highway Research Program, which
namic k value must be divided by 2 to obtain the static k is administered by the Transportation Research Board of the
value for use in design. National Research Council. The authors gratefully acknowl-
The engineer should be aware that this approach to deter- edge the NCHRP staff and the NCHRP Project 20-7/Task 39
mining the design static k value for PCC/AC design has some panel members for their guidance and assistance.
significant limitations. The k value nomograph in Part II of
the guide was developed using an elastic layer program, with-
out verification with field deflection data. Whereas it may REFERENCES
yield reasonable values in some instances, it may yield un-
1. Guide for Design of Pavement Structures. American Association
reasonably high values in other instances. Further research of State Highway and Transportation Officials, Washington, D.C.,
of the subject of support for PCC overlays, including deflec- 1986.
tion testing on in-service PCC/AC pavements and back- 2. M. I. Darter, R. P. Elliott, and K. T. Hall. Revision of AASHTO
calculation of effective k values, is strongly encouraged. Pavement Overlay Design Procedures. Draft Final Report, NCHRP
Project 20-7/Task 39, June 1991.
3. M. I. Darter, R. P. Elliott, and K. T. Hall. Revision of AASHTO
CONCLUSIONS Pavement Overlay Design Procedures, Appendix: Documf!ntation
of Design Procedures. NCHRP Project 20-7ffask 39, June 1991.
The revised AASHTO overlay design procedures use the con- 4. M. I. Darter, R. P. Elliott, and K. T. Hall. Revision ofAASHTO
Pavement Overlay Design Procedures, Appendix: Overlay Design
. cepts of structural deficiency, structural number for flexible Examples. NCHRP Project 20-7ffask 39, June 1991.
pavements, and future required structural capacity deter- 5. R. P. Elliott. An Examination of the AASHTO Remaining Life
mined from the AASHTO flexible and rigid pavement design Factor. In Transportation Research Record 1215, TRB, National
equations. These concepts were retained to maintain com- Research Council, Washington, D.C., 1989.
patibility between Parts II and III of the guide. 6. P. Ullidtz. Pavement Analysis. Elsevier Science Publishers B. V.,
1987.
Development of a more sophisticated mechanistic approach 7. P. Ullidtz. Overlay and Stage by Stage Design. Proc., Fourth
to overlay design was not within the scope of the revisions. International Conference on Structural Design of Asphalt Pave-
NDT is recommended for use in characterizing the existing ments, Ann Arbor, Mich., 1977.
pavement to the extent appropriate_ within the framework of 8. J. Boussinesq. Application des Potentials a I' Etude de l' Equilibre
et du Mouvement des So/ides Elastiques. Gauthier-Villars, Paris,
these empirical design procedures. 1885.
The AASHTO overlay design procedures were extensively 9. N. Odemark. Investigations as to the Elastic Properties of Soils
revised to make them easier to use, more adaptable to cali- and Design of Pavements According to the Theory of Elasticity.
bration by local agencies, and more comprehensive .. Key re- Meddelande 77, Statens Vaginstitut, Stockholm, Sweden, 1949
visions to the overlay design procedures include the following: (English translation by A. M. loannides, 1990).
10. E. S. Barber. Author's closure, comments on C. A. Hogen-
togler, Jr.'s discussion of Soil Displacement Under a Circular
1. Guidelines for overlay feasibility; Loaded Area by L. A. Palmer and E. S. Barber, Proc., Highway
2. Guidelines for several important considerations (pre- Research Board, Vol. 20, 1940.
overlay repair, reflection crack control, subdrainage, AC sur- 11. Guidelines and Methodolologies for the Rehabilitation of Rigid
Highway Pavements Using Asphalt Concrete Overlays. Pavement
face milling, shoulders, AC surface recycling, AC rutting, Consultancy Services/Law Engineering, June 1991.
overlay design reliability level, PCC durability, PCC overlay 12. K. T. Hall. Performance, Evaluation, and Rehabilitation of
bonding/separation layers, pavement widening, and PCC Asphalt-Overlaid Concrete Pavements. Ph.D. thesis. University
overlay joints and reinforcement); of Illinois at Urbana-Champaign, 1991.
 Hall et al. 45

13. W. P. Kilareski and R. A. Bionda. Performance/Rehabilitation the revision proposed by the paper chooses to ignore this
of Rigid Pavements, Phase II, Volume 2-Crack and Seat and important issue and the FRL factor totally. The paper adopts
AC Overlay of Rigid Pavements. Report FHWA-RD-89-143.
FHWA, 1989.
the traditional overlay equation SC0 L = SCy - SCeff through-
14. A. M. Schutzbach. Crack and Seat Method of Pavement Re- out. This writer considers the paper to be incomplete without
habilitation. In Transportation Research Record 1215, TRB, Na- addressing the remaining life concept related to the FRL factor.
tional Research Council, Washington, D.C., 1989. The only clue to why the authors have decided to ignore
15. R. C. Ahlrich. Performance and Structural Evaluation of Cracked Equation 24 is found in one misle.ading statement: "The re-
and Seated Concrete. In Transportation Research Record 1215,
TRB, National Research Council, Washington, D.C., 1989. maining life method as presented in the revised AASHTO
16. M. R. Thompson. NCHRP Synthesis of Highway Practice 144: overlay design procedure makes use of a thorough exami-
Breaking/Cracking and Seating Concrete Pavements. TRB, Na- nation of the relationship between remaining life and con-
tional Research Council, Washington, D.C., 1989. dition factor by Elliott." Elliott's work (4) did not produce a
17. Y. T. Chou. Asphalt Overlay Design for Airfield Pavements.
Proc., Association of Asphalt Paving Technologists, Vol. 53,
new condition factor CF expression as implied by the state-
April 1984, pp. 266-284. ment in the text. Instead, Elliott (4) addressed AASHTO
18. F. M. Mellinger and J. P. Sale. The Design of Non-Rigid Over- remaining life concept and FRL factor and concluded that (a)
lays for Concrete Airfield Pavements. Air Transport Journal, the appropriate value for FRL is 1.0 and (b) the AASHTO
Vol. 82, Number AT 2, May 1956. overlay design approach should be revised to exclude re-
19. M. S. Hoffman and M. R. Thompson. Mechanistic Interpretation
of Nondestructive Pavement Testing Deflections. Transportation maining life considerations. This discussion will show that
Engineering Series No. 32, Illinois Cooperative Highway and Elliott's sweeping conclusions are not justified and why it is
Transportation Research Series No. 190, University of Illinois at not wise to discard Equation 24 and revert to the use of the
Urbana-Champaign, 1981. traditional overlay equation.
20. Nondestructive Structural Evaluation of Airfield Pavements. ERES
Consultants, Inc., 1982.
21. P. T. Foxworthy. Concepts for the Development of a Non-
destructive Testing and Evaluation System for Rigid Airfield Pave- REMAINING LIFE CONCEPT AND F RL
ments. Ph.D. thesis. University of Illinois at Urbana-Champaign,
1985. Elliott's (4) recommendation to exclude remaining life con-
22. E. J. Barenberg and K. A. Petros. Evaluation of Concrete Pave-
ments Using NDT Results. Project IHR-512. Report UILU-ENG- sideration from AASHTO overlay design was based on the
91-2006, University of Illinois and Illinois Department of Trans- reasons that (a) AASHTO design produces inconsistent re-
portation, 1991. sults and (b) the appropriate FRL value is 1.0. The study by
23. A. M. Ioannides. Dimensional Analysis in NDT Rigid Pavement Fwa (3) shows that inconsistencies in AASHTO overlay de-
Evaluation. Transportation Engineering Journal, Vol. 116, No.
TEl, 1990. signs were due solely to a flaw in the formula for computing
24. H. M. Westergaard. Stresses in Concrete Runways of Airports. FRL· A corrected formula for FRL was derived according to
Proc., Highway Research Board, Vol. 19, 1939. the very concept of remaining life described in the AASHTO
25. Special Report 61 E: The AASH0 Road Test, Report 5, Pavement Guide. Using the corrected formula.for FRL, it was illustrated
Research. HRB, National Research Council, Washington, D.C., that consistent overlay designs were obtained with Equation
1962.
26. Research and Development of the Asphalt Institute's Thickness 24. Elliott's reason (a) is therefore invalid.
Design Manual (MS-1) (ninth edition). Research Report 82-2. Elliott's claim of FRL = 1.0 was based on an analysis using
Asphalt Institute, 1982. a "simple scale transformation" that relates RLyx to RLy for
a given RLx as follows:

DISCUSSION {Rd {~} (25)

T. F. FWA The transformation is artificial with no clear physical meaning.

Center for Transportation Research, Faculty of Engineering, National It is also controversial because RLx and RLyx of the old pave-
University of Singapore.
ment ·and RLy of overlaid pavement are computed from dif-
A major contribution of the 1986 AASHTO Guide (1) is the ferent base Nf values. Elliott said that Equation 25 was based
introduction of a remaining life concept in overlay thickness on the "philosophy" of "the man who each day walks halfway
design, where the overlay structural requirement is expressed to his destination," a "philosophy" many readers would find
in the following form: difficult to relate to overlay performance. Although not stated
by Elliott, Equation 25 actually assumes that the rate of de-
crease RLyx (of old pavement) is proportional to that of RLy
(of new overlaid pavement). This appears to be too strong
or an assumption with very restrictive application because it is
common knowledge that the structural capacities and hence
(24) the remaining lives of pavements of different ages decrease
at unequal (and nonproportional) rates. Incidentally~ Elliott's
An excellent description of the concept is found in Chapter assumption is similar to the condition for a lower bound over-
5 of Part III and Appendix CC of the AASHTO Guide. That lay design (with FRL = 1.0) independently identified by Easa
the concept is fundamentally correct and conceptually sound (2) and Fwa (3) as explained in the next section. Both Easa
has subsequently been ascertained independently by Easa (2) and Fwa also illustrated that there exists an upper bound
and Fwa (3). Unfortunately, without giving valid justification, overlay design and there are theoretically many possible so-
46 TRANSPORTATION RESEARCH RECORD 1374

lutions (with F~ values less than 1.0) between the two bounds. term and reverting to the traditional overlay equation is a
Elliott's analysis is therefore applicable to a very special case, move that is unwise and uncalled for. This writer hopes that
probably a highly unlikely one. Elliott's reason (b), based on the authors will make necessary amendments to their pro-
conclusions drawn from the limited analysis, is inadequate to posed revision before it is finalized.
justify his recommendations to set FRL = 1.0 and exclude
remaining life consideration from the AASHTO overlay de-
sign approach. REFERENCES

1. AASHTO Guide for Design of Pavement Structures (Vols. 1 and

2). American Society of State Highway and Transportation Of-
UPPER AND LOWER BOUND OVERLAY DESIGN ficials, Washington, D.C., 1986.
2. S. M. Easa. Extension of AASHTO Remaining-Life Methodology
Subsequent to Elliott's work (4) that pointed out inconsis- of Overlay Design. In Transportation Research Record 1272, TRB,
tencies in the AASHTO overlay design method, Easa (2) and National Research Council, Washington, D.C. 1990.
Fwa (3) separately confirmed the fundamental correctness of 3. T. F. Fwa. Remaining-Life Consideration in Pavement Overlay
Design. Journal of Transportation Engineering, Vol. 117, No. 6,
the AASHTO overlay design approach that incorporates the 1991.
concept of remaining life, and proposed different procedures 4. R. P. Elliott. An Examination of the AASHTO Remaining Life
to eliminate the inconsistencies caused by a flaw in FRL cal- Factor. In Transportation Research Record 1215, TRB, National
culation. Both Easa and Fwa defined a lower and an upper Research Council, Washington, D.C., 1989.
bound overlay solution. The lower bound solution corre-
sponds to the case with FRL = 1.0 (which is the maximum
possible value of FRL) where the rate of structural deterio- AUTHORS' CLOSURE
ration of an old pavement after overlay is assumed to be the
same as that of a new pavement. The upper bound solution The remaining life concept has not been discarded in the
is one with FRL ::; 1.0 where the old pavement after overlay proposed revisions to the AASHTO overlay design proce-
is assumed to continue to deteriorate at a ra~e as if no overlay dures. As described in the paper, three procedures are given
were applied. It is easy to see that the overlay solution that for estimating the effective structural capacity of an existing
represents the real-life situation will lie somewhere between pavement: a deflection-based approach, a condition survey
the two bounds. It is also easy to see that it is unwise to approach, and a remaining life approach.
set FRL = 1.0 (lower bound solution) because it would The basic concept of remaining life is that a pavement's
lead to an overlay solution that is underdesigned and thus past traffic and its total traffic-bearing capacity over its life-
unconservative. time may be used together to estimate the traffic the pavement
is capable of carrying for the remainder of its life. This concept
did not originate with the 1986 AASHTO Guide, but it has
TRADITIONAL VERSUS AASHTO OVERLAY been used in pavement evaluation for many years and is ap-
DESIGN plicable to any pavement design procedure based on a rela-
tionship between traffic and loss of structural capacity.
The traditional overlay equation is conceptually unsound and Indeed, this concept is intrinsic to the AASHTO design
inadequate because overlay thickness is derived on the basis methodology.
of the overlay requirement at the time of overlay application. The authors consider the basic remaining life concept to be
It does not include an analysis to examine whether the overlay valid. However, .the application of this concept in the pro-
provided is adequate during other stages of overlay service posed revisions to the AASHTO overlay design procedures
life. In terms of remaining life concept, the traditional overlay differs from the application presented in the 1986 guide.
design method is equivalent to setting FRL = 1.0 and assuming In the 1986 guide's overlay design procedures, procedures
an old pavement will deteriorate like a new pavement after were given for determining the effective structural capacity
being overlaid. In contrast, as explained in Appendix CC of (SCeff) of a pavement from deflection testing or distress ob-
the AASHTO Guide (1)' and demonstrated by Fwa (3), the servations. This effective structural capacity is expected to be
1986 AASHTO overlay design approach that incorporates less than the original structural capacity of the pavement when
remaining life consideration enables one to analyze the over- new (SN0 ) . However, the 1986 guide's overlay design .pro-
lay requirements for the entire design period and select an cedures then applied a traffic-based remaining life factor as
appropriate FRL value to compute from Equation 24 the over- a multiplier to the effective structural capacity determined
lay thickness needed. The value of FRL is equal to 1.0 if the from deflections or distress observations. This approach is
overlay requirement at the time of overlay application governs widely considered to penalize a pavement twice for the same
the design. In cases where overlay requirements at other stages past traffic.
of overlay service life are more critical, the value of FRL will Fwa has defended this double penalty with the reasoning
be less than 1. 0. that if a deteriorated pavement with a given effective struc-
tural capacity is overlaid, it will subsequently deteriorate at
a faster rate than a newly constructed pavement of the same
SUMMARY structural capacity that receives the same thickness of overlay.
This is a considerable distortion of the structural deficiency
This discussion shows that the authors' decision to discard the concept of overlay design. The essence of the structural de-
1986 AASHTO remaining life concept by ignoring the FRL ficiency concept is that a performance prediction model may
Hall et al. 47

be used to determine a required overlay, which will increase needlessly complex and poorly supported. For example, the
an in-service pavement's effective structural capacity to a procedure did not address the practical significance of a "neg-
structural capacity sufficient to carry the traffic expected over ative remaining life" computed for an in-service pavement.
the design period. The rate of deterioration of the overlaid The need to revise the application of the remaining life con-
pavement is thus predicted by the performance model used, cept in the 1986 guide's overlay design procedures was iden-
just as is the rate of deterioration predicted for new pavements tified by the AASHTO Joint Task Force on Pavements as
by the same model. Within the context of the AASHTO one of the high-priority revisions to the overlay design
design methodology, the flexible and rigid pavement perfor- procedures.
mance models presented in Part II of the guide are used to The authors have examined the work by FWA and by Easa
determine required future structural capacity (structural num- and have concluded that although they offer modifications to
ber of slab thickness), and the rate of deterioration is mea- the remaining life method as presented in the 1986 guide, they
sured by loss of serviceability as predicted by these models. do not correct "its major flaw. They also impose needless com-
If the two pavements described by Fwa have the same struc- plexity in the application of a simple concept.
tural capacity before overlay, and receive the same overlay, The authors have therefore recommended to the Design
then according to the structural deficiency concept their per- Subcommittee of the AASHTO Joint Task Force on Pave-
formance after overlay will be the same. One cannot correctly ments that the method developed by Elliott for considering
apply the structural deficiency concept of overlay design and remaining life be accepted as the best solution to the problems
at the same time conjecture a rate of deterioration of the associated with the application of this concept in the 1986
overlaid pavement other than the rate predicted by the perfor- overlay design procedures. It must also be clarified that de-
mance model used to define the structural deficiency. cisions concerning acceptance of this and other proposed re-
In the proposed revisions to the overlay design procedures, visions to the overlay design procedures are made not by the
a traffic-based estimate of remaining life is applied to a pave- authors but rather by the AASHTO Joint Task Force.
ment's original structural capacity (SC0 ) to estimate its current
effective structural capacity but is not applied to deflection-
based and condition-based estimates of the effective structural
capacity. In concept, these three approaches for estimating
The opinions expressed in this paper are those of the authors and not
sceff should yield similar results. necessarily those of AASHTO, FHWA, NCHRP, or TRB or of the
In addition to the conceptual flaw described earlier, the individual states participating in the National Cooperative Highway
1986 guide's remaining life computation was considered to be Research Program.
48 TRANSPORTATION RESEARCH RECORD 1374

Field Testing of AASHTO Pavement

Overlay Design Procedures
KATHLEEN T. HALL, MICHAEL I. DARTER, AND ROBERT P. ELLIOTT

The AASHTO pavement overlay design procedures were re- design procedures were evaluated by the highway agency per-
cently revised to make them easier to use, more adaptable to sonnel for clarity and ease of use, and many of their comments
calibration by local agencies, and more comprehensive. The re- were incorporated into the procedures.
vised procedures were extensively field tested using data from In addition, the overlay thicknesses indicated by the pro-
many actual in-service pavements located throughout the United
States. A total of 74 examples were developed for seven different cedures were evaluated with respect to state highway agencies'
categories of overlay and pavement types. State highway agency recommendations on the basis of their design procedures and
personnel provided the design, traffic, condition, and deflection experience with overlay performance.
data for the overlay design examples and participated in the de- Each of the example projects is identified by the region of
velopment of the examples. The revised AASHTO overlay design the United States in which it is located and by number within
procedures produce reasonable overlay thicknesses that are con- the region. The following regional identifiers are used: NE,
sistent with state highway agency recommendations. The exam-
Northeast; SE, Southeast; MW, Midwest; NW, Northwest;
ples ·illustrate the importance of selecting appropriate inputs for
overlay design, the use of nondestructive testing data and con- and SW, Southwest.
dition data in overlay design, the significance of design reliability Each of the regions is represented in the overlay design
level to overlay thickness, and the importance of preoverlay repair. examples for each pavement and overlay type to the extent
possible. Seven separate groupings of overlays designs are
included:
Chapter 5 of Part III of the 1986 AASHTO Guide for Design
of Pavement Structures (1) addresses overlay design for pave- Overlay Type Existing Pavement
ment rehabilitation. These overlay d~sign procedures were AC AC pavement
recently revised to make them easier to use, more adaptable AC Fractured PCC slab
AC and Bonded PCC JPCP and JRCP
to calibration by local agencies, and more comprehensive. AC and Bonded PCC CRCP
The proposed procedures are currently under consideration AC ACIPCC (composite)
by AASHTO. U nbonded PCC JPCP, JRCP, CRCP
This paper presents the results of the extensive field testing JPCP and JRCP AC pavement
of the revised overlay design procedures using data from many Lotus 1-2-3 spreadsheets were prepared for each of the
actual in-service pavements located throughout the United above overlap design procedures to aid in the calculations.
States. Darter et al. present the procedures in detail (2), Each example was prepared on a single-page spreadsheet
document the development of the procedures (3), and provide showing all of the inputs used and outputs obtained. The
complete results of the field testing (4). results obtained were also summarized for each of the seven
A total of 74 examples were developed to demonstrate and procedures.
validate the overlay design procedures. These results were Deflection data were used whenever available from the
extremely useful in verifying and improving the overlay design state agency. Typically one to five representative deflection
procedures. The example design projects may also be used basins were entered into a spreadsheet to keep the size of the
by future researchers to help verify improved overlay design output within reason. In some cases only a few deflection
procedures. basins were provided by the agency. In other cases a few
representative basins were selected for illustrative purposes
from a larger deflection data set provided by the agency. The
DESCRIPTION OF FIELD TESTING basins chosen are believed to provide overlay thicknesses close
PROCEDURES to the mean for the project. However, this does not imply
that any project should be represented by this small a number
The examples were developed for actual in-service pavements of basins. On the contrary, the procedures can be pro-
located throughout the United States. Design, traffic, con- grammed to handle any number of deflection basins and cor-
dition, and deflection data were provided for these projects responding overlay designs very efficiently.
by 10 state highway agencies. State personnel were actively
involved in developing these examples during the develop-
ment of the revised overlay design procedures. The overlay EXAMPLES OF AC OVERLAY DESIGN FOR AC
PAVEMENT
K. T. Hall, 1206 Newmark Lab, 205 North Mathews Ave., Urbana,
Ill. 61801. M. I. Darter, 1212 Newmark Lab, 205 North Mathews
Ave., Urbana, Ill. 61801. R. P. Elliott, 4150 Bell Engineering Center, Table 1 gives an example AC overlay design for an AC pave-
Fayetteville, Ark. 72701. ment (NW-1). For a range of reliability levels from 50 to 99
Hall et al. 49

