SCHOOL OF

CIVIL ENGINEERING

INDIANA
DEPARTMENT OF TRANSPORTATION

JOINT HIGHWAY RESEARCH PROJECT

FHWA/IN/JHRP-88/ 13
Final Report
LAYER COEFFICIENTS IN TERMS OF
PERFORMANCE AND MIXTURE
CHARACTERISTICS
Brian Coree
Thomas D. White

UNIVERSITY
JOINT HIGHWAY RESEARCH PROJECT
FHWA/IN/JHRP-88/ 13
Final Repor t
LAYER COEFFICIENTS IN TERMS OF
PERFORMANCE AND MIXTURE
CHARACTERISTICS
Brian Coree
Thomas D. White
 "

LAYER COEFFICIENTS IN TERMS OF PERFORMANCE

AND MIXTURE CHARACTERISTICS
FINAL REPORT

TO: H. L. Michael, Director September 5, 1988

Joint Highway Research Project
Project: C-36-55H
FROM: T. D. White, Research Engineer
Joint Highway Research Project File: 2-12-88

Attached is the Final Report on the HPR Part II Study titled, " Layer
Coefficients in Terms of Performance and Mixture Characteristics This .

report documents a study that examined in detail the origination of the AASHTO
layer coefficient. As a result, an analysis procedure was developed that
estimates the layer coefficient for asphalt mixtures used in Indiana. The
method is general and is applicable to most asphalt mixtures. Asphalt
properties, loading, rate, aggregate gradation characteristics and climatic
conditions are generally accounted for in the procedure.

Conditions and properties of materials at the road test were developed on a

distributive basis. A corresponding distributive analysis was conducted of
Indiana asphalt mixtures with respect to climatic regions in the state.
Previously the AASHO Road Test has not been examined on the basis of the layer
coefficient variation. Such an analysis provides a more logical evaluation of
the layer coefficient to be used in pavement design.

This report is forwarded to IDOH and FHWA in fulfillment of the objectives of

the study.

Sincerely yours,

Thomas D. White
Research Engineer

TDW/cab

A. G. Altschaeffl D. W. Halpin P. L. Owens

J. M. Bell R. A. Howden B. K. Partridge
M. F. Cantrall M. K. Hunter G. T. Satterly
W. F. Chen J. P. Isenbarger C. F. Scholer
W. L. Dolch J. F. McLaughlin K. C. Sinha
R. L. Eskew K. M. Mellinger C. A. Venable
J. D. Fricker R. D. Miles T. D. White
L. E. Wood
 Digitized by the Internet Archive
in 2011 with funding from
LYRASIS members and Sloan Foundation; Indiana Department of Transportation

http://www.archive.org/details/layercoefficientOOcore
 TECHNICAL REPORT STANDARD TITLE PAGE
1. Report No. 2. Government Acctmon No. 3. Recipient! Catalog Na

FHWA/IN/JHRP-88/13
4. Title and Subtitle 5. Report Dote

Layer Coefficients in Terms of Performance and September 5, 1988

Mixture Characteristics 6. Performing Organization Code

7. Authors) 8. Performing Organization Report No.

Brian Coree and Thomas D. White
JHRP-88/13
9. Performing Organization Name and Address 10. Work Unit No.
Joint Highway Research Project
Cvil Engineering Building 11. Controct or Grant No.
Purdue University HPR-Part II
West Lafayette, Indiana 47907 13. Type of Report and Period Covered
12. Sponsoring Agency Nome and Addres"
Indiana Department of Highways Final Report
State Office Building
100 North Senate Avenue 14. Sponsoring Agency Code
Indianapolis Indiana 46204
15. Supplemented Notes

Conducted in cooperation with the U.S. Department of Transportation, Federal

Highway Administration.

16. Abstract

A set of AASHTO Layer Coefficients has been derived for the ten (10) bituminous
mixtures currently specified by the Indiana Department of Highways (IDOH) This .

project was initiated by IDOH in response to a perceived need to optimize material

usage through the design process.

Layer Coefficients previously derived by other Agencies were found to be inadequate

for a variety of reasons. Indeed, Layer Coefficients as derived at the AASHO Road
Test were shown to be flawed in concept (Appendix A).

Recognizing that bituminous materials are very sensitive to temperature and time of
loading, a probabilistic approach was used to explicitly account for the range and
variety of environmental and traffic conditions encountered in Indiana. Equally,
in-place bituminous mixtures represent sample values of specification envelopes, or
tolerances: the "probable range of mixture parameters was used in the analysis to
derive Layer Coefficient Distributions rather than unique, deterministic values.

Two powerful methods were used in the analysis: the van der Poe l/ Ullidtz / Bonnaure
et_ a_l. method of predicting bituminous material stiffness, S , and the Rosenblueth
Point Estimate Method for dealing with variable distributions rather than mean, or
expected, values. It is believed that the resulting Layer Coefficients are more
realistic and represent the true range of behavior observed in practice.

17. Key Words 18. Distribution Statement

AASHO Layer Coefficients, Bituminous No restrictions. This document is available

Mixtures, Probabilistic Methods, to the public through the National Technical
Flexible Pavement Information Service, Springfield Virginia
22161

19. Security Clessif. (of this report) 20. Security Clossif. (of this page) 21. No. of Pages 22. Price

Unclassified Unclassified 104

 LAYER COEFFICIENTS IN TERMS OF PERFORMANCE
AND MIXTURE CHARACTERISTICS

by

Brian Coree
Graduate Research Assistant

Thomas D. White
Associate Professor of Transportation Engineering

Joint Highway Research Project

Project No.: C-36-55H

File No.: 2-12-88

Prepared as Part of an Investigation

Conducted by

Joint Highway Research Project

Engineering Experiment Station
Purdue University

in cooperation with the

Indiana Department of Highways

and the

U.S. Department of Transportation

Federal Highway Administration

The contents of this report reflect the views of the authors who are responsi-
ble for the facts and the accuracy of the data presented herein. The contents
do not necessarily reflect the official views or policies of the Federal High-
way Administration. This report does not constitute a standard, specification
or regulation.

Purdue University
West Lafayette, Indiana
September 5, 1988
 TABLE OF CONTENTS
Page
LIST OF TABLES iv

LIST OF FIGURES v

ABSTRACT vi

THE LAYER COEFFICIENT: ITS ORIGIN AND MEANING 1

The AASHO Road Test Analysis 2

STRUCTURAL LAYER COEFFICIENT DERIVATION 7

ODEMARK EQUIVALENT STIFFNESS CONCEPT 13

STIFFNESS MODEL FOR BITUMINOUS MIXTURES 17

MODEL INPUT PARAMETERS 23

AASHO Road Test 23
Initial Penetration 23
Time of Loading 24
Temperature 24
Volume Concentration of Binder 26

Volume Concentration of Aggregate 28

Road Test Variable Summary 29

IDOH Input 30

Initial Penetration 30
Time of Loading 32
Temperature 33
Volume Concentration of Binder, and Volume Concentration
of Aggregate 36
IDOH Variable Summary 37

LABORATORY RESULTS 39

Introduction 39

Materials Tested 40

Test Method 40

Results 44
 iii

SUMMARY 47

RECOMMENDATIONS 52

FURTHER RESEARCH 55

Crushed vs Uncrushed Aggregates 55

REFERENCES 57

APPENDICES

APPENDIX A
THE AASHO FLEXIBLE PAVEMENT DESIGN METHOD: FACT or FICTION? 60

APPENDIX B
AASHO ROAD TEST LAYER COEFFICIENT DISTRIBUTIONS 90

APPENDIX C
ROSENBLUETH POINT ESTIMATE METHOD 92
 .

LIST OF TABLES
Table Page

1 AASHO Road Test Layer Coefficients 9

2. Validity Limits on Ullidtz Equation 18

3. AASHO Road Test Initial Penetration 23

4. AASHO Road Test Air Temperature Distribution Parameters 24
5. AASHO Road Test Layer Thicknesses 26
6. AASHO Road Test Pavement Temperatures 26
7. AASHO Road Test Surfacing/Binder Thicknesses 27
8. AASHO Road Test Mixture Parameters 27
9. AASHO Road Test Volume Concentration of Binder 28
10. AASHO Road Test Volume Concentration of Aggregate 28

11. AASHO Road Test Summary of Variable Statistics 29

12. AASHO Road Test Mixture Stiffness Statistics 29
13. Indiana Specification AC-20 Binder Consistency 30

14. Indiana Binders Range and Mean Penetration 31

15. Indiana Binders Penetration Distribution Parameters 31

16. Speed Summary, Indiana Highways 1987 32

17. Indiana Time of Loading Characteristics 33
18. Temperature Statistics for 47 Cities in Indiana 35

19. Temperature ( F) Statistics for 47 Cities in Indiana 36

20. Indiana Pavement Temperature Distribution Parameters 36

21. Indiana Mixture Parameter Statistics 38

22. Indiana Summary of Variable Statistics 38

23. Materials. Summary of Extraction Tests 43

24. Comparison of Mixture Stiffnesses 44

25. Indiana Layer Coefficient Summary (North) 48

26. Indiana Layer Coefficient Summary (South) 49

27. Indiana Recommended Layer Coefficients 52

APPENDIX A
1. AASHO Road Test Measured Variables 63

2. AASHO Road Test Serviceability Data 66

3. AASHO Road Test Layer Coefficients 67

4. Retained AASHO Road Test data format 71

5. AASHO Road Test: Traffic and Failure Rates 75

 LIST OF FIGURES
Figure Page

1. AASHO Road Test - Layer Coefficient Distributions 11

2. Indiana - Mean Annual Temperature Contours 34

3. Load - Deformation Test Schematic 42

4. Computed vs Measured Moduli 45

5. Comparison of AASHO Road Test and IDOH Layer Coefficient

Distributions 50

APPENDICES

Appendix A

1. AASHO Road Test - Layer Coefficient Distributions 69

2. AASHO Road Test Serviceability Plots 72

3. AASHO Road Test Performance Schematic 74

4. AASHO Road Test Effect of Cracking on Serviceability 76

5. AASHO Road Test Section Serviceability Histories 77

6. AASHO Road Test Accumulated Serviceability Loss 79

7. AASHO Road Test ASL Plots 81

8. AASHO Road Test One-Year Survival Matrix 83
9. AASHO Road Test Survival Probability Plots 84
 ABSTRACT

A set of AASHTO Layer Coefficients has been derived for the ten (10)

bituminous mixtures currently specified by the Indiana Department of

Highways (IDOH). This project was initiated by IDOH in response to a

perceived need to optimize material usage through the design process.

Layer coefficients previously derived by other Agencies were found to be

inadequate for a variety of reasons. Indeed, layer coefficients as

derived at the AASHO Road Test were shown to be flawed in concept

(Appendix A).

Recognizing that bituminous materials are sensitive to temperature and

time of loading, a probabilistic approach was used to explicitly account

for the range and variety of environmental and traffic conditions

encountered in Indiana. Equally, in-place bituminous mixtures represent
sample values of specification envelopes, or tolerances: the probable
range of mixture parameters was used in the analysis to derive layer
coefficient distributions rather than unique, deterministic values.

Two powerful methods were used in the analysis: the van der Poel /
Bonnaure et al. method of predicting bituminous material stiffness, S ,

and the Rosenblueth Point Estimate Method for dealing with variable
distributions rather than mean, or expected values. It is believed that
the resulting layer coefficients are more realistic and represent the
true range of behavior observed in practice.
 i HE LAYER UUEFFICitNT: ITS ORIBIN AND HEANINS.

The layer coefficient has its origins in the AASHO Road Test
(1955-61). The 1972 edition of the AASHTO Interim Guide for Design
of Pavement Structures \1 5 states:

"2-^.h - Layer Coefficients. As mentioned in the previous

section? a coefficient must be assigned to each material used
in the pavement structure in order to convert structural
number to actual thickness. This layer coefficient expresses
the empirical relationship between SN and thickness, and is e

measure of the relative ability of the material to function as

a structural component of the pavement."

Consequently, the layer coefficient is an empirical measure of

contribution to structural performance. The overall structural

measure (capacity, adequacy), the Structural Number, SN, of a

pavement is given in the same publication as;

''2.H.3 - Structural Number. The solution of the design

equation in this guide is in terms of a structural number
(SN). The structural number is an abstract number expressing
the structural strength of pavement required for a given
combination of soil support value? total equivalent IS—kip
(BOkN) single axle loads, terminal serviceability index. and
r eg i ona I f ac tor .
!;

As a result, the layer coefficient is an empirical component of an

abstract number, both of which are inferred to be measures of

strenntfi. These definitions from the 1972 Interim Guide provide

 —

little guidance as to the means of measurement of either of these

parameters? indeed . they provide little substance as to their
signif icance.

in order to provide a more precise definition, recourse was made to

the full report of the AASHO Road Test (the several volumes of HRE
Specie! Report 61 i2) ) , in which the complete methodology and

analysis is givers in detail. Reference Has also made to a paper by

Carey and Irick (3) ? which provided the basis of the serviceability
measurement at the Road Test.

The AASHO Road Test Analysis

The following analysis is limited to the flexible sections of the

Road Test since the layer coefficient concept was not used in the

analysis of the rigid pavement sections.

In HRB 61E* the overall model of pavement oerformance is given ass

-c O -Cc O -c) fJLV (1)

( [pj

where p = the serviceability trend value-.

c = the initial serviceability trend value-

o
c = the serviceabi 1 i tv level at which a test section
l

was considered out— of -service,

w' = the accumulated axle load applications at the time
when d was observed.

p-.ft = functions of design (thickness) and load.

Conventional ly, equation (1) is rearranged in the following? more

familiar form:

loo <W) = ioa (p) + —G

io io ft

b = ioq
wherein c and p are synonymous, (in the original formulation; c
i '
i

referred to conditions and p to present serviceability).

Two variables in equation (23 are obtained by direct measurement 5

namely W_. the number of axle repetitions* and p , the serviceability,

d and d Here arbitrarily assiqned values of h,S and 1.5

o " t
""

resDectiveiy at- the Road Test).

Present serviceability index * p. is a composite parameter; measured

by direct evaluation of a select set of distress types (roughness?
cracking; patching and rut depth). The relationship which relates
serviceability to the direct measurements is the result of a

regression analysis of the observation of these distress types to

subjective ratings of pavement serviceability as perceived by panels

of both expert B.r$o non~e>^pert road users,

The ratine; of these panels, known as the present serviceability

ratines (PSR) •. is purely subjective. It is not a structural ratines.*

but more truly a "seat - of - the - pants" rankinp of the perceived ride
quality of a pavement. Specifically? it is the aggregated
perception of ride quality given by panels in the period 1957—1960,
There is no assurance that the results could be duplicated at any
lat-er time. Road user perception is not exoscted to be constant.
Consequently, the present serviceability index, p? is based on

perceptions and relationships founded in the late fifties/early

It is clear that the measurement of the serviceability index

provides an estimate of the ride quality of a pavement. It is not a

measure of strength.

At the Road Test, there were many concurrent experiments! out the
one which is of concern herein is the Main Factorial Experiment.
The Main Factorial Experiment consisted of 33E pavement sections?
each 100 feet in length. Forty eight (h8> sections in loop 1 of the
Main Factorial Experiment were not subjected to traffic: during the
Test, and served as unt-raff icked reference sections. Two hundred
and eighty four (E8m) sections were trafficked and included in the
-final analysis. Forty four (44) duplicated or replicate sections
were included in the experiment to provide a measure of statistical
robustness to the analysis.

The pavement sections were constructed of varying thicknesses of

three pavement materials.! a bituminous surfacing? a crushed dolomite

limestone base? and an uncrushed natural sand— gravel subbese. The
subgrade consisted of a 3 ft. thickness of an fi-6 \CL) soil placed
at slightly above optimum moisture content.

Each loop consisted of two lanes. Each lane was dedicated to

traffic of specific weight and axle configuration. Thus, the

measured axle repetitions, w, on each section is wholly specified by

two further parameters! L « the axle weioht
~ (in kips)* and L , axle
1 2
type (single or tandem).

Under this scenario, the present serviceability index? p< was

measured on each section at intervals of two weeks. Thus, for each
section, a history of traffic (W, L , L ) and present serviceability
1 2
index (p) was created.

it is reported in HRB SB61E that the serviceability data (i.e., the

present serviceability index) for each section ^as "smoothed".

Details of this are not given, but are assumed to be arithmetic
moving averages. Two records of traffic (w) were kept; unweighted
traffic (W5 was the true repetitions of axle (L L ) and was
1 2
unadjusted, while weighted traffic W was adjusted (2) in an attempt
to take into account seasonal variations in damage,

Takinc into account the assumed initial serviceability (p = c =

'
o o
h.E) and terminal serviceability (p = c = 1.5) conditions,
'
t t

equation CEj may be re-written:

—- = loa \p) +— id)

It mav be seen that this relationship is linear in loa (W) vs.

10
The parameters ioc (p) and lift represent loq (W) at p = i.5 and
10 10
the slope of the linear relationship respectively

Since the w vs. p history for each section was known by direct
measurement ? simple linear regression analysis provides estimates of
p and ft for each section. These parameters are defined as functions
of desiqn and load? and as such are defined 'a priori 7
as:

B B B
G = ft
O
+ B
O
(D+a5
4
iL+L?12 CL
2
3 ihbl

where i_
12 ? L are the a>;le weioht (kios) and tvoe?(i =

single? S = tandefii)

D is s composite parameter of pavement design?

the thickness index?
A ,6 are regression coefficients- and
L I.

ft is orovided such that the denominator on the

o
right hand-side of equation (3) may not become
zero

The thickness index. D? was further defined as the linear

combination of laver thicknesses? so that D = a

denote the thicknesses
112 t -§- a t
2
inches.
+ at
3 3
The
+

a ? where the t laver in

4 l.

coefficients a and ft were arbitrarily assiqned the values 1.0 and

4 O '

O.h respectively? (i.O is a nice number? and 0,4 is ''any small?

non-negative number" sufficient to prevent the denominator in

equation (3) from becoming zero (<•?> )

The previously estimated values of p and ft for each section were

then used in a regression analysis to determine the coefficients
A and B . This repression was initially undertaker! with the
0-3 0-3
coefficients a set to unity such that D = £t . Subsequently? a

variant regression was undertaken letting the values of the

coefficients a float. This latter repression vieidec D = SN =
Y,(a .i ) , where the thickness index D was termed the structural
1 I

number SN* and the coefficients a the layer coefficients.

Thus? it may be seen that the original definition of layer

coefficient (and structural number) gives no more significance than
secondary regression coefficients (p and ft being the primary
regression coefficients)

Two measured variables (W and d)< and five parameters (L

"12
? L *. t =

1 2 1

t i and t ) were used in the analysis, W* L and L are ourelv

2 3
measures of load (frequency (number of repetitions) 5 magnitude and
type) i the t are measures of pavement geometry and the

serviceability index p is a measure of the perceived pavement ride

qua! ity,

NU MEAiuRb Uh SiRENG;H wAb USED iH ThE HUAu TEST ANALYSIS

 STRUCTURAL LAYER COEFFICIENT DERIVATION

While the layer coef f icients, a , have been described as

i

"structural" from the original report SR61E up to and including the

1986 issue of the AASHTO Guide (-4) « this description 15 not

supported by an analysis of the original formulation. More rational

ar-.o accurate would be to ascribe to the layer coefficient the

attributes of "contribution to resistance to functional loss", in

other words to refer to it as the "functional layer coefficient".

Reference E8 describes a powerful reasoning as to why the original

Road Test analysis is flawed and why discussion of the true
significance of the layer coefficient is almost meaningless.

Conventionally j layer coefficients have been assumed to be

indicators of strength. Much effort has been expended in measuring
or estimating layer coefficients for new materials (or materials
other than those used at the Road Test )

Two predominant methods have been used to provide estimates of layer

coef f i c i ents s

i Direct: The layer coefficient, a , is related directlv to a

more conventional strenqth parameter? for example:

where A,a are experimentally derived constants, and

M is the Resilient Modulus (AASHTO T27-+)
R
 This type of definition is absolute? where the layer
coefficient is uniquely determined for the material by
reference to the chosen strength parameter.

Relative: The unknown laver coefTicient ? a . is Touno relative

to the "known" coefficient? a _, of a different material
ref
through the ratio o some strenoth narameter? for example;

I his type oT relationship will be shown to be analogous to

Odemark s eouivalent stiffness hvDothesis (5).

?
(M and B have
R
the same significance as in l5j above)

Since the direct type of layer coefficient measurement must in

itself assume equivalence at some point in the relationships the two

methods mav be considered identical! since for example:

ref
IV)
ef
are consxanT m tnis reiaxionsnip
ref

Using this concept;, structural or strength properties may be used to

estimate layer coefficients so long as the reference material is

defined both by its layer coefficient and the required strength

parameter

ihis is in its essentials the method advocated by the .966 AASHTO

buide? where the Resilient Modulus- H ? is the preferred strenoth
R
parameter. In the past? parameters such as the California Bearing
Ratio \kA5HT0 Ti93 ;
and Ha v snail Stability CASTH D1559) have been
used. Recent emphasis on a more fundamental measure of material
strenoth has led to the proposed use of Resilient Modulus (n ) in
R
preference to the more empirically derived measures used
previousl y
If this concept is accepted? then the problem of estimating layer
coefficients is considerably simplified. However, the question
remains as to what magnitude of the strength parameter should be
taken for the reference material? i.e.? under what conditions
(temperature? stress intensity or frequency? etc.) should the layer
coefficient - reference material relationship apply? This is a

particularly sensitive question in bituminous materials. For

example? the resilient modulus may vary by orders of magnitude

depending upon temperature? frequency of loading and mixture
parameters

Lonsiaering temperature alone.? it wiii De appreciated that the

resilient modulus of a bituminous material in August will be
substantially different from that in January. Not only will t'

values be different? but the ratio?

b i tumL nous

granular

(i.e.? the relative strengths of the bituminous surfacing and the

granular base course) will vary continuously throughout the year (or
climatic cycle).

In this light? the layer coefficients obtained at the Road Test are
seen to be time-averaqed values. The fact that the Road Test layer
coefficients are not unique deterministic values is well
demonstrated in Table 9 of HRB SR61E where the layer coefficients
for weighted traffic are seen to range as shown in Table 1.

Minimum Coefficient Ha j; i mum

. 33 . 78
"i

. 12 .23
2

07 Q 1 P
.
^3

Table 1. AA3H0 Road Test layer coefficients.

Range of values reported in HRB SR 61E (Table 9)
:imilar set of results was obtained from a detailed analysis of
 10

the raw AASHO data which provided the basis of the distributions
shown graphically in Figure i. The distributions in this figure
were constrained to be non-negative, since it is intuitively
unlikely that an irLcr&ased. thickness of a pavement layer would lead
to a Lesser strength.

Use of this distribution involves application of the Odemark

equivalent stiffness hypothesis (5) (see following section). As a

result? the following relationship may be stated:

h
C85
ref
ref

and s*1 denote tne iaver coetticient ana Hesnienr- noouxu;

R
x.

th
ot trie merer 'mixture) in the i
'

layer, respectively.

Recognizing that the layer coefficient of the reference material,

a is given, not by a single value, but bv the distribxition for
ref
the bituminous surfacing material shown in Figure 1, the

distribution of values of the layer coefficient of the unknown

material is defined if the distributions of the resilient moduli of

both the reference material and the "unknown" material are known.

The oniv reference for bituminous material is that of the surfacinc

materia* usee ax rne hhsi-iu hob.q i est » This approach was used in

this projscti with the difference that "mixture stiffness" was used
rather than resilient modulus*

The question arises as to which value should be used for the mixture
stiffness S , Should the summer value be used, when, due to the
m
thermal inertia of the pavement materials, the bituminous material
is in its "weakest" state, or should the spring values be used when
the bituminous materials are still relatively "strong" compared to

the weakened state of the subgrade and thus more prone to cracking?
This question becomes more complicated when comparing mixtures under
different climatic environments; how mav a bituminous material in
 11

OJ

I
(3

o
<
u_

jay

\ 01
r-

f H

i
oo !

• °= ;

'
' 1

i
> i

w /'
/

< i i

m /
m 1 i

\,
\
^V
OJ

< ^
m ^ .
-'
m
\
^,~-

oo r-- CO LO CO OJ

A0N3fl03Hd 3AIIV13H

Figure 1. AASHO Road Test - Layer Coefficient Distributions

 12

southern iexas (no freezing) be compared to the reference material

(Ottawa, Illinois) where freeze/thaw conditions are significant?

Traditional mathematical methodologies which might be used for such

analysis are complicated and require detailed knowledge of
high-power derivatives of each of the variables. Newer, recently
developed methodologies permit the comparison of material behavior
over the full range of expected conditions with only a knowledge of

cne Tirsx; Three sTamsi ice.i srsomenT-s ano 01 active correiati on

between variables. These techniques release us froi the need to

arbitrarily normalize material properties to some as ficiai state

(e.g. tvpically 70 F).

A further benefit of these methods is that a lack of precise

knowledge may be incorporated with advantage into the comoutat ions,
thus making maximum use of minimal information. Of particular
interest is the Rosenblueth point estimate method (6), which will be
used in the development of this report.

Using the Odemark equivalent stiffness method., the Rosenblueth

point estimate method and a. means for estimating mixture resilient-

moduli (or mixture stiffness), the distribution of layer

coefficients may be determined for any given bituminous mixture
under anv climatic cond i t i ons
 13

ODEMARK ELiUIv'ALEN i STIFFNESS CONCEPT

In 19^9j N. Odemark at the Swedish Highway Institute developed an

semi-empirical method for comparing different structures, while the

method is semi-empirical in fact, it has its basis in structural
enq i neer i nq

Consider a simpiv suooorted beam of lenqth L< of death t- * and of

1

unit width. From the theory of strength of materials? it is known

that the stiffness is qiven bv:

_ „ _ 3
hi
11 E
,

t
11

where E is the elastic (Young's) modulus of the beam

l

material
I is the second moment of area of the beam
l
cross-section about its neutral a>;is,

Consider a beam of a different material. E -, depth t , also of unit

2 2
width, acting under the same load? then its stiffness will be;

(10)
2 3
E t
2 2

By requiring that the two beams have the same (or equivalent)
stiffness s then:

1=2 _ ,3 _ 3 ,

t T- E t
11 2 2

wnicn may oe re— arranaeo- tnus:

 14

- 1/3

t = t (12)
2 3

iis is the Ddemark equivalent stitfness concept, which re-stated

ivs that t inches of material of modulus E has the same stiffness
i i

; t inches of a material of modulus E > where t is aiven fav:

2 2 2

lt an n— layer structure \t ? t , i=i;HJ is consiaerea. "epeaxeo

I L

application of the Udemark principle results in:

irsT. layer It transformed

eauivalent thickness ot the second layer material (E )
2

— -. 1/3
~1

Loncepruai ly , tne upj laver now consists or it- -t-

inches of E material

mis coispC'Site layer may suDsequenx-. transformed into

eauivalent thickness of the next laver (E «t

3 3

rr- -vi'3
 15

c) Hepeated application of this principle yields:

fE 1 (E "1
fE 1
1 z 3
r. = z — + t;
~
r + . .+ t 15)
E 2 E~ 3 E~
n n n

where t inches of material E is eauivalent (in stiffness) to the

n ii

original n— layer pavement. It- should be immediately apparent that

the shape, or form, of the above relationship is identical to the

structural number relationship, i.e.:

where SN is equal to the equivalent thickness (which is the general

definition of the Structural Number 5 5 a , the layer coefficient, is
L

equal to the ratio of the layer modulus to that of a reference

material raised to the one-third power.

By e>: :endino these relationships, the tol lowing assumpiii can oe

made:

-\ i/3
i doh

The desired isyer coefficient for an 1DCH

i-doh
asphalt mixture.
The "measured' AASHO Road Test asphalt layer
ciasho
coefficient
A reference modulus..

The modulus of the AASHO Road Test asphalt

aasho
mixture.

and bv division;

1/3
idoh Ldoh_ idoh
( 1 d
i idoh aasho
etas ho aashiQ aasho
mis reiaxionsniD provides me oasis ot the method used in this
 16

report. Since the layer coefficients (and their distributions) are

known for the AASHO Road Test (as indeed are all other pertinent
mixture and performance data)? it requires only a knowledge of the

parallel parameters of the IDOH mixtures to estimate their layer

coeff ic ients,
 17

SilFFNESS MODEL FOR BITUMINOUS MIXTURES

In the above development of the Odemark equivalent stiffness method?

use was made of Young's elastic modulus? E. in the remainder of

this reoort? use will be made of binder stiffness. S < and mixture
b
stiffness? S = The reasons for this btg two-fold. Bituminous
materials are inherently visco— elastic? and as such their
stress-strain behavior is both non-linear and influenced greatly by

temperature and time (or frequency) of loading. While it is

possible to measure Young's modulus in bituminous materials? there

does not exist a good method to relate mixture parameters (binder
content? void content? etc.) to measured results^

Making use of binder arid mixture stiffnesses? which are analogous to

elastic moduli? permits use of existing models relating engineering
mixture properties to stiffness? and through use of a Creep test,

allows laboratory verification of the estimates.

Van der Poei(7) showed that the properties of bituminous mixtures-

are primarily influenced by the properties of the bituminous binder

and by the volume concentration of the mineral aggregate within the

mixture. The influence of these components on the properties of

mixture models has been confirmed by Alien(S').

A nomograph which yields the behavioral properties of bituminous

binders was published by van der Poei . With this nomograph? the
binder stiffness S ? mav be estimated throuqh a knowledge of the
b
tonder penetration (ASTM D5) ? binder softemnq point ;ASTM D365?
penetration index ilOnll), frequency (or duration) of loading and
temperature

i he accuracv of binder stiffness S ?

'
estimated by this metho
b
compared to laboratory results is reported by Pel 1(9) to be within
 18

factor of two. Granted that this level of accuracy is poor compared

to strength measurements of other more conventional engineering
materials, it- nonetheless provides a good measure of the material
property and is widelv accented and used within the industry.- At

the same time the accuracy of results obtained from laboratory

tests, particularly for bituminous materials, is hardly precise? and

may or may not truly reflect the performance of such materials in

service.

ullidtz \Q) ? provided a regression equation based upon the van der

Poei nomograph which permits the estimation of binder stiffness

within 10 percent of the nomograph values.

RB

the binder stiffness (MPa)

the time of loading (seconds)
the Penetration Index \lO,2 2 )

the ASTM (D36; Ring and Bali Softening Point ( C)

the ambient? or material test temperature ( C)

uiiio-cz reports t-nax tl ; above equation yields XiU percent

correspondence with the ,'an der Poel nomograph when the variables
are within the Aim a* 1
given in Table 2, If the allowable
temperature range (T
RB
D is increased to 15 C to Z C? then the
corresDondina ace i ±20 Der cent.

Minimum Variable Maximum

. i 1 1 sec . so

E0 C ^ i — \ ) 60 °C

-i FT 4-1

iable d, validity Limits on Uiiidtz Equatio;

Effects of loading are incorporated in the above relation by t,

beinq the time of ioadina or the inverse of the freauency of

 19

loading. The frequency (f=i/t) provides a good approximation to the

vehicle speed ikm/h). Ambient climatic conditions are covered by
the temperature? T C. Material properties of the binder are
represented by the penetration index? PI; and the softening point
temperature? T
RB

Bituminous material properties used in road construction are not-

constant during the life of the pavement, Effects such as heat

during mixture production? and long term oxidation of the binder in

place can affect the material's properties. while changes in

property can not be accurately predicted? the methods of van der

Poel ? Ullidtz and others have provided the means whereby the

in-service (or recovered) properties of bituminous materials may be

reasonably estimated from initial properties.

An estimate mav be made of the recovered binder penetration? Pen ?

r

from a knowledge of the initial penetration? Pen (ASTM D5)

Pen = 0.65 Pen (E0>

me recove? sot'temna ooint (ASTM B36.> froi

db.Szs ioa \ Fen >

"lO r

ana the recovered oenetrst ion maex rroit:

S7 loo"10 (Per

while these relations (developed by Ullidtz (Q) ) are now frequently

used and accepted by the paving industry (although more commonly in

Europe) ? no published statement of accuracy or range of validity has-

been found. However? applying these results to a wide range of

tests conducted by the Bureau of Public Roads (BPR) \12, 13); the

overall agreement was found to be excellent (r >0„95 in all cases).

The BPR tests covered penetration-grade binders ranging from 60-70
penetration to 150—200 penetration? produced variously by vacuum
distillation? steam distillation? biowina (oxidation)? blending-
 20

propane fractionation? fluxing? and combinations of these processes,

from crudes obtained from throughout the USA and overseas.