TABLE 1 Example AC Overlay Design for AC Pavement

REVISED CHAPTER S AASHTO DESIGN GUIDE OVERLAY DESIGN
====================================================--============================
NW-1 AC OVERLAY OF CONVENTIONAL AC PAVEMENT
=================================================================================
EXISTING PAVEMENT DESIGN
AC SURFACE 4.2S inches SUBGRADE SANDY SILT, SANDY GRAVEL
GRAN BASE 8.00
GRAN SUBBASE 0.00
TOTAL THICKNESS 12.2S
Future design lane ESALs 2400000 (FLEXIBLE ESALs)
DETERMINE SNf
Vary trial SNf until computed ESALs equal future design ESALs.
SNf MR,psi R Z So Pl P2 ESAL
3.60 S634 so 0 0.4S 4.2 2.S 2417312
4.14 S634 80 0.841 0.4S 4.2 2.S 2430778
4.44 S634 90 1. 282 0.4S 4.2 2.S 2429228
4.69 S634 9S 1. 64S 0.4S 4.2 2.S 2408097
S.19 S634 99 2.327 0.4S 4.2 2.S 240324S
TRIAL INPUT INPUT INPUT INPUT
DETERMINE SNef f BY NOT METHOD
Vary trial Ep/MR until computed DO equals actual value.
ACTUAL SUB GRADE TRIAL COMPUTED
LOAD,lbs DO,mils Dr,mils MR,psi C FACTOR Ep/MR DO,mils Ep,psi SNef f
9000 12.80 3.SS 16901 3 8.4S 12.80 142817 2.88
r = 36 inches
Check r > 0.7 ae = 17.9S inches
DETERMINE SNef f BY CONDITION SURVEY METHOD
LAYER STR COEF DRAIN m SNef f
AC SURFACE 0.3S 1.00 1.49
BASE 0.14 1.00 1.12
SUBBASE 0.00 1.00 0.00
SNef f = 2.61
=================================================================================
DETERMINE SNef f BY REMAINING LIFE METHOD
Past design lane ESALs = 400000 (FLEXIBLE ESALs)
LAYER THICK,in NEW ST CF SNo
AC SURFACE 4.2S 0.44 1.87
BASE 8.00 0.14 1.12
SUBBASE 0.00 0.00 0
TOTAL 12.2S 2.99
SNo MR, psi Z So Pl P2 Nl. S RL, % CF SNef f
2.99 S634 0 0.4S 4.2 l.S 1140161 6S 0.93 2.78
INPUT INPUT INPUT
DETERMINE OVERLAY THICKNESS AC OL structural coefficient 0.44
DESIGN NOT CONDITION REM LIFE
RELIABILITY METHOD, in METHOD,in METHOD,in
so 1.63 2.26 1. 8S
80 2.86 3.48 3.08
90 3.S4 4.16 3.76
9S 4.11 4.73 4.33
99 S.2S S.87 S.47

percent, the required future structural capacity of the pave- plate (2). A check is included to· ensure that the distance is
ment, SNr, is determined by varying trial SNf values until the greater than the minimum distance required for accurate de-
ESALs computed using the AASHTO flexible pavement de- termination of MR. The backcalculated subgrade modulus is
sign equation (from Part II of the guide) match the design then divided by a factor of three to obtain the design subgrade
ESALs for the overlay. The overall standard deviation S0 and modulus used in determining SNr (3).
initial and terminal pavement serviceability values Pl and P2 The existing pavement's structural capacity, SNeff> may also
may also be varied. In the examples, these three inputs were be determined by the NDT method, by varying the ratio of
set at 0.45, 4.2, and 2.5, respectively, unless other values were pavement modulus to subgrade modulus (Ep/MR) until the
given by the state highway agency. computed maximum deflection d0 matches the deflection mea-
The subgrade resilient modulus MR is backcalculated from sured beneath the load plate. The pavement modulus (that
a deflection some distance away from the center of the load is," the effective modulus of all pavement layers above the
50 TRANSPORTATION RESEARCH RECORD 1374

subgrade) may be computed µsing the backcalculated subgrade by dividing the structural deficiency (SNr minus SNeff) by an
modulus, and SNeff is computed as a function of the pavement AC layer coefficient value of 0.44.
modulus. In general, the AC overlay thicknesses for AC pavement
SNett may also be determined by the condition survey method, indicated by the revised AASHTO procedure agree with state
in which a structural coefficient is assigned to each pavement recommendations, as shown in Figure 1. Some of the differ-
layer above the subgrade. The layer coefficients used to de- ences are due to the lack of consistent data from some of the
termine SNerr should be less than or equal to the values that examples. For example, some projects had thicknesses that
would be assigned to the layer materials if new and should varied widely along their length, and the exact thicknesses at
reflect the quantity and severity of distress present and evi- the locations of the deflection basins provided were unknown.
dence of pumping, degradation, or contamination by fines. Errors in assumed pavement thickness are reflected directly
In the examples, the layer coefficients used for the existing in errors in estimating SNeff by the NDT method.
pavements were those provided by the state highway agencies. The overlay thickness designs based on NDT are generally
When layer coefficients were not provided, pavement con- consistent with those based on the condition survey method.
dition information obtained from the state were used to assign Figure 2 shows a comparison between overlay thicknesses at
reasonable layer coefficients. the 95 percent reliability level determined by the NDT method
The third method of determining SNeff for flexible pave- and condition survey method.
ments is the remaining life method. This method requires the The subgrade resilient modulus has a large effect on the
past ESALs accumulated in the design lane since construction. resulting overlay thicknesses. Therefore, it is of utmost im-
Layer coefficients appropriate for new pavement are assigned portance to obtain an appropriate modulus value to enter into
to each layer material in order to compute SN0 , the structural the AASHTO flexible pavement design equation. Use of too
capacity of the pavement when new. The AASHTO flexible high a subgrade modulus in design will result in inadequate
pavement design equation is then used to determine the al- AC overlay thickness. The reduction in backcalculated mod-
lowable ESALs to a terminal serviceability level of 1.5 for a ulus by a factor of three appears to be reasonable (3). Some
50 percent reliability level. The difference between the past data available from one state permit a direct comparison be-
traffic and the allowable traffic, expressed as a percentage of tween laboratory and backcalculated modulus values:
the total traffic to "failure," is the remaining life. The existing Backcalculated
pavement's structural capacity SNeff may be estimated by mul- Project Lab MR (psi) MR (psi) Ratio
tiplying the original structural capacity SN0 by a condition NW-2 6,000 13,483 2.25
factor, CF, which is a function of the remaining life. The past NW-3 6,000 19,608 3.27
traffic data required for the remaining life method of SNeff NW-4 4,150 14,085 3.39
determination was typically very difficult for state highway NW-5 4,500 14,286 3.17
Average 5,163 15,365 3.02
agency personnel to obtain. As a result, the remaining life
method could be used for overlay thickness design for only Each agency will need to evaluate this ratio, as well as other
three of the examples submitted. factors, to tailor the design procedure to its own conditions.
For each reliability level considered and each of the SNeff The design reliability level is very significant. The example
methods used, the required AC overlay thickness is obtained AC pavement projects ranged from collector highways to

95 % REL OL THICKNESS. IN
10,--~~~~~~~~~~~~~~~~~~~~~~~~~~

9 ................................................................. ··························· ......................... .

6 ........................................................................................ .

5
*
.......................................................

4 ·············*· .. ··f····*······ *
3 ····································*··············································································································

2 ·············*····*····*··· ·····*································

..........., I· . .
* AC/AC

2 3 4 5 6 7 8 9 10
AGENCY OL THICKNESS. IN

FIGURE 1 Comparison of AASHTO AC overlay thicknesses and agency

AC overlay thicknesses for AC pavements (95 per~~nt reliability).
Hall et al. 51

CONDITION SURVEY OL THICKNESS, IN

10.--~~~~~~~~~~~~~~~~~~~~~~~~-,,

9 ·········································

6 ·················································

5 .......... .

4 ................................ .

3 * *
·····························*·············-.*···················*···································································································

2 .

o~~~~~_.__~___._~~-'-~--''---~---'-~~-'--~--'-~~-'-~--'

0 2 3 4 5 6 1 8 9 10
NDT OL THICKNESS, IN

FIGURE 2 Comparison of AC overlay thickness determined by NDT and

condition survey procedures for AC pavements (95 percent reliability).

heavily trafficked Interstate-type highways. A design relia- psi) had very thin overlay thickness requirements. It is be-
bility level of approximately 95 percent usually produced rea- lieved that the subgrade modulus is too high for this design.
sonable overlay thicknesses. The design reliability level is very significant. For these
projects, a design reliability level of 90 to 95 percent appears
to provide reasonable overlay thicknesses and in general agrees ·
EXAMPLES OF AC OVERLAY DESIGN FOR with agency recommendations.
FRACTURED PCC PAVEMENT

Table 2 gives an example AC overlay design for a fractured EXAMPLE AC AND BONDED PCC OVERLAY
PCC slab pavement (SW-6). The design procedure is similar DESIGN FOR JPCP AND JRCP
to that used for AC overlays of AC pavements, with the
notable difference that the subgrade modulus backcalculated Table 3 gives an example of AC overlay and bonded PCC
before the slab was fractured is divided by a factor of six, overlay design for a PCC pavement (MW-7). For a range of
rather than three, to account for the increase in subgrade reliability levels from 50 to 99 percent, the required future
stress state after fracturing. The condition survey method is structural capacity Dr is determined by varying trial Dr values
the only method for determining SNeff for fractured PCC until the ESALs computed using. the AASHTO rigid pave-
pavement. For this example, the state agency recommended ment design equation (from Part II of the guide) match the
a 4.2-in. AC overlay plus a crack relief fabric after cracking design ESALs for the overlay. The overall standard deviation
and seating the pavement. S0 , initial serviceability Pl, and terminal serviceability P2 were
Only seven examples could be developed for AC overlays set at 0.35, 4.5, and 2.5, respectively, unless other values were
of fractured PCC slab pavements, so it is difficult to judge given by the state highway agency.
the adequacy of the design procedure. The limited results The effective dynamic k value and PCC elastic modulus
show that the required AC overlay thickness of fractured slab were backcalculated whenever deflection data were available.
PCC appears reasonable for most projects and generally agrees The static k value used to determine Dr was obtainecl by
with the state recommendations. A comparison of AASHTO dividing the dynamic k value by a factor of two (3). The PCC
overlay design thicknesses at 95 perc~nt reliability versus over- modulus of rupture was estimated from the backcalculated
lay thicknesses recommended by s~ate agencies is given in PCC elastic modulus unless another value was given by the
Figure 3 along with data points from the conventional AC state agency.
overlays previously shown. Three rubblized designs in the The two methods available for determining the effective
Southwest show thicker overlays than state recommendations structural capacity Deff of the existing pavement for bare PCC
even when the layer coefficient was at its maximum 0.35 for pavements are the condition survey method and the remaining
crack/seat, which may indicate that a thinner AC overlay is life method. However, past traffic data were not provided for
adequate in warm climates for fractured slab pavements. any of the PCC pavement examples submitted, so the re-
The backcalculated subgrade moduli were all divided by 4 maining life method could not be applied. For the condition
(C = 0.25), which is apparently needed to give appropriate survey method, P:ett is determined by multiplying the existing
overlay ~~icknesses. One section in the Northeast that had a slab thickness by a joints and cracks condition factor Fie• a
CBR of±? (and a correspond~ng estimated modulus of 12,000 fatigue factor Fr~o and a durability factor Fctun which are se-
TABLE 2 Example AC Overlay Design for Fractured PCC Pavement
REVISED CHAPTER 5 AASHTO DESIGN GUIDE OVERLAY DESIGN
========--=======- """""'~=---===============
SW-6 AC OVERLAY OF CRACKED/SEATED JPCP (PROJ STN 353)
========================--====- ------- -- -=============
EXISTING PAVEMENT DESIGN
RUBBLIZED PCC 8.20 inches
C.T.BASE 3.70
SUBBASE o.oo
TOTAL THICKNESS 11.90
Future design lane ESALs 7370000 (2/3 OF 11000000 USED AS FLEXIBLE ESALs)
-------------- -=---=============================
DETERMINE SNf
Vary trial SNf until computed ESALs equal future design ESALs.
SNf MR,psi R Z So Pl P2 ESAL
4.50 4350 50 0 0.49 4.5 2.5 7364787
5.15 4350 80 0.841 0.49 4.5 2.5 7516147
5.50 4350 90 1. 282 0.49 4.5 2.5 7452560
5.80 4350 95 1. 645 0.49 4.5 2.5 7401524
6.40 4350 99 2.327 0.49 4.5 2.5 7354079
TRIAL INPUT INPUT INPUT INPUT
=================================================================================
DETERMINE SUBGRADE MR BY NOT METHOD
Vary trial ~p/MR until computed DO equals actual value.
ACTUAL SUBGRADE TRIAL COMPUTED
STATION LOAD,lbs DO,mils Dr,mils MR,psi C FACTOR Ep/MR DO,mils Ep,psi
8952 6.31 3.43 17399 4 44.00 6.32 765574
r = 36 inches
Check r > 0.7 ae = 29.70 inches
================--================================================================
DETERMINE SNef f
LAYER STR COEF DRAIN m SNef f
RUBBLIZED PCC 0.35 1.00 2.87
C.T.SUBBASE 0.15 1.00 0.56
SUBBASE 0.00 1.00 o.oo
SNef f = 3.43
~==============================================================================
DETERMINE OVERLAY THICKNESS AC OL structural coefficient = 0.44
DESIGN CONDITION
RELIABILITY METBOD,in
50 2.44
80 3.92
90 4. 72
95 5.40
99 6.76
===============---- ------ -===============--====================================

95 % REL OL THICKNESS, IN
lQr-~~~~~~~~~~~~~~~~~~~~~~~~

9 ··························································································································································· ················

6 ················································

5 ..... ···························
z *
4 ...
······*···~···*·
3 ························· ······················ ······*·············· ································································································

2 ..... *
z * AC/AC
Z AC/Fractured PCC

0 2 4 5 6 7 8 9 10

AGENCY OL THICKNESS, IN
FIGURE 3 Comparison of AASHTO AC overlay thicknesses and agency
AC overlay thicknesses for fractured PCC pavements (95 percent
reliability).
Hall et al. 53

TABLE 3 Example AC Overlay and Bonded PCC Overlay Design for JRCP and JPCP
REVISED.CHAPTER 5 AASHTO DESIGN GUIDE OVERLAY DESIGN
========== -==-===================
MW-7 AC AND BONDED PCC OL OF EXISTING JRCP
==================== ===============================
EXISTING PAVEMENT DESIGN AND FUTURE TRAFFIC
Slab thickness 10.00 (in)
Future design lane ESALs = 10000000 (10 YEARS)
============--======--
BACKCALCULATION OF Kef f AND Ee
INPUT INPUT INPUT INPUT INPUT RADIUS
LOAD DO Dl2 024 036 AREA RELSTIFF Kdyn SLAB Ee
(lbs) (mils) (mils) (mils) (mils) (in) (in) (pci) (psi)
11144 4.39 3.97 3.49 3.01 30.51 36.16 239 4.8E+06
10864 4.90 4.57 4.18 3.70 31.96 45.36 133 6.6E+06
10928 4.51 4.09 3.69 3.14 30.88 38.12 206 5.1E+06
10824 4.SS 4.17 3.77 3.30 31.29 40.S8 179 S.7E+06
189 S.6E+06
================================================================~========
DETERMINE Of
Vary trial Of until computed ESALs equal future design ESALs.

INPUT INPUT INPUT INPUT INPUT INPUT

Kef f J Sc Pl P2 Ee So LOS Cd
(psi/in) (psi) (psi)
9S 3.S 730 4.S 2.S 5.6E+06 0.3S o.oo 1. 00

TRIAL COMPUTED
Of R z ESALs
(in) (millions)
8.70 so 0 9973718
9.70 80 0.84 10214S87
10.30 90 1.282 10619093
10.80 9S l.64S 108Sll07
11.60 99 2.327 1009S41S
DETERMINE Def£
INPUT Fjc = 0.97 (10 FAILURES/MI UNREPAIRED)
INPUT Ff at = 0.9S (SO MIDSLAB WORKING CRACKS)
INPUT Fdur = 1.00

Deff (in) = Fjc * Fdur * Ffat ~ Dexist = 9.22

=============================--===============================================
DETERMINE OVERLAY THICKNESS

RELIABILITY PCC BOL PCC to AC- AC OL

LEVEL THICK FACTOR THICK
so o.oo 0.00 o.oo
80 0.48 2.1s 1.04
90 1.09 2.07 2.24
9S 1.59 2.01 3.18
99 2.39 1.91 4.S6
=====--======================================================================

lected on the basis of the reported condition of the existing For this example, the overlay design procedures yield an
pavement. The ranges and recommended values for these AC overlay thickness of 3.3 in. at the 95 percent reliability
factors are described elsewhere (2). level. The state policy design is a 3.25-in. AC overlay for a
For each reliability level considered, the required bonded 10-year design life.
PCC overlay thickness is equal to the structural deficiency, The revised AASHTO overlay design procedures produce
obtained by subtracting the effective structural capacity Deft reasonable conventional AC overlay and bonded PCC overlay
froin the required future structural capacity Dt. The required thicknesses for jointed PCC pavements that are consistent
AC overlay thickness is equal to the bonded PCC overlay with state recommendations, as shown in Figure 4. The pro-
thickness multiplied by a factor that is a function of PCC cedures are also consistent with state recommendations in
thickness deficiency. This factor decreases as the PCC thick- identifying when no overlay is required for a pavement.
ness deficiency increases and so is different for each reliability Specific difficulties in AC and bonded PCC overlay thick-
level considered. ness design include the sensitivity of the J factor for load
54 TRANSPORTATION RESEARCH RECORD 1374

95 % REL 01 THICKNESS. IN
lOr-~~~~~~~~~~~~~~~~~~~~~~~~-,.,

9 ································································································· ························································

8 ....... ································

7 ...... .

6 ·························································································

5 ............... . *
4 .................................. *
···*····t·············*·······

* AC/JCP
0 Bonded PCC/JCP

5 6 7 8 9 10
AGENCY 01 THICKNESS, IN

FIGURE 4 Comparison of AASHTO and agency AC overlay thicknesses

and bonded PCC overlay thicknesses for JRCP and JPCP (95 percent
reliability).

transfer and the necessity of imposing practical minimum and lays, a reliability of 99 or greater is needed to match state
maximum values for the PCC elastic modulus, the PCC mod- recommendations. Figure 5 shows the comparison between
ulus of rupture, and the effective k value. design overlay thickness and agency recommendations for
The design reliability level is very significant. Most of the these levels of reliability.
projects were Interstate-type highways. A design reliability The examples illustrate the importance of condition data
level of 95 percent appears to be reasonable for AC overlays and deflection data for overlay design. The condition factor
of JRCP and JPCP. F;c, which indicates the amount of pavement deterioration left
A few examples yielded overlay thicknesses that appeared unrepaired before overlay, has a significant effect on the over-
to be excessive. These examples were located in the Southwest lay thickness requirement. Agencies will find that much
region, in a state with a very mild climate, which may have greater overlay thicknesses are required to meet desired per-
a significant effect on improving overlaid pavement perfor- formance lives if overlays are placed without adequate
mance and reducing overlay thickness requirements. This could preoverlay repair. Most agencies specified thorough repair
be addressed by using a lower design reliability level or by for the CRCP examples submitted.
using a lower J factor to determine Df. The design reliability level is very significant. Most of the
projects were Interstate-type highways. A design reliability
level of 95 percent appears to be reasonable for AC overlays.
EXAMPLE AC AND BONDED PCC OVERLAY Bonded PCC overlays appear to be designed at a 99 percent.
DESIGN FOR CRCP reliability level.

Table 4 gives an example of AC overlay and bonded PCC

overlay design for CRCP (MW-9). The state agency's design EXAMPLE AC OVERLAY DESIGN FOR ACIPCC
procedure indicates a 6.2-in. AC overlay is needed for this PAVEMENT
pavement. However, the state's policy design is a 3.25-in. AC
overlay. Table 5 gives an example AC overlay design for an existing
The design procedure for AC and bonded PCC overlays of AC/PCC pavement (MW-15). The AC modulus was deter-
CRCP is the same as for JPCP and JRCP. The key difference mined from diametral resilient modulus tests on AC cores
is that a lower J (load transfer) factor is needed to produce from the pavement, adjusted to account for the difference
a reasonable overlay thickness. The appropriate J factor also between the laboratory testing frequency and the FWD load-
seems to vary from state to state, so each agency needs to ing frequency. The resilient modulus tests at 70°F and 90°F
determine its own value for J. were used along with the deflection data to assign an ap-
The revised AASHTO overlay design procedures produce propriate AC mix temperature to each of the deflection
reasonable AC overlay and bonded PCC overlay thicknesses basins. Then, using the backcalculation procedure described
for CRCP consistent with state recommendations, provided elsewhere (2), the maximum deflection d0 and deflection
different reliability levels are used. For AC overlays, a reli- basin AREA of the PCC slab were computed and used to
ability level of 95 percent produces overlay thicknesses com- backcalculate the effective dynamic k value and PCC elastic
parable with state recommendations. For bonded PCC over- modulus.
Hall et al. 55

TABLE 4 Example AC Overlay and Bonded PCC Overlay Design for CRCP

REVISED CHAPTER S AASHTO DESIGN GUIDE OVERLAY DESIGN

MW-9·AC AND BONDED PCC OVERLAY OF EXISTING CRCP

EXISTING PAVEMENT DESIGN AND FUTURE TRAFFIC

------------------------
Slab thickness 8.00 (in)
Future design lane ESALs = 18000000 (S% ESAL GROWTH RATE, 10 YEARS)
=====================::=========
BACKCALCULATION OF Kef f AND Ee

INPUT INPUT INPUT INPUT INPUT RADIUS

LOAD DO Dl2 D24 D36 AREA RELSTIFF Kdyn SLAB Ee
(lbs) (mils) (mils) (mils) (mils) (in) (in) (pci) (psi)
9000 4.0S 3.60 3.04 2.48 29.3S 31.22 280 6.1E+06
9000 4.16 3.49 2.7 1.91 26.61 23.63 471 3.4E+06
9000 3.49 3.04 2.S9 2.14 29.04 30.12 349 6.6E+06
9000 S.29 4.84 4.16 3.38 30.2S 34.93 172 S.9E+06
318 S.SE+06
============================================================================
DETERMINE Df
Vary trial Df until computed ESALs equal future design ESALs.

INPUT INPUT INPUT INPUT INPUT INPUT

Kef f J Sc Pl P2 Ee So LOS Cd
(psi/in) (psi) (psi)
1S9 2.2 727 4.S 2.8 S.SE+06 0.3S o.oo 1.00

TRIAL COMPUTED
Df R z ESALs
(in) (millions)
7.40 so 0 177431SO
8.40 80 0.84 1881SS26
8.90 90 1.282 18776S3S
9.30 9S l.64S 18436227
10.10 99· 2.327 17988918
============================================================================
DETERMINE Def f
INPUT Fjc = 0.96
INPUT Ff at = 0.98
INPUT Fdur = 0.8S ("D" CRACKING)
Deff (in) = Fjc * Fdur * Ff at * Dexist 6.40
============================================================================
DETERMINE OVERLAY THICKNESS
RELIABILITY PCC BOL PCC to AC AC OL
LEVEL THICK FACTOR THICK
so 1.00 2.08 2.08
80 2.00 1.96 3.92
90 2.SO 1.90 4.76
9S 2.90 1.86 S.40
99 3.70 1. 79 6.63
=======----------=====--=--===========--===--=================

For PCC pavement with an existing AC overlay, the only decreases as the PCC thickness deficiency increases and so is
method for determining Deff is the condition survey method. different for each reliability level considered.
The joints and cracks condition factor Fie• durability factor Only five examples could be developed for AC overlays of
Fdun and AC quality factor Fae are selected on the basis of AC/PCC pavements, so it is difficult to judge the adequacy
available distress data. The ranges and recommended values of the design procedure. The limited results show that the
for these factors are described elsewhere (2). In computing revised AASHTO overlay design procedure produces rea-
the effective structural capacity of the existing AC/PCC pave- sonable second AC overlay thicknesses that are consistent
ment, the AC surface thickness is divided by a factor of two with state recommendations. The reliability level required to
to convert it to an equivalent thickness of PCC. match the state recommendations is variable, however. This
For each reliability level considered, the required AC over· is not too surprising since agencies have little performance
lay thickness is equal to the structural deficiency (obtained experience with second overlays.
by subtracting the effective structural capacity Deft from the All of the condition factors significantly affect overlay thick·
required future structural capacity Df) multiplied by a factor ness, indicating that the amount of pavement deterioration
that is a function of PCC thickness deficiency. This factor left unrepaired before overlay has a significant effect on the
56 TRANSPORTATION RESEARCH RECORD 1374

95 % REL OL THICKNESS~ IN
10.--~~~~~~~~~~~~~~~~~~~~~~~~~

9 ···························································································································································

8 ......................................................................................................................................... .