The binder stiffness, S ,

"
is onlv one component of the behavior of
b
the mixture, Heukeiom and Klompil 4) developed a procedure to

combine the aggregate volume concentration (volume of aggregate as a

percentage of the total volume) and binder stiffness to estimate the

mixture stiffness? S . However, the Heukeiom and Kiomp procedure is
in

basse on a single air voids content (3 percent) and a restricted

range of mixture types. Other works by Samier and Basin (f 5),
Verstraetend S) and others have provided alternate relationships,
but also limited in application, usually to single mixture types.

In 1977, Bonnaure ex al.Cf?"), conducted extensive laboratory

experiments using both laboratory prepared specimens and cores
recovered from in-service pavements. The range of mixture types was
large, five surfacing mixtures (two asphaltic concretes, a German
Gussasf al t , a British Hot Roiled Asphalt and a British Open-graded
mix), five base-course mixtures (two coarse asphaltic concretes, a

Dutch gravel-sand asphalt, and two French bitumen-stabilized sands),

an asphalt grouting mixture and a f i 1 ier/bi turners mastic. The binder
contents ranged from h to 24 percent by weight of aggregate, and the
voids from to 3B percent.. They proposed a methodology "which

estimates the mixture stiffness, S ? valid for a wide range of

mixture types and loading conditions within the accuracy of the van
der Poel nomograph

The Bonnaure et aL method requires the computation of four mixture

parameters, each of which was found to have an influence upon the
mixture stiffness.

100
V . OC - 1 . Jtc 1 —r-, r— ( £3 a

B . + 5 68 x . i v + S . 1 35 x i v ( h:3b
 21

2
.37 V - 1
b
,6 I DO ( E3c
1.33 V - 1

where V is the volume concentration of aggregate (percent)

a
V is the volume concentration of binder (percent)
b

These parameters and the binder stiffness? S ? e,r& used to estimate

the mixture stiffness, S ? in the following relationships:

D.U X S.K
1
< S < 1U Pa? then:
b

log's J = 5 + S 4- d.UV6 \5 ) ', iDQt: ( E'HD )

m y v : y

ihuss tor any civer, mixture* trafjic speed and ambient temperature?
the stiffness of the mixtures S > may be estimated. It will be
b"
appreciated that there are five independent variables in this
analysis: initial penetration? time of loading? temperature? volume
concentration of binder and volume concentration of aggregate. What
values of these variables should be used? If the mean values are
chosen? then information at the extremes is lost? and such
information may be of particular importance. By virtue of using the
mean values? the final magnitude of the mixture stiffness will be
exceeded fifty percent of the time? - equally the fiiaqnitu.de will be
overestimated fifty percent of the time.

Methods exist whereby the distributions of the input variables may

be used to provide an estimate of the output distribution.
Classically? since the relationships are non-linear? recourse might
be had to the First— Order* Second— Moment (FOSM) method ( ± S') . However
fc>r the rel at ionsh ids involved in this analysis and the lack of
 22

precision or accuracy relative to several of the variables; the FOSH

methodology would be incredibly cumbersome? if indeed it is

tractable. Alternatively? the Rosenblueth Point Estimate Method

\f5*2€>) was adopted? which is in itself much more flexible in

application and more forgiving in terms of the quality of the input

information which may be used.
 23

SilFFNESS MODEL INPUT PARAMETERS

As mentioned previously} five input parameters need to be defined in

order to complete the stiffness model? these parameters are; initial

binder penetration, time of loading , temperature, volume
concentration of the binder and the volume concentration of the
aggregate, These variables have to be defined for both the
reference case, i.e., the AASHO Read Test, and for the conditions
prevalent within the State of Indiana.

AASHO Road iest

Initial Penetration

iqure Vj of HRB ooeciai hseDt 61B < 1 9 3 . g i ves cumulative

frequency plot of the binder penetration at the Road Test (based on
82 samples taken during the course of construction). The mean
penetration and standard deviation from this data are Given in Table

Mean Penetration 90.5

Standard Deviation 1.98

Table 3. AA3HU Road Test

Initial Penetration

The specitied binder was an 85—100 pen asphalt. Based on iable 3j

compliance was good, and the material of uniform consistency.

 24

Time of Loading

Traffic at the Road Test was held closely to a speed of 35 mph (5t>.3
kmph). No information is given in the records of the Road Test
about the degree of compliance with this figure. An assumption is

made that this stated speed is deterministic and unique. Since the

time of loading is approximately equal to the reciprocal of the

running speed (kmph)j the time of loading at the Road Test may be
taken as 0.0178 seconds.

Temoerature

Appendix B of the HRB Special Report ME tabulates the mean maximum

and mean minimum air temperatures for each of the 55 two-week
periods of the Test. Using this data the air temperature
distribution throughout the period of the Test was synthesized, by

summing 55 uniform distributions? each bounded by the mean maximum

and mean minimum recorded temperatures. The statistics of the
resulting distribution are given in Table h.

Mean Air lemp era tune i Fi C) 50 S . 10.M-4

Standard Deviation ( F, C! 1
SO 7 . il .^8

Skewness \(3 ) -0=19

ia.Dje h. hh&hU Koao iesx

Air Temperature Distribution Parameters

in order to make use of the Odemark eauivalent stiffness methodology

outlined above, the temperature of the pavement is reguired rather

than the air temnerature. No information of this kind has been
found by the Authors in the published Koad Tes- i tpra cure or

comouter data-base.

In order to synthesize the pavement temperature distribution

the air temperature distribution, recourse was made i
 25

relationship reported by Wi tczak {20) which relates the mean monthly

pavement temperature (MMPT) at any depth? z? to the mean monthly air

temperature (HHAT). In this project, predicting the temperature at

any depth at any particular time is of no interest. It is the
expected (or time averaged) distribution of pavement temperature
that is required. On this basis? it was feit that the level of

accuracy provided by this means would be adequate.

Since Witczak's formula yields the expected temperature at any given

depth t? recourse was made to the mean-value theorem ( £'7 > to find the

expected temperature in a layer of z— thickness (i.e.; in a layer

starting at the surface (2=0) and extending to a depth of z = t

inches)? from which:

3h loo" -——— -

1Q l' z+H) J
T
a
1.0 - ioq -
-ioltz+"jJJ
——
x = A o - + I (25)

;ne exDecT-ec na\ lent temperature surrace

layer z inches thick, i t-

is the expected value of air temperature (

The pavement surfacing thickness z? at the Road Test was not

constant. Design thicknesses were 2?3?h?5 and 6 inches depending
upon the section considered (disregarding any variation in

constructed thickness). Since the above relationship indicates that

the mean pavement temperature is a function of the layer thickness?
mo T-ne oinaer i +-F nes; is a function lavement
b
Temperature ? tnen tne use ot a singie vaiue of tne layer coerficient
for different thicknesses of the surfacing material may be
questioned

The AASHO method? however? does indeed assign a single value of

layer coefficient to the surfacing material? regardless of the layer
thickness. The distribution parameters of surfacing layer
thicknesses at the Road Test are qiven in Table 5.
 26

Mean Layer Thickness (in) H 00.

Standard Deviation ( i n i = US

iabls 5, hAsHQ Road Test

Layer Thicknesses

Applying Rosenbiueth' s point estimate method to Equation (25) , and

using the distribution parameters for air temperature (Table H) and

layer thickness (Table 5) ? the mean layer temperature and standard
deviation are given for the AASHO Roaci Test pavement surfacing in

Table 6=

Mean Layer Temperature \F • C) Pi .V 1 i . 05

Standard Deviahiin F ? C5 24, H 13.53

acie e>= hhsmu kobo i est

Pavement Temoeratures

routine uoncentratiors or sinoi

bvaiuaxion ox tne omaer volume concenxrat-ion DistriDution n

somewhat complicated by the fact that the bituminous material
characterized by the layer coefficient a at the Road Test was not ;
i

unique mixture. The Road Test "surfacing" actually comprised twc

distinct bituminous mixtures laid in combinations of differing

thickness. The surfacma mixture (S) was qeneraUy laid in a layer

i.5 inches thick, and the binder layer (B) varied in thickness tc

make up the reauired total surfacino thickness, as shown in Table 7,

 27

Surf ac i ng Th i c kness (in)

2 3 H 3 6

Surface Mix E . 1.5 1 .5 1 .5 i .5

Binder Mix - i . 5 E.5 3.5 4.5

Frequency i/iE I/H 1/3 i lh 1 / i E

Table 7. AASHu Road Test

Surf ac i ng /Binder Th i c knesses

Since each mixture? Surface <S) and Binder (B) possessed distinctly
different mix parameters, the combined effect was different for each
total thickness? (another reason why a unique layer coefficient is

not reasonable). However? recognizing the variation of thickness

and mixture properties allows an explicit treatment through the
Rosenbiueth Point Estimate Method by developing the required
distribution for the combined mixtures and thicknesses.

HRB Special Report 6iB reports the volumetric mixture parameters

from the density and extraction tests conducted on both bituminous
materials. The distribution characteristics are shown in Table S.

3 U"! face
-
B i rider

Mean Std .Dev Mean Std Jev .

Air voids ( %S/} 3 i £> 0.-3 4 . 0.55

Voids Fi 1 led (XF) 77.9 E.E1 66 E . s 70

Table a. AAbHU Road Tesi

Mixture Parameters

Denoting xne proportion ot ea.cn surTacing tniCKness taken up by tne

surface six \S> as p. and 11— p'; for the corresponding b:-.^der mix?

then the composite volume oercentaae of binder Cv 5 mav be Given as

b
fol lows:
 l.'H

CVCSjxFCB] ) , ( V[B3 xF[B1 )

1-pj
(1 - FlSj) ( 1 - FCBj 3

In Equation [263, VTj denotes the volume percentage of ear voids?

and F[] the volume percentage of voids filled with binder? S and B

relate to the surfacing mixture and the binder mixture respectively.

Using the Rosenblueth point estimate method yields the distribution
parameters given in Table 9 for the combined thicknesses of

bituminous surfacing. Note that in this case the skewness \'s is

significant

Mean V i% ) 11. 32
b

bra, L'ev, 1 .58

Skewness ft 0,58

labie ?. HHbriu rioaa ies-T-

Volume Concentration of Binder (V

volume ivOncentration or nqoreqate

In a similar way* the combined effects of thickness? air voids and

voids filled may be resolved by the Rosenblueth method to give the

values in Table 10j where

VL3] )
1

V = p
I - rib.

Mean v 34 h0 .

CL

- _i n n
=xd >

. uev 1 .79

Skewness ft
—0 20 .

iabie iu= hhshu Koaa lest

'/olume Concent at ion -

cri
1
'
Aqqreqete W
 29

Road Test variable Summary

All the required parameters for the distributions of the five

significant independent variables needed to derive the distribution
of the AASHO Road Test Surfacing mixture are thus available. These
are summarized in Table 11.

\
V3.r i ab 1 e Mean Std -Dev Skewness

Penetration Pen 90.5 1 .98 *

L

Time of Loading (sec) t 0.0178 * ¥r

Temperature ( C) T 1 1 . 05 13 .58 *

Volume Concentrati on
of Binder \% V 11.32 1,58 . 58
b

volume Concentration
of Aggregate (55 ) y S^ . H-0 1.78 -0.20
a.

Table li. AASHO Road Test

jmmary of variable Statistics

Utilizing these values with the Rosenbiueth point estimate method;

the following values for the distribution parameters for the

variables required by the Bonnaure et al. relationships of the

AASHT0 Road Test Surfacing were obtained. Since the stiffness S ir

m
equations Sh is given as a logarithm, the distribution parameters ir

sole 12 are oiven for the logarithm of S , loo (3 ).

rn "
"iO b

Mean loaiS ) 8 9238

.
m
t>T;a, Dev. . 6902

Skewness ft 2.2801

Table 12, AASHO Ho ad Test

nixture Stiffness (S ) Statistics
 30

IDOH Input

Initial Penetrati on

Historically. the State of Indiana has specified 60-70 grade

penetration binder. During the 1^70's, the asphalt industry changed
over to viscosity graded binders, this resulted in the State
specifying AC—SO grade binder. while these grades were not the only
grades specified* they have both, by far ? comprised the greater
percentage in actual use.

On any given contract, the specified grade of binder proposed by the

Contractor? and accepted by IDOH will? under normal circumstances,
be used throughout the Contract. At the same tifne? the consistency
of the binder provided to the Contractor will normally be closely
controlled within narrow limits by the manufacturer. A separate
Contract in another part of the State might use binder supplied by a
different manufacturer, or by the same manufacturer at a different
date. Each of the binders supplied by the manufacturer (s> may well

comply with the State Specifications, and be well controlled.

However, over a period of years the distribution of the consistency
parameters (penetration or viscosity) taken overall will be wider
than might be expected on any single Contract.

Binders specified under the previous penetration grading system, and

the current viscosity grading system have generally complied with
the specifications. Fortunately, it is only necessary to analyze
the distribution of the binder penetration rather than individual
data.

The 1935 IDOH Specification requires that the AC— HO grade binder
comply (inter alios) with the consistency criteria given in Table
13.

viscosity poise 1600 - S400 "able 13. Indiana Specification

kC-Eu render Consistency
Penetration 50 -11
 31

The effect of the change in primary consistency variable

(penetration -» viscosity) is in the wider range of penetration -For

an acceptable viscosity. In actual practice, the range of measured

penetrations is much narrower than that permitted in the
Specifications. IDOH no longer routinely test binders for

penetration? however McConnaghay Emulsions Inc. still routinely tes 4-

all binders supplied to them for penetration. They state that the
binder in most common use (i.e. 5 AC-EO) is still essentially
classifiable as a 60-70 penetration binder. However the range is

somewhat wider, more nearly 55-30 penetration.

Using this information? the distribution of the initial penetration

of binders in current use may be synthesized ? and are given in Table
is-.

Kange 5C < Penetration < 110

Mean 64

Table in. Indiana Binders

Range and Mean Penetration

using the beta distribution iSO) (for bounded data)? the distribution
parameters given in Table 15 ars developed.

Mean Penet ration 6h

iaoie 15. Indiana fanners
Standard Deviation L Penetration Distribution Parameter

Skewness ft .
3°
1

The resulting distribution complies with all the statements made by

IDOH Materials & Testing personnel and McConnaqhay personnel with
regard to the current distribution of penetration.
 32

Time of Loading

Whereas? at the AASHO Road Test, the traffic speed was tightly
controlled? the spectrum of speeds observed on the State Highways
and Interstates is much wider, Purdue University maintains an

annual Speed-Study program for the State for purposes of Federal

Speed Regulation compliance estimation. Consequently, there is a

solid basis for estimating the time of loading parameter within

Indiana, The data in Table 16 was abstracted from the 1937 Traffic
Soeed Report \21 ) ,

Mean Ebeed Std Dev

, vn xi
i

Urban Interstate 57 99 mph

. 3. SO 3. 7565

Rural Interstate 60.99 mph h.15 k. 5A88

A - lane State 57.21 mph A. 70 h. 120A

S - lane State 55 A i mph

. 5. 25 5. 5560

AI I Highways 57.67 mph A 98

. *

laDie is. speea summary, maiana mqnways ivts

(VHT = vehicle miles travelled)

The figures given in Table 16 for !

'AIi Highways" are the speed

parameters for the whole state obtained by weighting the observed
values by VMT, Expressing these data in terms of the time of
loading which is the parameter of primary interest? the values given
in Table 17 are found.

iriere is really insufficient Difference between the results for the

individual classes of highways to require them to be treated
separately. Consequently, the values given above for "All

Highways" will be adopted.

 33

Mean ( sec Std.Dev

Urban Interstate 0.0107 . 0007

Rural Interstate '--

.0103 u . OuU

h - lane State . 1 09 . 0009

2 - lane State 0.0112 . 00 1

All Highways 0.0102 . 0009

laaie I/, maiana

Time of Loading Characteristics

Temoer a ture

The National Oceanographic and Atmospheric Administration (NOfiA)

maintains el imatological data for each state and publishes both

daily, monthly and summary statistics for many towns av.d cities
within each state. Detailed summary data for h7 cities and towns in
Indiana are available \22) and have been used to provide statistics
of air temperature (30yr statistics). These data are summarized in

Table 18. On the basis of these statistics? the state of Indiana

was sub-divided into two distinct temperature zones? labelled North
and South? (Figure 2 shows the temperature contours throughout the

state and indicates the dividing line between North and South).
Converting the temperatures for the two zones and for the whole
state from Fahrenheit to Celsius and using the same method as for

the AASH0 Road Test to estimate the pavement temperatures? the

summary statistics are given in Table 19.

On the basis of these statistics? the state of Indiana was

sub-divided into two distinct temperature zones? labeled North and
South. The summary statistics are given in Table 19 for both the
North and South zones as well as for the state as a whole.
 34

87.0 86.0
42.0
88.0
— — I
85.0
42.0

41.0 41.0

North

South

40.0 40.0

39.0 39.0

38.0 38.0

Location of cities
used in study

88.0 87.0 86.0 85.0

Figure 2. Indiana - Mean Annual Temperature Contours.

 35

City/Town Zone Mean F Std Dev Beta ft

Anderson N 51 .1793 17.8599 -0 03 29.

Berne N 50.9975 18. 1346 -0 1 36

.

Brookvi lie N 5 1 34 58
. 17.9767 0.01 U6
Cambr i d q e C i t N 50.03 92 18.1831 00 23.

Columbia City N 49 .57 03 18.4568 -0.0132

Columbus S 53 . 4 1 35 1 7 85 1. — .01 68
Crawfordsvi i le N 50 . 34 58 18.5126 -'J.Od i i
Deiph N 51 .5375 1 8 2027.
— . 04 05
El wood N 49.84 58 13.379=? -0.0221
Evansvi lie B 57.53 08 1 S520
7 . -0.01 88
Frankfort N 50 39 58
. 13.3836 — .0331

Goshen N 49 53 75
. i 8 2235. -0.02 19
Greencast ie 3 53 . 38 83 18.5189 -0::046l
Greenfield N 51 .6250 18.3169 -0.03 63
Greensburg N 51 94 58
. 17.4352 -0.01 99
Hobart N 50 65 43
. 18.4627 -0 . 04 23
Huntington N 50 . 70 00 18.2297 -0.01 74
Kent land N 5 i 35 00
. 1 3 8562. -0 05 66.

LaPor te XH 49 73 08
. 18.5631 -0.0386
Marion N 49 .55 00 13.4438 —0 00 83.

Mount Vernon S 55 58 ci3

. 17.7614 —0 .0331
New Castle N 50.19 17 18.2610 00 24
.

North Vernon S 54.3458 17.3633 -0 03 58.

Oqden Dunes N 50 39 58
. 13,1794 —0 04 25.

Go I i t i c s 53.1250 17.6508 —0 . 03 68
Paol i c-
53.00 83 17.7968 — 1 97
.

Plymouth N 50 38 75
. 1 8 7433. -0.0102
Princeton 3 55 49 58
„ 1 7 5530. —0.0331
Richmond N 50 48 75
. 17.6566 -0 00 58.

Rochester N 43 93 75
„ 18.774S -0 03 44.

Rockvil Ie g 53 07 08
. 8,c :
":2i -0.0579
Rusnvi lie N 50 87 9H
. 18. 0099 -0 04 37.

Scottsburg 3 53 .71 67 17.8155 —0 1 6. 3

Seymour 3 53.6625 17.8888 -0.0185

Shelbyvi lie N 5c:U4 17
„ 13.1413 —0 04 42.

Spencer S 53.31 E5 18.3644 —0 . 1 26

Tell City 5 55 76 25
. 17.3456 —0 00 78.

Terre Haute s 53 60 43
. 18.8972 -0.0565
Valparaiso N 49.5135 18.3281 — 06 55 .

Vevay 3 54.89 17 17.3463 00 77.

vincennes q 54 09 58
. 1 8 208 . — 02 69.

Wabash N
1

hB '"^3 "?3
B 18.5948 -0.0191
Wash i not on S 55.6135 1 7 38 1. -0.04 71
Waterloo N 49.20 42 18.4533 —0 00 66.

West Lafavette N 50 . 03 75 18.5902 — 06 dB .

Wi namac N 49.79 17 18.4330 — .051

Winchester N 49.74 17 1 8 2 1 60
. -0 03 90.

BD 1 Er It?, ,iti bs i n i no baib.

 36

North Zone N 50 .3639 i8 3243 . -0 0294

.

South Zone s 54 . 0355 17.3495 —0 0347

.

Ail N+S 51 .6919 i 3 3395

. -0 0375
.

Table 19= Temperature ( F3 statistics to r 47 Lities in Indiana

JonvertiRC tie temperatures tor the two zones and the whole state
"enneiT r = i = ius ana usmc
AASH0 Road Test to estimate the pavement temperatures the. vain

Table 30 are obtained.

Mean ( C) std Dev .

Nortn 1 1 . B1 11.96

South 1 4.E4 11.66

All 13 = 47 13. IS

Table 20.- Indiana Paveme mperature Distribution Parameters

me conversion to pavemenx temperatures reiies on

that the range of layer thicknesses used throughout
Indiana is similar to that used at the fififil
j!-iu no at

inches with a mean of 4 inches.

'olume Concentration of Binder* and Volume Uoncentration of

Aggregate

The 1985 ID0H Specification covers ten mixtures for highway paving
from relatively coarse base course mixtures (maximum aggregate size
1+ inches) through sand surfacing mixture (maximum aggregate size
3/3 inch). Each of these mixtures is characterized in the
Specification by a gradation envelope and an acceptable range of
binder content. The binder content on any given Contract would be
tightly controlled, however? the overall variation for a given
mixture over a number of years is likely to be quite large.
Consequently « the distribution of binder content for each mixture
 37

has been characterized by the uniform distribution between the

specified binder content limits.

In the Supplemental Specifications to the 1985 Standard

Specifications ( 198?) (23) > the Percent air voids is specified to be
in the range h to 3 percent for all mixtures. Again, for the same
reasons* the air voids have beer: assumed to be uniformly distributed
between the limits.

The binder specific gravity was assumed to be constant at 1.081- and

the total aggregate specific gravity was assumed to be a constant
8.65 (aqqregates? both individually and in a mixture? vary in terms
of their specific gravity « however the error introduced by using
8.65 is assumed to be small when compared to other sources of
uncertainty). Table Si summarizes the volume concentration of
aoaisoste iv ) and volume concentration of bmdev 3
Cv ) cava for al <

a. b
of the IDQH mixtures.

It may be clearly seen that the two Ease mixtures are identical in

terms of the parameters of interest. However? in terms of their

gradations they are mutually distinct. The implication in this is

that while their strengths may be identical. their workabilities-

might be quite different.

IDQH Variable Su mmary

The distribution parameters for the five significant independent

variables needed, are qiven in Tables 81 and 88.
 38

MI V */.
aVJ ) c ( '-J ) pev^
a a '
b

BASE

5 83 6u . 1 , 22B 10.34 0. 688 — \) u

5B 83.65 1 . 288 10.34 . 683 — Q .5

B I ND ER

B 83 . 44 1 . 288 10.56 .683 -c ,5

9 83 OS . 1 . 220 1 . 98 - 679 -0 5

i i 83 60 . 1.214 11. 40 . 67B —0 ,5

SURF AC ING

8 81 .98 1 . 273 12.02 0.798 -0 .5

9 81 .26 1 . 296 1 2 . 74 . 842 -0 ,5

1 i 80 . 1 1.218 13.86 0. bb 7 -0 .5

12 79.74 1.31 14. 26 . 888 -0 .5

SAND 77 66 , 1 » 106 16.34 0.598 - . 5

Table 81. Indiana Mixture Parameter Statistics

ie standard deviation; p the correlation coeff icierri

i Variable Mean 8 to. Dev Skewness

Penetration Pen. 6h . 4 . 00 0.39

Time of Lo ad it) q ( sec ) t 0.0108 0.0009 *-

Temperature ? C)
North 1 1 , 67 *
South 1 4 24., 66
1 1 . *
12=12

22. Indiana Summary of Variable Statistics

 39

LABORATORY RESULTS

introduction

Some form of verification of the proposed method is required. Due

to the deficiencies in the original AASHO method as outlined in

sections i and E? and in reference E9? it is not possible to obtain

a measure of layer coefficients directly? nor is it feasible to run
laboratory tests which duplicate the time-averaging effects inherent
in the layer coefficient concept.

Two aspects of the prooosed model lend themselves to verification;

the Odemark equivalent stiffness method and the van der

Poel/Ul 1 idtz/Bonnaure et al . relationships. The Odemark equivalent

stiffness method has been adequately verified by others (e.g. (9))?
and has its basis in structural mechanics. On the other hand? the
van der Poel /Ul 1 idtz/Bonnaure et al. relationships avs less well

supported. The van der Poel nomograph is well supported? but as

previously mentioned is quoted as having an accuracy within a factor

of two. The uilidtz relationships reduce the van der Poel nomograph
to equation form within an accuracy of 10 percent. The Bonnaure et

al . relationships rely on the van der Poel nomograph and were

extensively verified by the originating authors? however the
applicability of these relationships to mixtures other than those
originally tested is not known.

The laboratory phase of this project was specifically undertaken to

provide some verification of the ability of the van der Poei/Ullidtz

/Bonnaure et ai. relationships to adequately predict the modulus? or
stiffness S ? of mixtures in Indiana.
 40

Materials Teste d

Six samples were prepared for testing in the Laboratory. One sample
•was made up in the laboratory to comply with ths 1985 IDOH Standard
Specification for sand mix surfacing. The other samples were cut
from core samples taken in the course of a recent JHRP research
pro ject \S5) . One of these (F-15) was cut to obtain a sample of fine
surfacing mixture? however since no single surfacing thickness could
be found of adequate thickness for testing (3 inches or greater)
this sample contained two distinct surfao.-q layers (appro..-. 3-h-

inch and S I/'h inch cut). The remaining four samples were selected
to represent coarser base mixtures. The basis of selection for the

base mixtures was to select visually comparable paired samples such

that one pair represented similar (crushed aggregate) mixtures from
both North and South zones? and similarly the other pair contained
rounded aggregate.

In this way? it was hoped that the gradation range (fine to coarse)
and the material range (rounded and angular) would be represented in
the testing program. At the same time, the contractor bias could be
reduced by selecting samples from different areas of the State*

Test Method

The uniaxial Static Creep test recommended bv Bolk et al , (24) was

selected to measure mixture stiffness. Other forms of creep testing
are available? but the majority rely on repeated cyclical loading
over two or more minutes. These latter are designed to simulate the
mechanism of rutting? while the uniaxial static creep test is more
relevant in the case of single load applications and provides
information in the time interval of typical traffic wheel passage.

Samples for the creep test were cut from the selected cores using a

diamond rock saw. The ends were checked for parallelism? and any
corrections made with a thin layer of Piaster of Paris. The samples-

were conditioned for at least 24 hours to uniformly attain the

required test temperatures (5 C? IB C and 39 O? a dummy sample with
 41

an internal thermometer was held in the conditioning chamber as a

reference to control the sample temperatures.

The samples were placed between the platens of the MTS apparatus. A

double layer of heavy-duty polythene sheeting was placed between

both ends of the sample and the platens to reduce end friction
effects, in addition the interface between the polythene sheets was
coated with silicone grease. A load of 265 lbs was applied as a

step function to the sample with the MTS machine, and load and
deformation measurements from the MTS 500 lb load ceil and vertical
LvDT were recorded by the automated data acquisition system
(LAB-TECH Inc. on an IBM PC-XT) at intervals of i/1000 seconds.

After testing, the samples were returned to the conditioning chamber

to'equi 1 ibrate to the next temperature. The sequence of temperatures
(5 C -* 18 C -» 30 0) was chosen to minimize the effect of sample
disturbance

The stiffness B of the mixtures was computed at load durations of

fli

1/100 second and 1/10 second, to emulate the effects of vehicular

loading at approximately 60 tnph and 6 mph_- Tne sample lengths L,

and diameters (for cross-sectional areas) were measured by

micrometer to 0.1 mm? five measurements of each provided mean
values). The stiffness was computed as;

Loaa/Hrea .,__ .

(2Ba>
Strain Deformation/bamDie Lenctr

<2Bb;
m 6(t,T)/L A*<5it 5 T;

me Designation ottii) denotes the deformation unoer a loac Duration

of t seconds at a temperature of T C. Results of the stiffness

calculations are piven in Ta:Hie E3. An idealized plot of the data
is cjiven in Finur e 3, showinq the reiationshin between load-
displacement and t i rr- = -
 42

300
Actual Idealized
(265) jL.
IT

200 -

<
O 100

_i i i i i_ _/ i i_ time
0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 10 0.11 (sec)

Background
system noise

O
<
a:
O
u_
LU
Q time
0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10 0.11 (sec)

Figure 3. Load-Deformation Test Schematic.

 43

Sarjnip

Sieve Sand F-15 V-9 F— 18 L-12 S-12

1 1 /£ 1 00 1 00 100 100 100 1 00

3/4 1 00 1 00 77 77 30 32

3/S i oo 79 h5 41 h7 52

# 4 96 57 33 27 27 33

i$ u 77 4E 28 21 P 27

# 16 M-9 31 21 17 16 20

i?" 30 33 32 14 14 13 13

# 50 19 1 6 3 7 5

# too IP. 7 2 3 3 2

.•? 200 5 4 i i 1 1

IT
% Bind 7 4 . 5.3 H.9 4.7 4.8 ^

MIX Sand ??? D Z) D 5/5B

lanis da, natenais, dt txrraci; ion lest;

Bulk density (AASHTO T166) was obtained by weighing the samples dry
and m water. Specific gravity was determined by Rice's method
(AASHTO T20S) . The binder content was determined by centrifuge
extraction (AASHTO T164= Method A) with tr ichioroethyiene and the
washed gradation (AASHTO T30) found. This information permitted the
calculation of the parameters needed for the Bonnaure et al.

relax ionshios (V and v. ), and allowed an estimate to be made of the

a. b
original mixture designations which are shown in Table 23.
 44

C 0.01 seconds 0.10 s IS CO nds

r Hix 1 TemP

Measured Computed Measured Computed

1307000 317000
1 BAND 1

I 1
1S 545000 609000 23h0;>0 321000
I I 30 f 133000 1 48000 70000 79000

2261 000 2140000 150S000 1440UOU

1131000 969000 646000
I \ 30 2S3000 274000 1 68000 156000

2 1 46000 2607000 2146000 1 83 1 000

\\i -9 t
\ 1 073000 1 248000 715000 "11000
1

1 V
3( 36 1 000 159000 206000

2196000 do66u0v 2196000 1 883000

I f— i"3 I
1464000 1293000 732000 /4du0u
3 48B00U 383000 r^ /*fOOO 221 000
\ \

\ L-1H \ 2176000 2451000 1451000 l699uw

i 1 ^ 1088000 1 1 60000 725000 664000
3 363000 339000 207000 194000
\ 1

IS- 13 1
'
2173000 2707000 i 4^3000 1916000
i

I
l 1086000 l.cf 99000 62 1 000 / J d'-.-'OO

1 3 278000 367000 150000 207000

icibie i^ariscn D f Mixture Sti"

ired vs Co>T;DLited (asi )

Result;

\e i esuits or nation and extraction tests are aiven i

ble 23, as areti^ates of the mixture designations, TabI

gives the res the laboratory stiffness tests and th
• ctea stiffness from the estimation orocedure, h number o
immary obsei> may be made. There is little difference i

r cne fneasLUomputed stiffnesses between the mixture

rounded 3.C5 its? i_-icf) and Tnose with crushec
erall agreeirisnt bet-ween the oisasurec
 45

B~-3

LI. J

O
"n
LL

£ "c sd) i ]["npo|»| \j<b 4n dm 03

Figure 4. Computed vs. Measured Moduli.

 46

using as a measure of agreement the ratio "measured/computed" ?. the

mean value is 0.96 and the standard deviation 0.152 which gives a

probable range 'mean plus/minus 3 standard deviations'* of 0.50 to

1.4is this agrees well with the van der Poei and Bonnaure et ai,

estimates of the accuracy of the comnonent parts of the model.

it snouio de no Leo "cnaL ax tns z> l tesi temperanLsrej

def ormat ,M>rss we< e c y the order of 0,00002 i "ch?