7 ··········································································*········································ ····················································

5 ···································································

3 ················································· ···············X·······································································································

* AC/CRCP
X Bonded PCC/CRCP
o--~~~~-'-~----'~~-'-~~'--~-'-~~-'--~----'~~-'-~--'

0 2 3 4 5 6 7 8 9 10
AGENCY OL THICKNESS. IN

FIGURE 5 Comparison of AASHTO and agency AC overlay thickness (95

percent reliability) and bonded PCC overlay thickness (99 percent
reliability).

overlay thickness requirement. Some existing AC/PCC pave- thicknesses that are consistent with state recommendations,
ments are very badly deteriorated due to PCC durability as shown in Figure 6 for a reliability level of 95 percent. Only
problems. six unbonded overlay design examples could be developed
The design reliability level is very significant. A design from the project data submitted.
reliability level of 90 to 95 percent appears to be reasonable The unbonded overlay thicknesses were obtained using the
for second AC overlays. original Corps of Engineers equation developed for airfields.
An improved design methodology can and should be devel-
oped in the future to replace this empirical equation.
EXAMPLE UNBONDED PCC OVERLAY DESIGN The design reliability level is very significa~t. Most of the
FOR PCC PAVEMENT projects were Interstate-type highways. A design reliability
level of 95 percent appears to be reasonable.
Table 6 gives an example unbonded PCC overlay design for
a PCC pavement (SW-19). The required future structural ca- EXAMPLE PCC OVERLAY DESIGN FOR AC
pacity Dt is determined using the PCC elastic modulus, PCC PAVEMENT
modulus of rupture, and J load transfer factor of the unbonded
overlay. The design static k value used to determine Df is ihe Table 7 gives an example PCC overlay design for an AC
backcalculated effective dynamic k value of the existing pave- pavement (SE-5). The required PCC overlay thickness is equal
ment, divided by a factor of two. to the required future structural capacity Dt. The design static
The effective structural capacity Deff of the existing pave- k value used to determine Dt is determined from the nomo-
ment is obtained by multiplying the existing slab thickness by graph in Part II of the guide, using the total thickness of the
the joints and cracks condition factor Ficu· For any given quan- existing pavement layers and the subgrade resilient modulus
tity of unrepaired deteriorated joints and cracks per mile, the and effective pavement modulus backcalculated from deflec-
Ficu factor makes a smaller adjustment to the slab thickness tions measured on the existing AC pavement.
than the Fie factor, which is used for bonded PCC and AC For the example AC pavement, the state's design method
overlay design, because unbonded overlays are much less sen- indicated that a 6.4-in. PCC overlay was needed. The state
sitive to deteriorated joints and cracks in the existing slab constructed experimental sections of 6, 7, and 8 in. State
than these other overlay types. recommendations were not available for the other PCC/AC
.For each reliability level considered, the unbonded PCC examples developed .
overlay thickness required is the square root of the difference The sensitivity of PCC overlay thickness to k value is small,
between the square of Dt and the square of Deff· For the as illustrated for one example project:
example pavement, the overlay design procedure yields 8.0
PCC overlay
in. at the 90 percent reliability level and 8.7 in. at the 95 thickness (in.)
percent reliability level. The state's design procedure indicates k value (psi/in.) (R = 90 percent)
that an 8-in. unbonded PCC overlay is needed. 147 9.9
Overall, it appears that the revised AASHTO overlay de- 147 * 2 = 294 9.5
sign procedures produce reasonable unbonded PCC overlay 147 * 4 = 588 9.0
TABLES Example AC Overlay Design for AC/PCC Pavement

REVISED CHAPTER S AASHTO DESIGN GUIDE OVERLAY DESIGN

======--=======================
MW-lS AC OVERLAY OF EXISTING AC/JRCP (I-74)
===========================----==============
EXISTING PAVEMENT DESIGN AND FUTURE TRAFFIC
AC layer thickness 3.00 (in)
Slab thickness 10.00 (in)
Future design lane ESALs = 10000000 (20 years)
- - ====================================
BACKCALCULATION OF Kef f AND Ee
AC temp (deg F)
AC modulus 1,626,000 (psi) from lab tests of cores
AC/PCC 0 (0 for bonded, 1 for unbonded)

INPUT INPUT INPUT INPUT INPUT AC PCC PCC RADIUS

LOAD DO 012 024 036 AREA DO AREA RELSTIFF Kdyn SLAB Ee
(lbs) (mils) (mils) (mils) (mils) (in) (mils) (in) (in) (pci) (psi)
9000 S.19 3.99 3.40 2.79 26.31 S.14 26.49 23.39 389 l.4E+06
9000 3.82 3.20 2.8S 2.38 28.74 3. 77 29.02 30.06 324 3.1E+06
9000 4.0S 3.SO 3.09 2.6S 29.4S 4.00 29.72 32.6S 2S9 3.SE+06
9000 3.84 3.19 2.80 2.41 28.48 3.79 28.76 29.19 341 2.9E+06

328 2.7E+06
=====--=======--==========================================================~===========
DETERMINE Of
Vary trial Of until computed ESALs equal future design ESALs.
INPUT INPUT INPUT INPUT INPUT INPUT
Kef f J Sc Pl P2 Ee So LOS Cd
(psi/in) (psi) (psi)
164 3.2 606 4.S 2.S 2.7E+06 0.39 o.oo 1.00
TRIAL COMPUTED
Of R z ESALs
(in) (millions)
8.S9 so 0 10066278
9.73 80 0.84 10036274
10.37 90 1.282 1003670S
10.92 9S l.64S 10034620
12.02 99 2.327 10048S32
====================================================================================
DETERMINE Def f
INPUT Fjc 0.90 (SO unrepaired areas/mile)
INPUT Fdur = 0.90 (localized failures from "D" cracking)
INPUT Fae 0.9S (fair AC mixture)
Thickness of AC to be milled 0. SO (in)
Dae = original Dae - milled Dae = 2. SO (in)

Deff = (Fjc*Fdur*Dexist) + (Fac•Dac/2.0) = 9.29 (in)

====================================================================================
DETERMINE OVERLAY THICKNESS

RELIABILITY PCC BOL PCC to AC AC OL

LEVEL THICK FACTOR THICK
so o.oo o.oo o.oo
80 0.44 2.16 0.9S
90 1.08 2.07 2.24
9S 1.63 2.00 3.26
99 2.73 1.88 S.13
==================================--~=============================================
TABLE 6 Example Unbonded PCC Overlay Design for PCC Pavement

REVISED CHAPTER 5 AASHTO DESIGN GUIDE OVERLAY DESIGN

===========================================================================
SW-19 UNBONDED JPCP OVERLAY OF JPCP
===========================================================================
EXISTING PAVEMENT DESIGN AND FUTURE TRAFFIC
slab thickness 8.20 (in)
Future design lane ESALs = 11000000
=========================================.==================================
BACKCALCULATION OF Kef f

INPUT INPUT INPUT INPUT INPUT RADIUS

LOAD DO 012 024 036 AREA RELSTIFF Kdyn SLAB Ee
(lbs) (mils) (mils) (mils) (mils) (in) (in) (pci) (psi)
9144 3.89 3.37 2.85 2.40 28.89 29.62 329 5.4E+06
9088 3.89 3.33 2.81 2.31 28.50 28.40 355 4.9E+06
9104 3.94 3.33 2.81 2.36 28.29 27.78. 366 4.6E+06
9128 3.94 3.42 2.85 2.40 28.75 29.17 334 5.1E+06
346 5.0E+06
===========================================================================
DETERMINE Of
unbonded overlay modulus of rupture (psi) = 700
Unbonded overlay modulus of elasticity (psi) = 4900000
Vary trial Of until computed ESALs equal future design ESALs.

INPUT INPUT INPUT INPUT INPUT INPUT

Kef f J Sc Pl P2 Ee So LOS Cd
(psi/in) (psi) (psi)
173 4.0 700 4.5 2.5 4900000 0.35 0.00 1.00

TRIAL COMPUTED
Of R z ESALs
(in) (millions)
9.40 50 0 10972879
10. so- 80 0.84 11282235
11.10 90 1.282 11337203
11.60 95 1.645 11285326
12.60 99 2.327 11235624
===========================================================================
DETERMINE Def f
INPUT Fjc = 0.94 (assume 100 deteriorated transverse cracks/mi)
INPUT Fdur = 1.00
Deff (in) = Fjc * Fdur * Dexist = 7.71
===========================================================================
DETERMINE OVERLAY THICKNESS
RELIABILITY UBOL
LEVEL THICK
50 5.38
80 7.13
90 7.99
95 8.67
99 9.97
===========================================================================
 95 % REL 01 THICKNESS, IN
10 .

9 ········································································································································~·············· .............. .

8 ...... . ············8··············

6 ······································································································ ··························· ········································

5 .

4 ................................. .

3 ........ ······································· ···················································

......, 6. Unbonded PCC/PCC 1···

0 2 3 4 5 6 8 9 10
AGENCY OL THICKNESS, IN

FIGURE 6 Comparison of AASHTO unbonded PCC overlay thicknesses

and agency unbonded PCC overlay thicknesses for PCC pavement (95
percent reliability).
TABLE 7 Example PCC Overlay Design for AC Pavement

REVISED CHAPTER S AASBTO DESIGN GUIDE OVERLAY DESIGN

==================================================================================
SE-S JPCP OVERLAY OF AC PAVEMENT (US l)
==================================================================================
EXISTING PAVEMENT DESIGN AND FUTURE TRAFFIC
EXISTING PAVEMENT DESIGN
AC SURFACE 2.00 SUBGRADE: SAND
CR STONE BASE a.so
SUBBASB 12.00
TOTAL THICKNESS 22.so
Future design lane ESALs 1100000
===================================================================z===============
DETERMINE Kef f
Vary Ep/Mr until actual MR*DO/P matches computed MR*DO/p.

SUBGRADE ACTUAL TRIAL COMPUTED

STATION LOAD DO,in Dr,in MR MR*DO/P Ep/MR MR*DO/Ep Ep
(lbs) (mils) (mils) (psi)
9000 12.96 1.86 24604 3S.43 0.80 3S.63 19683
r = 47.2
Check r > 0.7 ae lS.19
Using Figure 3.3, Part II:
Keff (dynamic) = 1200 psi/in INPUT
Keff (static) = 600 psi/in
===================================================================================
DETERMINE Of
INPUT
PCC overlay modulus of rupture (psi) 63S (mean)
PCC overlay modulus of elasticity (psi) 4000000 (mean)
Vary trial Of until computed ESALs equal future design ESALs.
INPUT INPUT INPUT INPUT INPUT INPUT
Kef f J Sc Pl P2 Ee So LOS Cd
(pci) (psi) (psi)
600 3.2 63S 4.2 2.S 4000000 0.3S o.oo 1.00
TRIAL COMPUTED
Df R z ESALs Dol
(in) (millions) (in)
3.80 so 0 1173786 3.80
S.30 80 0.84 1127398 S.30
S.90 90 1.282 1114201 S.90
6.40 9S 1.64S 1108802 6.40
7.40 99 2.327 1162870 7.40
==================================================================================
60 TRANSPORTATION RESEARCH RECORD 1374

Additional work is needed to investigate effective k values panel members for their guidance and assistance and the many
for PCC overlays of AC pavements, including deflection test- state highway agency personnel who provided the project data
ing on in-service PCC/AC pavements. and participated in the development of the examples.
The design reliability level is very significant. Most of the
projects were Interstate-type highways. A design reliability
level of 95 percent appears to be reasonable for most projects. REFERENCES

1. Guide for Design of Pavement Structures. American Association

CONCLUSIONS of State Highway and Transportation Officials, Washington, D .C.,
1986.
2. M. I. Darter, R. P. Elliott, and K. T. Hall. Revision of AASHTO
Overall, the revised AASHTO overlay design procedures yield Pavement Overlay Design Procedures. Draft Final Report, NCHRP
reasonable overlay thicknesses that are consistent with state Project 20-7/Task 39, June 1991.
highway agency recommendations. The following major points 3. M. I. Darter, R. P. Elliott, and K. T. Hall. Revision of AASHTO
Pavement Overlay Design Procedures, Appendix: Documentation
are made in regard to the field testing of the procedures. of Design Procedures. NCHRP Project 20-7/Task 39, June 1991.
Reliability level has a large effect on overlay thickness. On 4. M. I. Darter, R. P. Elliott, and K. T .. Hall. Revision of AASHTO
the basis of the examples developed, it appears that a design Pavement Overlay Design Procedures, Appendix: Overlay Design
reliability level of approximately 95 percent gives thicknesses Examples. NCHRP Project 20-7/Task 39, June 1991.
comparable with those recommended for most projects by the
state agencies. Exceptions.to this are bonded PCC overlays,
which appear to be designed for a somewhat higher structural DISCUSSION
reliability. There are, of course, many situations for which it
is desirable to design at a higher or lower level of reliability. T. F. FWA
Some overlay projects were designed for huge traffic load- Center for Transportation Research, Faculty of Engineering, National
ings (more than 25 million ESALs). Whereas thick concrete University of Singapore.
overlays should be able to handle this level of traffic, thick
This paper provides comparisons between overlay designs by
AC overlays may rut before their structural design life is
a revised AASHTO approach and state highway agency rec-
achieved.
ommendations. Among the conclusions drawn are (a) the
Designing AC overlays by the NDT method and designing
revised AASHTO overlay design procedures yield reasonable
AC overlays by the co.ndition method produced similar re-
overlay thicknesses which are consistent with state highway
sults. However, the NDT method is believed to be the more
agency recommendations, (b) the reliability level has a large
accurate method and is highly recommended. The condition
effect on overlay thickness (a design reliability level of 95
survey method, coupled with materials testing, can be de-
percent appears to be reasonable for most projects), and (c)
veloped to give adequate results.
because of climatic and geographic differences, each agency
For the very few example projects for which past traffic
will need to test the procedures on its pavements and deter-
data were available, the remaining life method produced over-
mine their reasonableness and required adjustments. This dis-
lay thicknesses comparable with those produced by the NDT
cussion highlights an important weakness in the proposed
and condition survey methods. However, the remaining life
revised procedures, recommends the use of the original 1986
method has some very significant limitations (2) and should
AASHTO overlay equation with a corrected formufa for re-
be used with caution. Perhaps the most significant limitation
maining life factor FRL, and suggests procedures for using field
of the remaining life method is that it cannot take into account
tests to calibrate overlay design equations that include re-
the benefit of preoverlay repair.
maining life considerations.
It is apparent from the field testing results that different
climatic and geographic regions require different overlay
thicknesses, even if all other design inputs are exactly the
same. The AASHTO guide does not provide a way to deal BASIC OVERLAY DESIGN EQUATION
with this problem. Therefore, each agency will need to test
the procedures on its pavements and determine their reason- Details of the authors' revisions to 1986 AASHTO Design
ableness and required adjustments. There are many ways to Guide (1) are given in a companion paper in this Record by
adjust the procedure to produce desired overlay thicknesses them. An important deviation of the proposed revised pro-
(e.g., reliability, resilient modulus, J factor, etc.). cedures from the 1986 AASHTO Guide is the reverting to
the use of traditional overlay equation SC0 L = SCy - SCxeff
and discarding the original remaining life concept that intro-
ACKNOWLEDGMENT duced a remaining life factor FRL in the overlay equation. In
a discussion of the companion paper, the following comments
This work was sponsored by the American Association of were made: (a) the authors' recommendation to set FRL =
State Highway and Transportation Officials in cooperation 1.0, thereby reducing the 1986 AASHTO overlay equation
with the Federal Highway Administration and was conducted to the traditional overlay equation, was based on an analysis
in the National Cooperative Highway Research Program, which that had a restrictive and weakly founded assumption on over-
is administered by the Transportation Research Board of the lay performance and an incorrect procedure of computing
National Research Council. The authors gratefully acknowl- FRL, (b) the traditional overlay equation yields a lower bound
edge the NCHRP staff and the NCHRP Project 20-7/Task 39 overlay solution that underdesigns and is unconservative, and
Hall et al. 61

(c) the traditional overlay equation is conceptually unsound some changes in the conclusion about the level of reliability
and inadequate because overlay thickness is derived on the are expected.
basis of the overlay requirement at the time of oyerlay ap-
plication. It does not include an analysis to examine whether
the overlay provided is adequate during other stages of over- SUMMARY
lay service life.
Esa (2) and Fwa (3) separately confirmed the fundamental The original 1986 AASHTO overlay equation with remaining
correctness of the AASHTO overlay design approach that life factor FRL should be used instead of the traditional overlay
incorporates the concept of remaining life, and proposed dif- equation. The AASHTO remaining life approach is funda-
ferent procedures to eliminate inconsistencies caused by a flaw mentally correct and technically sound, and recent studies
in the FRL calculation. Both identified the case with FR.L = show that it yields meaningful and consistent results. Field
1.0 to be the lower bound overlay solution that assumes the tests reported in the paper can be used to calibrate overlay
rate of structural deterioration of an old pavement after over- equations for design procedures that incorporate the AASHTO
lay to be the same as that of a new pavement. Computation- concept of remaining life.
ally, the traditional overlay equation is the same as this lower
bound solution. The upper bound solution is one with FRL s;
1.0 where the old pavement after overlay is assumed to con- REFERENCES
tinue to deteriorate at a rate as if no overlay were applied.
Easa adopted the original 1986 AASHTO design equations 1. AASHTO Guide for Design of Pavement Structures (Vols. 1and2).
as the upper bound solution. He proposed using a linear com- American Society of State Highway and Transportation Officials,
Washington, D.C., 1986.
bination of the lower and upper bound solutions (by mean_s 2. S. M. Easa, Extension of AASHTO Remaining-Life Methodology
of weighting factors 'A. and 1 - 'A., 'A. s; 1) to represent actual of Overlay Design. In Transportation Research Record 1272, TRB,
overlay performance and considered those 'A. values that pro- National Research Council, Washington, D.C., 1990.
duced consistent designs as feasible solutions. 3. T. F. Fwa. Remaining-Life Consideration in Pavement Overlay
Design. Journal of Transportation Engineering, Vol. 117, No. 6,
Fwa (3) identified the flaw in the AASHTO formula for 1991.
computing FRL and derived a new FRL expression in accor-
dance with the concept of remaining life. When substituted
into the 1986 AASHTO overlay equation, consistent results AUTHORS' CLOSURE
are obtained, and these represent the upper bound overlay
solutions. Since the actual overlay requirement lies between The remaining life concept has not been discarded in the
the lower and upper bounds, a linear combination of the two proposed revisions to the AASHTO overlay design proce-
solutions (by means of weighting factors a and 1 - a, a s; dures. As described in the paper, three procedures are given
1) was proposed. Since both the lower and upper bound so- for estimating the effective structural capacity of an existing
lutions produce consistent overlay designs, the full range of pavement: a deflection-based approach, a condition survey
a values give feasible overlay solutions. approach, and a remaining life approach.
The lower and upper bound analyses performed by Easa The basic concept of remaining life is that a pavement's
and Fwa provide a rational basis for overlay design that in- past traffic and its total traffic-bearing capacity over its life-
corporates the fundamentally correct remaining life co~cept time may be used together to estimate the traffic the pavement
of AASHTO. Their studies also show that "the appropriate is capable of carrying for the remainder of its life. This concept
value of FRL is 1.0" (see the companion paper) is not a valid did not originate with the 1986 AASHTO Guide, but it has
claim. In the light of these findings, it appears logical for the been used in pavement evaluation for many years and is appli-
authors to reconsider their decision to discard the 1986 cable to any pavement design procedure based on a relation-
AASHTO overlay equation. ship between traffic and loss of structural capacity. Indeed,
this concept is intrinsic to the AASHTO design methodology.
The authors consider the basic remaining life concept to be
USE OF FIELD TEST DATA valid. However, the application of this concept in the pro-
posed revisions to the AASHTO overlay design procedures
It is interesting to note that Easa (2) and Fwa (3) independent- differs from the application presented in the 1986 guide.
ly proposed very similar concepts of representing the dete- In the 1986 guide's overlay design procedures, procedures
rioration of existing pavements after being overlaid, although were given for determining the effective structural capacity
their methods differ in the way the upper bound solutions are (SCeff) of a pavement from deflection testing or distress ob-
derived. Both methods contain a weighting parameter that servations. This effective structural capacity is expected to be
requires calibration using field performance data. The field less than the original structural capacity of the pavement when
tests reported in the paper offer a good opportunity for this new (SN0). However, the 1986 guide's overlay design pro-
purpose. cedures then applied a traffic-based remaining life factor as
As far as the three conclusions of the paper cited at the a multiplier to the effective structural capacity determined
beginning of this discussion are concerned, it is believed that from deflections or distress observations. This approach is
they would still hold because as Easa and Fwa have illustrated widely considered to penalize a pavement twice for the same
in their studies, the trends of variations of the feasible solu- past traffic.
tions are similar to the trend of the lower bound case (which Fwa has defended this double penalty with the reasoning
is the solution given by the authors in the paper). However, that if a deteriorated pavement with a given effective struc-
62 TRANSPORTATION RESEARCH RECORD 1374

tural capacity is overlaid, it will subsequently deteriorate at based and condition-based estimates of the effective structural
a faster rate than a newly constructed pavement of the same capacity. In concept, these three approaches for estimating
structural capacity that receives.the same thickness of overlay. sceff shoultl yield similar results.
This is a considerable distortion of the structural deficiency In addition to the conceptual flaw described earlier, the
concept of overlay design. The essence of the structural de- 1986 guide's remaining life computation was considered to be
ficiency concept is that a performance prediction model may needlessly complex and poorly supported. For example, the
be used to determine a required overlay, which will increase procedure did not address the practical significance of a "neg-
an in-service pavement's effective structural capacity to a ative remaining life" computed for an in-service pavement.
structural capacity sufficient to carry the traffic expected over The need to revise the application of the remaining life con-
the design period. The rate of deterioration of the overlaid cept in the 1986 guide's overlay design procedures was iden-
pavement is thus predicted by the performance model used, tified by the AASHTO Joint Task Force on Pavements as one
just as is the rate of deterioration predicted for new pavements of the high-priority revisions to the overlay design procedures.
by the same model. Within the context of the AASHTO The authors have examined the work by Fwa and by Easa
design methodology, the flexible and rigid pavement perfor- and have concluded that although they offer modifications to
mance models presented in Part II of the guide are used to the remaining life method as presented in the 1986 guide, they
determine required future structural capacity (structural num- do not correct its major flaw. They also impose needless com-
ber or slab thickness), and the rate of deterioration is· mea- plexity in the application of a simple concept.
sured by loss of serviceability as predicted by these models. The authors have therefore recommended to the Design
If the two pavements described by Fwa have the same struc- Subcommittee of the AASHTO Joint Task Force on Pave-
tural capacity before overlay, and receive the same overlay, ments that the method developed by Elliott for considering
then according to the structural deficiency concept their per- remaining life be accepted as the best solution to the problems
formance after overlay will be the same. One cannot correctly associated with the application of this concept in the 1986
apply the structural deficiency concept of overlay design and overlay design procedures. It must also be clarified that de-
at the same time conjecture a rate of deterioration of the cisions concerning acceptance of this and other proposed re-
overlaid pavement other than the rate predicted by the perfor- visions to the overlay design procedures are made not by the
mance model used to define the structural deficiency. authors but rather by the AASHTO Joint Task Force.
In the proposed revisions to the overlay design procedures,
The opinions expressed in this paper are those of the authors and not
a traffic-based estimate of remaining life is applied to a pave- necessarily those of AASHTO, FHWA, NCHRP, or TRB or of the
ment's original structural capacity (SC0 ) to estimate its current individual states participating in the National Cooperative Highway
effective structural capacity but is not applied to deflection- Research Program.
TRANSPORTATION RESEARCHRECORD 1374 63

Overlay Design Procedure for Pavement

Maintenance Management Systems
ARIEH SIDESS, HAIM BONJACK, AND GABRIEL ZOLTAN

A methodology for flexible pavement rehabilitation and devel- erties derivation of the pavement layers by empirical analysis,
opment of overlay thickness design curves for pavement main- using maximum deflection (DMD), surface curvature index
tenance management systems (PMMSs) is presented. The meth- (SCI), and base curvature index (BCI). (5) or structural anal-
odology is based on nondestructive testing of deflection basin ysis, based on the multilayer theory, which includes elastic
measurements and on the rational approach that characterizes
pavement response to major deterioration criteria, such as fatigue moduli derivation by backcalculation programs (7-9); and (c)
and rutting. Within the general framework, subgrade and pave- determining the relation between layer material properties
ment were classified into three categories of strength: weak, me- and pavement performance in terms of pavement life by means
dium, and strong according to the measured deflection D 6 (at a of empirical deflection models (2-6,16,19) or semimechan-
distance of 1.80 m from the loading plate) and the surface cur- istic models to predict pavement response and its resistance
vature index parameter. With reference to these categories, a to deterioration criteria such as fatigue and rutting (7-9).
structural index (SI) was defined. Between the SI and overlay
thickness there is a dependence that can be expressed by a design In pavement maintenance management systems (PMMSs)
curves system related to different traffic categories and subgrade the rehabilitation solutions are subsystems within the whole
type. The present methodology can readily be incorporated as a system where additional indices exist, such as visual deteri-
subsystem within the general PMMS, and it enables fast solutions oration, pavement condition index (PCI), distress rating (DR),
at the network level for economic evaluation and rehabilitation and roughness, for the purpose of economic evaluation and
priority order determination of extensive road systems. rehabilitation priority determination. Such a systems appli-
cation for road system management of hundreds of kilometers
There are several main approaches to determining the overlay requires fast and reliable representation of rehabilitation so-
thickness of rehabilitated flexible pavements: (a) engineering lutions at the network level. In this case the rational ap-
experience and judgment, which in many cases is still being proaches, although reliable and advantageous, do not apply,
used to design overlays (J); (b) the standard overlay thickness because of the amount of work required by the tests to de-
for a given existing pavement type, traffic level, and other termine layer thickness, layer moduli derivation, and stresses
factors; (c) the empirical approach, which is based on the and strains calculation to determine the rehabilitated pave-
limited deflection criteria or correlation between the elastic ment resistance to deterioration criteria.
deflection and pavement life (2-6); (d) the rational approach, This paper presents a methodology for the design of flexible
which is based on the major deterioration criteria, such as pavement rehabilitation and development of overlay thickness
fatigue and rutting [in this approach the pavement perfor- design curves for PMMS. The methodology is based on NDT
mance and its load response is expressed in terms of strains deflection basin measurements and on the rational approach
and stresses (7-9)]; and (e) the mechanistic approach, which for predicting the main deterioration criteria of fatigue and
is based on principles of fracture mechanics, to which the rutting. The rationale behind this inethodology is that it is
Paris law (JO) is being applied that associates crack propa- possible to characterize and express structural strength of the
gation with the stress intensity factor (11-15). The first two rehabilitated pavements by means of an index called structural
methods do not require evaluation tests of pavement materials index (SI), which is determined according to the measured
and therefore are faster but not reliable. By contrast, the other deflection basins. Between this index and the required overlay
methods demand destructive and nondestructive testing (NDT) thickness there is a mutual dependence, which can be ex-
for the evaluation of the properties and the performance of pressed by curve systems for traffic categories and different
the pavement being rehabilitated. subgrade types. The advantage of this method is that it can
The use of NDT is more common and widespread and readily be incorporated as a subsystem within the overall PMMS
nowadays is an integral part of the overlay design procedures to furnish reliable and fast solutions at the network level of
applied in many institutions in the world (4,6-9,16-19). The extensive road systems for economic evaluation and rehabil-
design process carried out by these methods usually includes itation priority order determination as shown elsewhere (20,21).
the following stages: (a) deflection basin measurements and
their correction according to standards of load level, tem- EVALUATION PARAMETERS AS INPUT DATA
perature, load frequency, and so forth (1,2,16-18); (b) prop-
The rehabilitation design methodology and overlay thickness
design curves development are based on a data base of de-
A. Sidess and G. Zoltan, Yariv Civil Engineering, 1 Remez St.,
Givataim 53242, Israel. H. Bonjack. Technology and Management, flection basin measurements in different rehabilitated roads
Moshav Shoeva, 59 D. N. Harey Yehuda, Israel. in Israel totaling 450 km in length and moduli derivation
;t
of
64 TRANSPORTATION RESEARCH RECORD 1374

the layers by means of backcalculation programs. The mea- the narrow load range and close to the reference load, the
sured sections represent a wide range of pavement thicknesses linear relation adopted in Equation 1 is precise.
and subgrade types. The deflection basins were measured at The correction function of the central deflection, D 0 ac-
100-m intervals by a Dynatest FWD 8002 model, which de- cording to standard temperature of 30°C, was based on nu-
livers an impulse load of 7 to 110.kN (1.15 to 24.2 kips). In merical analysis in which the modulus of the asphalt layer was
each basin seven deflections were measured under an average determined in the temperature range 10°C to 50°C according
load of about 75 kN (within the range of 70 to 80 kN) dis- to the relationship recommended by Uzan et al. (25). This
tributed uniformly on a circular load plate 300 mm in diam- function was corrected according to the temperature range
eter. The deflections were measured at a distance of 0, 0.30, 10°C to 20°C (the temperature range in which the deflection
0.60, 0.90, 1.20, 1.50, and 1.80 m (D 0 , D 1 , .•. , D 6 , respec- basins were measured) by means of the measured deflection
tively) from the center of the load plate. At the same time basin and moduli derivation of the pavement layers. The
the asphalt layer temperature of each section was measured. regression function for the central deflection correction, FT(D0 ) ,
Since measurements were performed at the end of the winter, is expressed as follows:
temperatures ranged between 9°C and 22°C.
The moduli derivation of the pavement layers and the FT(D 0 ) = 1.694 - 3.155 X 10- 2 TP + 3.286
subgrade (the three-layer model) was carried out by a back-
calculation program, DEFMOD (20). This program derives x 10- 4 ~ - 1.667 x 10- 6 ~ (2)
the moduli according to the data base approach (22) similar
where T P is the asphalt layer temperature at the time of mea-
to MODULUS (23). The e~aluation results of this software surement in degrees Celsius.
were compared with results of BISDEF (23,24), MODULUS
This correction was carried out only on the central deflec-
(23,24) and ELMOD (22) and were found to be correct and
tion D 0 (in addition to the load correction) of all the deflection
reliable. The modulus of the asphalt layer, which was derived basins as follows:
from the deflection basin, was corrected to fit the standard
temperature of 30°C (86°F), according to the relationship pro-
(3)
posed by Uzan et al. (25).
The measured deflections and layer moduli derived from The central deflection is highly sensitive to variations of tem-
them were used as input to develop a structural index for the perature, whereas the other deflections are not affected by
pavement and overlay thickness design curves according to the temperature or the effect is negligible (25).
this index.

Subgrade Characterization and Classification

PROPOSED APPROACH FOR OVERLAY DESIGN
PROCEDURE One approach to subgrade quantitative and qualitative eval-
uation from NDT measurements is by means of empirical
NDT Correction to Fit Standard Conditions indices, such as spreadability (18), SCI (5), BCI (5), and basin
area (8). Figure 1 shows the relation between the modulus of
The development of overlay design curves according to the the subgrade Es derived by DEFMOD backcalculation and
proposed approach must be based on standard condition mea- the measured deflection D 6 , corrected to standard load of 75
surements. Therefore, the deflections were corrected by two kN (see Equation 1). The figure shows that irrespective of
factors for load level and temperature. the subgrade type or pavement thickness, there is a correlation
All the FWD measurements were performed within the between the subgrade modulus and D 6 that allows the char-
load range 70 to 80 kN. Therefore, all the deflection basins acterization and classification of the subgrade by that deflec-
were corrected according to the standard load of 75 kN in a tion. This fact is well known and is commonly used for subgrade
linear manner as follows: modulus derivation (19). The regression that fits the rela-
tionship shown in Figure 1 is as follows:

(1) D6 = 8.588 x 103 Es -i.o55 (R 2 = 0.961) (4)

where Es is expressed in MPa and D 6 in µm.

where
On the basis of local experience, the variety of subgrade
Dr0 r = corrected deflection for the standard load of 75 kN materials, and Equation 4, the subgrade was classified ac-
for the ith sensor, cording to the recommended methodology into three quan-
Dr = measured deflection at Pm load for the ith sensor, titative and qualitative categories-weak, medium, and strong
and subgrade, as indicated in Table 1. Categories and values sim-
Pm = load at the measurement time. ilar to those given in Table 1 were proposed by Scullion (26).
Linear modification of Scullion's results for standard load of
Because of the nonlinear behavior of the pavement layers, 75 kN leads to 93 µm and above for weak subgrade, 56 to 93
the measured deflection versus load level is also nonlinear. µm for medium subgrade, and below 56 µm for a strong
However, since most measurements were performed within subgrade.
 Sidess et al. 65

-...
c
0
200
180 0 = 8.588 x 103 Es1·
055

g
u
160
SCI = 2.533 x 105 Ep1'099
CID 140 06 ~ 105 micron

0
e 120
~ 1,000
-
T""

ca
c
0
100
80
+:< 60
u
Cl)
;;:::

-
Cl)
Cl) 40 u
c ca
.::J..
2025 50 75 100 125 150 175 200 en 1oqoo 1,000
Subgrade Modulus, Es (MPa)
Equivalent Modulus, Ep (MPa)
FIGURE 1 D6 deflection versus subgrade modulus E•.
FIGURE 2 SCI versus equivalent modulus EP for weak
subgrade.
Pavement Characterization and Classification

The structural condition of the pavement is a combination of tween EP and SCI expressed by means of the following equa-
the thickness and strength of its l~yers. Therefo:r:e, a quan- tions:
titative and qualitative characterization of the pavement ac- For a weak subgrade (D 6 ;:.::: 105 µm),
cording to NDT measurements requires the presentation of
the above factors by means of a single index. For that purpose SCI = 2.533 x 105 Ep -i.o55 (R2 = 0.936) (7)
the SCI parameter and equivalent pavement modulus EP ex-
pressed by the following equations were adopted: For a medium subgrade (55 :s; D 6 < 105 µm),

(5) SCI = 3.425 x 105 Ep - u 52

(R2 = 0.902) (8)

For a strong subgrade (D 6 < 55 µm),

(6)
SCI = 1.974 x 105 EP -i.o52 (R2 = 0.890) (9)
where
where EP is expressed in MPa and SCI in µm.
D0 = central deflection corrected to standard load of On the basis of local experience and a variety of materials,
75 kN and temperature of 30°C; . the pavement was classified according to the following three
D 1 = measured deflection at 0.3 m from the loading categories: weak pavement, EP :s; 200 MPa; medium pave-
plate corrected to the standard load of 75 kN;
EA, E 0 = moduli of the asphalt (at standard temperature
of 30°C) and the granular layers, respectively;
and
hA, h 0 = thicknesses of the asphalt and granular layers, 5
respectively. SCI = 3.425 x 10 Ep1"152
Figures 2 through 4 show the relation between the equiv- 55 ~ 06 < 105 micron
alent modulus EP derived from the deflection basin analysis
and the SCI values for the three subgrade types, characterized
respectively. The figures show that there is a correlation be- 1,000

TABLE 1 Subgrade Classification Based on NDT

Elastic Db
Subgrade 8
Modulus C.B.R 6
Classification (MPa) (%) (micron)
Weak ~ 65 40 4 105
Medium 65-120 4-8
:;;!!:
55 -105
1oqoo 1,000
Strong > 120 >8 < 55 Equivalent Modulus, Ep (MPa)
a
bBased on the correlation Es=(15-16)CBR. FIGURE 3 SCI versus equivalent modulus EP for medium
According to Eq. 4. subgrade.
66 TRANSPORTATION RESEARCH RECORD 1374

'2 10,000
0 11.... 1.0
CJ 5 -1.052 06 ~ 105 micron
SCI= 1.974 x 10 Ep 3 0.9
§, CJ)
0.8
0 06 < 55 micron ><
Cl>
CJ)
"C
0.7
><~
Cl>
.E o.~
"C ca 0.5
.E
-
11....
1,000 ::s
Cl> 0.4
-
()

-
11....
::s ::s
ca
11....
CJ)
0.3
>
11.... 0.2
::s
CJ 0.1
Cl>
0
·~oo
CJ
ca 300 400 500 600 700 800 900
't:
::s Surface Curvature Index, SCI (micron)
CJ) 1oqoo 1,000
Equivalent Modulus, Ep (MPa) FIGURES Siw as a function of SCI for weak subgrade.

FIGURE 4 SCI versus equivalent modulus EP for strong

subgrade. remaining life in terms of number of load application to fail-
ure, Nt according to the proposed criterion, is 30 percent
relative to pavement of identical thickness with an SI of 1.0.
ment, 200 MPa < EP :::;; 400 MPa; and strong pavement, EP > 3. The remaining life of the actual pavement layers (the
400 MPa. three-layer model) is determined according to the criterion
Classification of the pavement according to the measured of compressive strain at the top of the subgrade. For that
SCI determined by Equations 7 through 9 leads to identical purpose, the model of Verstraten et al. (27) was adopted.
results irrespective of the subgrade classification. Therefore, Since the SI refers to the measured deflection basin, the load
uniform criteria for pavement classification were determined that served for calculation was 75 kN distributed on a circular
as follows: weak pavement, SCI ;::::: 750 µm; medium pave- plate with a diameter of 300 mm. Accordingly, the SI is used
ment, 350 µm:::;; SCI< 750 µm; and strong pavement, SCI< as follows:
350 µm.
Table 2 gives the criterion values for both subgrade and SI = Nt(SCI > 350 µm)
pavement classification according to the correction deflection (10)
Nt(SCI = 350 µm)
basin measurements.
In Figures 5 through 7 the relationship between the SCI pa-
rameter and the SI for different subgrade categories is shown.
Structural Index The SCI of any pavement is determined according to Equa-
tions 7 through 9. These figures show that it is possible to
To express the pavement classification in a quantitative man- define the structural index depending on SCI by three distinct
ner, SI was adopted. This is an index within the range 0 to ranges as follows:
1.0 that expresses the performance condition of the rehabil-
itated pavement in terms of remaining life. SI determination 1. For SCI values smaller than 350 µm (strong pavement),
for any subgrade type was based on the following assumptions: the SI value equals 1.0. This range was determined by the
assumption.
1. For· any pavement in the strong category (SCI :::;; 350
µm), SI is equal to 1.0.
2. The SI of a pavement in the other categories expresses
1.0 55 ~ 06 < 105 micron
its remaining life relative to identical pavement with an SI of
1.0. For instance, an SI of 0.3 means that the pavement's :!ii 0.9
en>< 0.8
.g 0.7
TABLE 2 Subgrade and Pavement Classification Based on NDT c: 0.6
(micron)
~ 0.5
Pavement Subgrade Classification
-
::s
() 0.4
Classification

Weak
Weak
06 ~105
Medium
SSE; 0 <105
6
SCl;i.750
Strong
06<55
SCI ~750
-
::s
11....
CJ)
0.3
0.2
SCI ;i.750
0.1
06 ;;i.105 55< 06<105 06<55
0
·~oo
Medium
350~ SCI <750 350E; SCI <750 350E; SCI <1s() 300 400 500 600 700 800 900
06 ;;i:.105 55~ 06<105. 06<55 Surface Curvatur~ ll"'!dex, SCI (micron)
Strong
SCI <350 SCI <350 SCI <350
FIGURE 6 SIM as a function of SCI for medium subgrade.
 Sidess et al. 67

Overlay Thickness as a Function of SI

1.0 D6 < 55 micron
Cl) 0.9 The rational approach (7-9) for overlay thickness design re-
en 0.8 lies on the resistance of the rehabilitated pavement to the
><~
Q)
main deterioration criteria such as fatigue and rutting. The
0.7
"C SI indicates the strength of the pavement and provides eval-
.E 0.6 uation about the structural performance of the pavement. It
...
ii 0.5

-...
::::I
() 0.4
is therefore possible to present rehabilitation curves on the
basis of the rational approach depending on the SI index. The
-::::I
en
0.3
0.2
• development of overlay thickness design curves was based on
the following principles:
0.1 1. Several dozen structures within the weak, medium, and
0
·~oo 300 400 500 600 700 800 900
strong pavement categories with a wide range of SI were
adopted. The total pavement thickness was between 300 and
Surface Curvature Index, SCI (micron)
600 mm with asphalt thickness of 60 to 200 mm.
FIGURE 7 Sis as a function of SCI for strong subgrade. 2. The subgrade modulus (Es) representing the weak
subgrade groups (Es s 65 MPa) was determined as 650 MPa
(D6 = 115 µm according to Equation 4). The modulus repre-
2. For SCI values between 350 and 750 µm (medium pave- senting the medium subgrade groups (65 µm <Es s 120 µm)
ment), the SI values decrease in a nonlinear manner as the was determined as 90 MPa (D 6 = 75 µm). For the strong
SCI parameter increases. subgrade groups (Es> 120 MPa) the representative modulus
3. For SCI values greater than 750 µm (weak pavement), was determined as 120 MPa (D 6 = 55 µm).
the SI value is constant. 3. For each case the pavement and subgrade were classified
according to D 6 , and SCI and SI were calculated according
The range definitions for all the subgrade types were ex- to the principles demonstrated in previous sections.
pressed by the following regression equations. 4. A constant load distribution representing the rural mixed
For a weak subgrade (D 6 =::: 105 µm), traffic in Israel was adopted. The load range is between 20
and 180 kN for a single axle (20). The effect of that mixed
Siw traffic to determine overlay thickness was taken into account
according to the Miner hypothesis (28). After having calcu-
0.2 SCI=::: 750 µm
= 2.361 x 105 sCI- 2 · 112 (R2 = 0.987) 350 :s SCI< 750 µm lated the overlay thickness and to express the mixed traffic
{ levels in the overlay design curves (Figures 8 through 10) by
1.0 SCI< 350 µm
(11) means of a single traffic number (for easier utilization), the
various load levels were transformed to a number of equiv-
For a medium subgrade (55 s D 6 < 105 µm), alent applications of a 130-kN single-axle load, which is the
design axle load in Israel. The transformation was performed
SIM by means of the load equivalency factor of the AASHO method
(1).
0.25 SCI~ 750 µm 5. Two deterioration criteria were adopted, fatigue and
= 4.243 x 104SCI-t.819 (R 2 = 0.975) 350:sSCI<750µm compressive strain at the top of the subgrade, used as rutting
{
1.0 SCI< 350 µm criteria. The fatigue criterion was adopted according to the
(12) model proposed by Finn et al. (29) with some modification
referred to by Uzan and Gur (30). As a rutting criterion, the
For a strong subgrade (D 6 < 55 µm), model of Verstraten et al. (27) was adopted. The overlay
thickness for each rehabilitated pavement according to Min-
0.3 SCI=:::750µm er's law is the thickness that fits the critical criterion between
Sis= 1.046 x 104 SCI-i.sso (R2 = 0.988) 350sSCI <750µm the two.
{
1.0 SCI<350µm
Figures 8 through 10 show overlay thickness design curves
(13) depending on SI for a weak, medium, and strong subgrade,
respectively. The overlay thickness is presented for various
There is no relation between the numerical value of SI in equivalent 130-kN load applications. The required overlay
the different subgrade categories. Identical SI values in dif- thickness of 20 mm was defined as the surface treatment clas-
ferent subgrade categories are not equivalent and do not ex- sification and the 10 mm as distress repair.
press identical strength of the pavement. The reason for this
lies in the definition and determination of this index according
to Equation 10. APPLICATION AND LIMITATION OF THE
A similar index for subgrade and pavement categories iden- PROPOSED METHOD
tical to those shown was developed by Scullion (26), who
assigned each subgrade category six separate structure strength The proposed methodology including all its principles and
index groups for all the pavement categories. fundamental assumptions can be incorporated within PMMS
68 TRANSPORTATION RESEARCH RECORD 1374

-E 280
.§. '
06 ~ 105 micron
280
e 200
.§.
06 < 55 micron 200

> 240 240 > 160

0 160
0
J: 1 x 107 J:
200 200 x 107
"'"' 160
Cl)
c
7.5 x 106
5x106
160
"'"' 120
Cl)
c
.!II::
1

l,5 X706
120
.!II::
(,) 2.5 x 106 (,) S X7~6
:c 120 1 x 106
120 :c
t-
80 80
t- . 2.5 x 106
> >
<1S 80 5 x 105 80 <1S
'i:
1x106
'i: Cl) 40 40
Cl)
> 40 1 )( 105 Surface Treatment 40 > Distress Repair
0 Distress Repair 0
0 0 'b.3 0.4 0.5 0.6 0.7 0.8 0.9 1.8
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Structural Index, Slw Structural Index, Sis

FIGURE 8 Overlay thickness versus Slw for weak subgrade. FIGURE 10 Overlay thickness versus Sis for strong subgrade.

to provide fast rehabilitation solutions at the network level. 9. Determine the required overlay thickness according to
The system was successfully applied to a road system totaling subgrade type, SI, and traffic analysis (Figures 8 through 10).
900 km in length (20,21).
Figure 11 shows a general flowchart of the process for over- The design curves shown in Figures 8 through 10 have two
lay 'thickness determination. The process stages may be sum- main limitations, which under certain circumstances may lead
marized as follows: to conservative results or overdesign with reference to the
rational approach. The two limitations mentioned in the fun-
1. Choose the rehabilitated section, measuring the deflec- damental assumptions may be summarized as follows:
tion basin at suitable distances and asphalt layer temperatures.
2. Correct the central deflection D 0 for standard load of 75 •The curves were developed to different structure thick-
kN (Equation 1) and standard temperature of 30°C (Equa- nesses up to 600 mm. The overlay thickness is affected con-
tions 2 and 3). siderably by the pavement thickness in the case where the
3. Correct D 1 and D 6 for standard load of 75 kN. critical criterion is rutting. In instances where the thickness
4. Calculate the SCI parameter according to corrected de- is above 600 mm and the deterioration criterion is rutting,
flections D 0 and D 1 • one must expect overdesign. These instances are characteristic
5. Determine the representative D 6 and SCI parameters of of pavements based on weak subgrade.
the section by statistical analysis. • The subgrade modulus Es representing the different
6. Classify subgrade and pavement by D 6 and SCI param- subgrade categories was constant at 60, 90, and 120 MPa for
eters (Table 2). a weak, medium, and strong subgrade, respectively (D 6 of
7. Calculate the SI according to the subgrade and pavement 115, 75, and 55 µm, respectively). Therefore in cases where
classification (Equations 11 through 13). the subgrade modulus presented by the deflection value
8. Do traffic analysis for a design period. D 6 (see Equation 4) was radically different from the repre-
sentative values, one must expect conservative results or
overdesign.

e 240 55 ~ 06 < 105 micron 240 These limitations are partially solved by statistical analysis

-
J:
E
> 200
0
~
~2x 10 1"1""
130KNESWL
200
for determining the representative basin properties or by cal-
culating the overlay thickness for every single point within
the section and choosing the required overlay thickness on
160

~
"' 160
"'
Cl)

(,)
120
---------====
:::::::------.:

-
1x1o7

7.Sx7 0 6
Sx1o6 120
the basis of statistically homogeneous units (1).
For the validation of the proposed procedure, a comparison
of the overlay thickness calculated by the proposed procedure
:c -2.sx106 and by the rational approach was carried out. The results of
t- 80 80 the analysis showed good correlation between the solutions.
>
<1S
'i: The deviation of the results were in the range 0 to 20 mm,
~ 40 40 which is acceptable for solutions at network level.
0 The proposed method is not intended to replace the rational
'b.2 0.3 0.4 0.5 0.6 0. 7 0.8 0.9 1.8 approach for pavement rehabilitation. To establish the ap-
Structural Index, SIM proach it is necessary to develop modification factors to the
said limitations and calibrate the method with extensive lab-
FIGURE 9 Overlay thickness versus SIM for medium oratory and field tests. However, its advantage lies in pro-
subgrade. viding fast rehabilitation solutions at the network level in
Sidess et al. 69

Rehabilitation Link for predicting the main deterioration criteria such as fatigue
and rutting. The findings indicate that it is possible to classify
the subgrade and pavement by the seventh deflection, D 6 ,
and the SCI parameter and to express the structural pavement
condition by the structural index, SI. Between this index and
the overlay thickness there is a dependence, which makes
itself evident in the overlay curves presented here. The ad-
Asphalt layer vantage of the proposed procedure is in its application pos-
Tem erature sibility as a subsystem within a general PMMS. This enables
fast and reliable solutions at the network level of extensive
road systems. Such solutions are decisive for economic eval-
uation and rehabilitation priority order determination.
Do - Correction For
Standard Load Level

ACKNOWLEDGMENTS

D1, D6 - Correction For The methodology reported in this paper was developed under
Standard Load Level a project financed by the Public Work Department of Israel.
(Eq. 1) Special thanks are extended to the departments of Road
Maintenance and Material and Road Research, which par-
ticipated in the steering committee of the project.