Thus, a reading of O.O'j.'OP inch might reai.sOcailv

value between 0.000015 and 0.000025 inches = the

_ C' _ o _
ot the same type at the IB U and 3v L tef

correspond inaiv smal ler

me laboratory test results provide reesonaoie veriT ication tot

van dev Poel/Ul lidts/Bonnaure et ai. relationships. Then
certainly little evidence that the relationships are invalid.
 47

ine a istt idu 1 1 co ; parameters Tor one RHoiiu Koaa iest s 1UUM
miMtures have been aooiied to the van der Poel/Bonnaure aI

relationships. As a result, three possible approaches to estimating

layer coefficients for lEifJH bituminous mixtures may be considered.
These approaches address deterministic , a combination of

deterministic and probabilistic'! and probabilistic concepts,

i » Deterministic. (i.e.? only using the mean values of each

parameter or variable). In this approach? the expected
values of each of the final distributions were used alone,
such that:

1/3

C a
tdoh ^ l_aashoJ

11. nixec ueie 1 " "ni n : st i c/ Prodaoi 1 1 Stic . i ne value usbd

Hfi&riU laver coefficient- a is treated determinist icai ly
aasho
and assigned its mean value. The two stiffness variables,
and S are treated as distributions, such
>T. m
i-doh 'aasho
that

^ idoh J
idoh [^aashoj 8
of ™ 1
I aasho

et-ic. in wis approscn, an "cnree final variables;

and S , were treated as distributions
n m
ciasho Idoh
 48

Dp,,
doh
r
^ i J

idoh ^ aashoj
T^ aashoj1
(^

D oenotes r-hax; me oietriDUtionai parameters Kfj

of these variables have been used.

NuR H i

Fl I Deterministic Hi xed Probabilistic

He an Mean Mean a

HHist

-
5/5B 0.466 . 4y3 0.114

B I NDER

BB .h64 .4^8 . 1 00 . 8 W
9B 460 4S9 o 01
Si
. . . i

1 IB 0.456 . 486 . 1 1

SURFACING

. 438 . 544
. 4 44 . 477 . i 08 . 538
.435 . 1 03 . 530 . ~d 86
IBS .^3 i . 468 . 1 05 . 587
0.414 = I 04

ab^e 85. Indiana Laver Coefficient Summarv North,

hs may be seen tron the values TaDUiareo in sanies 'cZ> arm £o, i-ne

computed mean values of layer coefficient for each type

(deterministic, deterministic/probabilistic and probabilistic) are
distinct, and in fact differ from each other by an almost constant
amount. For the mixed deterministic/probabilistic model and the

probabilistic model? the standard deviations are essentially

constant in each, O.i and 0.3B respectively. There is not a large
variation between the coarsest, nsixture <5&5B> and the finest mixture
(BAND) in each climatic zone, 0.466 to 0.414 in the Northern zone
and 0.449 to 0.396 in the Southern;: and yet i>?itn reference to the
 49

AASHO mean (0.427). these represent a range of 109 percent to 97

percent in the North and 105 percent to 93 percent in the South.

A word of caution should be sounded here in relation to the results

of the probabilistic approach, since this relies on accepting the

distribution of the AASHO iaver coefficient a as beina true. It is

l

believed that while the distribution shown in Figure 1 is correct

for the AASHO model as or iqinai iy derived . it is not a true
reoresentat ion of the iaver coefficient in the oavemeni.

SuL! i H

MIX Deterministic Mi ;ed . Probab ilistic

Mean Mean a Mean a

BASE

5 / 5B . 449 . 4-77 0.0 98 0.539 0.2 89

BINDER

hB . 447 . 475 » VB 0.536 "?-

1

9B y . 4-4-j . 472 . 93 u tx33

. ,2 36
1 IB 0,439 . 469 . 98 . 530 .2 85

3LSRFAC I I;

8S 0,434 .464 . 99 0.524 0.282

9S .42 a 0.460 0.099 .5 B 1 2 79
.

1 IS . 4 1 . 452 . 1 00 0.509 .274

iES « 4 1 .450 . 1 00 0.506 0.2 73
Sand 0.396 ,436 .101 a490 d. 66
.

moiana Layer L-oeTT icier-r. cuvimarv 'souxn

the AASHO F;oao Test, (ii) the 5&:5B Base mixture in the Norther
zone? and On; the S^B mixture ?.r> the Southern ?.r;~e. These latte
were chosen to represent the "stiffest" and the "weakest" of th
Indiana mixtures. It will be seen that;
 50

A
i8
CD
Z
te™H

s I
~
TO u
//
Q n
2: tf5
«*
o (A \,
£

"' "5

, o

LO _S

^r

J L_ J L I I I I T-- t.

—o
I I I I I

O 'ID CO r-- iX? LO co o-j cd oo r- LO

O
X^uenbeJd oA8|iiey

Figure 5. Comparison of AASHO Road Test and IDOH Layer Coefficient

Distributions
 51

i. the 'envelope' of IDOH mixtures (represented by the two

extreme distributions mentioned above) is relatively narrow-

and

ii. that this envelope is considerably higher than the AASHO Road
Test distribution, i.e.-, 'stronger 7
,

Generally these may be explained thus;

i. The mixtures represented in the IDOH specification represent a

relatively narrow sub-set of asqhaltic concretes. Truly coarse

( open— qraded ) mixtures and truly fine mixtures (tending to

mastic) are not represented. The two most significant

variables within this envelope are the binder type? and the

temperature regime. In Indiana the binder type is essentially

homogeneous (60-70 pen or AC— SO) and the temperature
distribution is not dissimilar from that at the AASHO Road
Test

ii. Given that the temperature distributions at the Road Test and

in Indiana are similar, then the major discriminant variable is

the binder type. At the Road Test, the binder was

characterized as 85—100 pen, while in Indiana the predominant

binder (both historically and currently) may be taken as 60-70
pen. Thus the Indiana binders are 'stiffen 7
than that used at

the Road Test. This explains the major shift in the layer
coefficients which is not compensated by the generally slightly
warmer temperatures in Indiana.

The apparent shift between the deterministic and probabilistic

sets of values may be explained by the assymetrical \(3 &Q)
distributions involved. Had the distributions of a

S and S been symmetrical, then the tabulated values

in m
vdoV, aasho
of a would have been identical reoardless of the model,
Ldoh
 52
REVISED
July 9, 1990

RECOMMENDATIONS

The values tabulated in previous section represent the layer

coefficients for the ten IDOH specified mixtures as calculated by
the van der Poel, Bonnaure et al., Rosenblueth method and
approximately corroborated by Laboratory testing on real mixtures,
(one laboratory prepared sample, and five core samples).

The absolute magnitudes selected for use by IDOH based on these

results are a matter of in-house policy decision, however, the
relative magnitudes are considered to represent in-service reality.

It is recommended that the layer coefficients for the base and

binder mixtures be combined and the surfacing mixture layer
coefficients by divided as shown below:

North South

i ii lil 1 il 1
Hi
Mix Mean Mean a Mean a Mean Mean Mean

Base/Binder 0.46 0.49 0.10 0.55 0.30 0.44 0.47 0.10 0.53 0.29

Surface 8 & 9 0.45 0.48 0.10 0.54 0.29 0.43 0.46 0.10 0.52 0.28

Surface 11 & 12 0.43 0.47 0.10 0.53 0.28 0.42 0.45 0.10 0.51 0.27

Surface Sand 0.41 0.46 0.10 0.51 0.28 0.40 0.44 0.10 0.49 0.27

Table 27. Indiana Recommended Layer Coefficients

(NB. i, ii, iii, refer to Deterministic, Mixed and Probabolistic)

Any one column (necessarily the same North and South) of layer
coefficients may be used as long as it is consistently used. The
 53

mixed type of layer coefficient (column (ii) above) represents the

"safest' compromise between the uncertainty of the original AASHQ
layer coefficient distribution (Figure 1>? and the simplistic
deterministic layer coefficients (column (i) above).

The use of the word 'safest' above? is relative? and relates to the
uncertainty of information, A deterministic value may be considered
as having no uncertainty? while information contained in a

distribution is inherently uncertain. The 'flatter' a distribution

(i.e. $ the larger the variance with respect to the mean) the greater
the uncertainty. The trite value of the variable mav lie OLrt,\?it?h.&jrfz?

w.'.thm the limits of the distribution? but its exact value cannot be

known, Ta«inq the #8 surfacing mixture from the Souther" zone? it

will be seen that the probable range of the layer coefficient (mean
± 3 standard deviations; is 0.167 to 0,761 for the mixed tvpe? and
-0,322 to 1,370 for the orobab? 1 istic tvoe,, In this case, et'er; l r

an assximecl value "oris? ? the abso lute error in the mixed case will
be less than in the orobabi 1 istic case.

Using the mixed type of layer coefficient is also the more

conse-rvo.i ive option? since the magnitude is less than the
probabilistic value. In this case a marginally thicker layer will
be required by design? thereoy providing a measure of safety margin,

''re aAbHj fT'sihud dce^ not exu 1 1 c 1 1 1 v oerr-it the use of" other than a

single coefficient to represent the asphaitic component in a

pavement? however? considering the linear relationship between

the laver thicknesses,. SN = a =t + a .t
id
11 2 2
+ 3 »t there is no valid reason whv this mav not be transformed
3 3

11 "ll 12 ±2

.th

1 aver

Keqaraiess 07 which column ot ia.Die d/ mignt oe aaopten? it; is

co^siaered vital That the overall pavement thickness requirements as

 54

outlined in the Reference 28 be taken into account. The authors feel

that this requirement overrides the need to optimize the pavement

structure within the total pavement thickness; at least insofar as

the total pavement thickness has been shown to govern the life
capacity of the pavement, and given an adequate total pavement
thickness, the thickness design of the individual layers governs the

level of maintenance required throughout that life.

 55

l-umntM KhstHhLn

while the Authors believe that the objectives of this re 1

project have been adequately addressed? many questions have

raised? and potential applications of the resulting snethos

perceived

Crushed vs Uncrushed Aqoreoates

The method used did not differentiate between the performance of

crushed vs uncrushed aggregates, The form of the creep test used
was such that the duration of load was very short (0,01 to 0,10
seconds). It is likely that a Lariger load duration is needed to

discriminate between responses of the different aggregate particle

shapes.

The effect of particle shape in a bituminous mixture is further

influenced b/ the eve? ail gradation (grave! vs sand;, and may be

greatly influenced by the particle shape of the sand particles.

Particle shape and mechanical soundness are of greater importance in
open— graded mixtures and bituminous macadams? than in continuously
graded dense asphaitic concretes.

The effect of particle shape in a bituminous mixture should bs

studied, but only in a ful Istudy of such fete tors as stone/sand
ra t i os , aggre-ga teSb i nder ra t i os arid fill er/b i rider ra i os t .

Since all of these factors are closely related with each other 5 and
with the effects of voidage? it is very important that all of
these be studied toqether in a coherently Dlanned croiect.
 56

Mi xtures Studied

In the current study? the range of mixtures studied comprised a

sinal i subset of the full range of competent bituminous mixtures*

Since the study provides a means whereby the strength
characteristics of mixtures may be estimated based upon the mixture
characteristics? existing mixtures may be examined in terms of
sensitivity to variations in binder content? air voids? VMA end
voids filled. Equally? nsw or proposed, mixtures may be examined to

oredict octentxai behavior.

The van der Poel/Ul I idtz/Bonnaure et ai . relationships are based

upon unmodified residual bituminous binders. Increasing use of
rubber and polymer-modified binders may sufficiently alter the
behavior of mixtures such that these relationships no longer hold.
The power of this method is such that an effort to extend their
qeneraX validity to modified binders would be well <ust\f:ted.=
 57

REFERENCES

AASHTD Interim Guide for Design of Pav ement Structures ?

American Association of State Highway and Transportation

Off i c i a i s 1 97E
?

The AASHO Road Test - Rep ort 5 - Pave ment Research

Special Report 61E? Highway Research Board? 196S

Carey? t-j, and Irick? P.? "The Pavement Serviceability-

Performance Concept', HRB Record £'50, Highway Research
Board? I960

AASHTO Guide for Design of Pave ment St ructures American

?

Association of State Highway and Transportation

Off i c i a 1 s i 986
?

Odemark? N. "Undersokning av
? eiastici tetseqenskaperna
hos oiika jordarter samt teori for berakning av
belagomnga; enligt eiast ».cj. tetsteorm ?
i!
Statens
vaginstitut meddeiande 77? 194-9

Rosenbiueth E. ? "Point Estimates

? for Probability
Moments"? Proceedings National
? Ac ademy of Sc iences ?

USA vol 7E? £10? 1975

?

Poel ? C, van der "A ? General System Describing the

Visco—Elastic Properties of Bitumens and its Relation to
Routine Test Data"? She 1 B i tumen ? Reprint No. 9? Shell
Laborator ium-Koninki i ike, Amsterdam, 1954

b. mien? u, L, ? atitrness or solid— Liquia nixtures:

Theoretical Considerations"? Kentucky DOT? Research
Report h-om-? Division of Research? Lexrvngtrn? KY? January
1979

9. Ullidtz? P.? "A Fundamental Method for Prediction of

Roughness? Rutting and Cracking of Pavements" AAPT vol
? ?

hS, 1979

10. Pfeiffer? J. P.? "The Properties of Asphaitic Bitumen"?

Elsevier Press Inc.? Houston? TX? 1950

11. Pfeiffer? J. P.? and van Doormal P. M.?? "Rheological

Properties of Asphaltic Bitumen"? Journal of the
Institute of Petroleum Technologists? vol £E, 1936

12. welborn? J. V'., Haistead? \4. J. and Boone? J.

"Properties of Highway Asphalts - Part I? 35-
penetration grade"? Pub lie Roads vol 30? ?Mo. 9?
197— £07 August 1959
5
 58

13« Welborn, J. Y. Halsteadj W. j J. and Boone? J. G. ,

•"Properties of Highway Asphalts - Part II.. various

penetration grades"* P ub lie Roads vol 31 5 No. A, pp ?

73-99 October i960

?

\"t Heukelom, w. and Klomp, A.

, J. G=, ''Road Design and
Dynamic i_oadvnqS Proceedings A APT? vcl 33, Ann Arbor ;,

MI , 196

15. Samier J., "Module complete des enrobes

5 bitumirieu;-;:
evaluation du module de rigidite a partir de parametres
faci lenient accessible"? Revue Generale des Routes at des
Aerodromes? vol AE1 , May 1967

16, verstraeten, J. "Comportement en flexion repetee

, des
melanges bitumineuK - Module et lois de fatigue". Centre
des Recherches Rentiers? vol RR 156/ JV/ 1971 Brusseilsj ?

Mnnaure, r.« best, b,, Bravais, h,, ana uge, r., h new
method of predicting the stiffness of asphalt paving
mixtures*' Proc AAPT, vol A6, 1977
, .

Harm, G. J... and Shapiro? S= S., "Statistical Models in

Engineering", Wiley? NY, 1967

The AAShO Rr-ad Test - Reoort 5 - Pavement Research ,

Special Report 61B, Highway Research Board, 196E

Mitczakj M. w= "Design of Full Depth Asphalt Airfield

,

Pavements", Proc Third I n t e r n a t i o n a I Conference on thja

.

Structural Design of As phal P avements " vol 1, pp ,

550-567 , 1 97E

Traffic Speed Report Joint , Highway Research Project,

Purdue University, 1987

CIi mates of the Uni ted States 1? , pp 333-353, Gale

Research Company, Detroit, HI, 1985

Supplemental Spec i f i c a t i ons to the 1 985 Standard

Spec if icat ions Indiana Department of Highways, 1987
.,

Bolkj H« J, N. A,, "The Creep Test", Study Centre for

Road Construction, Record #5, Arnhem, Holland, 1981

Lindlyj J« K.) White, T. D., "Development of an Overlay

Design Procedure for Flexible Pavements in Indiana",
IDOH/Purdue university Joint Highway Research Project^
October 1987

Harr, M. E. "Reliability-Based
, Design in Civil
Engineering", MGre.w-Hill, 1987

Jeffrey, A, "Mathematics for Engineers and Scientists",

Barnes Eh Noble, Inc.,, New York, 1969
 59

Coree, B, J., and White? T. D.

, "The AASHO Flexible
Pavement Design Method: Fact or Fiction?", Paper
presented at the 67 Annual TRB Meeting. Washington,
DC. Jan. 198S
 APPENDIX A

THE AASHO FLEXIBLE PAVEMENT DESIGN METHOD:

FACT or FICTION?
 60

THE AASHO FLEXIBLE PAVEMEN1 DESIGN METHOD:

FACT or FICTION?

[This Appendix comprises the te;:t of a paper presented to the 7En.d

Annual Transportation Research Board Meeting? Washington DC? January

19881

i. INTRODUCTION

The American Association of State Highway and Transportation Officials?

(AASHTO)? flexible pavement design method is now well established:
indeed the final edition of a series of Interim Guides was recently
(1986! adopted by AASHTO. This document (i) adds significant volume and
some additional factors to the design process. Specifically, the new

AASHTO Guide incorporates the original development of the Road Test Data
with more recent additions relating to sub-surface (internal) drainage?
materials? reliability and others.

One interesting observation is that the developments in the new Guide

apnear to accept the original AASHTO formulations as a start inci point.
Taking this approach ignores the wealth of information contained in the

original AASHO Road Test data. New analytical techniques? additional

years of experience with pavement performance? and much hard thinking on
the results of the Road Test provide an opportunity that was missed in

developing the new Guide. In lieu of a fundamental approach? the

developments presented in the new Guide that affect the design procedure
were evolved by a combination of theoretical and empirical models. The
cause and effect of the actual AASHO Road Test performance is not
incorporated
 61

The AASHTG methodology is used hy a number of states in some form or

another. In many cases, states have modified details of the method to

better suit their peculiar circumstances and experience. Other states

have made modification only to the various constants in the original
formulation. Overall, some of the statistical constants and coefficients
originally developed in the analysis of the Road Test results have taken
on physical meanings that are not valid.

A recent research project initiated by the Indiana Department of

Highways <IDOH' T
required the Authors to determine the "layer
coefficients" for the ten bituminous mixtures currently being specified
in Indiana. IDOH uses the original AASHO formulation with their own set
of layer coefficients. However, IDOH has recognized that while the range
of mixtures specified represents a vast array of performance
characteristics, the assignment of a single layer coefficient to

characterize them is both unrealistic in engineering terms and

inefficient in financial benefit.

As part of the preliminary research task, a literature review was

undertaken: In the review of the literature, a specific effort was made
to establish the definition of layer coefficient and the original method

for calculating this parameter. The results of this otherwise simple

task initially proved to be highly disturbing- Subseouently , further
analysis highlighted some very interesting possibilities. This paper
deals with the findings and conclusions of this preliminary part of the
project

Much of this preliminary research is definitive., while some is still

speculative, since the project is on-going

2. LAYER COEFFICIENT

Considerable disagreement is apparent as to both the definition and the

recommended method of measurement of layer coefficients.

In the 1966 AASHTO Guide(l), the following statements are Quoted:

 62

I.E... ''The st rue tura.1 number is an abstrac t number. . . . converted.

to actual thickness of surf acing, basea.ndsubba.se, by means of
appropr iate I ayer coeff ic i en t s represen t i ng t he re lati ve s t reng t
of the construction materials"

"In ef fee t he I ayer coeff ic i en t s are based on t ?s.e e I as tic

t ,

moduli M and have been determined based on s tress and s trail

p.

calculations in a mul t i layered pavement system"

E.3.3 . it is no t es sent ial t ha t el as tic mc>du I i of t hes e

, ." . .

mater zals are charac terized. In general, layer coef f icients

deriued from, test roads or sat el lite sections are preferred. "

At the International Conference on the Structural Design of Asphalt

Pavements (E), Shook and Finn stated that:

"It isbelieved that the coef f icients a a a are tunc t ions of

3 12
, ,

the strengths of the various layers involved. At the present time

C 1 9622> , however , no en t i re I y sa t i s fee t ory t echn i ques are
available for defining or measuring these strength fee tors"

Perusal of existing and current literature reveals two predominant

methods have been adopted for estimating the layer coefficients of
bituminous materials; (a) A power law relating the layer coefficient to

the Resilient Modulus. M Ce.o. (1), Fio. S.53. and (b) based on
R
Oderoark's equivalent stiffness hypothesis (3)j an analogous relationship

is used, wherein the one-third power of the ratio of the material

modulus to that of a reference material (whose layer coefficient is

presumed known) gives the ratio of the unknown layer coefficient to that
of the reference material.

Assuming a relationship between strength and the layer coefficient is a

surprising extrapolation? since no measure of structural strength or
 63

adequacy was included in the data used to calibrate the AASHO model. The
only variables used in the AASHO model (4-) are given in Table i.

It is readily apparent that none of these variables are measures of

strength, although it is conceded that the layer thicknesses may be
indirect strength indicators. Further, no cognizance was given to the
annual cyciic subgrade strength. Instead? the subqrade strength was
assumed to be uniform throughout the Road Test, as indeed was climate.

Table 1. AASHO Road Test Measured Variables

FACTOR VARIABLE DESCRIPTION VARIABLE UNITS

TRAFFIC i ) Number of Axle Repetitions W #
i i ) Axle Weight L kips
. l
i i i Axle Type u 1 = single
2
S = tandem

PAVEMENT iv) Surfacing Thickness- t in.

1
v) Base Thickness t in.
2
vi ) Subbase Thickness t
3
m.
2 2
DISTRESS vi i Extent of Cracking c ft / 1000 ft
Extent of Patching p Tt /I 000 ft
Slope Variance (Roughness Sv
x i ) Rut Depth RD in.

The AASHO statistical flexible pavement design model was set up to

integrate the effects of (i) traffic? (ii) pavement (materials and layer
thicknesses) and Ciii) serviceability (or level of distress). These
effects were measured either directly or indirectly through the factors
in Table i. The development of the AASHO model is given below. A review
of this development shows that the layer coefficient is a secondary
parameter? whose significance is as a regression coefficient? or a

calibration constant. If any physical significance is attributed to the

layer coefficient? the significance should be relative to the 2~e-sisia.ri.cS'

to disiress?. rather than to the structural capacity of the layer

material
 64

3. AASHO MODEL

3. 1 Background

The basis of the AASHO model is a decay curve, wherein it is assumed

that the condition of a pavement will deteriorate? or decay? with
accumulated traffic (this implies a certain element of time). In order

to implement this model? the concept of functional pavement

serviceability was developed. This resulted in the composite measure
known as the present serviceability index? CPSI)?which results from a

regression equation relating the distress measurements given in Table i

to an aggregated subjective rating of the adequacy of the pavement by a

panel of adjudicators. A scale of to 5 was assigned to the panel

rating and subsequent qualitative serviceability index* A serviceability
of 5 is the ideal? or perfect? pavement. Pavement serviceability has
several important characteristics;

Initial Serv iceabi i i ty ? p ? is considered to be the

serviceability of the freshly constructed? untraf f icked pavement. The
ideal pavement has to be rare. In fact, newly constructed flexible
pavements at the Road Test reflected an average serviceability index

value of 4.E.

Terminal Serviceability? p ts considered to be that level of

serviceability at which the pavement is deemed to be no longer
oerformina its reauired function. The lower? limitina value of p at the
t

Road Test was 1.5.

?sent Ser v iceabi I itv ? d? is the measured? or estimated:

level of serviceability at any time during the lite of the pavement.
under normal circumstances then? o > d > d .
'
o ' t

Usinq these definitions? the AASHO model may be stated;

where w is the number of axle (18 kin) repetitions which will reduce the
 65

serviceability from p to p. p represents the number of axle repetitions

at terminal serviceability, p , and ft is a shape factor. Simply stated
the cumulative traffic at any time as a proportion of the traffic
capacity of the pavement is represented as a power function of the
proportion of usable serviceability consumed.

By a simple maxnemaxicai re-arrangement. , tne more t ami liar Torm

AASHO relationship is obtained:

loq
t
loo (W) = loo (p) + loq (p) + -
"10 10 10 ft

where G represents the logarithm of the serviceability ratio.

3.E Calibration

On each section of the Road Test. the present serviceability p, and

traffic M, were determined at intervals of two-weeks throughout the life
of the section. The unknown parameters, ft and p, were obtained by

regression analysis. The summarised data used for the original

repression analysis is in the form shown in Table 2. In Table 2> the
loqaritnm ox Trail ic is given tc serviceabilities of 3.5, 3.0= 2.5,
and 1.5. These five £53 points then provided the data for the regression
used to obtain estimates of ft and p. It should be noted that on a

log— log plot of w vs. S the relationship in equation (S) is linear.

Having obtained the two parameters, ft and p, for each section. it was
assumed that these parameters were functions of the section design
(i.e.? thickness) and traffic type. On this basis the following
functional relationships were assigned to ft and p:

p = A
o
* CD + i

ft — . h- "* B • ( L ) (3b)
 66

O 1 03CDO 1 ODO ^CDO * OD O T "JCDO^ffiO tj XO « XOO

oooooooooooooooooooooooooooo
c 0003^^n^«0>fl000n ,
^ 'i^>o«'0 3
,
00^n ,
n>fl'0

OJCVOJMOJOJOJfyCVCVr^KXfO^r-IKliw-if--,^^-^ ^ t^ C ^ ^T ^

13

E s i
** is c

c- ^ o
\o -^ ,\j m r^ ^ \o r- ^ o m o a, <r o o cc* o c\ o»HO"r\t;'0^
oto^cO'-''C^ co-nj", X!^o\irj'C' T <r ^ £
-
,
xt k\ ^ r\ 3) m m
x-oxcocococ^coaN xxo^o»ccoM^c--or- xxooxor-wo
•^•Ki-qijijsr^Ti a/j'c^'tf'ff^ij^^^ t -c -c ^ u; ji
-i) tf)

OHDH\OCOHt'i)On'!r^xfOHHOKO^O:^CO'0\'J",
OM-^tnroo-nnj[ coJx^O\m ^'rvj tT\o\oc03'JC^>.n^
£ < v- - ,

b~
< a
C*-iO'X)rD'DfDO\ro'J\aDXCnO\COO\0\r-0\f -OXXOj\Of-*-fO
'C "^ t flTT^ri^-n "^ tx *v
,

T'CTrLO^j'iDm^j'triO'rJi-OLT-.T

ea N > ^ 1DOHCCC^rlr(C>OTlO\^fl3fl^ 1 fl3sjiOl/1'OC

voouoLOco'^oJ^o»-t ocs-rHioojc\iinc-t-tyDinc-x ^o^ ^o\o\

v f t

c^W!i>cDr-a)CDaj5\r-xj^j\a33\tDfH>oc-xm7\3\?vr- ^ :>

X)-COXC-- O3\tHX'-0C-O
v

XX C~ X C^^t^'^XJOx'^X

HHHriHdHHHHHHHHHHriHtHHHrlHHrlHriW

IHHriHHHHHrlWrl H H ri r< ri H H ri rt rt H H

n^^i^^^nni^ni^nn r^K\(^t ^nfiniT^Kii^Kiio^^

Table 2. AASHO Road Test Serviceability Data

 67

In these eauations L i L and D (where D is the total pavement

1 2
thickness) were known for each section. The eicjht unknown constants A
0-3
and B were obtained by reoressioi analysis. A variant repression was
O—
conducted in which the thickness index u, was qiven by D = a +
11
• t a
2
* t
2
+ a *t
3 3
i such that ". . these coefficients were per mi t ted to vary
so thai three elements of the pavement structure might
the
each enter into the thickness index D, wi th a different weight ,

per unit thickness" t (3). This linear combination of the layer

thicknesses provided a better regression and was retained in the
model; the transformed thickness ( thickness index? D ) becoming
better known as the Structural Number- SN. and the coefficients a
1-3
the layer coefficients. The very well known dues for the layer
coefficients obtained from the Road Test are qiven in Table

abie 3. AhoHQ Road Test Laver Coefficients

3 "2
1 3

weighted Traffic . HH- . 1m- . i i

Unweighted Traffic . 37 . 1 H- . 1

the materials to which these coefficients relate are : a - Asohaitic

1

concrete? a - crushed limestone base? and a - sand-gravel base. As

2 3
listed? the raw traffic data was referred to as "unweighted"? while the
"weighted" traffic was adjusted in an empirical fashion to attempt to

take into account the varying effect of the annual climatic cycle upon
the subqrede. This technique was the only concession originally given to
the effect of climate or variable subgrade support.

3.3 Interpretatioi

The review of the AASHO Road Test analysis showed the "layer
coefficient" to be no more than a regression coefficient with no truly
ascribable physical or engineering meaning other than being a form of
scaling or normalizing constant. Certainly, no connotation of strength
could be assumed since no measurement of strength had been used in its
derivation. As a result of this situation? some concern deveiooed about
 68

being able to discriminate the strength and performance characteristics

of the asphalt mixtures being specified by IDOH. At this stage the
Authors deemed it wise to determine the variability (distribution) of
the layer coefficients found at the Road Test. The distributions were
obtained by using the original equations (3a and 3b) j but solving for

the layer coefficients section— by-section rather than globally. Thus the
iaver coeff icients, a ; for each section were determined. Since the
1-3
materials in each layer were designed to be uniform throughout the Road
Test? variation in the value of the layer coefficients was expected to

be a minimum.

The resulting distributions are shown in Figure 1. The only adjustment

made to these distributions was to constrain the layer coefficients to

be non-negative. Such a constraint may be explained on the basis that

regardless of whether strength or serviceability was to be considered,

an increasing layer thickness must lead to an increase in the overall
strength or serviceability. Consideration of the layer coefficient
distributions in Figure 1 caused considerable concern. The degree of
variation, particularly in the surfacing (the most closely controlled
material)? was surprising. In a specific review of the literature, a

measure of confirmation was provided by the values given in Table 10 of

HRB SR 61E, in which it was shown that the value of a varied from 0.33
i
to 0.83 (weighted), and from 0.33 to 0.73 (unweighted )

Considering the degree of control exercised at the Road Test. further

research and analysis was demanded to resolve the apparently fatal layer
coefficient variability represented in Figure 1. The term fatal is

emphasized because discriminating the structural quality of different

asphalt mixtures when the sinqle asphalt mixture at the Road Test

resulted in sucn great variability seems funis. Consequently , after

checking both the methodology and the computations which were found to-

be substantially correct, a copy of the original Road Test data for the
flexible pavement sections was obtained from the Transportation Research
Board. This detailed data-base provided qualitative improvement in the
data being analyzed.
 69

CO

1
»

O
o
<
m
05

\ r^jij
co
LL.
'
.1

• UJ
i
- O
/
1

o !

O UJ
>-
1 :

HI /'
/

< /
/

en /'/
tn

fy

~\ <\J

V UJ ^^' ,s
<
m S* '

.
-'

\
/
00 r^ CO LO CO OJ

A0N3fK)3Hd 3AIIV13H

Figure 1. AASHO Road Test - Layer Coefficient Distributions

 70

3.H Remarks

In the remainder of this papers the raw? or unweighted traffic data will
be used? since the weightina function is somewhat arbitrary end would be
difficult to transform for application to sites or climes other than
Ottawa. Illinois.

For refernce. the final relationships used by AASHG after the final
regressions are given for the standard 18— kip single a;:ie (standard
axle) configuration:

w'eiohted:
- p~l
tn.B
ioo
to
(w; = 9.61*Iog (bN +
"10
i > - u.t +
. ,
*—----
lOVti
'
— (ta>
O.H +
1P
(SN + 1)="

unweighted;

ioaJ
io
(W) = 8.94* loq
~io
(SN + 1> + U.d5 +
- ,
i3«Ei±i 1 4-0.1.55
v.h +
8
(SN + i)

<+. ROAEi TES i DATA

The remaining Road Test data held by the Transportation Research Board
is a reduced set of the original data. Many files? folders and card
decks which had not been used have been deleted, the remainder,
transferred to computer tape have been preserved. Of particular value?
the results from the full factorial experiment have been retained.

The data from the full factorial experiment were reduced to permit
easier manipulation? the data fields qiven in Table h were maintained.
 71

Table 4. Retained AASHO Road Test data format

Field # Description Field # Description

i Lood No. 7,8 Extent of Cracking

2 Lane No 9,10 Extent of Patching
3 Section No. 11,12 Slooe Variance
H Surfacing Thickness 13, 14 Rut Depth
5 Base Thickness 15 PS I
& Subbase Thickness 16 Accumulated Traffic
17 AASHO Day

In this file, complete observational data is available for each section

for each AASHO day (1 AASHO day = in real days). Not only can the PSI
(present serviceability index) vs. Traffic history for each section be
reconstructed, but the component parts of the PSI (cracking, patching,
slope variance and rut depth) can be examined. Not well recoqnized are
the facts that the amount of patching at the Road Test was essentially
negl igeable, and equally, but of great importance there was no r-oxttine
rrcai nter tance tmd.ertG.heTL on the pavement sec t ions such as might be
expected on in-service hiqhwav pavements.