Statistical Analysis
For Determination REFERENCES
Representative
D6& SCI 1. AASHO Interim Guide for Design of Pavement Structures. Amer-
ican Association of State Highway and Transportation Officials,
Washington, D.C., 1972 (revised 1981).
2. Asphalt Overlays for Highways and Streets Rehabilitation. Manual
Series No. 17, The Asphalt Institute, 1983.
3. C. K. Kennedy and N. W. Lister. Prediction of Pavement Perfor-
mance and the Design of Overlay. Report 833. TRRL Laboratory,
Structural Index - SI 1978.
Calculation (Eq. 11-13)
4. A. C. Bhajandas, G. Cumberledge, and G. L. Hoffman. Flexible
Pavement Evaluation and Rehabilitation. Presented at ASCE
Symposium on Pavement Design, 1975.
5. D. E. Peterson et al. Asphalt Overlay and Pavement Rehabili-
Traffic Analysis For
tation Evaluating Structural Adequacy for Flexible Pavement
Design Period Overlays. Report 8-996. Utah Department of Transportation,
1976.
6. N. K. Vaswani. Method for Separately Evaluating Structural
Overlay Thickness Performance of Subgrades and Overlaying Flexible Pavement.
Determination HRR 362, 1971, pp. 48-62.
Fig. 8-10 7. A. I. M Classen and R. Ditmarsch. Pavement Evaluation and
Overlay Design. Proc., 4th International Conference on Structural
FIGURE 11 Flowchart of the overlay design Design of Asphalt Pavements, Ann Arbor, Mich., 1977.
procedure. 8. M. S. Hoffman and M. R. Thompson. Backcalculating Nonlinear
Resilient Moduli from Deflection Data. In Transportation Re-
search Record 852, TRB, National Research Council, Washing-
ton, D.C., 1982, pp. 42-51.
PMMS· for economic evaluation and rehabilitation priority 9. K. Majidzadeh and G. Ilves. Flexible Pavement Overlay Design
Procedure. Final report. FHWA/RD-811033, FHWA, 1981.
order determination of extensive road systems (20,21). The
10. P. Paris and F. Erdogan. A Critical Analysis of Crack Propa-
fast solutions at the network level save a lot of work required gation Laws. ASME Journal of Basic Engineering, Vol. 85, No.
by the field tests in the determination of layer thickness, layer 4, 1963.
moduli derivations, and stress and strain calculations to pre- 11. R. L. Lytton and U. Shanmugham. Analysis and Design of Pave-
dict pavement response. ments To Resist Thermal Cracking Using Fracture Mechanics.
Proc., 5th International Conference on Structural Design of As-
phalt Pavements, 1982.
12. F. P. Germann and R. L. Lytton. Methodology for Predicting
SUMMARY the Reflection Cracking Life of Asphalt Concrete Qverlays. Re-
search Report 207-5. TTI, Texas A&M University, 1981.
13. K. Majidzadeh, C. Buranarom, and M. Karakouzian. Applica-
This paper presents a methodology for flexible pavement re-
tion of Fracture Mechanics for Improved Design of Bituminous
habilitation design and development of curves to determine Concrete. Report FHWA-RD-76-91, Vol. 1, 1976.
the overlay thickness in flexible pavements. The methodology 14. M. Perl, A. Sidess, and J. Uzan. Numerical Simulation of Fatigue
is based on NDT measurements and the rational approach Creep Crack Growth in Visco-Elasto-Plastic Material I. Pre-
70 TRANSPORTATION RESEARCH RECORD 1374

sentation of the Model. Engineering Fracture Mechanics, Vol. Testing of Pavement and Backcalculation of Moduli, ASTM STP
23, No. 2, 1986, pp. 321-331. 1024, 1988. -
15. J. Uzan, M. Perl, and A. Sidess. Numerical Simulation of Fatigue 23. J. Uzan and R. L. Lytton. General Procedure for Backcalculating
Creep Crack Growth in Visco-Elasto-Plastic Material II. Exper- Layer Moduli. The First Symposium on Nondestructive Testing
imental Validation and Application. Engineering Fracture Me- of Pavement and Backcalculation of Moduli, ASTM STP 1024,
chanics, Vol. 23, No. 2, 1986, pp. 333-344. 1988.
16. Asphalt Concrete Overlay Design Manual. California Department 24. J. Uzan, T. Scullion, C. H. Michalek, M. Parades, and R. L.
of Transportation, 1979. Lytton. A Microcomputer Based Procedure for Backcalculating
17. H. F. Southgate, G. W. Sharpe, and R. C. Deen. A Rational Layer Moduli from FWD Data. Research Report 1123-1. TTI,
Thickness Design System for Asphaltic Concrete Overlays; Re- 1988.
search Report 507. Kentucky Department of Transportation, 1978. 25. J. Uzan, M. Livneh, and I. lshai. Thickness Design of Flexible
18. R. W. Kinchen and W. H. Temple. Asphaltic Concrete Overlay Pavements with Different Structures. Journal of Australian Road
of Rigid and Flexible Pavement. Report FHWNLA-80/147. Lou- Research, Vol. 10, No. 1, 1980.
isiana Department of Transportation, 1980. 26. T. Scullion. Incorporating a Structural Strength Index in Texas
19. R. L. Lytton and R. E. Smith. Use of Nondestructive Testing in Pavement Evaluation System. FHWA/TX-87, 409-3F, 1987.
the Design of Overlays for Flexible Pavements. In Transportation 27. J. Verstraten, V. Veverka, and L. Franken. Rational and Prac-
Research Record 1007, TRB, National Research Council, Wash- tical Design of Asphalt Pavements To Avoid Cracking and Rut-
ington, D.C., 1985, pp. 11-20. ting. Proc., 5th International Conference on Structural Design of
20. Engineering and Economic Evaluation for Maintenance Solutions Asphalt Pavements, Vol. 1, 1982.
in the Israeli Network Level Roads (in Hebrew). Yariv Civil En- 28. M. A. Miner. Cumulative Damage in Fatigue. Journal of Applied
gineering Ltd., 1991. Mechanics, 1945.
21. G. Zoltan, A. Sidess, and H. Bonjack. A Rational Method for 29. F. Finn, C. Saraf, R. Kularni, K. Nair, W. Smith, and A. Ab-
Selecting Maintenance Treatment Alternatives Based on Dis- dallah. The Use of Distress Prediction Subsystems for Design of
tress, Structural Capacity and Roughness. Presented at the 71st Pavement Structures. Proc., 4th International Conference on
Annual Meeting of the Transportation Research Board, Wash- Structural Design of Asphalt Pavements, Vol. 1, 1977.
ington, D.C., 1992. 30. J. Uzan and Y. Gur. Flexible Pavement Design in Local Urban
22. K. M. Chua. Evaluation of Moduli Backcalculation Programs for Streets in Israel (in Hebrew). Pub. 89-143. Transportation Re-
Low Volume Roads. The First Symposium on Nondestructive search Institute, Technion Israel Institute of Technology, 1989.
TRANSPORTATION RESEARCH RECORD 1374 71

Pavement Evaluation and Development of

Maintenance and Rehabilitation
Strategies for Illinois Tollway East-West
Extension
ELIAS H. RMEILI, KURT D. JOHNSON, AND MICHAEL I. DARTER

The procedures used in the field evaluation and development of •What are the extent and severity of the deterioration?
maintenance and rehabilitation strategies for the Illinois Tollway • What are the causes of the deterioration?
Authority's 70-mi, four-lane East-West Extension are presen~ed. • Is the load-carrying capacity adequate for future opera-
The extension consists of 14-in. portland cement concrete (PCC) tion or is structural improvement needed?
placed directly on the silty clay subgrade with random joint spac-
ing varying between 12 and 17 ft. Although the pavement is only • What are the feasible rehabilitation alternatives?
17 years old, it is experiencing rapid decrease in serviceability. • What are the expected life cycle costs of each feasible
The field evaluation consisted of a visual condition survey, non- alternative?
destructive deflection testing including void detection and trans-
verse joint load transfer efficiencies, coring, soil evalu~tion, de-
termination of slab and soil properties, and petrographic analysis BACKGROUND
of the PCC. After the pavement evaluation was completed, sev-
eral maintenance and rehabilitation strategies were developed for The original design plans called for the typical pavement sec-
different life expectancies varying from 2 to 20 years. All of the
strategies were evaluated using EXPEAR (EXpert system for tion to be built as a 10-in. concrete slab on a 4-in. cement-
Pavement Evaluation And Rehabilitation). stabilized base on the subgrade. However, during construc-
tion, the typical section was modified to 14-in. concrete slab
placed directly on the subgrade. The typical pavement sec-
The results of an extensive pavement engineering evaluation tions consist of a nonreinforced concrete slab lying directly
and rehabilitation analysis of the Illinois Tollway East-West on the subgrade and randomly skewed joints with joint spac-
Extension on I-88 are presented. The East-West Extension ing ranging from 12 to 17 ft. The concrete shoulders are tied
begins approximately 4 mi west of the Fox River near Aurora to the mainline slabs and decrease in thickness from 8 to 6
and extends in a westerly direction for 69 mi to its terminus in. as they extend out from the traffic lanes. Portions of the ·
1 mi east of Rock Falls. It was constructed during the 1972, East-West Extension were paved as one wide monolithic slab
1973, and 1974 construction seasons and opened for traffic in including the shoulders, whereas other portions were paved
fall 1974. with the traffic lanes separate from the shoulders. The cen-
On the basis of the structural design and the traffic level, terline joints and the longitudinal joints in the monolithic
the East-West Extension was expected to provide up to 30 sections were formed by using a polyethylene tape embedded
years of satisfactory performance with only moderate reha- in the concrete.
bilitation. However, it is experiencing a rapid decrease in the
level of service, and a rehabilitation program will need to be.
implemented soon. DATA COLLECTION
Interesting and useful information, which will help reduce
future portland cement concrete (PCC) pavement design fail- The data collection efforts started with a visual survey of the
ures, was gained from this study. pavement surface. After the survey, the cause of deterioration
was not clearly defined, and it was determined that a complete
PCC evaluation program must be implemented to clearly de-
OBJECTIVE fine the problem. The program included traffic information,
nondestructive deflection testing (NDT) using the falling weight
The objective of this investigation was to develop a cost- deflectometer (FWD), pavement coring and soil boring, and
effective pavement rehabilitation program. The objective was laboratory testing of the soil and PCC.
achieved through a comprehensive engineering evaluation that
answered the following critical questions about the current
condition of the East-West Extension pavement: Traffic Information

The current average daily traffic (ADT) is 10,000 and the

ERES Consultants, Inc., 8 Dunlap Court, Savoy, Ill. 61874. average daily truck traffic (ADTT) is 1,125 in two directions
 100

90

80

~-
0
70
u
ti)

~
..... 60
E-<
.....
i::I

~ 50
~

~
ti)
40

~
30

~
0
20

10

0
61000 71000 81000 91000 101000 111000 121000
MILE POST

FIGURE 1 Eastbound visual rating.

100

90

80

~ 70
0
u
ti)

~
..... 60
E-<
.....
i::I
z
0
u 50
~
~
::::>
40
ti)

~
30

~
0
20

10

129000 119000 109000

MILE POST

FIGURE 2 Westbound visual rating.

Rmeili et al. 73

along most of the route (trucks include vehicles of 6 tons or The distress severity associated with the longitudinal joints
greater). The total number of truck loadings on the East-West is one of the most severe because both concrete durability
Extension was estimated using traffic data from 1988, 1989, spalling and longitudinal cracking exist. In several locations,
and 1990. Assuming a yearly growth since 1974 of 7 percent, the severity level of the centerline joint is so severe that it
the average yearly transaction growth on the East-West Ex- can interfere with vehicles going back and forth across the
tension over the last 6 years, it is estimated that the East-. joint during lane changes.
West Extension has had approximately 3,895,000 truck ap- Another major problem with the pavement was the spalling
plications or 5,452,441 equivalent single-axle loads (ESALs) of the transverse joints. The severity varied from low to high,
in both directions (85 percent of the traffic in the driving lane and in some locations the spalling was as wide as 12 in. on
and 15 percent fo the passing lane). either side of the joint and as deep as 8 in. Figures 3 and 4
show the combined condition of the longitudinal and trans-
verse joints throughout the project.
Visual Condition Survey The entire pavement structure exhibits a nearly total failure
of the transverse and longitudinal joints sealant.
The visual survey consisted of a detailed survey of the first The average faulting that was measured in the field was
500 ft beyond the even mileposts in the direction of travel 0.123 in., and in some locations the measurement was as high
from Milepost 61 through Milepost 128. Information outside as 0.27 in. A faulting of less than 0.25 in. is considered low
the 500 ft was gathered by driving at low speed on the shoul- severity.
der. Conditions such as heave cracking, patched areas, broken
slabs, and areas of notable in-slope erosion were visually
inspected. NDT
Several types of distress were present with the extent and
severity varying significantly throughout the entire area. The NDT was performed using an FWD. The FWD is an impulse
predominant types of distress present were spalling of the device that exerts a force similar in magnitude and duration
transverse joints and spalling and cracking of the longitudinal to a moving vehicle tire load. By varying the weight and height
joints in the driving lanes and on the shoulders. The overall from which it is dropped, the magnitude of the load can be
condition scores for each direction are shown in Figures 1 and changed. The resulting pavement deflection is measured by
2. A score of 100 means the pavement is in excellent condition, seven seismic deflection transducers, one of which is at the
a score of 50 means the pavement is in fair condition, and a loading plate and the others at preset intervals from the load-
score of 0 means the pavement is failed. ing plate.

0::: - r

\ I \\J\
0 -
~

0::: -
~
~ ~ \_
~
~
,_
,_
J
~ ~
fl
\
Q
~ -
~ -
-- A v0vJ
~-
Q_
......:l
-
-
J
-
-
-

61 )00 71 )00 81 JOO 91 )00 101000 111000 121000

66 )00 76 )00 86( )00 96( )00 106000 116000 126000
MILE POST.

FIGURE 3 Eastbound visual rating of joints.

74 TRANSPORTATION RESEARCH RECORD 1374

~ ,..._
c..? t-

~
(
N
t-

~ ~

~ V1
r'-
~
N
~ IN
t-
~

..:I ~ t-

~
~
..:I

~
~
~
~

\ ,I
\J
'\ v- ' '-
~ ~

~
~ t-

'
en
-
~
0
~
-
-
-
.J
v· '
-
t-

t-

129000 119000 109000 99 )00 89 100 79000 691)00

124000 114000 104000 94100 84J00 74)00 64000
MILE POST

FIGURE 4 Westbound visu~ rating of joints.

The initial pattern called for each test site to have five center transverse joint load transfer, loss of support underneath the
slab tests, three lead corner tests, and three lag corner tests slab, and foundation support (effective K-value beneath the
as shown in Figure 5. However, after 3 days of testing with PCC slab).
very little variability between test readings, the pattern was
changed to four center slab, two lead corner, and two lag Coring
corner tests.
For the East-West Extension testing program, the. weights Concrete coring was performed to determine the extent of
were dropped from three heights to produce loads of ap- the joint deterioration, properties and thickness of the PCC
proximately 8,000, 11,000, and 14,000 lbf. The deflections
under the loading plate, 12 in. behind the plate, and at 12,
24, 36, 48, and 60 in. in front of the plate were recorded Direction of Traffic
during testing.
Pavement load-deflection data were used to estimate the
PCC slab modulus of elasticity, PCC modulus of rupture, Inside Shoulder

Direction of Traffic

Inside Shoulder

Outside Shoulder

• Typical Core Location

Outside Shoulder

• Typical NDT Plate Location

FIGURES FWD testing pattern. FIGURE 6 Concrete coring locations.
 Rmeili et al. 75

TABLE 1 PCC Thickness and Aggregate Type tom of the PCC slab) were backcalculated from deflection
MILE POST THICKNESS, in. AGGREGATE TYPE basin measurements. A closed-form backcalculation proce-
67.5 EB 14.5 Buff and white crushed dolomite dure was used, which is based on a theoretically rigorous
76 EB 15.5 Buff and white crushed dolomite approach using the principles of dimensional analysis as well
92 WB 16 Gravel as the concept of deflection basin area. The backcalculation
98.5 WB 15 White dolomite method was developed and automated by Ioannides through
100 EB 14 White dolomite the computer program ILLI-BACK. The approach models
112 EB 14.5 Buff and white crushed dolomite the pavement system as an elastic medium-thick plate resting
116 WB 14 Buff and white crushed dolomite on a dense liquid foundation.
125WB 14 Gravel The backcalculated mean values and ranges were as follows:
E = 4,235 ksi (3,300 to 6,000 ksi) and K = 276 psi/in. (120
to 440 psi/in.).
slabs, air voids system analysis, and petrographic examina- The static K-value is estimated (J,2) to be approximately
tions. A total of 53 concrete cores were taken, in groups of 276/2 138 psi/in., which is a typical value for this type of
4 to 8, at eight different milepost locations, four eastbound soil.
and four westbound. The sites for testing were chosen to get
representative samples from each soil type, pavement distress
level, and construction zone on the East-West Extension. PCC Modulus of Rupture
The coring pattern called for a core to be taken at the
longitudinal, transverse, and lane/shoulder joints, center slab, The .PCC modulus of rupture is most accurately determined
and comer slab. Additional cores were take_n at approximately by sawing standard-sized beams (6 by 6 by 30 in.) from several
6 in. from the longitudinal joint, on the shoulders, and in slabs and subjecting them to third-point loading tests. This is
frost heave locations. Figure 6 shows the location of the cores expensive and time-consuming, however. The PCC modulus
within a site, and Table 1 gives the thickness of the PCC and of rupture can be estimated fairly well by using the indirect
the aggregate type at specific mileposts. tensile strength of recovered 6-in.-diameter cores or even from
After coring, the drilling water remained in the holes (no the compressive strength of the cores. ·
drainage), showing a very low permeability of the subgrade. The PCC flexural strength or modulus of rupture (MR) can
The water was manually removed, and the holes were patched also be estimated approximately from the PCC modulus of
with Set 45 concrete. elasticity backcalculated from the FWD test results. The fol-
lowing relationship was developed at the University of Illinois
Soil Sampling to obtain an approximate flexural strength of pavement slabs
nondestructively (J):
At 14 of the 53 cores taken, borings were made using a truck-
mounted drill rig with the bore holes being advanced by con- MR = 43.5(E/106) + 488.5
tinuous auger flight methods. Samples were taken according
to the ASTM D 1586 procedure for split-spoon sampling of where MR is the modulus of rupture of the PCC and E is the
soils. Representative portions of the split-spoon samples were modulus of elasticity of the PCC.
placed in glass containers with screw-type lids and taken to The calculated mean third-point modulus of rupture value
the laboratory for examination and testing. Laboratory work is 673 psi.
consisted of water content determinations for most of the
samples, with unconfined compression strength tests being
performed on representative samples. Approximate measure- Transverse Joint Load Transfer
ments of unconfined compression strengths were made for
some of the samples using a calibrated pocket penetrometer. The deflection load transfer across the transverse joints was
The pocket penetrometer is an indirect method for evaluating measured by FWD testing. The load transfer is computed as
the compressive strength of a clay soil. All of the slab sections the ratio of the unloaded slab's deflection to that of the loaded
sampled rested directly on a fine-grained soil subgrade. slab's deflection. A small correction is applied to this load
transfer efficiency to allow for natural slab bending, which
would occur even though no joint existed between the first
FWD ANALYSIS and second sensor. The following equation was used to cal-
culate the percent load transfer:
FWD analysis was performed using finite element techniques
and procedures developed by the University of Illinois and Load transfer efficiency ( % ) = ~L x 100
the Corps of Engineers. A

where DA is the deflection in the approach slab and DL is the

Modulus of Elasticity and Modulus of Subgrade deflection in the leave slab.
Reaction The FWD testing was conducted during cool temperatures
when the joints were not excessively tight. The mean load
The PCC slab modulus of elasticity (E) and the effective transfer of transverse joints is 95 .6 percent, which is very high,
dynamic modulus of subgrade reaction (K-value) (at the bot- indicating good aggregate interlock across the joints. The high
76 TRANSPORTATION RESEARCH RECORD 1374

A LINE A - NO VOIDS
• LINE B - SMALL VOID
e LINE C - LARGE VOID

15

10

/
// /
/ / /
5
/ /· /
/ /
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40

MAXIMUM PAVEMENT DEFLECTION {mils)

FIGURE 7 Load versus deflection void detection plot.