After reviewing the data file all observations were removed for sections
after overlay, Data from overlaid sections was not used in the original
layer coefficient determination. Subsequently « plots were prepared for
PSI vs. Traffic, PSI v oo (Traffic) and PSI vs. AASHO dav. Examples
of these plots are given in Figures Ha, 2b and 2c. Examination of these
figures reveals that the AASHO model provides a very poor predictive

model. In addition, both sets of data (observed and predicted) were

submitted to analysis through the SPSS statistical package. The "Runs
Test" of this statistical package returned a z— statistic of 16,

(typically a value of less than 2.0 might be expected). Thus while the
overall regression might appear reasonable in terms of the r*st-atistic,
me model is ccnerwise inappropriate, une values ot tne r sxatistic
pertaining to the Road Test are not easily found in the literature,
however the Authors believe that the values are 0,23 for unweighted
traffic and 0,49" for weighted traffic).
 72

AASHO ROAD TtST

Model Investigation

139
5

f 577
1
fo_
2
4<

^f&
n«.
f£pt>ft» jj °
»
a«aaoU ° ODOD
DO
•»
\ -°^ « 453
•
, ^
3
£
* AA « 455
^
•AAA A ^
2 AAA* # 303
O

C1 200000 400000 600000

Axle Repetitions
(a)

AASH0 ROAD TEST

Model Investigation

139
5 Figure 2. AASHO Road Tes:
a t 577 Serviceability
*

^"*«B
a. Plots
4 f pvr** ntS.
gs&ffl,«
^*£2 5^ * 453
1-1
*
3
a # 455
aa
2 £ « 303
O
1

2 5 3.0 3.5 4.0 4.5 5.0 5.5 0.

Log (Axle Repetitions)

(b)

AfiSnO tiunD TcST

Model Investigation

t 139

t 577
«fca u
£&»**$,

e 3
••'
*

r
'oo
b? poa

A
jjd Oj

bt"
*
cucPc a a
M '
« 453
*
# 455
~A^
AA^A
2 v
« 303
O

() 5 10 15 20 25 30 35 40 45 50 55

AASHO Day
(c)
 73

A review of Figures 2a, 2b and 2c reveals that the shapes of the curves
are not as would have been expected. Instead of being smoothly
downcurvmg ? the data appeared to turn sharply downward at either one or
two reasonably well defined events. This was apparent in both the PSI

vs. Traffic and PSI vs. loq (Traffic) plots* although more pronounced
~io
in the former. In the PSI vs. AASHO day plot the locations of these two

critical events were particularly noticeable? and were found to

correspond very closely with the periods of spring-thaw.

Not all sections showed this tendency; however in those that did net? a

downward step was noted occurring at the same times. On the PSI vs.

AASHO day plots the curves apoeared to be generally piecewise linear,

i.e.. the PSI decreased linearly at one rate and then linearly at a much

increased rate. In some cases, this decrease in F'SI took the -form of a

step function.. the original slope being re-established: these

observations are shown schematically in Figure 3. Figure 3 oulines three
observed performance patterns:

a) P-^A-*!.? (rarely observed except on very thick sections and/or

very light axle weights), exhibits no significant distress. Figures 2

The slope of the line (rate of decay) is clearly independent of traffic

volume (axles/day); it is equally independent of climatic (freeze/thaw)
events

b) P -» A-»C-» D->E , a number of sections showed this pattern. where a

distinct and sudden loss of serviceability was noted (A'+k' coinciding

with the first spring/thaw. Some of these sections cracked but the
majority survived the first year uncracked. Those sections which
survived cracked failed at the onset of the next winter. while the

uncracked survivors generally failed during the following freeze/thaw

event

c) P-'A^E' a large number of sections exhibited this behaviour,

failing rapidly and catastrophically during the first freeze/thaw event.
 74

co

oa ,
o
CO

LU
1—
co
o
- o
o Cvj <c
•H
a
<c
+J
ItJ
LU
>-
o E O
0)
X
o
X o
co
CO

CO <:
<c
<

- O
o li_

a.

I I
I 1 ! o
u) ^- CO C\J t- c3 <
M
CO
a •H
(

Li-

Figure 3. AASHO Road Test Performance Schematic

 75

The overall linearity of these plots was particularly puzzling since no

suggestion had been made that time was a significant factor. which is

implied by this result. In fact. axle repetitions (traffic) were

considered the primary forcing variable. A check confirmed that the rate
of trafficking (axles per AASHQ day) over the period of the Road Test
was not constant, however the changes in trafficking rates did not

coincide with the changes of slope in the plots. Table 5 shows the rate
of section failure (d < 1.5) is not well correlated with traffic. but
'
t

is highly correlated with the season of the year. Thus, time as measured
by the number of spring/thaw events appears to be of greater
signif icance than previously suspected.

Table 5. AASHO Road Test: traffic and Failure Rates

Season % Traffic *'*

Failures

Winter 33 38
Spring Eh 60
Summer 30 3
Fail 23 10

In extending the piscewise linear hypothesis from the trafficked

sections to the untraf f icked sections of Loop 1. only three particularly
thin sections displayed the same pattern. As a reult. the Authors
concluded that the interaction of time and traffic might be significant.
the relatively sudden serviceability loss observed during the
spring/thaw periods was addressed by examining the component parts of
the FBI. The examination revealed that this phenomenon was paralleled by
the initial observation of Class 3 (alligator) and/or Class 3

(granulated) cracking (Figure h) . The full set of observed data from a

number of sections was plotted (Figure 55 (instead of only those at p =

3.5 5 3*0. 2.5, 2.0 and 1.5 as used in the original AASHO Road Test
analysis). This full set of data clearly reflects the piecewise linear
relationship contrary to the expectation of Eqn (3). The points of
intersection of the linear portions of the plots closely match the
initial observation of Class 3 and/or Class 3 cracking* In practical
terms? two populations of pavement performance have been observed; (a)

integral? uncracked and still sealed pavement? and (b) cracked,

disintegrating pavements. Since at the Road Test, most Class 2 and/or
 76

Figure 4. AASHO Road Test Effect of Cracking on Serviceability

 77

oi
co
r-
r-
CO
£
to O
£2
t- d m <

o
ID

a^Sa,
H A_i_
H
10
LO

< c

-I
LO
c to

h-
o
•H > *>> J
r 4
c
o
H
00 4J
+>
la
•H
«»>JL* Li
T
LO •H

a
<:
•M
LO
* D «
ai
Q.
o
oc
0)
>
4
CU
or

o
X Mc 1_B o ai

CO i— X
<: CD <c
"O
<c
o o«
n en
o
LO

OC D CO

-^ -*-«-
o
CO

LO

CSJ 00 o CNJ

r i l" r
I

CD

Figure 5. AASH0 Road Test Section Serviceability Histories

 78

Class 3 cracking was initially observed at, or about, a PSI of 3.0 to

3.5, it must be stated that the final AASHO equations (^a and hd) are
calibrated to failed, or failing pavements.

5. CUMULATIVE SUM (Cusum) PLOTS

If, using the observed data from any section, the cumulative
seviceabi 1 ity loss (ASL) is plotted against AASHO day, a piecewise
linear graph will be observed (Fig. 6). (In Figure 6 the serviceability
i =t
loss is given as (5.0-p(i)), and the Cusum as £ (5.0— p(i)), where t is
i. =i
the number of AASHO days from the start of trafficking to the point in

time considered). In the following analysis, serviceability loss is

defined as (p -p ( i ) ) . Close inspection indicates that the relation

rapidly approaches linearity on each leg, i.e., that the data exhibits
linearity asymptotically.

If the expected serviceabi I itv loss at any time t is represented by a

re-arrangement of Equation (1)., then:

p - pviJ = Serviceability loss = \p - p i'\ 1

(5?

Then, re-creating the Cusum plot mathematically:

AASHO Day Incremen ta 1 Loss Accumulated L oss

1 \p - p )• f ,

(p - p
.
)•
wan"

E \p - p )• (p - p i •

° l
L p J ° l
\

I J3 J

. fW(n)")^ , .
fW( i)^+W(E )^+..+ W(n)n

In order to approach linearity, (Cusum(n+1) - Cusum(n)) must approach a

constant value in the limit as n increases. Assuming for simplicity that

the traffic rate is constant at w axles per AASHO day (i.e., that
W(n)=n*w), then:

Cusum(n+1) - Cusum(n) = (p - P )• .„"

J-j (6)
 79

AASHO ROAD TEST

Model Investigation
Loop 6
75 1 i i

70 o 6/332/5-9-12
65 .. . — o
I

""'" " ,9
60
t
bb
50
I

d
k i

45
I
o
d
i

i
rf
j 40
)
L.

o\ ^
« 35
i

< I a
30 ,

25 \p

20 \
0° I

15
i oi° i

I o
10

5
*e£3
^
J

10
0°
y*jr

15 20 25 30
i

35 40 45 50 55
laosho/bjc Aasho Day

AASHO ROAD TEST

Model Investigation
Loop 6

o 6/332/5-9-12

•°=f<t\005-1
* i i

tTi-tTh^ i

i
U
i

i
iM i

I I

J_

5 15 20 25 30 35 40 45 50 55
aasho/bjc Aasho Day

Figure 6. AASHO Road Test Accumulated Serviceability Loss CASR

 80

If this expression is to be constant? then it may be simplified thus:

n
n' = constant (?)

This expression is only true in the limit for all n, as n approaches

infinity if ft is vanishinqiy small. For a range of structural number,
SNj of 1 to 6j the corresponding range of ft in Equation (4b; is found to

be O.h < ft < 300= Thus? the piecewise linear relationship observed in

Figure (6) cctrtnot be supported by the AASHO model.

The implication of this analysis is that on linear portions C/3~03 of the

Cusum plot, traffic (axle weight and repetitions) has no effect? and it

is only during those periods of time when the Cusum is in transition

from one linear portion to another (ft&Q) that traffic has a significant
effect on sers'iceabil ity

The mathematics of the Cusum transformation have not been developed

sufficiently by the Authors to permit the derivation of the relationship
which gives rise to the observed asymptotically linear Cusum function.

The smoothing effect of the Cusum transformation can clearly be seen in

Figure 6 and was used by the Authors to help identify the critical
points (changes of slope) in the behavior of each section.

Figure 7 superimposes the Cusum plots for a number of different sections

on the same graph? the axle weight was the same for all of these
sections. While not overly evident, it can be seen that (i) until
cracking (a change of slope) separates a section from the main plots the
behavior of all of the sections is essentially the same, and (ii) in

general, the thicker (total pavement thickness) pavements survive the

longest before cracking, and Eiii) within pavements of the same or

similar total thickness, those with thicker surfacinq t survive lonq-er.

i

6 . SURv I vAL PROBAB I L I TY

A probabilistic analysis was made in en attempt to quantify, or model,

the step function reported above. which was identified as being
triggered, or initiated, by a spring/thaw event. On any given lane of
 HI

-+ CD
1
OJ
'+

B +
..
1

'0 '+

E i + i

n -i- LO
1

1
OBH
1 — ^i\ !

^r 00 o y
CO o
i
! i

eg CO CO
1

«?;\
•;£''# <
J-
1

OJ
1

^r
1

co
1

a
C0 1 1
-
Mi

CD ~— CO
LO IO CO CO li
SO
H D
1.
<
LU
o i 1

o o o ^T 00 CO
-CO CO
1

1
CO o CO
1

— -....
% LO

2 ^
1 1
1 1

CO CO

1^ CD — CO LO \
LO LO LO •*
- , *
*
-

* B +
i

\
oo
. 1
I

o
1

O
CO
LO
CVI
O
OJ
LO to

ssoi Ainiev3oiAaas 3AiJLVinwno o

Figure 7. AASHO Road Test ASL Plots

 82

the Road Test (constant axle weight and type) pavements of various
composition either failed or survived the first full seasonal cycle of
spring/thaw. By trying to relate the elements of structure to the
probability of failure (or survival)? certain conclusions could te
drawn, i.s, ? while the thicknesses of each individual layer provided a

small measure of correlation with the probability of survival? no factor

was found t-o be as significant as the effect of the total physical

thickness (surfacing + base + subbase) , regardless of the layer
thickness combinations.

while total pavement thickness was particularly significant in this

analysis? it is evident that for pavements of the same total thickness,

those with thicker sur+acing have an enhanced probability of survival.
This has been demonstrated by plotting? but the mathematical and

statistical analyses are not yet complete.

An example of the first year survival matrix is given in Figure 8 for

Lane i of Loop h '18-kip single axle), where i = survival and =

failure (did not survive), (the italicized numbers give the full

pavement thickness for each section), the composite survival regression

curves are given in Figure 9 for ail the axle weights and types used in

the main factorial experiment.

These observations serve to validate the concept of the US Army Corps of

Engineers method for frost design wherein the pavement is first designed
from purely structural considerations, and subsequently the total

pavement thickness is checked against the anticipated depth of frost

penetration.

From the limited data, available (two seasonal cycles) the data tends to

support the possibility that among survivors? the probability of

survival is Markovian, thus if a pavement has a 0.95 probability of

first-year survival? then its two— year survival probability is (0,95) ,

and its n—year survival probability is (0.95) .

These curves (Fig. 9) are of course only applicable to the climatic and
subqrade conditions of the Road Test. If the subqrade were
 83

AASHO ROAD TEST

1 — Yr Survival Matrix

LOOP: 4...

LANE: 1....

LOAD:..18S...

|Sub-b ase
I
Base
^ XX >
\\\
.
I Surfacingx
4j 12

// 15
w 14 1 18
13 17 21
12 16
11 00 15 Iff

14 16 22
13 17
12 16 20
1 1 15 19 23

1/2

7 8 9 10 11 12 13 14 15 IB 17 10 19 20 21 22 23

Total Thickness (in)

-Figure 8. AASHO Road Test One-Year Survival Matrix

 84

., ,

\
\
X
|

*
V
^*,
1 o
'
X <^ I

•-J?

^^
J
*"**••«
£^ '----
---^.
•-.^
^^T"
v
A
*^
•**.
"'
X
\
V
CO \

> \X ^i*-^
;
\

o
>
\\

x^x
^
5S|
- ^
"
""

3
CO ^^ :y
^"-x.J

>-
v. \\
»\
^s
£«x>
*5 V*.
^=js-
—SSJ ^J> _
\
in
X
X 1
X
**
«w
I

"**""
».... \
1

'""-x X \\
X
X *
\
X
\
\
UD
-

o
o com ^rocvj o
—
od oo r- - ° ^
O CD CD O O O o o o o o JO

Figure 9. AASHO Road Test Survival Probability Plots

 85

free-draining? not frost-susceptible or if there were no frost., then it

might be reasonably expected that these curves would translate

significantly to the left.

7. SUMMARY AND CONCLUSIONS

7.1 AASHO node I

It has been shown that the AASHO model does not represent the observed
behavior of the Davements trafficked at the Road Test, The AASHO model
(Eon, (1)3 is biased to more closely represent the behavior of crach&d.
pavement. Within the AASHO model? the laver coefficients are shown to be
seconaajy regress ion co&fftc tents with no direct physical significance
To attribute to them a significance as indicators of strength is

spur i o us , Instead? the layer coefficients are indicators of resistance

to serviceability loss.

In the original development of the AASHO model? no cognizance was given

to the effect of climate (including its effect on the subgrade support
characteristics)? i.e.? the effect of climate was assumed to be constant
from section to section. A tacit assumption was made at the Road Test-

that ail deterioration in pavement serviceability was due to the

composite effects of traffic (axle weight and frequency) and pavement
structure (materials and layer thicknesses).

However, the present analysis suggests that at Ottawa? Illinois? the

effect of climate was in fact decisive. The initiating event of all
significant deterioration in pavement serviceability was inevitably
linseed to spring/thaw, the subsequent performance of the trial sections
was found to be critically dependent upon the observation of Class 3

and/or Class 3 cracking.

The effect of traffic (frequency) is most difficult to define. However?

results of the current analysis indicate that the frequency of loading
is critical only during the spring/thaw periods. At other times of the

year? the effect- of traffic (axle weight) is explicitly clear in the

AASHO model (Eqns. 3a and 3b)? however the effect of the axle weioht is
 86

seen to be negligible except in relation to the survival probability,

which is significantly effected by the spring/thaw events.

7.2 Alternative Analyses

7.2.1 Cusum analysis. The Cusum analysis outlined in this paper

provides a clear method whereby the performance of Road Test pavements
may be anal/zed. The surprising smoothing effect and piecewise linearity
lend themselves to the identification of changes in performance. It is

anticipated that the study of the Cusum plots and their mathematical
basis will provide a more defensible foundation from which to build an

alternative pavement performance model.

7 2 E
= . Probabilistic analysis^ The variability in the factors and

parameters- associated with pavement design and performance ? however well

controlled, lends itself readily to a probabilistic analysis. The first

analysis presented here clearly demonstrates the power of such an
approach.