load transfer explains the low severity faulting of the trans- In addition to the NDT, an epoxy/core test procedure de-
verse joints even though the joints are not doweled. veloped at Purdue University was used to detect voids at
selected locations. The test was performed on several loca-
tions where voids were thought to exist on the basis of visual
Corner Void Detection observation. No evidence of voids was detected at any of the
locations tested.
A void can be described as an area of loss of support beneath
the slab corners. This occurs because of heavy loads deflecting
the slab corners, causing some pumping of fines with impulse LOAD-CARRYING CAPACITY
water pressure. Any loss of support, even less than 0.10 in.
thickness beneath the slab, will result in greatly increased slab The load carrying capacity was evaluated using ILLISLAB,
stresses, causing corner breaks or diagonal cracks. The most a computer-based finite element model, to determine the crit-
common location for these voids is slab corners along the ical stresses in the PCC slab created by the truck loads. The
lane/shoulder joints. The loss of support can be determined modulus of elasticity of the PCC and the effective K-value
using a procedure developed at the University of Illinois for
the FWD (3). An examination of the load versus deflection
at a slab corner can provide a rapid and simple indication of
the existence of a void beneath the slab corner. The corner
of each slab tested was loaded at three load levels and the fil 14 ............. ..
p..
corresponding deflections measured. Corners that have a load g 12
versus deflection plot that crosses the deflection axis near the 'O
origin of zero deflection, such as Line A in Figure 7, do not ~ 10 .................. ..
0
have voids beneath the slab. However, a line that passes ....:I
significantly to the right of the origin, such as Lines B and C
in Figure 7, is indicative of a void or loss of support beneath .... , .... ···i

the slab corner. Figure 8 shows actual data from the East-
West Extension.
After reviewing all of the load versus deflection plots, it
was concluded that no voids or loss of support existed beneath
the slabs tested except at Milepost 62.0153 eastbound. This 2 3 4 5 6 7 8 9 10
was probably caused by the full-depth asphalt patch of the Maximum Pavement Deflection (mils)
shoulder across the joint. Most likely the subgrade soil was
disturbed during patching. FIGURE 8 Actual load versus deflection detection plot.
Rmeili et al. 77

were used along with the traffic characterization and weight plasticity indices (Pis) ranged from 10 to 30. The cohesive
to determine the critical stresses. The axles were positioned soils varied from hard to very tough, whereas all of the cohe-
so that the wheel was placed directly on the slab edge along sionless soils were firm. A thin layer of organic topsoil was
the lane/shoulder joint, the most critical location for stresses occasionally encountered, with characteristically high mois-
and crack development. ture contents and low dry densities. Occasional lenses of fill
The number of edge loads allowable to failure (defined as directly beneath the pavement were well compacted, with
the point at which 50 percent of the slabs suffer cracking) is moderate moisture contents.
calculated on the basis of the strength of the slabs and the
critical stress produced by the traffic loading. The following Analysis of Air Void Systems
equation form developed using Corps of Engineers field data
was used: Properly air-entrained PCC is protected from the increase of
hydraulic pressure during freeze-thaw -cycles. The air void
log 10 (axles) -
-
a( stress
MR )b bubbles would _act as pressure relief valves or "safety valves"
that allow the excess water to escape to them and freeze
where without damaging the PCC (4). ·
To evaluate the air void parameters of the project, 10 cores
MR = modulus of rupture of PCC slabs (psi), were selected for testing. The results are given in Table 2.
stress = edge stress under traffic load (psi), ACI Committee 345 has recommended that the number of
a, b = empirically derived constants developed from Corps voids per lineal inch be significantly greater than the per-
of Engineers field data, and centage of air, that the specific surface (a) be greater than
axles = number of edge loading axles to failure (defined 600 in. 2/in. 3 and that the spacing factor (L) be less than 0.008
as 50 percent slab cracking). in. For the %-to 1-in. maximum aggregate size in the I-88
This analysis showed that the number of edge loads to failure cores, these requirements should be met with an air content
for an 18-kip axle was 5.45 x 1018 load applications. This . of about 4V2 to 7% percent. However, although essentially all
indicates a very structurally sound PCC pavement, and no 10 of the cores tested had total air contents within this range,
transverse cracking from traffic loadings is expected. only one (125-4) met the requirements listed. The causes for
this are not determinable. But the undesirable air void pa-
rameters measured are certainly sufficient to explain the cyclic
LABORATORY STUDIES freezing ·damage observed in most of the cores subjected to
petrographic examination.
A total of 51 cores were received. Several of those cores were In general, the entrained air voids of the entire project are
used for petrographic examinations and air void parameter not well distributed. They tend to forin in clusters and around
determinations. The petrographic studies were done accord- coarse aggregate particles; thus, the air void systems are not
ing to ASTM C 856, and the air content studies were done well developed. In addition, the air in many of the cores occurs
according to ASTM C 457. The modified point count method as large entrapped voids, which furnish a minimum protection
was used. from cyclic freezing damage. This explains the premature fail-
Detailed, petrographic studies were ·conducted on 1 core ures of the transverse and longitudinal joints.
from each section of roadway, plus 2 additional cores from
selected area, making a total of 10 cores studied. Air void
studies were also done on one core from each section plus SUMMARY OF EVALUATION RESULTS
two from selected areas. Cursory petrographic examinations
were conducted on the remainder of the cores. This section presents the findings and engineering ·analyses
Three distinct coarse aggregate types were used in the con- on which maintenance and rehabilitation strategies were de-
cretes. They consist of a buff and white crushed dolomite, a veloped_. The results of the evaluation study are as follows:
white limestone/dolomite, and a gravel consisting of dolomite, 1. The concrete slab is 14 in. thick or more. Most pavement
granite, chert, and basalt. All cores contained natural sand, cores were greater than 14 in., and those that were not were
which did not contribute to the deterioration. within 1/4 in.
The deterioration observed in the cores fell generally into
two distinct categories: (a) freeze-thaw damage to the port- TABLE 2 Results of Air Void Analysis
land cement binder and (b) D-cracking of the coarse Mile Post Total% Air Voids/in Specific Surface (a) Spacing Factor (L)
aggregates. in 2/in 3 in
Freeze-thaw damage occurs in concrete that is sufficiently 67.S EB 7.0 3.9 224 0.014
saturated and does not contain an adequate air void system. 76 EB 6.2 3.3 210 0.017
D-cracking results from a combination of a specific aggregate 92 WB 5.0 3.8 303 0.012
type, the presence of moisture, and freeze-thaw cycling. 98.5 WB 5.5 4.3 313 O.D15
98.S WB 4.4 4.1 367 0.014
100 EB 5.6 4.8 342 0.009
Soil Results
112 EB 5.3 3.6 271 0.014
116 WB 6.7 4.4 262 0.016
All of the soils were fine-grained (sand, silt, and clay) ranging
125 WB 5.1 7.8 621 0.006
in AASHTO classification from A-2-6 to A-7-6; most were
125 WB 4.9 3.8 308 0.014
A-6. The liquid limits (LLs) ranged from 20 to 51 and the
78 TRANSPORTATION RESEARCH RECORD 1374

2. Estimated total air contents range from 4.4 percent at egies were evaluated using EXPEAR (EXpert system for
Milepost 98.5 WB to 7.0 percent at Milepost 67.5 EB, and Pavement Evaluation And Rehabilitation) (5-7). EXPEAR
much of this was entrapped air. In general, the entrained air is a practical and comprehensive computerized system to eval-
voids are not well distributed. The voids tend to form in uate concrete highway pavements, develop rehabilitation al-
clusters and around coarse aggregate particles; thus, the air ternatives, and predict the performance and cost-effectiveness
void systems are not well developed. In addition, the air in of the alternatives. EXPEAR was originally. developed for
many of the cores and in spalled pieces of PCC observed in FHWA by the University of Illinois and it was recently mod-
the field occurs as large entrapped voids, which furnish a ified by ERES Consultants, Inc.
minimum of protection from cyclic freezing damage. This is On the basis of the recommendation of the Illinois Tollway
believed to be the major cause of the spalling. Authority, strategies with different life expectancies were de-
3. Some of the aggregate is susceptible to D-cracking dam- veloped. These strategies were grouped into three different
age. This has resulted in some of the spalling in limited specific groups (short-term, intermediate-term, and long-term perfor-
sections along I-88. · mance) so that the Illinois Tollway Authority can base its
4. The average dynamic modulus of subgrade reaction is choice of strategy on budget constraints and performance pe-
276 psi/in. (static value of about 276/2, which is typical of this riod. The strategies were evaluated by life cycle cost analysis,
type of soil). which makes it possible to compare alternatives with different
5. The average modulus of elasticity of the PCC interim initial construction costs on the basis of equivalent annual
slabs is 4,200,000 psi, which is considered sound concrete. cost. The unit costs information was obtained from several
6. The average modulus of rupture of the PCC estimated local areas and is included in Table 3. Table 4 gives all of the
by NDT testing was 673 psi, which is considered good. strategies, life expectancy, total initial costs, and equivalent
7. The average load transfer across the transverse joints is annual costs.
95 percent. Any value 70 percent or more is considered good. The overlay thicknesses in Strategies 4, 5, 6, 7, and 9 are
8. The average transverse joint faulting is 0.123 in. In sev- only approximate and must be verified through an engineering
eral locations, the faulting was as high as 0.27 in. design analysis.
9. The PCC slab is structurally sound and can carry a large
number of repetitions before load fatigue cracking. Existing 1. Spray patch (1): This strategy includes removing loose
cracking is caused from other causes such as frost heaving or materials from the center longitudinal and transverse joints
inadequate longitudinal joint construction. and spray patching the joints with emulsion and aggregate.
10. There were no indications of any loss of support beneath 2. Spray patch (2): This strategy indudes cleaning all of the
any of the slabs tested. joints except the shoulder joints and spray patching them with
11. The soils found on the extension were generally emulsion and aggregate. .
AASHTO Classficatiort A-6. 3. Repair spalls with PCC: This· strategy includes milling
12. All of the joints throughout the entire project are ex- out only the center longitudinal and transverse joints 2 ft wide
periencing deterioration caused by damage from cyclic freez- by 6 in. deep, r·epairing them with PCC, and then sawing and
ing, D-cracking of the coarse aggregate particles, or a com- sealing the joints.
bination of the two. 4. AC overlay (1): This strategy includes repairing the joints
13. The pavement has become very rough from the severe with AC and applying a 5-in. AC overlay.
joint spalling. 5. AC overlay (2): This strategy includes repairing the joints
with PCC and applying a 5-in.-minimum AC overlay.
6. Crack and seat: This strategy includes breaking all of the
MAINTENANCE AND REHABILITATION PCC slabs of Lanes 1 and 2 into small pieces (1 to 2.5 ft),
STRATEGIES seating the pavement, and applying a 5-in. AC overlay.
7. Rubblizing: This strategy includes breaking all of the
Several maintenance and rehabilitation strategies were con- PCC slabs into rubble (less than 12 in. in size), compacting
sidered for the East-West Extension project. All of the strat- the rubble, and applying a 7-in. AC overlay.

TABLE 3 Unit Costs Information

DESCRIPTION UNIT COST(dollars) UNIT

Unbonded PCC Overlay 3.05 per inch SY

AC Overlay 33.22 TON
Crack and Seat 1.00 SY

Rubblizing 1.65 SY
AC Partial Patch 1.31 LINEAR Ff
(Spray Patch)
11.79 SY

PCC Partial Patch 15.00 LINEAR Ff

68.00 SY

PCC Recycling + Base 39.06 SY

Longitudinal Subdrains 2.46 LINEAR Ff

Reconstruct Heaves 50.00 SY

Rmeili et al. 79

TABLE 4 Summary of the Rehabilitation Strategies

Performance Strategy Life Expectancy Total Initial Cost Total Annual Cost
I
j

Period I (Years) (U.S. Dollars) (U.S. Dollars)

1. Spray Patching (1) 2 4,115,160 2,087,940

Short-Term 2. Spray Patching (2) 2 6,024,252 3,056,700

3. Patching with PCC 5 29,158,986 6,181,572

4. AC Overlay (1) 7 33,938,064 5,288,574

Intermediate- 5. AC Overlay (2) 10 58,933,728 6,707,628
Term 6. Crack & Seat/ AC OL 10 34,919,658 3,974,400

7. Rubblizing/ AC OL 10 52,246,662 5,946,558

8. Reconstruct Outer Lane 15 45,315,198 3,685,290

& Restore Inner Lane

Long-Term 9. Unbonded PCC Overlay 20 84,640,230 5,523,450

10. Reconstruction 20 107,454,390 6,985,284

8. Recycling and restoring: This strategy includes recycling Low-Severity Sections

Lane 1 into a 4-in. stabilized base with a 10-in. PCC pavement
and the joint spacings matching those of Lane 2. Also, all of The low-severity areas should be addressed as soon as possible
the joints in Lane 2 would be repaired with PCC. to prevent further disintegration of the pavement surface. The
9. Unbonded PCC overlay: This strategy includes applying recommendation is to clean and reseal all the longitudinal and
a 1-in. AC separation layer over Lanes 1 and 2 and then transverse joints. There will also be a need to do limited slab
placing a 9-in. jointed plain PCC overlay with 15-ft joint spac- replacements and some joint repairs in isolated areas. There
ing (8). The shoulders will be overlayed with 9 in. AC. are locations within the low-severity areas with few broken
10. Reconstruction: This strategy includes removing the PCC slabs due to frost action in the soils. Those could be replaced
from the mainlines and shoulders and rebuilding the pavement and the soil removed and replaced. However, this recom-
with 4-in. base, 12-in. PCC surface with 15-ft joint spacing, mended treatment will not address the profile of the
longitudinal edge drain, and 7.5-in. PCC shoulders. pavement.

CONCLUSIONS AND RECOMMENDATIONS Medium- and High-Severity Sections

The engineering analysis conducted to determine an effective Before recommendations are made for repair in these areas,
rehabilitation plan for the East-West Extension had many it must be recognized that everything cannot be done im-
associated complexities. A major effort was made to deter- mediately from budget, project development, and overall toll-
mine the primary cause of the current surface distress affecting way system planning aspects. Interim alternatives to address
the user's satisfaction aild riding comfort. After the pavement the current effort required to maintain the surface profile and
engineering investigations and analysis were completed, it was provide user comfort should be performed. It was recom-
concluded that the main problem was the spalling of the trans- mended to the Tollway Authority that the transverse and
verse and longitudinal joints. The spalling of the joints was longitudinal joints be repaired temporarily using the spray
caused primarily by damage from cyclic freezing of the cement patch procedure until a perm~nent rehabilitation strategy is
paste and D-cracking of the coarse aggregate particles. Even implemented.
though the air content in the concrete was within specifica- The overall rehabilitation of the project will be based on
tions, the size and distribution of the voids were not within the service life the Tollway Authority wishes to obtain, budget
specifications to protect the concrete from freeze-thaw cycles, constraints, and construction time and lane closures. The
which caused the premature failure of the joints. patching alternatives are not favorable due to insufficient life.
The pavement itself is structurally sound and has a re- The other alternatives pose a trade-off between longer service
maining life left in it as woulq be typically expected. The life artd lower initial construction cost and increased future
primary concern is rehabilitatioin of the transverse and lon- rehabilitation costs. For example, if the Tollway Authority
gitudinal joints and restructuring the pavement's resistance favors a long-term (20 or more years) strategy, an unbonded
to infiltration of surface moisture. By accomplishing this, any PCC overlay is recommended. If the Tollway Authority favors
chances of further damage from freeze-thaw will be reduced, a 10-year strategy, an AC overlay after rubblization or crack
and the service life of the pavement will substantially increase. and seat is recommended.
The preferred rehabilitation for the East-West Extension
is primarily composed of three components. These are rec-
ommended on the basis of effectiveness, performance, and ACKNOWLEDGMENTS
user satisfaction. Annual budget constraints have also been
recognized along with the cost-efficiency of interim mainte- The authors would like to thank Jam es Mack and Philip Miller
nance operations. of Envirodyne Engineers and Steve Gillen of the Illinois Toll-
80 TRANSPORTATION RESEARCH RECORD 1374

way Authority for their assistance on the East-West Extension 4. S. Mindess and J. F. Young. Concrete. Prentice-Hall, Inc.,
Englewood Cliffs, N .J.
project. 5. K. T. Hall, J. M. Connor, M. I. Darter, and S. H. Carpenter.
Rehabilitation of Concrete Pavement, Vol. III-Concrete Pavement
Evaluation and Rehabilitation System. FHWA-RD-88-073. FHWA,
REFERENCES 1989.
6. M. I. Darter and K. T. Hall. Structural Overlay Strategies for Jointed
1. P. T. Foxworthy. Concepts for Development of a Nondestructive Concrete Pavements, Vol. -IV-Guidelines for the Selection of
Testing and Evaluation System for Rigid Airfield Pavements. Ph.D. Rehabilitation Alternatives. FHWA-RD-89-145. FHWA, 1990.
thesis. University of Illinois at Urbana-Champaign, Urbana, 1985. 7. K. T. Hall and M. I. Darter. Structural Overlay Strategies for Jointed
2. Special Report 61E: The AASHO Road Test, Report 5, Pavement Concrete Pavements, Vol. VI-Appendix A-Users Manual for
Research. HRB, National Research Council, Washington, D.C., the EXPEAR Computer Program. FHWA-RD-89-147. FHWA,
1962. 1990.
3. J. A. Crovetti and M. I. Darter. Void Detection Procedures. (Ap- 8. G. F. Voight, S. H. Carpenter, and M. I. Darter. Rehabilitation
pendix C). Final report, NCHRP Project 1-21, TRB, National of Concrete Pavements, Vol. II: Overlay Rehabilitation Tech-
Research Council, Washington, D.C., 1984. niques. FHWA-RD-88-072. FHWA, 1989.
TRANSPORTATION RESEARCH RECORD 1374 81

PARES-An Expert System for

Preliminary Flexible Pavement
Rehabilitation Design
TIMOTHY Ross, STEPHEN VERZI, ScoTT SHULER, GORDON Mcl<EEN,
AND VERNON SCHAEFER

The development of a knowledged-based expert system to assist problems and offer alternative solutions for obtaining the best
~he New Mex~co State Highway and Transportation Department pavement rehabilitation scheme for a given situation.
m the evaluation and design of rehabilitation schemes for flexible This paper summarizes a recent New Mexico study (J) to
pavements is described. The system uses information provided develop a PAvement Rehabilitation Expert System (PARES)
by users to establish preliminary rehabilitation schemes that would
be reasonable and cost-effective. A cost-estimate module for for the preliminary rehabilitation design of New Mexico flex-
~anking th~ rehabilitation schemes according to relative cost is ible pavements. Good reviews of expert system technology
mtegrated mto the system. The need for such a system in New in transportation and other civil engineering disciplines are
Mexico and the knowledge base used to construct the IF-THEN- available in the literature (2). The paper presents some rel-
ELSE type rules in the expert system are described and the evant previous efforts on the application of expert systems
d.istress condi~ions addres~ed and the rehabilitation strat~gies con- technology 'in highway pavement management, discusses the
sidered are discussed. The system is rich in the sense that it also
distinguishes among distress situations requiring routine main- current practices in the state of New Mexico, addresses the
tenance ~s opposed to rehabilitation requiring more extensive particular features and utility of the PARES code for use in
~onst~ctlon efforts. An example session using the expert system flexible pavement rehabilitation, and concludes with an ex-
1s provided. ample application of the system.

The selection of pavement rehabilitation alternatives depends

on distress type present in the pavement, ride quality, traffic
RELATED EXPERT SYSTEMS FOR HIGHWAY
volume, structural section, maintenance history, and other
PAVEMENTS
factors. Although the manual process of determining reha-
bilitation schemes has been effective, a computerized
An expert system originally developed for the Washington
knowledge-based expert system would allow a more detailed
State Highway system by Ritchie et al. involves the area of
preliminary estimation of rehabilitation needs such that costs
flexible pavement rehabilitation using a code called SCEPTRE
could be better ascertained. This would contribute to a more
(3). The SCEPTRE code is used to provide a user with several
accurate identification of the number and extent of projects
rehabilitation strategies based on the existing condition of a
to be scheduled for rehabilitation.
roadway and the user-specified service life of the desired re-
Such an expert system could also be used to reduce the
habilitated pavement. SCEPTRE is based on "IF-THEN"
time required for new personnel to develop an adequate level
type rules and uses a backward-chaining inference method
of on-the-job experience. More experienced personnel may
(reasoning goes back from known facts to a hypothesis).
use the expert systems to make their own designs a more
Haas and his colleagues (4-6) have developed expert sys-
expeditious and economical process. The expert system for
tems for flexible pavement management, pavement distress
preliminary rehabilitation design will immediately benefit the
analysis, and pavement condition data inventory. One of these
New Mexico State Highway and Transportation Department
systems, PRESERVER (4), assists field engineers and su-
(NMSHTD) by providing assistance to personnel responsible
pervisors in analyzing pavement distress data and proposes
for estimating initial costs of rehabilitation projects with ex-
routine maintenance strategies. This system is similar to
pert guidance regarding the most cost-effective alternatives
SCEPTRE, except that it proposes maintenance rather than
~sing available information. The expert system could be quer-
rehabilitation strategies.
ied by users for details on the construction or rehabilitation
In other ~evelopments, Hall et al. (7) have developed an
problem of concern, with the output used to identify potential
expert system, called EXPEAR, to assist the design engineer
in the evaluation and preliminary rehabilitation design for
T. Ross, Department of Civil Engineering, and S. Verzi, Department jointe~ reinforced, jointed plain, and continuously reinforced
of Computer Science, University of New Mexico. Albuquerque, concrete pavement (JRCP, JPCP, and CRCP): EXPEAR uses
N.Mex. 87131. S. Shuler, Asphalt Institute, Box 14052, Lexington,
Ky. 40512. G. McKeen, New Mexico Engineering Research Institute,
information provided by pavement engineers to determine
B?~ 25, ~lbu9uerque, N.Mex. 87131. V. Schaefer, Department of the type and cause of distress so that an appropriate reha-
CIVIi Engmeermg, South Dakota State University, Brookings, S.Dak. bilitation strategy can be selected.
82 TRANSPORTATION RESEARCH RECORD 1374