The new AASHTQ Guide (1986) advocates the use of reliability concepts,
the application of the principles of reliability to empirically derivied
deterministic formulae (the AASHO model 5 is fraught with problems in

implementation and interpretation. The application of probabilistic

methods to pavement design would be better served by ^ total r^—arcctlysis
of the Road Test data (and other data bases) from a probabilistic basis?
in this fashion the full model (and its sub-models) wo7\uld be
internally consistent and far more transportable.

7.3 Conclusion

The Authors have presented the preliminary results from a study which
has highlighted shortcomings in the AASHO model and the interpretation
of the Road Test data. As far as is possible? they have sought to

provide both observational and mathematical justification for each point

raised

The Authors s t rangly recommend that the AASHO Road Test data be closely
scrutinized and re— analyzed in the light of twenty five years of
 87

hindsight) newer pavement technology tools and the more recent concepts
of probabilistic analysis and reliability. In this way new (and betterl
models and sub-models of pavement behavior and performance may be
developed
 REFERENCES

AASHTO Guide for Desiqn of Pavement Structures * 1986

Shook, J.F., ar.ti Finn, F.N., "Thickness Desiqn Relationships fm

Asphalt Pavements", Proc if I nternationa l C onference on the
Structural Design of Asphalt Pav ements Ann Arbor, MI, 196E,

Odemark, M. "Lindersikning av eiast ici tetsegenskaperna hos

, oliki
jordarter , samt teori for berakning av belaggningar eniigi
elast ici tetsteorin" Staters vaqirsti tut meddelande 77, 19<+9.
,

me HHbriu h;oao i esx ; h;eport d - ravemerrc Kesearcn

Research Board, Special Report fclE, IvfcEl
 APPENDIX B

AASHO ROAD TEST LAYER COEFFICIENT DISTRIBUTIONS

 8-9

AASHU RuAD TEST LAYER COEFhICIENT

DISTRIBUTIONS

i'he AASHU model mav be stated;

nunasr dt reoexiTionB dt a;:ie wsicirc i_

and axle type L (1 = single* S = tandem)

serviceability index.

pf.ft = parameters

j'ased on the AASHO Roao Test analysis tor weighted traffic

parameters p« /3 srs given by:

ft
= 0.4 + — .-. -.-,, .

"^
.,
^
,5. i£>
V
,
,3.

3 .
23

23

(B35
,4.7P
\ t_ + L )
1 2

"2 "33
i'i "2
-^. .th
xne layer costt icieiTC dtr rne
.. ,
material ,iaver
, . . .

iU _ . , ,. .th . , . _ .. . .
 wo

Thus, the comolete model may be written:

= 5.93 + 9.36* logJ (D + 1) + h.33* looJ (L )

iu io z

i<H 2.7 J
l 7* 1 '"
Q
U.VBi V L.

1 V , 3.23
1_
2

>°U^]
,,S.1P , 3. 23
(
D + i> L
2

usinq the values of Ion w and d tabulated for each section in HRB
SR 6iE: Reoort 5? Append!;; A? pp Eh<+ — Sh8 ? the expression in

equation B5 above may be solved for by a non-linear least-squares

regression analysis (e.g.? IMSL routine ZZSSGn . Thus for each
section .
: en estimate of the thickness index (D)j as well as the
individual laver thicknesses (t ) a ,_ e known.

A linear regression >SFSb routine KhbKESSIUN) of the estimated

thickness vnde -
(D; . or the Structural Number (SN = D) . acHinst a t
l i

wnere tns z are Known j :he entire 28^ sections

2 2 3 3
was then performed. This yielded the estimates of the layer

cobtt Kserrtb; nd provided estimates of tne standard deviations

c(a 5 .
 APPENDIX C

ROSENBLUETH POINT ESTIMATE METHOD

 91

RQSENBLLiETH POINi ESiiMATE METHOD

Conventional iy ? a distribution is represented by the mean value i/j)

and the s ia.ndard deviation \a) . However, if the distribution is not

symmetrical * or departs significantly from the normal, or gaussiarij

two further parameters snav be used; the sfeewness {ft 5 and the
i

kur tosis '.ft >

mese measures or cisxr louxion arise rrom me vanoui a .-; iosix ic

definitions of orobabilitv measure. Thus:

ffxl-dx = =
J 1
J
bquaxion L-I states that tne total prooabiiiry ot an event is ani^
or Probithe event will occur] + ProbEthe event will not occur] = 1

This definition is of primary axiomatic importance in portability-

+ co n

j».™-d..„-^.«

Equation C2 defines the mec;^! or e-y.pec tat ion? of the distribution,

and provide a measure of location. If only a single measure ie

available to describe the distribution? the mean? /u? provides thai

measure.

r , i z _ , , , z v

me variance? Derinea in equation i^j provices a measure ot xns

dispersion of the distribution about the mean value. The square
root of the variance is termed the sta.nda.rd. deviation. A true
Gaussian distribution is fully defined by the definitions C* thru

C3.
 92

Distributions which exhibit skew? or asymmetry? require a further

descriptive statistic? which normalized is given by:

+ co n
< , . 3 _ . , , _ 3 r , ,3
J 1 /_,
'
I
-co 1

The parameter ft is called the s, ?e-w? !

and is a measure of the degree
of asvmmetrv of the distribution. If ft — the distribution is
i

symmetrical 5 if ft > 0? the distribution is skwewed rioht (i.e.

i

lone~ tail to the rioht

~ of the isean! i and if ft < 0? the distribution
i

is skewed left (long tail to the left of the mean).

+ CO n
f . , 4 _ , , = 4 V- ,4

- CC 1

Similar! 1-',
the parameter f? is a measure of the fLatness* or
2
p&czfc.edne'Ss of the distribution? and is termed the coefficient of

kiirtcs;5, Vf ;3 < 3, then the d:tstr ibu.r-:- on as flatter than the

2
Gaussian distribution, and if /? > 3 the distribution is more peaked
2
than the qaussian.

The distribution parameters defined above are sufficient to define

most? if not si?.? p-Hctic^l -ummodal 5 distributions. However.; in

o ? and verv occasional Iv? the skew, ft , are known.

i

If these parameters are known for a variable? say x? (i.e.? fj\x) <

pixii ft '
x ) and ft ' '<'> ) ? and it is reouired to estimate the
'
1 2
distribution parameters of a function of x, fCxJ, then
conventionally? recourse must be had to a truncated Taylor series-

expansion of the function. This method is termed the First Order

Second Moment method (FQSM). For a function of a single variable?
this results in:
Z
_,. „- . i d fCx? z_. ._.
+ -•
. .,
-a CxJ
.

b[/(x?J = /j(x; \Vb)

 93

and

ft C xJ> -1
2
• <" x'J + a k x.J t
-fer-J

Z
d/Cx) d fCxy
+ /? CxJ-o CxJ-
r>

Thus it is seen that to make use of the FOSn technique? not only
must the distribution parameters he known, but also the first and
second derivatives of the function. This is not always easy or

feasible I

With a multivariate function, say fCx yJ, ? nc triji' iiiusx me

distribution parameters be known for both x and }
but also the

first and second derivatives of the function with respect to both x

and y, and the covariance COvCx,yJ cr the correlation coefficient

P>-,v . The resulting equations needed to define the mean and

» of the function fCx y5 ere
y not trivial.

The Rosenblueth Point Estimate Method (REM) provides a convenient

and computationally attractive method to achieve the same end

results. This method accounts for correlations between variables

the distributed vertical loading on a rigid, simply-supported

horizontal beam. The magnitude i;f the " ifjrjcp is defined by equation

Ci above, ard the center of gravity, cs ,! t o;d or ioca'ioi of the

resultant, of the loading diagram? is located by using equation CE
above. Equations C3 thru C5 nay be recognized as forms of a general

= a =
J "nM> ft
n-Z
-
lv<\
when Tt—5.: the term ft • a (or variance o -
3 represents the radius of
n—
oyration of the loading diaaram about the resultant. Other values
 94

ti-.
of n lead to representations of the n moment about the mean. or
resultant
Since the "beam" is simply supported} the the foi lowing

relationships may be stated (using the discrete forms of CI - Ch):

{ C?3 )

(C9b)

\ -a - ;j<. x fjC = rv CX> ( C9c

- i,r
Uf. v
XJ
~>
) "I" D * \ X - C xj • o' '-.
xj v C9d 5

where p AT'.d d reoresent either the maoni tunes of the "beam"

T-ne to i i dw i no

relet ionshios are found:

/? fx;
1 +
L i

; qn used <±) ;.e the opposite i'o that of /5 Cx)j

i

= LICXJ + O-C
/^JK
- !>( X > -

snots uhet it ine Qis"criDUT-ion m CxJ = n

these become;

= /jCaO + c<xJ> ', L i i D .

= L,CX^ - ctCx; (Ci ic'

lx snouid be rioted mat no information has oeen ios*;, it the

distribution is characterized by the mean ar\c variance? then two
'pieces' of information are available; two ^pieces' of information
remain after the above transformation! i.e.? Co >-. 5 and (p ? h 5-
 95

Example 1

Given that /j<TxJ> = 35 and cCxS> = 5 , what is the distribution 0'

<pCx2 = sinCxl? ?

Given only this information, we mav assume that ft C xj = 0. then d

i •
<-

d
:
—
=0.5, and ;:
+
= 35 +5 = h-0 • and x
—
=35 - 5 = 30 . Bo tha

cb = cbC x .y = sin (40 ) = 0.6486

-t +
rp = (pCx = sin(30 > = 0.5000
E[0] = fj (<£.) = p. <p + p 4> = 0.5* (0.6428 + 0.5000) = 0.5714
_ 2 2 2
Ll<£ J = d <£ 4- p = CJ

y
z
\co-
—
= tL<p
,a
2 -
j

aicp

ihus the distribution parameters for the function cfi

= f'CxS> have beev

found? i.e., /j <p) = 0.5714% and c\<£) = 0,0714-.

ample c

biven that i_<C~xl> = cs5 and aCxS> = 5 and ft Cxl> = 1 « what is

i

distribution of <pC x„> = sinCxO ?

Given this information- then bv ClOa: d = 0.H764 and o — 0.7J

+ ' —

+ ' +
cp = fC(p j = sm(3i ,91) = 0.5236

Ei'.cpl = fj\cb) = p cp + P $ ~ 0.2764*0.6831 + 0.7236 'CSESb =0,571:

^-,2- .2 ,Z . , ,, . ,,__,2 - _, _ ,_,,_, , „„..„
tL<z> j = D <p -f D - <Z> = U.d7fc<+ • O.tddi + i.j
. /cab ' v.jdBti = O.ddic
:

+ + ' —
a 2,
\q>)
.,
= r-i-J.2-
ti0 j
"J.- 2 = u.j3ic;
- tL0j r- ---0-" - -- =T"-0= 2
i0.cs/i3j -
= u.Uu-Ha .------

aisxr lDUT-ion paramaexers tot me Tuncuion or

An extension of the same principles leads to the solution of the

case of multivariate functions, say z. = /Cx,yJ> . As before-, the

 96
individual point estimates (i.e.? >; j x ? y and y ) are found usinq
equations CIO above. The the relationship may be written:
n,
,_,.
tL2J = DS n +D2
'++ ++ — +- '
H
n
-i-ps n
-+ - +
+D2 n

in inis case? s ;:

±±
= fCx ± >y ± l> ^d P++ = P __ =
1 +
—p •. <

D = D where p is the correlation coefficient

between x and y

If the dependent variable is a function of n variables. then S

values are required to provide the distribution parameters j and the

correlated p values ars found as follows (hers n=3)=

r, = d

p12

The above is a very simple outline of the Rosenblueth Point Estimate

Method, h more complete development is give'; in ::
Re 1 lab:-. 1 i ty-3aseri

Design in Civil Engineering"' by M. E, Harr? M~Sraw— Hill Book

EXAMPLE from PROJECT

tquanon iV gives The uiiinx: jt xne

relationship for binder st if fries: .5 aiven

JS_„ -7 ,0.3.58 -p:

, *
^

is the Binder Stiffness (MPa

is the time of loading (sec)

 97

PI is the Penetration Index

T is the Ring and Bail Softenina Point ( C)

RB
T is the material temperature ( C)

From Table 17? the IDOH time of loading (All Highways) is given a;

0.0108 seconds, with a standard deviation of 0.0009 seconds.

: ne fsnetratior, inaex, hi f ana rang ana ee i i softening Poittcj

R B
may be found from equations SO thru 22? which are functions of

initial oenetration? pen , From Table 15? the distributee;

i.

parameters of the initial penetration are given as: EEqen 3 = 64

Js:<ng the pavement temoevature d-.str:<bution parameters (A 1

Trom zaoiB d<j*. i.e.: tL i j = ii ana ciu = i d . i t

ins two varisoiesj time ot loading, r- ? ana pavement temperature?

are assumed svmfTietr ir = l (since 5Kew;~ess = fi
1
• is ze~o ) and mav
reduced to their ;-; and x values? thus:
t = EEtj + ci t 3 = 0.010B + 0.0009 = 0.0117
+
t = EEt-3 - citj = 0.0108 - 0.0009 = 0.0099

and
T = EET3 + cET3 = 12.47 + 18.12 = 24.59
+
7 = EET^ - cE t j = 12.47 — 12.12 = 0„ 35

>mce rne initial Denerraiion? oer. element ot skew {ft

,./ d /p
pen j*y

= 64 + w o j = db. at
 98

/ j. ,043/0.5957 = 60. 7(

since rne Deiisira- ; on innex* n , arc me BOTteniinc poim;. i

RB
• are

functions cf initial penetration) pen ? the Following itsv be fou ~;r

bv direct substitution ot the above results in Equations E0 thru EEi

l+ r+ rb+ +

and

feWi to; j.Dwina the Ecsenblueth PB-1 method above; (all S are in HPa)

„ , j_„ ..-.-7,.-. ,.,,„,-0.3c'>8 O. 1435,;.;. , „, , 5 _ <Q ^„„i^

~'b+++
_ , . ,_,, .-.-7,... ._ ., , ^ -O. 3<3S
, _0. 1-435 „_ _. , „ ... 3-=-'« 5 — -D=r: ce^"
"b++-
_-7,_
.
JJ _.-O.3<50
, . O. 1221, „, -„ _._ .5
=
_. -„_ .

3 = l.ia'/xiU vu.Uii'/) *e ij/.U/ - ci^.bVi cn.hitiDn

b+-+
. „ „_ 5 _ . ....
, O/ - U.JS) = dV^t.M-llH
,

.--7,. 0.1435.^^ __ 3
= l „i —
„ -,_^, -O. 3<58 .„ _._ . _, - _ . ._
3 j'/ i. i.ii.uijvv! v: *B
i i dj , c j - dt-DYi i = c! l .'-_•' -3 1
"b-++
.,.
M _-7 . -O. 3<58 0.1435.^^ ,_ -
, . __, 5 __, _ ,_ . .

3 = I.IdyxiO iO.OOVVj
,

*e Ijj.cj O.-Jb; = js/6. /V it

b+-t -
„ , ,_„ .,.-.-7,^, .-,-._.„.. -O. 3<5S O. 1221, ,_,, .-,„ „, r.-./S _ Q= . -_-,,_,

\>+-+
,.-,, -O. 3<58 O. 1221

-,2 ,2
(Loa (h
% b
'"'
"' '
"b
'

b
5

19.5775 3b3, E^ , 1 . E9 I X . 6686

354 3E55 . 1E5546, 56 E.5494 6,4995

£4 . £856 5S9, .79 1 .3853 1 .919E
394,41 14 » 5556(5, 35 d b95¥
. 6 7390
.

El .0315 44E 32 . 1 . 3EE9 1 . 7500

376 . 79 1 1 6.1 Wl'x 76 E 5761
cf5 . Bc:tj4 666 ,95 1 .41E0 1=9939
4-19.4190 17591 E. 30 E.0EE6 B7 83
-

i 635 6673 . 60 1 07 3 .31 15 756u = 34 084B

.

i-roii! wnicn:
ElS .', = E04-.5 HPh 0.4 Dss- cCS D = 1.84.6 MPa \'i. ,.27 psi
 <)<)

EtLoqCS )3 = 1.9695 .^[Lcg(S )! = 0.6178

b

Two facts lead to the choice of the logarithmic solution? (i)

equations 54a and 2hd require the use of Loo(S > rather than 8 . and
' b b
(ii) use of the arithmetic option presumes that S can be negative
ifj - 3<y - -349 MPa), while the logarithmic option prevents this

absurd i tv

This latter example of the Point Estimate method is oniy one smsn
part of the full analysis? but it clearly desionstre bes the power of

the method* and highlights- the overall computations

used with an intrinsical iv non-iinea:' multivariate ; >: or ess ion.