Aougab et al. (8) have developed an expert system, PAMEX, tion history, and other data, if available. Rehabilitation al-
for maintenance management of flexible pavements. Ritchie ternatives are compared on the basis of initial and long-term
(9) has developed an expert system, termed OVERDRIVE, cost-effectiveness for a design period of 10 years.
to assist local engineers in designing the structural thickness Although the manual system in New Mexico was effective
of asphalt concrete overlays. Haas and Shen (4,6) have de- in determining appropriate rehabilitation alternatives, an ex-
veloped PRESERVER, an expert system for the Canadian pert system will be advantageous in the assessment of initial
province of Ontario to help field engineers and supervisors rehabilitation costs for at least two reasons. First, the initial
analyze pavement distress information to propose routine cost estimate for prioritized projects would be significantly
maintenance strategies. Hajek et al. (10) have developed more accurate. Second, much of the iterative process involved
ROSE, an expert system for recommending routing and seal- with comparing the preliminary design with initial estimates
ing of asphalt concrete pavements in cold areas of Canada. made by planning personnel would be reduced because the
Tandon and Sinha (JJ) have developed an expert system to initial estimate involves the same reasoning that is included
estimate highway pavement routine maintenance needs and in the preliminary design procedure.
expected costs at the subdistrict level. And finally, to under-
score the growing emphasis and importance of expert systems
in pavement management, Barnett et al. (12) have published
THE EXPERT SYSTEM-PARES
a Federal Highway Administration report that provides guide-
lines to the states for the development and distribution of
PARES was implemented in EXSYS Professional (14), a com-
highway-related expert systems.
mercially available expert system shell. EXSYS has been used
extensively in other applications (15). EXSYS allows for both
backward-chaining (goal driven) and forward-chaining (data
NEW MEXICO PAVEMENT REHABILITATION
driven) inferencing. The IF-THEN-ELSE structure is the gen-
SYSTEM
eral form of the rules of an EXSYS knowledge base. This
structure is used for the rule base irrespective of whether the
Performance of each of 3,000 evaluation sections in the New
rule is chained in a forward or backward manner. All portions
Mexico pavement network is documented periodically through
of the IF clause must be satisfied before the conclusion (THEN
visual condition surveys and roughness measurements. Re-
clause) of a rule is activated. If a single portion of the IF
habilitation procedures for flexible pavements in New Mexico
clause is disproved, the ELSE portion is activated.
are intended to provide 10 years service with routine main-
tenance; however, the routine maintenance required during
this interval will vary depending on the rehabilitation method
selected. The repair strategies vary in effectiveness, cost, and THE INFERENCING SYSTEM-EXSYS
intended purpose.
The current New Mexico pavement management system Each of the conditions in the IF portion of a rule is specified
consists of a very detailed description of the roadway to be by a Boolean-valued formula that will evaluate to either true
evaluated. Information collected from the field is transferred or false. The formula can be composed of mathematical var-
to a computer system to present the user with seven types of iables in a logical relation ( <, >, = , etc.) or propositional
inquiries regarding the roadway segment: (a) pavement data, variables in a predicate calculus relation. When all IF con-
(b) condition data, (c) planned projects, (d) project history, ditions have a truth assignment, the rule can be invoked, and
(e) traffic data, (f) road safety data, and (g) distress detail. either the actions of the THEN portion or the ELSE portion
In New Mexico distress is quantified on the basis of American are carried out. Actions in the THEN or ELSE portions of
Public Works Association (APWA) guidelines (13). the rule can perform many different functions, such as exe-
A priority ranking system based on field condition surveys cution of external programs (e.g., the PARES cost module),
and traffic volume has been developed by NMSHTD to assess manipulation of mathematical variables (e.g., calculation of
which of the sections should be rehabilitated or reconstructed. mill depths and overlay thickness), setting of conditions for
This system has been developed such that a priority assign- the IF portions of other rules, and selection of final rehabil-
ment indicates that rehabilitation is necessary. Therefore, itation strategies. In EXSYS the ELSE portion of the rule is
pavements requiring routine maintenance theoretically would optional.
not receive a priority value and therefore would not be con- In a forward-chaining inference, an existing knowledge base
sidered for rehabilitation. An exception to this might include is used to invoke as many rules as possible, where the actions
pavements with escalating maintenance costs, which a partic- from these rules are used as the conditions for new rules and
ular highway district judges as requiring more than routine the invocation proceeds forward until no more rules can be
treatment. invoked. Backward chaining proceeds by selecting a rule in
After the priority assignments are made an ini!ial estimate which it is desired to have one or more of the actions executed
of cost for rehabilitation is made. The preliminary cost esti- in the THEN portion of the rule (goal). For the action to take
mate is used to determine the number and extent of projects place, all of the conditions in the IF portion of the rule must
to be considered for rehabilitation depending on the funds be satisfied (i.e., evaluated to be true). Backward chaining
available. After the projects to be rehabilitated are identified, proceeds in a "depth-first" manner through the rule base,
a more comprehensive preliminary design is initiated. This searching the rule base for rules whose actions will enable the
design is based on a visual survey by the design engineers, firing of rules that have already been considered in the chain
results of the condition survey and roughness data, construe-. of rules.
Ross et al. 83

DEVELOPMENT OF THE KNOWLEDGE BASE algorithmic form. Both the distresses and plausible rehabili-
tation strategies were classed into special categories that would
The PARES knowledge base was developed from procedures be useful in relating the distresses to the rehabilitation strat-
documented in New Mexico state highway agency manuals, egies. The research team acted as additional experts to de-
some AASHTO procedures, and a group of NMSHTD pave- velop categorizations of distresses and rehabilitation strategies
ment rehabilitation experts. The state highway experts pro- as well as the logic to tie those together into a multilevel logic
vided the heuristic rules used to formulate the knowledge tree. The experts decided that 21 rehabilitation strategies (given
base. The experts were particularly important to this work, in Table 2) captured the experience in the past of rehabili-
but the input from different individuals invariably resulted in tations to New Mexico road surfaces and, to a lesser extent,
some conflicts of opinion. The resolution of these differences plausible strategies not frequently used in New Mexico. The
was addressed by the research team in selecting among the multilevel logic tree relating the distress situations to the po-
available alternatives. tential rehabilitation strategies is shown in Figure 1, where
The first step in the development of the knowledge base the encircled numbers represent the strategies given in
was to construct a list of the data to be entered by the users Table 2.
into the expert system. Rehabilitation of roadway surfaces is To categorize the pavement distresses, the research team
necessitated by the existence of certain types and levels of classified the possible distresses into five distress type sets.
distress. The development of the PARES code used standard Some of these sets may be empty for a particular design sit-
distress types as documented in the APW A Pavement Con- uation, indicating that no distresses in that category have been
dition Index for Asphalt Pavements (13). The expert system seen on the existing pavement surface. These five distress
PARES considers 23 types of distress and 3 levels of distress categories are general maintenance distresses, localized main-
severity for each distress type. The distresses and levels are tenance distresses, surface-mix distresses (due to the asphaltic
given in Table 1. material), surface cracking distresses, and subgrade (subsur-
The expert system also considers the extent and the severity face) distresses. When all known distresses have been spec-
of each distress type. For most distresses, the extent is entered ified by the user, PARES inferences on the rule base to pro-
in terms of the percentage of the road covered by the partic- vide possible rehabilitation strategies to the user.
ular distress severity. For longitudinal cracking and transverse
cracking the extent is entered as the number of lineal feet of
cracks per project. For depressions at bridges and railroad SPECIFICS OF THE RULE BASE IN PARES
crossings, the extent is the number of these distresses of a
particular severity present in the project. These two are in- There are 278 rules in the PARES rule base. These rules, in
herently localized distresses and, as such, can ,be treated sep- conjunction with the geometric information about the road-
arately from the rest of the distresses. way surface provided by the user, compose the knowledge
The second step in representing the asphalt pavement re- base for a particular application of the PARES code. Road-
habilitation knowledge was the creation of hierarchical struc- way length and width (without shoulders) are the only geo-
tures (a logic tree). The logic tree forms the shell in which metrics used in PARES. Rules 1 to 73 in the PARES rule
knowledge-based rules stating the declarative and procedural base embody the logic used to segregate all the input data
knowledge are inferenced. and the distress conditions provided by the user into the five
Interviews were conducted with five expert New Mexico general distress type sets. Each time a distress type set is
pavement designers. From these interviews, a multilevel logic confirmed in a rule a counter is incremented so that PARES
tree was conceived to capture the generality of possible dis- can use the number of distresses in a particular distress type
tress conditions. This logic tree attempts to structure the set in logic deeper in the logic tree. Rules 1-73 are inferenced
knowledge of the experts that typically is not reducible to

TABLE 2 Rehabilitation Strategies

TABLE 1 Distress Types and Severity Levels for
PARES 1. Do nothing
2. Crack seal
Distress Types 3. Chip seal
4. Asphalt Overlay
Alligator (fatigue) Cracking Patch Deterioration 5. Crack seal + overlay
Bleeding (flushing) Polished Aggregate 6. Asphalt rubber interlayer+ overlay
Block Cracking Potholes 7. Geotextile fabric strip +overlay
Bumps and Sags Pumping and Bleeding 8. Geotextile fabric sheet + overlay
Conugation Railroad Crossings 9. Cold in situ recycling+ chip seal
Depressions Raveling and Weathering 10. Cold in situ recycling + overlay
Depressions at Bridges Rutting 11. Hot recycling + chip seal
Edge/Center-line Cracking Shoving 12. Hot recycling+ overlay
Joint Reflection Cracking Slippage Cracking 13. Heater scarification +overlay
Lane/Shoulder Drop-off Swelling 14. Cold milling+ overlay
Lane/Shoulder Separation Transverse Cracking 15. Cold milling+ interlayer+ overlay
Longitudinal Cracking 16. Cold milling + cold in situ recycling + overlay
17. Pulverization + overlay
Distress Severity Levels 18. Portland Cement Concrete
Low Severity 19. Reconstruction
Moderate Severity 20. Dig out and patch
High Severity 21. Patch surface
84 TRANSPORTATION RESEARCH RECORD 1374

START

OO~NG

REHABILITATION

ADDITIONAL NO ADDITIONAL
STRUCTURE NEEDED STRUCTURE NEEDED

SUBSURFACE SURFACE
PROBLEMS CRACKING

FIGURE 1 Multilevel logic tree for PARES.

in a forward-chaining manner since classification is inherently THEN localized distress set contains alligator cracking
a data-driven task. In PARES, the main body of rules (Rules AND increment localized distress set counter by 1.
74-249) are inferenced in a backward-chaining mode, where (Note: as described later, PARES will query the user about
the output consists of rehabilitation strategies for roadway the "type" of alligator cracking present.)
repair for the current input situation. Finally, Rules 250-278
are inferenced using a forward-chaining mode, since these RULE NUMBER: 28
rules calculate specific values for mill depth and overlay thick- IF distress type set contains lane/shoulder drop-off
ness for the rehabilitation strategies chosen in Rules 74-249. AND lane/shoulder drop-off severity is low
Examples of the first 73 rules follow: AND the extent of lane/shoulder drop-off is greater than
10 percent,
RULE NUMBER: 1 THEN subsurface distress set contains lane/shoulder drop-
IF distress type set contains alligator cracking off
AND the alligator cracking is type C. AND increment subsurface distress set counter by 1.
Ross et al. 85

RULE NUMBER: 72 potholes are usually found in low extent, and so can be han-
IF distress type set contains alligator cracking dled using maintenance strategies. However, if these same
AND the alligator cracking is Type A OR Type B types of distress are apparent over a significant extent of the
AND distress type set contains rutting project surface and the average daily traffic is high enough,
AND rutting severity is moderate OR high the pavement would tend to require rehabilitation. For sim-
THEN subsurface distress set contains alligator cracking plicity, it was decided that when a generally nonlocal distress
AND rutting type occurs with generally local distress types, if the local
AND the rutting is Type B (surface mix material failure) distresses occurred in low severity, they could be handled by
AND increment subsurface distress set counter by 1. maintenance strategies. But, if the extent of local distresses
becomes too high and if the average daily traffic is too high,
The categorization of rehabilitation strategies was accom- rehabilitation would be required rather than maintenance.
plished at different levels within the logic tree. At each major The numerical thresholds for this transition are shown in
branch point in the logic tree in Figure 1, decisions have to Figure 2.
be made on the basis of available evidence provided by the In PARES there are four maintenance strategies (Reha-
user. It is instructive to list here some of the rules that affect bilitation Strategies 2, 3, 20, and 21 in Table 2) that can be
some of these decisions. For example, there is the strategy of recommended if the distress conditions are not sufficient to
do nothing, which is an "escape route" strategy for the expert warrant new construction, or these same four maintenance
system when no distresses are entered by the user. If PARES strategies may be recommended by PARES as "additional
receives no distress types, it must assume that no rehabili- construction" (see typical example later) if they are associated
tation is required. The rule governing this escape is with recommended rehabilitation strategies as explained above.
These maintenance strategies are also considered escape routes
RULE NUMBER: 74 from PARES, since the primary purpose of PARES is re-
IF all distress type sets contain nothing, habilitation. Typical maintenance rules follow:
THEN do nothing.
RULE NUMBER: 84
To decide whether a distress is maintenance only or re- IF the type of strategy needed is maintenance only
habilitation, a typical decision rule is OR maintenance with rehabilitation
AND maintenance distress set contains edge/centerline
RULE NUMBER: 82 cracking
IF subsurface distress set contains nothing THEN crack seal.
AND surface cracking set contains nothing
AND surface mix distress set contains nothing RULE NUMBER: 91
AND average daily traffic is less than 5,000 IF the type ·Of strategy needed is maintenance only
AND total extent of maintenance distresses present is OR maintenance with rehabilitation
less than 50 percent, AND localized distress set contains alligator cracking
THEN the type of strategy needed is maintenance only. THEN dig out and patch.

The difference between a maintenance strategy and a re- Whether additional pavement strength is needed is deter-
habilitation strategy is not clear in many repair situations. mined by the difference between the required new pavement
Although it is easy to determine that railroad crossings and structural number (SNnew) and the existing structural number
bridge depressions can almost always be addressed with main- (SN 01 d). If SNnew minus SN01 d is greater than 0.30, an overlay,
tenance strategies, other distress types that are typically non- or recycled pavement, is recommended because additional
local phenomena may· not be easily categorized as mainte- strength is necessary. In PARES the new SNnew is calculated
nance problems. Generally local distress types such as bumps using the 1972 AASHTO guide (16) and SN01ct is provided by
and sags, corrugation, depressions, patch deterioration, and the user. The choice between strategies that rehabilitate the

> 70%
® ® ® ®- Maintenance only will handle
distresses
Maintenance <
Extent
703
® ® ® ®- Rehabilitation is needed (possibly
with maintenance)

< 60%
® ® ®
< 50%
® ® ®
< 1000 >1000 > 5000
< 5000

Average Daily Load (AOL)

FIGURE 2 Transition zones between maintenance and rehabilitation.

86 TRANSPORTATION RESEARCH RECORD 1374

pavement surface only and strategies that rehabilitate both The surface cracking matrices mentioned in this rule em-
the subgrade and the surface is made on the basis of whether body the knowledge from experts used to assess issues as-
a subsurface distress condition is present. sociated with surface cracking problems. The surface cracking
If a subsurface distress condition is present (e.g., swelling, problems are addressed by a variety of rehabilitation strate-
Type B alligator cracking, or Type A rutting) in significant gies, depending on the severity of the cracking and the number
extent, then a rehabilitation strategy that treats the subgrade of cracking-related distresses.
would be recommended. There are exceptions (such as if the In PARES, the surface cracking rehabilitation strategies
moisture condition of the pavement is stable) to such a rec- are divided into three categories (see Figure 1): small (crack-
ommendation, but the general rule is that a rehabilitation ing is not a problem), medium (cracking could be handled
strategy resulting from a subsurface distress will override re- with an interlayer), and large (cracking should be eliminated),
habilitation strategies resulting from other types of distress which are the output from the second matrix shown in Figure
conditions, giving rise to rules of the following form: 3 (S, M, or L). A choice among the three categories is made
through the use of the surface cracking matrices in Figure 3.
RULE NUMBER: 110 There are two surface cracking matrices. The first matrix
IF the type of strategy needed is rehabilitation shown in Figure 3 is designed to be used with each non-
AND structural strength of the existing road is not ad- maintenance, nonsubsurface distress type, where a repair level
equate for future design (L, M, S) is determined on the basis of severity (low, medium,
AND subsurface distress set contains swelling, high) and extent (in percentage, 0-100) of the distress type
THEN remove and replace with asphalt (reconstruction). in question. In other words, the first matrix addresses the
different types of cracking that can happen (e.g., alligator
Some rehabilitation strategies are used when no additional cracking, transverse cracking, etc.), and it also addresses those
strength is needed (i.e., either to improve the surface course distresses that can appear along with the cracking (e.g., rav-
or to improve the subsurface as well as the surface). A rule eling, rutting, etc.). The second matrix in Figure 3 is designed
to determine whether an increase in the structure of the road to take the repair level outputs from the first matrix (S, M,
is necessary is L) and produce an overall repair level (S, M, or L corre-
sponding to small, medium, and large), which is then used in
RULE NUMBER: 102 the rule base to choose among the three categories of strat-
IF (SNnew - SN01d) > 0.3 egies. The interface code used at the beginning of the expert
THEN structural strength of the existing road is not ade- system to get pertinent information from the user uses ~rrays
quate for future design for keeping track of the distresses and calculates the necessary
ELSE structural strength of the existing road is adequate overlay depth. Typical rules for surface cracking repair strat-
for future design. egies follow:

If no additional strength is required, PARES determines RULE NUMBER: 112

whether improvements to the asphalt concrete are required IF the surface cracking matrices should be used
through either recycling material or reconstruction. An ex- AND distress type set contains alligator cracking
ample of a rule in this situation is (note: rehabilitation strat- AND alligator cracking (fatigue Type A) severity is low
egies are conclusions at this point) AND extent of alligator cracking is less than 10 percent,
THEN repair level is small.
RULE NUMBER: 238 (Note: in this rule PARES would query the user about the
IF the type of strategy needed is rehabilitation type of alligator cracking.)
AND structural strength of the existing road is adequate
for future use
AND surface mix distress set contains bleeding
THEN hot recycle + chip seal
OR cold mill + overlay.

If additional strength on the roadway surface is required,

100
M
l L L
3 or more

3
L

M
L

L
L

L __
PARES determines whether the rehabilitation should be based EXTENT
50
,....... ,
-
-
,

M ~M~ L 2 M M L
# Distress
on subsurface problems or surface cracking problems. The 10
Types
decision whether additional strength rehabilitation is due to s (s·) M s M
subsurface or cracking problems is given in the following rule: 0 ·-- 1 L

Lo Med Hi s M L
RULE NUMBER: 111 SEVERITY REPAIR LEVEL ....
IF the type of strategy needed is rehabilitation
AND structural strength of the existing road is not ad- (M';.
' ...._.
Becomes L for ·Alligator Cracking and Rutting
equate for future design .'$··, Becomes L for Alligator Cracking and Rutting
' .. ·'
AND subsurface distress set contains nothing,
THEN the surface cracking matrices should be used, FIGURE 3 Surface cracking matrices developed in the
ELSE the surface cracking matrices should not be used. knowledge base.
 Ross et al. 87

RULE NUMBER: 139 RULE NUMBER: 252

IF the surface cracking matrices should be used IF crack seal
AND distress type set contains edge/centerline cracking AND overlay
AND edge/centerline cracking severity is moderate AND overlay + crack seal, all are recommended strategies,
AND extent of edge/centerline cracking is between 10 THEN crack seal + overlay will cover all the situations.
and 50 percent,
THEN repair level is medium.
ADDITIONAL FEATURES OF THE KNOWLEDGE
Finally, when PARES is at the level of abstraction below BASE
"surface cracking" in the logic tree (at the numbered circles
in Figure 1), its rule base recommends rehabilitation strate- If the user specifies that alligator cracking is one or more of
gies. Typical recommendation rules follow: the distresses, PARES will ask the user to specify the type of
alligator cracking present and to determine whether the dis-
tress is primarily related to surface or subgrade problems.
RULE NUMBER: 210 PARES asks the user this question whenever the first rule
IF existing cracking should be addressed with an interlayer involving alligator cracking is addressed in the inferencing
AND surface cracking set contains alligator cracking, process. A typical question would be the following:
THEN asphalt rubber interlayer + overlay
OR geotextile fabric sheet + overlay The alligator cracking is (choose one or more of the following)
OR heater scarification + overlay.
1. Type A [alligator cracking in the surface only (i.e., due
RULE NUMBER: 224 to loading fatigue)]
IF existing cracking should be destroyed 2. Type B (alligator cracking as a result of a subgrade
AND surface cracking set contains alligator cracking, problem)
THEN cold in situ recycle + overlay 3. Type C [localized alligator cracking (i.e., low extent in
OR hot recycle + overlay a small portion of the roadway surface)].
OR cold mill + overlay
OR pulverization + overlay PARES can also query the user as to the type of rutting,
OR remove and replace with asphalt (reconstruction). if rutting is a distress indicated by the user. More information
on rutting is available from Pavlovich et al. (17). A typical
question is
RULE NUMBER: 237
IF existing cracking should be destroyed The rutting is (choose one or more of the following)
AND distress type set contains swelling
AND the moisture condition of the pavement is stable, 1. Type A (indicates a subgrade problem)
THEN cold mill + overlay. 2. Type B (indicates a surface mix material problem).
(Note: if this rule is invoked, PARES will query the user
about the moisture condition of the road.) Both of the preceding user questions involve visual inspection
of the road surface, and the user is assisted in PARES with
Another escape route designed into PARES is the situation a schematic illustrating Type A and B rutting.
requiring a rigid pavement rehabilitation scheme. Since PARES To assist in determining milling depth, if milling is neces-
is an expert system for flexible pavements, it treats a situation sary, or to determine the amount of crack seal to apply, if
requiring a concrete pavement as a special situation. The crack seal is necessary, PARES will query the user about the
following rule governs in this situation: depth of surface cracking. A typical question is

The depth of cracking is known to be_ (inches) (user fills

RULE NUMBER: 254
in the blank).
IF reconstruction is a recommended rehabilitation strategy
AND the average daily traffic exceeds 30,000
An additional feature in PARES is that it has the capability
AND the expected design life desired is greater than 10
to calculate mill depths and overlay depths for specific re-
years
habilitation strategies used in New Mexico; hence it is a design
AND the time since the last repair on this road is less
tool. If the user has provided PARES with surface cracking
than 10 years,
depths, these values are used for the milling depths. If the
THEN go concrete.
user does not know these values, PARES estimates the depths
(Note: if this rule is invoked, PARES will query the user
on the basis of the level of severity of the cracking (e.g., low
about the time since last repair.)
severity implies mill depth = 1 in., medium severity implies
mill depth = 2.5 in., etc.). The layer (structural) coefficient
PARES even has logic built into it to avoid overlapping (SC) for the old pavement is input by the user (see example
repair strategies. An example of this logic is the following later), and this value, along with the difference between SNnew
rule: and SN01d, is used to determine the new overlay thickness, if
88 TRANSPORTATION RESEARCH RECORD 1374

required. In PARES, new or recycled asphalt material carries with a layer coefficient (SC) of 0.2 for the existing surface
an SC of 0.4, and existing pavement carries a value for SC and other parameters as shown. The user specifies the input
between 0.1 and 0.4 depending on user judgment. that is known (PARES can run with incomplete input data).
Such quantities as the equivalent single-axle load (ESAL) and
the required new structural number, SNnew' are calculated
OPERATION OF THE PARES CODE from information provided by the user (see Figure 4a). The
rules used within a typical PARES session are a function of
Complete details of the physical operation of the PARES code the distress information provided by the user. Recommended
are addressed elsewhere (1). A single screen data entry in- rehabilitation strategies ranked according to cost of the
terface (see Figure 4a) minimizes the user's interaction with constructed surface (exclusive of shoulders) are shown in
EXSYS. Built-in error checking is accomplished on the in- Figure 4b.
formation entered by the user. The interface also contains an In this example, six strategies are recommended with the
explanation window, which is designed to be helpful to the top two being very competitive in cost according to New Mex-
user in terms of additional on-line help to explain what is to ico practice. To illustrate how PARES selects various reha-
be entered in each field (or section) for the infrequent user. bilitation strategies, the inferencing involved in the first strat-
In PARES a cost module was implemented to rank rehabil- egy, cold in situ recycling 2.5 in. + overlay 1.75 in., will be
itation strategies. The costs estimated are for construction of described. Only a few of the rules listed in this example session
the roadway surface only. The cost module was designed to of PARES appear in this paper.
be interactive with the user and to have considerable flexi- Since in this example alligator cracking was provided twice
bility. The user can use a default unit cost file, call his own by the user as two of the distress conditions, PARES queries
unit cost file, or change unit costs during run time in a PARES the user as to the types of alligator cracking; the user answers
session. "Types A and C." PARES uses this information to invoke
rules 2, 34, and 69 to classify the distress into surface cracking,
and it counts three surface cracking distress conditions. Then
TYPICAL EXAMPLE OF A PARES SESSION it invokes rules 76 and 77 to infer that rehabilitation is needed,
not maintenance alone (see Figure 1). PARES invokes Rule
A typical session of the PARES code is shown in Figure 4. 102 to determine that the existing strength of the pavement
This session was compared with an actual rehabilitation test is inadequate and that the pavement needs more strength.
job done in New Mexico in early 1988 near Cimarron. Figure Rule 111 then determines that the surface cracking matrices
4a shows the input for a pavement segment 0.2 mi in length, (Figure 3) should be used to estimate the needed repair level.
Rules 113, 148, and 190 £1.re then invoked for alligator crack-
(a) ing, longitudinal cracking, and transverse cracking, respec-
tively, to determine that a medium repair level is needed.
Number of Lanes: 2 Roadway Length: 0.20 miles Rule 198 uses the results of the recommended repair level
Roadway width: 24 feet Regional Factor: 1.5
Design Life: 10 years ESAL: 2920000 and the fact that three surface cracking conditions exist to
SC for old AC: 0.2 Serviceability Index: 2.0 determine that destruction of the existing cracks is necessary
Estimated R-Value: 45 Existing SN: 1.8
before any new overlay. Finally, Rules 224, 228, and 229 use
ILH..,ln&:.'-''-'"'"'
~ ~ Extent these results and conduct simple calculations to recommend
Transverse Cracking Low 1000 feet that the rehabilitation strategy should first cold in situ recycle
Transverse Cracking Medium 300feet
Alligator Cracking Low 15% 2.5 in. of the old pavement structure (this repairs the existing
Alligator Cracking Medium 5%
Longitudinal Cracking Medium 300feet cracks), then add 1.75 in. of new overlay (this reinforces the
pavement structure to the recommended strength).
Calculated future SN: 3.068 Exit (Y/N')?:

CONCLUSIONS
..... this window provides the user with explanations and instructions for any of the fields
in which the computer cursor is awaiting data entry ..... .
An expert system for preliminary pavement rehabilitation de-
sign for flexible pavements in New Mexico has been described.
(b) Its implementation when compared with an actual New Mex-
Rehabilitation Strategies ~
ico rehabilitation project is illustrated. The rehabilitation
strategy used on the actual project was one of the recom-
Cold insitu recycling 2.5 inches+ overlay 1.75 inches 13,636
Hot recycling 2.5 inches+ overlay 1.75 inches 13,693 mended strategies developed by PARES. The system cur-
Cold mill 2.5 inches + overlay 4.25 inches 26,660
Cold mill 2.5 inches + interlayer + overlay 4.25 inches 33,700 rently is in use in New Mexico and has been shown to be both
*Pulverization+ overlay 59,523 a rapid initial estimator of rehabilitation job costs and a tool
*Reconstruction - remove and replace asphalt 69,661
for new engineers to understand and learn current procedures
Additional Construction used by expert designers.
Dig out + patch 5,333
(These are supplemental maintenance strategies)
ACKNOWLEDGMENTS
*This strategy includes an estimated overlay pavement depth

FIGURE 4 (a) Typical input screen and (b) typical results The authors wish to thank Doug Hansen, Richard Lueck,
screen with cost ranking in PARES. Robert Olivas, James Stokes, and John Tenison of NMSHTD
Ross et al. 89

for their efforts in contributing pavement rehabilitation ex- 7. K. Hall, M. Darter, S. Carpenter, and J. Connor. EXPEAR: A
pertise in the formulation of the rule base and for their sug- System To Assist the Design Engineer in Concrete Pavement Eval-
uation and Rehabilitation. University of Illinois, 1987.
gestions to the project team for improvements to the rule
8. H. Aougab, C. Schwartz, and J. Wentworth. Expert System for
base. Appreciation is also expressed to George Luger and Pavement Maintenance Management. Public Roads, Vol. 53, No.
Carl Stern for their thoughts on the structure and imple- 1, June 1989, pp. 17-23.
mentation of the rule base. This project was supported by 9. S. G. Ritchie. A Knowledge-Based Approach to Pavement Over-
the NMSHTD Planning and Research Bureau under HPR lay Design. In Transportation Research Record 1145, TRB, Na-
tional Research Council, Washington, D.C. 1987, pp. 61-68.
Project 88-03. 10. J. Hajek, R. Chong, R. Haas, and W. Phang. Knowledge-Based
Expert System Technology Can Benefit Pavement Maintenance.
In Transportation Research Record 1145, TRB, National Re-
REFERENCES search Council, Washington, D.C., 1987, pp. 37-47.
11. R. P. Tandon and K. C. Sinha. An Expert System to Estimate
1. T. Ross, S. Verzi, S. Shuler, G. McKeen, and V. Schaefer. A Highway Pavement Routine Maintenance Work Load. Journal
Pavement Rehabilitation Expert System (PARES) for Preliminary Civil Engineering Systems, Vol. 4, 1987, pp. 206-208.
Design. Report FHWA-HPR-NM-88-03. 1990. 12. D. Barnett, C. Jackson, and J. Wentworth. Developing Expert
2. M. L. Maher (ed.). Expert Systems in Civil Engineering: Tech- Systems. Publication FHWA-TS-88-022. FHWA, 1988.
nology and Applications. ASCE Publications, New York, 1987. 13. Paver: Pavement Condition Index Field Manual for Asphalt Pave-
3. S. Ritchie, C. Yeh, J. Mahoney, and N. Jackson. A Surface ments. American Public Works Association, 1984.
Condition Expert System for Pavement Rehabilitation Planning. 14. D. Huntington. EXSYS® User's Manual. EXSYS Inc., Albu-
Journal of Transportation Engineering, ASCE, Vol. 113, No. 2, querque, N.Mex., 1987.
1987. 15. T. J. Ross, H. C. Sorensen, S. J. Savage, and J. M. Carson.
4. C. Haas and H. Shen. PRESERVER: A Knowledge-Based Pave- DAPS: Expert System for Structural Damage Assessment. ASCE
ment Maintenance Consulting Programmer. 2nd North American J. Computing in Civil Engineering, Vol. 4, No. 4, 1990, pp. 327-
Conf Managing Pavement, Toronto, Canada, 1987. 348.
5. C. Haas, B. Ritchie, and J. Shelley. An Expert System for Pave- 16. AASHTO Guide for Design of Pavement Structures. American
ment Distress Data Analysis. University of Waterloo, 1984. Association of State Highway and Transportation Officials, 1972.
6. C. Haas, H. Shen, W. Phang, and R. Haas. An Expert System 17. R. Pavlovich, S. Shuler, and G. McKeen. Investigation of Math-
for Automation of Pavement Condition Inventory Data. 2nd North ematical Models for Asphalt Pavement Rutting. FHWA-HPR-
American Conference, Toronto, Canada, 1985. NM-84-92. New Mexico, 1989.
90 TRANSPORTATION RESEARCH RECORD 1374

Abridgment

Interlayers on Flexible Pavements

HoNG-JER CHEN AND DouGLAs A. FREDERICK

A study was initiated to evaluate the effectiveness of stress-relieving be used. This paper summarizes 7 years of evaluating perfor-
interlayers in reducing reflective cracking on asphalt overlays over mance of interlayers between asphalt pavements and overlays.
existing flexible pavements. Six lane-wide interlayers were in- Construction details were reported in the study's interim re-
stalled on three construction projects under New York's two stan-
dard overlay thicknesses (1and2Y2 in.). Strip interlayers 1 ft wide port (3).
were also placed on two additional construction projects to cover
individual transverse cracks. The strip applications failed within
1 year and were considered inappropriate for future use. Per- INVESTIGATION
formance of full-lane sections was monitored for 7 years. From
statistical analysis it is concluded that overlays with interlayers Materials
have lower average crack returns than those without them. Coring
showed that half the interlayers at cracked areas did not remain
intact. Resplts of a simplified life cycle cost analysis indicated Materials selected for this study may be classified into two
that interlayer treatments were not cost-effective compared with general categories: applied full-lane width and applied in strips
normal overlays. However, a 1-in. overlay with interlayer was over single cracks (Table 1). They were designed to provide
shown to be more economical than a 2Y2-in. overlay without one. stress-relieving overlay reinforcement and an impervious
Interlayer products should continue to be considered as experi- membrane to prevent water intrusion. Strip materials to cover
mental features. individual transverse cracks were supplied in 1-ft-wide rolls.
Full-width materials to cover an entire lane were used over
The purpose of an overlay is to extend the service life of an more extensively cracked pavements.
existing pavement by restoring its riding quality and correcting
its structural deficiencies. Reflective cracking caused by prop-
agation of existing cracks or joints in the original pavement Test Sites and Construction
up through the new surface, however, is a problem that has
long troubled highway engineers. New York conducted a study Three sites were selected for full-lane applications and two
(J) to address reflective cracking in asphalt overlays on rigid for strips. The two strip applications were on Routes 156 and
pavements. Methods investigated included bond breakers, 9, both near Albany. Route 156 and most of Route 9 were
membrane reinforcement, sawing and sealing, breaking and conventional flexible pavements, but portions of Route 9 were
seating, and thicker overlays. The sawing-and-sealing method composite (i.e., asphalt over concrete). Full-lane interlayers
[sawing joints in the new surface directly over those in the were installed on I-481 in Syracuse, Route 10 in Schoharie
original pavement and sealing them, expecting cracks to re- County, and Route 12 in Jefferson County. All were flexible
flect through the sawed joints (2, p. 47)] was found to be the pavements. Figure 1 shows a Route 12 cross section and its
most effective. For severely deteriorated pavements where structural components. New York's two standard overlay
slabs were not intact, breaking-and-seating or "rubblizing" thicknesses (1 and 2Vz in.) were both placed on Routes 10
methods were recommended. and 12; only the 1-in. overlay was placed on I-481. Figure 2
The reflective cracking problem, however, is not unique to shows layouts of the test sites. In all, there were 28 control
rigid pavements, but occurs over flexible pavements as well. sections and 22 treated sections. I-481 had both temperature
Reflected cracks lead to premature failure of overlays by al- and load-associated cracking. On Route 10, wheelpath alli-
lowing water to enter the subbase and cause loss of support. gator cracking was the predominant distress, plus some areas
In the early 1970s manufacturers promoted use of geotex- of edge cracking. Block cracking was extensive on Route 12.
tiles as stress-relieving interlayers. Proprietary stress-relief All test sites were overlaid in 1980 and 1981. Several problems
systems using rubberized asphalt made from waste tires were were encountered, including improper application ;ate, wrin-
also developed. These various materials received extensive kling during placement, and insufficient overlay thickness.
national attention, and it was decided to initiate a study to I-481 was given another 1-in. overlay course in 1982, when
determine whether they could be cost-effective in reducing the 1980 overlay was found to be only 3/s to Y2 in. thick. In
reflective cracking in overlaid asphalt pavements in New York some areas on Route 10, severe rutting and edge cracking
State. resulted in truing-and-leveling courses of up to 5 in. being
The benefits claimed for interlayers were (a) increased overlay placed before overlay.
service life and (b) cost savings because thinner overlays could

Performance Evaluation
H.-J. Chen, Engineering Research and Development Bureau, New
York State Department of Transportation, State Campus, Albany,
N.Y. 12232. D. A. Frederick, New York State Department of Trans- Cracks were sketched on survey sheets by surveyors walking
portation. 84 Holland Avenue, Albany, N.Y. 12208. along shoulqers. Individual cracks were measured in linear
Chen and Frederick 91

TABLE 1 Summary of Materials Tested

Brand Name Manufacturer Description

FULL-WIDTH APPLICATION
Petromat Phillips Fibers Corp. Non-woven polypropylene fabric
Mirafi Celanese Fibers Marketing Co. Heat-stable polyester and
polypropylene woven fabric 3
Typar/ReePavb E.!. duPont de Nemours & Co. Spun-bonded polyester fabric 3
Bidimc Monsanto Textile Co. Non-woven polyester fabric (Style C-22)
Propex Amoco Fabrics Co. Non-woven polypropylene (Style CEF4545)
Arm-R-Shield Arizona Refining Co. Blend of reclaimed rubber and modified
asphalt cement applied as a binder coat
with a subsequent layer of st~ne chips
STRIP APPLICATION
Bituthene W.R. Grace and Co. Polypropylene woven mesh laminated to
a layer of self-adhesive rubberized-
asphal t
Roadglas Owens-Corning Fiberglas Corp. Composite system of high-strength fiber
glass woven roving and a proprietary
hot asphalt binder (Roadbond)
Polyguard Polyguard Products Inc. Rubberized asphalt waterproofing
element with a polypropylene mesh
laminated to the outer surface

~ 1980 and 1981 fabric weights differed.

Introduced experimentally in 1980 as Style T-323 under the brand name "Typar";
reintroduced at a different weight in 1981 as Style T-376 under the brand name
"ReePav. 11
c Manufacturer discontinued production of this engineering fabric in 1981.

feet and alligator cracks in square feet. They were summed providing the extent of cracking in each section regardless of
for each test section and reported in linear feet. All sections its length.
were surveyed before overlay and annually through 1987. A Benkelman beam deflections were measured to examine
measure of performance called crack return percentage was structural adequacy and uniformity among sections at each
obtained by dividing the amount of cracking in 1987 by the site. Several cores were taken in 1987 from Route 12 to check
amount existing before overlay. Crack density was also cal- the manufacturer's claim of the fabric's ability to remain intact
culated, defined as linear feet of cracking per 100 lane-ft, and keep water from entering the subbase.

New Asphalt Concrete Overlay, I" and 2~" RESULTS AND DISCUSSION

Strip Applications
2~" Existing Asphalt Concrete Surface Course
Cracks in all treated and control sections reflected through
6" Existing Asphalt Concrete Base Course the overlay in the first winter. Cracks over the fabrics required
more maintenance than those without fabrics. Overlays with
interlayers raveled and delaminated, thus requiring patching.
12" Existing Gravel Subbase Course
Cracks in control sections only needed sealing.

Full-Lane Applications
Sub grade
After 7 years average crack return was about 20 to 30 percent.
FIGURE 1 Pavement cross section on Route 12. Crack returns were generally lower on interlayered sections
92 TRANSPORTATION RESEARCH RECORD 1374

1-IN. RESURFACING ON I 481

MM MM MM
2087

_ _ c3 I
~.,....,....&.-C-1--
C2

I• S88'
Bi dim •I Mirafi

1-IN. RESURFACING ON RTE 10

MM MM MM
1388 1391 1393
I

-
cs ~
Cl C2
l NB
~
CJ

l..2s8·J 11 246~1
Bidim Petromat

2~-IN. RESURFACING ON RTE 10

MM MM MM MM

--::-~
13

I ~,-----.-?_' I ~)f
13

CZ C3

I .. Jso·
Bidim
... .
I 300' 1
Mirafi
...
1 300' 1 300' 1
··~
Petromat DuPont
•·
1-IN. RESURFACING ON RTE 12
MM MM MM MM MM MM
3336 3338 3340 3342 3344 3346

Cl C4 c~

Cl C4 C6 NB

1. .I
so1·
Petromat
1.. AMOCO
498' .I 1..Miraf
498' .I
i DuPont

2~-IN. RESURFACING ON RTE 12

MM MM MM MM MM MM
33S6 33S8 3360 3362 3364 3366

cs ~B

cs
~ C6
~
~l . 499·.I
Petromat AMOCO
J ~o·I ~
AC-20 Mirafi
~
DuPont
w Rubber

FIGURE 2 Test section layouts.

than on the controls. Crack-return ratios of interlayered to and 2Y2-in. overlays. Average crack returns of control sections
control sections, after eliminating some extreme cases, ranged on Route 12for1- and 2Y2-in. overlays were 52 and 14 percent,
from 40 to 70 percent. Route 10 had a nonuniform condition respectively. The average for 1-481, after eliminating one ex-
due to the previously mentioned localized distress and edge treme section, was 26 percent.
failures. Deflection measurements also showed this nonuni- A t-test was used to assess effectiveness of interlayers in
formity among sections. Route 10 data thus were discarded reducing reflective cracking. The test sections are assumed to
from the analysis. Condition on Route 12 was relatively uni- represent typical conditions and are random samples. Testing
form and hence offered consistent results. Because of the was performed separately for 1- and 21/z-in. overlay sections.
added 1-in overlay on 1-481, the overlay (with an actual thick- Detailed testing [given in this study's final report (4)] shows
ness of about 11/z in.) had a performance between that of 1- that for 1-in. overlay sections on 1-481 and Route 12, the null
Chen and Frederick 93

hypothesis of mean percentage of crack return on treated of overlaying shoulders were included. Results indicated that
sections equaling or exceeding that on control sections is re- the 1-in. interlayer option is cheaper.
jected at the 0.025 significance level. On sections with 2:Y2-in. This simplified analysis could be viewed as only an ap-
overlays ~n Route 12, the null hypothesis can be rejected at proximate assessment of relative benefits among treatments.
a significance level of 0.05 and the alternative hypothesis fa- For interlayer treatments on pavements receiving both over-
vored. This supports the conclusion that interlayers are ef- lay thicknesses, the 3-year extended life does not warrant the
fective in reducing crack return. Using the same testing pro- expense. Interlayers are more cost-effective for 2Yz-in. over-
cedure on Route 12 for percentage of crack return on sections lays than for 1-in. overlays. Because the shoulder is involved
with 2Yz-in. versus 1-in. overlay, crack return for 2Yz-in. cover in the cost, the 1-in. interlayer alternative is more economical
is significantly less than with 1-in. cover, justifying the con- than the conventional 2Yz-in. overlay. The complexity and
ventional approach of using thicker overlays. difficulty involved in determining the lives preclude general
A comparison of average crack return between this study conclusions as to benefits of interlayers. Engineers should
and two previous New York State studies (5 ,6) shows that continue to consider this option on an experimental basis.
crack return on 1-in. overlays in this study is similar to those
of the two other studies, but those on 2Yz-in. overlays have
lower return percentages.
Other Considerations
Besides overlay thickness and interlayer treatment, a third
factor-crack density of the pavement before overlay-was
Factors other than performance and economy should be con-
found to affect return of cracking. An "increase-decrease"
sidered for overlay projects. If distresses other than cracking
relationship was found between these two variables for control
are present-for example, rutting, edge cracking, or local
sections with 1-in. overlays. This could probably explain the
depressions-they may call for a truing-and-leveling course
lower crack return percentages in 2Yz-in. sections just dis-
before overlay. This additional asphalt thickness would also
cussed because both Routes 10 and 12 had very high initial
reduce reflective cracking. Other construction procedures are
crack densities-202 and 484, respectively. Possible expla-
also available for cracked pavements, such as milling before
nations for this phenomenon were found and are discussed
overlay and cold in-place recycling. These procedures may be
in the final report (4).
more cost-effective than interlayers.
Results of coring on Route 12 pavements showed that about
half the interlayers did not remain intact when cracks reflected
through them. The benefit of keeping water from entering
the original pavement is under discussion. If water enters CONCLUSIONS
cracks in the overlay and is retained by an interlayer, it may
cause pumping, stripping, or huge pressure buildups. Whether 1. Test sections treated with strips all failed. These appli-
this is an advantage or disadvantage was not examined. cations consequently should not be considered for further use.
2. Statistical analysis indicated that overlays with full-lane
interlayers had lower average crack return percentages than
those without them. Testing also confirmed that 2:Y2-in.
Economic Analysis overlays had significantly less cracks reflected than 1-in.
overlays.
Route 12 was chosen for economic analysis to represent the 3. An increase-decrease relationship was found between
whole study. Overlay lives of the four alternatives were de- crack densities on overlaid pavements and crack return per-
termined by defining a failure criterion and extrapolating the centages on overlays. Possible explanations were also found.
reflective crack progressing trend (4). The resulting lives are 4. Coring results indicated that some fabrics did not remain
as follows: intact. The benefit of keeping water from infiltrating into the
Years in subbase is questionable.
Sections Service 5. Simplified life cycle cost analyses performed for the four
1-in. control 8 alternatives on Route 12 showed that interlayers were not
1-in. treated 11 cost-effective compared with normal overlays for both overlay
21/z-in. control 12 thicknesses. The 1-in. interlayer option was cheaper than the
21/2-in. treated 15
normal 2:Y2-in. option, but this analysis was limited in scope
On the basis of these overlay lives, 1987-1988 bid prices and based on many assumptions that may be subject to
on interlayers and asphalt concrete, and a discount rate of 4 discussion.
percent assumed by New York State Department of Trans-
portation, simplified life cycle costs were analyzed to see In summary' there is no question regarding abilities of stress-
whether interlayers were cost-effective alternatives (4). Three relieving interlayers, if installed properly, to reduce or delay
analyses were conducted: (a) 1-in. overlay with ;md without reflective cracking on overlays over flexible pavements. Their
interlayer, (b) 2:Y2-in. overlay with and without interlayer, and cost-effectiveness, however, depends on how long they can
(c) 1-in. overlay with interlayer versus 2Yz-in overlay. Results delay cracks from occurring. For heavily cracked Route 12,
of the first two analyses showed that interlayers offered no interlayers were more effective for 2Yz-in. than for 1-in. over-
cost benefits for either 1- or 2Yz-in. overlays,. given the relative lays. Because mixed results were obtained, engineers should
life of each alternative. The third analysis examined substi- continue to consider using various interlayer products pri-
tution of 2Yz-in. overlay for 1-in. interlayered overlay. Costs marily on an experimental basis. Other techniques should also
94 TRANSPORTATION RESEARCH RECORD 1374

be considered when a flexible pavement overlay is being velopment Bureau, New York State Department of Transporta-
designed. tion, 1983.
2. L. G. O'Brien. NCHRP Synthesis of Highway Practice 153: Ev-
olution and Benefits of Preventive Maintenance Strategies. TRB,
National Research Council, Washington, D.C. 1989.
ACKNOWLEDGMENTS 3. D. A. Frederick. Stress-Relieving Inter/ayers for Bituminous Re-
surfacings. Research Report 113. Engineering Research and De-
velopment Bureau, New York State Department of Transporta-
This abridgment is based on the results of a research project tion, 1984.
conducted by the New York State Department of Transpor- 4. H. J. Chen, D. A. Frederick, and J. M. Vyce. Inter/ayers on Flex-
tation in cooperation with the Federal Highway Administra- ible Pavements. Engineering Research and Development Bureau,
New York State Department of Transportation (in preparation).
tion, U.S. Department of Transportation. John M. Vyce was 5. J. M. Vyce. Bituminous Resurfacings on Flexible Pavements. Re-
the project supervisor, and Rickey L. Morgan helped collect search Report 31. Engineering Research and Development Bu-
field data. Acknowledgment is also extended to Luis Bendana reau, New York State Department of Transportation, 1975.
for his major contribution in determination of failure criteria. 6. K. C. Hahn. Effects of Preventive Maintenance on Pavement Ser-
viceability. Engineering Research and Development Bureau, New
York State Department of Transportation (in preparation).

REFERENCES Trademarks, products, or manufacturers' names appear in this paper

only because they are considered essential to the object of this document
1. J. M. Vyce. Reflection Cracking in Bituminous Overlays on Rigid and do not constitute an endorsement by the Federal Highway Admin-
Pavements. Research Report 109. Engineering Research and De- istration or the New York State Department of Transportation.